<%BANNER%>

A Spatial, Temporal and Deterministic Analysis of the Economics of Conflict in Haiti

Permanent Link: http://ufdc.ufl.edu/UFE0045083/00001

Material Information

Title: A Spatial, Temporal and Deterministic Analysis of the Economics of Conflict in Haiti
Physical Description: 1 online resource (201 p.)
Language: english
Creator: Engelmann, Jens
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: acled -- analysis -- conflict -- disaggregated -- haiti -- spatial -- temporal
Food and Resource Economics -- Dissertations, Academic -- UF
Genre: Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This research examines the spatial and temporal patterns of conflict and political violence in Haiti’s 133 communes during the time period 1997 to 2010. Using the publicly available Armed Conflict Location and Event Dataset, statistical and geospatial techniques tested for patterns and causal relationships at a disaggregated level. The temporal analysis of conflict identifies two unique temporal periods of conflict between 1997 and 2010. Haiti experienced a civil war period with elevated levels of political conflict from September 2003 to June 2005. Prior to and after the civil war period, Haiti experienced continuous low-intensity conflict with political conflict remaining part of Haitian living conditions. The spatial analysis of conflict indicates that political conflict and violence is spatially clustered, and primarily occurs in the urbanized areas of greater metropolitan Port-au-Prince, Gonaïves, Saint-Marc,Cap-Haïtien, and Petit-Goâve. Furthermore, as conflict intensified spatial diffusion of conflict occurred, with conflict intensity rising in the north of Haiti. Within the studied time period, demographic, political, and military strategic factors primarily impact conflict propensity in Haiti. The determinants of conflict are time variant and the set of determinants for continuous low-intensity conflict as compared to the determinants of conflict during civil war are dissimilar. The low-intensity continuous conflict has a mass population effect, where conflict is greater in areas with higher population. The determinants of the civil war are demographic, political, and military strategic factors. Contrary to other disaggregated studies of conflict, average wealth levels in a location do not explain conflict propensity in Haiti.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jens Engelmann.
Thesis: Thesis (M.S.)--University of Florida, 2012.
Local: Adviser: Sterns, James A.
Local: Co-adviser: Burkhardt, Robert J.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0045083:00001

Permanent Link: http://ufdc.ufl.edu/UFE0045083/00001

Material Information

Title: A Spatial, Temporal and Deterministic Analysis of the Economics of Conflict in Haiti
Physical Description: 1 online resource (201 p.)
Language: english
Creator: Engelmann, Jens
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: acled -- analysis -- conflict -- disaggregated -- haiti -- spatial -- temporal
Food and Resource Economics -- Dissertations, Academic -- UF
Genre: Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This research examines the spatial and temporal patterns of conflict and political violence in Haiti’s 133 communes during the time period 1997 to 2010. Using the publicly available Armed Conflict Location and Event Dataset, statistical and geospatial techniques tested for patterns and causal relationships at a disaggregated level. The temporal analysis of conflict identifies two unique temporal periods of conflict between 1997 and 2010. Haiti experienced a civil war period with elevated levels of political conflict from September 2003 to June 2005. Prior to and after the civil war period, Haiti experienced continuous low-intensity conflict with political conflict remaining part of Haitian living conditions. The spatial analysis of conflict indicates that political conflict and violence is spatially clustered, and primarily occurs in the urbanized areas of greater metropolitan Port-au-Prince, Gonaïves, Saint-Marc,Cap-Haïtien, and Petit-Goâve. Furthermore, as conflict intensified spatial diffusion of conflict occurred, with conflict intensity rising in the north of Haiti. Within the studied time period, demographic, political, and military strategic factors primarily impact conflict propensity in Haiti. The determinants of conflict are time variant and the set of determinants for continuous low-intensity conflict as compared to the determinants of conflict during civil war are dissimilar. The low-intensity continuous conflict has a mass population effect, where conflict is greater in areas with higher population. The determinants of the civil war are demographic, political, and military strategic factors. Contrary to other disaggregated studies of conflict, average wealth levels in a location do not explain conflict propensity in Haiti.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jens Engelmann.
Thesis: Thesis (M.S.)--University of Florida, 2012.
Local: Adviser: Sterns, James A.
Local: Co-adviser: Burkhardt, Robert J.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0045083:00001


This item has the following downloads:


Full Text

PAGE 1

1 A SPATIAL, TEMPORAL AND DETERMINISTIC ANALYSIS OF THE ECONOMICS OF CONFLICT IN HAITI By JENS ENGELMANN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Jens Engelmann

PAGE 3

3 To my parents and my sister

PAGE 4

4 ACKNOWLEDGMENTS I thank God. I also thank my parents and sister for their encouragement and patience. I am grateful for And res Garcia, and his persisten t advice and support I thank Dr. Sterns for his enthusiasm, and strong belief in me. Also, I want to acknowledge Dr. in his advice and guidance without I would have not been able to complete this thesis. I want to thank Dr. Burkhardt for the support and help he has given throughout my academic career.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 Overvi ew ................................ ................................ ................................ ................. 13 Motivation ................................ ................................ ................................ ............... 15 Problem Statement ................................ ................................ ................................ 16 Objectives ................................ ................................ ................................ ............... 17 2 LITERATURE REVIEW ................................ ................................ .......................... 19 A Brief Overview of Haiti ................................ ................................ ......................... 19 The Historical Context ................................ ................................ ............................. 21 Armed Conflict Location and Event Dataset ................................ ........................... 41 Country level Studies of Violence, Conflict, and Civil War ................................ ...... 46 Disaggregated Studies of Violence, Conflict, and Civil War ................................ .... 68 3 TEMPORAL AND SPATIAL PATTERNS IN THE HAITI ACLED ............................ 87 Temporal Patterns in the Haiti ACLED ................................ ................................ .... 87 Simple Temporal Frequency Analysis ................................ .............................. 88 Exponentially Weighted Moving Average Statistical Process Control ........... 90 Spatial Patterns in the ACLED Event Dataset ................................ ........................ 94 Analysis Structure ................................ ................................ ............................ 95 Simple Spa tial Analysis ................................ ................................ .................... 96 Geographic pattern of conflict and violence in Haiti between 1997 and 2010 ................................ ................................ ................................ ........ 96 Geographic pattern of conflict and violence in Haiti for three distinct time periods ................................ ................................ ............................ 97 Geographic pattern of conflict and violence in Haiti for different event types ... 98 Spatial Autocorrelation ................................ ........................ 99 ................................ ................................ 100 ................................ ................................ .......... 100 ................................ ................................ .......... 103

PAGE 6

6 Mean Center and Standard Deviational Ellipse ................................ .............. 104 4 ANALYZING LOCAL PATTERNS OF CONFLICT IN HAITI ................................ 136 Disaggregated Study of Conflict in Haiti ................................ ................................ 137 Determinants of Conflict in Haiti ................................ ................................ ............ 139 Population Size in a Given Location ................................ ............................... 139 Urban Population Percentage in a Given Location ................................ ......... 140 Male Population Percentage in a Given Location ................................ ........... 142 Adult Population Percentage in a Given Location ................................ .......... 143 Distance from Political Center ................................ ................................ ........ 143 Elevation Data in Haiti ................................ ................................ .................... 144 Border with the Dominican Republic ................................ .............................. 146 Departmental Capital in Haiti ................................ ................................ .......... 146 Distance to Route Nationale ................................ ................................ ........... 147 DHS Wealth Index Score ................................ ................................ ................ 148 Statistical Analysis ................................ ................................ ................................ 151 Covariate Selection ................................ ................................ ........................ 152 Model Validity ................................ ................................ ................................ 153 Determinants of Conflict in Haiti for Various Time Periods ............................. 155 5 DISCUSSION AND CONCLUSION ................................ ................................ ...... 178 Summary of Results ................................ ................................ .............................. 178 Suggestions for Future Research ................................ ................................ ......... 181 APPENDIX A EWMA METHODOLOGY ................................ ................................ ..................... 184 B ................................ ................................ ......... 186 C LOCAL INDICATOR OF SPATIAL ASSOCIATION SCORE ................................ ................................ ................................ ................. 188 D MEAN CENTER AND STANDARD DEVIATIONAL ELLIPSES ............................ 189 E NEGATIVE BINOMIAL GENERALIZED LINEAR MODEL ................................ .... 191 LIST OF REFERENCES ................................ ................................ ............................. 192 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 201

PAGE 7

7 LIST OF TABLES Table page 3 1 Simple temporal frequency analysis of Haiti ACLED ................................ ........ 109 3 2 EWMA statistic for Haiti ACLED ................................ ................................ ....... 110 3 3 ACLED event count for communes in Haiti between 1997 and 2010 ............... 114 3 4 ACLED event count for communes in Haiti between 1997 and 2002 ............... 115 3 5 ACLED event count for communes in Haiti between 2003 and 2005 ............... 116 3 6 ACLED event count for communes in Haiti between 2006 and 2010 ............... 117 3 7 ACLED event type distribution ................................ ................................ .......... 118 3 8 Spatial diffusion of protest in Haiti ................................ ................................ .... 118 3 9 Spatial diffusion of violence against civilians in Haiti ................................ ........ 119 3 10 Spatial diffusion of battles with no change of territory in Haiti ........................... 120 3 11 ................................ ......................... 120 3 12 types of conflict events in Haiti ............................... 120 3 13 ................................ ......................... 12 1 4 1 Correlation matrix of covariates and dependent variable ................................ .. 162 4 2 Variance inflation factor ................................ ................................ .................... 162 4 3 ........... 163 4 4 Determinants of conflict events in Haiti for various time periods ....................... 164 4 5 Determinants of Conflict Events in Haiti for Various Time Periods ................... 165

PAGE 8

8 LIST OF FIGURES Figure page 2 1 Political Map of Haiti. ................................ ................................ .......................... 86 3 1 Temporal frequency chart for Haiti Armed Conflict Location and Event Dataset between 1997 and 2010. ................................ ................................ ..... 122 3 2 EWMA chart for monthly Haiti ACLED event data. ................................ ........... 123 3 3 ACLED event count for the 133 communes between 1997 and 2010. ............. 124 3 4 Percentage of total events for the 133 communes betwee n 1997 and 2002. ... 125 3 5 Percentage of total events for the 133 communes between 2003 and 2005. ... 126 3 6 Percentage of total events for the 133 communes between 2006 and 2010. ... 127 3 7 Percentage of total protest events for the 133 communes between 1997 and 2010. ................................ ................................ ................................ ................ 128 3 8 Percentage of total violence against civ ilians events for the 133 communes between 1997 and 2010. ................................ ................................ .................. 129 3 9 Percentage of total battles no change of territory events for the 133 communes between 1997 and 2010. ................................ ................................ 130 3 10 ................................ ............................... 131 3 11 ............................ 132 3 12 Mean center and standard deviational ellipses for different time periods. ........ 133 3 13 Standard distance of north south direction in kilometers over time .................. 134 3 14 Mean center and standard deviational ellipses for different con flict events from 1997 to 2010 in Haiti. ................................ ................................ ............... 135 4 1 Population size per commune in Haiti in 2003. ................................ ................. 166 4 2 Urban population percentage per commune in Haiti in 2003. ........................... 167 4 3 Male population percentage per commune in Haiti in 2003. ............................. 168 4 4 Adult population percentage per commune in Haiti in 2003. ............................ 169 4 5 Mean elevation per commune in Haiti in meters. ................................ .............. 170

PAGE 9

9 4 6 Standard deviation of elevation per commune in Haiti in meters. ..................... 171 4 7 Route Nationale highway system in Haiti and commune centroids. ................. 172 4 8 Approximate DHS survey locations in Haiti ................................ ...................... 173 4 9 Estimated wealth distribution in Haiti using ordinary Kriging. ........................... 174 4 10 U.S. census database population estimate for Haiti. ................................ ........ 175 4 11 Estimated wealth value per commune in Haiti. ................................ ................. 176 4 12 Randmonized Quantile Residual. ................................ ................................ ..... 177

PAGE 10

10 LIST OF ABBREVIATION S ACLED Armed Conflict Location and Event Dataset CD Convergence Democratique DHS Demographic and Health Surveys EWMA Exponentially Weighted Moving Average FL Fanmi Lavalas FLRN Front National pour le Changement et la Democratie GDP Gross Domestic Product GIS Geographical Information System HNP Haiti National Police Lower control limits LISA Local indicators of spatial associati on MAGIC Mid American Geospatial Information Center MIF Multinational Interim Force MINUSTAH United Nations mission S Upper control limits VIF Variance Inflation Factor

PAGE 11

11 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Master of Science A SPATIAL, TEMPORAL AND DETERMINISTIC ANALYSIS OF THE ECONOMICS OF CONFLICT IN HAITI By Jens Engelmann December 2012 Chair: James Sterns M ajor: Food and Resource Economics This research examines the spatial and temporal patterns of conflict and political violence in Haiti Using the publicly available Armed Conflict Location and Event Dataset, statistical and geospatial techniques tested for patterns and causal relationships at a disaggregated level. The temporal analysis of conflict identifies two unique temporal periods o f conflict between 1997 and 2010. Haiti experienced a civil war period with elevated level s of political conflict from September 2003 to June 2005. Prior to and after the civil war period, Haiti experienced continuous low intensity conflict with political conflict remaining part of Haiti an living conditions The spatial analysis of conflict indicates that p olitical conflict and violence is spatially clustered, and primarily occurs in the urbanized areas of greater metropolitan Port au Prince, Gonaves, Sain t Marc, Cap Hatien, and Petit Gove Furthermore, as conflict intensified spatial diffusion of conflict occurred with conflict intensity rising in the north of Haiti Within the studied time period, d emographic, political and military strategic factors primarily impact conflict propensity in Haiti The determinants of conflict are time

PAGE 12

12 variant and the set of determinants for conti nuous low intensity conflict as compared to the determinants of conflict during civil war are dissimilar The low int ensity continuous conflict has a mass population effect, where conflict is greater in areas with higher population. The determinants of the civil war are demographic, political, and military strategic factors. Contrary to other disaggregated studies of con flict average wealth levels in a location do not explain conflict propensity in Haiti

PAGE 13

13 CHAPTER 1 INTRODUCTION Overview Conflict and economic development are interrelated terms of human suffering and socioeconomic decline is well known, and conflict is (United States Agency for International Development 2005) Conflict begets lower l evels of economic development, and lower levels of economic development increase the probability of further conflicts and civil wars. With the end of Cold War era, many experts on international security and conflict expected the level of conflict and civ il war in the world to decrease. The Cold War, in the opinion of experts, had propelled conflict and made it feasible to engage in conflict in the world. The United States and the Soviet Union provided arms and financial support enabling for armed conflict and civil war. Hence, most experts believed that w ith the end of Cold War, and its underlying struggle between communism and capitalism both countries would be less interested in supporting civil wars and conflict. However, the frequency of conflict and civil war has not necessarily decreased, after the end of the Cold War era. superpower competition in the developing world witnessed not less armed conflict but new and deadlier forms (Arnson and Zartman 2005) The root of conflicts and civil war cannot be easily determined for many of the observed post Cold War era conflicts and civil wars, but the most frequent explanation s of conflict are economic, social, political, religious, or institutional in nature.

PAGE 14

14 In 1804, Haiti declared its independence fro m France, the reigning colonial power in Haiti. Haiti has now been an independent and sovereign nation for over two hundred years. The birth of the Haitian nation was similar to the birth of the United States of America. Nevertheless, the contrast of what unfolded in the last two hundred years, in both nations, could have not been starker. Democracy, prosperity, and stability developed in the United States. Chaos, poverty, and instability developed in Haiti. Mats Lundhal, a renowned Haitian history scholar, th (Lundahl 1989) developed. The elite in the country function was altered from serving the needs of people to providing economic rent to the elite. As the state turned into a predatory state, the control over the state became exceptions, sitting governments lasted only for short periods. Coups, insurrections, and ci vil war took place with amazing frequency more than a hundred time s (Lundahl 1989) In 1957 Franois Duvalier was elected as president of Haiti Duvalier developed developed the predatory state into a full fledged reign of t error, using sheer violence to (Lundahl 2008) Duvalier abandoned democratic processes, stifled any opposition him, and led the country into serious economic trouble. In 1971, after the death of Franois Duvalier, his son Jean Claude Duvalier took over the office of president. Jean ly brutal and suppressive as his with focusing primarily on maintaining a stronger power base

PAGE 15

15 and suppressing any antagonism. 20 th century was characterized by brutal dictatorships committing horrendous acts of violence and oppression. Conflict and violence remained part of the Haitian experience in the beginning of the 21 st century as political instability and tension in the country remained high. Sadly, conflict, violence, and civil war have now long been part of the Haitian identity. Motivation There is an emerging literature on the economics of conflict, and the related structur al variables causing or correlating with violence, conflict, and civil war Researchers have investigated the underlying relationships between economic development, the emergence, continuation and cessation of violent conflict, the economic costs and benefits associated with violent conflicts, and policy interventions. Political s cientists, geographers, and economists attempt to broaden and deepen our understanding of the causal structures affecting the likelihood of intrastate conflict and violence in society. The World Bank spearheaded these renewed efforts to understand intras tate and violence in a globalized world, which has significantly been altered after the end of the low income countries and, evidently, reduces income even further Henc e, it is of concern for an organization whose mission is poverty reduction (Collier and Sambanis 2002) A discussion about economic development must encompass a deep understanding of the roots of conflict and civil war, the processes of civil war and conflict, the remedies agains t conflict and civil war, and possible interventions both prior and post conflict.

PAGE 16

16 (Heine and Thompson 2011) The level of poverty in Haiti is extreme. The infrastructure, education system, transportation system, and health care are all in disarray. The causes f or underdevelopment in Haiti are political instability, a predatory elite, and outburst s of violence, conflict, and civil war. The fragility of the state and the viole nt climate of Haiti must be one of the root causes of underdevelopment in Haiti. Hence, an economic development strategy for Haiti should encompass a robust understanding of conflict and civil war in the Haitian society Problem Statement I will examine and attempt to identify the determinants of violence and conflict in Haiti by identifying if and when various demographic, geographic, economic, and institutional factors are associated with different levels of violence. Furthermore, the spa tial and temporal structure of violence and conflict will be examined to uncover trend s and structure s of violence in the context of Haiti. The research uses the Armed Conflict Location and Event Dataset (ACLED) (Raleigh et al. 2010) The ACLED is designed for disaggregated conflict analysis and crisis mapping. This dataset codes the location of conflict ev ents in 50 countries from 1997 to early 2010. The data set contain s information on the date and location of conflict events, the types of events, and their specifics. The data set can be used in any spatial analysis or mapping program. The advantage of the A CLED, as compared to any datasets similar to it, is that it allows for a study of conflict and violence at a local level. The country is not seen as a homogenous entity, in which conflict is similarly likely in all locations, but the country is a heterogen eous entity with differen t risks of conflict and

PAGE 17

17 civil war within its national borders The ACLED provides disaggregated data for Haiti between 1997 and 2010, which enables a temporal, spatial, and causal examination of conflict and violence in Haiti. To examine the causal structures of violence in Haiti, the ACLED dataset will be linked with localized demographic, institutional, economic, and geographic determinants. Objectives This research is guided by the following objectives, Examine the temporal structure of the Haiti ACLED data. Assessing the variation in the ACLED data for various time periods. Examine the spatial structure of the Haiti ACLED data in order to understand the local, disaggregated context of violence in Haiti. Empirically test t he relationship s between determinants and the occurrence and variance of violence between 1997 and 20 10 Findings to date within the Economics of Conflict literature have identified various relationships and causalities between various determinants and vi olence and conflict in countries. Our research on violence and conflict in Haiti seeks to identify the particular determinants correlated with violence and conflict in Haiti. The goal is to extend the understanding of the dynamics of violence and conflict, while emphasizing the changing temporal and spatial aspects to violence. A broader understanding of violence and conflict in Haiti will assist policy makers in their attempt s to foster economic development in Haiti. Conflict and violence have historicall y been high in Haiti, and an understanding of the patterns of violence in Haiti will enable policy makers to preemptively reduce the risk potential for violence and conflict. Since conflict and violence are a determinant of economic development, the

PAGE 18

18 struct ure of conflict and violence must at least be considered in policy discussion for economic development in Haiti.

PAGE 19

19 CHAPTER 2 LITERATURE REVIEW The literature review in this chapter encompasses three distinct parts. The historical context of violence in Ha iti focuses in particular on the time period between 1997 and 2010. Violence and conflict occurs within a social, political, and economic context that has emerged over time Hence, any understanding of causal structure necessitates a deep knowledge of the historical background. The second part of the literature review focuses on the ACLED, which details the intensity of conflict and violence for different time period s and locations. Lastly, a review of the conflict and civil war literature in developing cou ntries is necessary. It provides the theoretical background to the studies of localized violence and conflict. A Brief Overview of Haiti Haiti is a island nation located in the Caribbean, being in western one third of the island of Hispaniola (Central Intelligence Agency 2012) as well as several nearby islands in the archipel (e.g. Gonave, and Tor tue) ; Figure 2 1 shows a map of H aiti. The Dominican Republic lies in the eastern two thirds of the Hispaniola Island and the Dominican Republic is the only country Haiti shares a border with. Haiti is a relatively small country; it is slightly smaller th an Maryland in size (Central Intelligence Agency 2012) in of Haiti is mountainous (Central Intelligence Agency 2012) The lowest elevatio n level in Haiti is at 0 meters, and the greatest elevation point is 2680 meters. Haiti is also frequently hit by hurricanes and massive storms. In recent years, Haiti has had to suffer through various natural catastrophes, such as hurricanes, earthquakes, and massive flooding In addition to natural catastrophes, Haiti is also one of the countries with the greatest

PAGE 20

20 deforestation in the world. The geography of Haiti poses significant challenges to the country. Environmental degradation the rugged terrain, and the overall climate pose significant challenges to the country. fractio population is black, while the remaining 5 perce nt of the population is mulatto and white (Central Intelligence Agency 2012) Frequently, developing countries have various large separated ethnic groups; h owever, this is not true for Haiti. Furthermore, there is little religious fractionalization in Haiti. 80 percen t of the population is catholic, and 16 percent is protestant (Central Intelligence Agency 2012) though a pproximately half of the population actively practices voodoo (Central Intelligence Agency 2012) (Central Intelligence Agency 2012) afterward. Haiti struggled with p olitical instability, and was ruled by various authoritarian regimes. To understand Haiti's authoritarian and turbulent politics only 7 of its 44 presidents have served out their terms, and there have been only 2 peaceful transitions of power since the b eginning of the republic (Fatton 2005) In 1986, Jean Claude Duvalier, known as Be be Doc, was ousted as President of Haiti. He had inherited the political power from his father Franois Duvalier. Both of them were authoritarian rulers and 1986, at least in part, marked the ending of authoritarian rulers in Haiti Yet tical pr ocess, since then has not been stable. A turbulent political culture has remained to this day, though Haiti is a democracy that has been led by several democratic elected Presidents since 1986

PAGE 21

21 Haiti, with some 8 million people, is the poorest country i n the Western Hemisphere and (Verner 2008 ) of the Haitian households lived in absolute poverty with 20, 56, and 58 percent of the households in metropolitan, urban, and rural areas, respectively, being poor based on a (Verner 2008) vast array of factors (i.e., low educ ation levels, high corruption, lack of natural resources, environmental degradation frequent natural catastrophes, weak institutions, etc. ). There are man y reasons for the development of these factors Haiti is a country with a long history of political i nstability, governance crisis, weak institutions and low investment as will be detailed in the following section The Historical Context four presidents have served out their terms, and there have been only two peaceful transi tions of power since the beginning of the (Shamsie and Thompson 2006) Haiti has been marred over the last two hundred years by political instability, the rule of the wealthy and powerful, and the existence of severe poverty. Political and social power has been acquired in the history of Haiti primarily as an avenue for individuals to gain economic wealth and conce ntrate social power. Once political power ha s been acquired, the goal of those in power has been to maintain such power, and the rents received from it. To protect their power, position (Shamsie and Thompson 2006) Political violence towards regime opponents and civilians has been common throughout the history of Haiti. In 1957 Franois Duvalier came to power in Haiti. He can be best described as military dictator par excellence, who ideologically framed himself as a black nationalist.

PAGE 22

22 Duvalier challenged the power of the military, which had played a crucial role in the political history of (Shamsie and Thompson 2006) With the help of the macoutes, (Hallward 2007) Duvalier reign did nothing to promote either political freedom or economic development. Franois Duvalier died in 1971, and his son Jean Claude Duvalier succeeded him as president. Throughout th e 1970s, Jean Claude Duvalier enacted economic reforms in Haiti, which were backed and supported by the international community, in particular the United States, and a short period of economic development occurred. ses of the macoutes, tolerated some dissent, (Shamsie and Thompson 2006) In the early 1980s, Jean Claude Duvalier stopped economic liberalization and Duvalier embraced the use of political violence again. The brutality of the Duvalier regime prompted the rise of another important political institution that has influenced the history of Haiti ever since. In the early 1980s, s (Hallward 2007) emerged in (Shamsie and Thompson 2006) materialized from the Catholic Church, and vehemently fought for social justice a nd the rights of all people, opposing the Duvalier rule and the macoutes. Both the (Hallward 2007) In 1985, the wave of protest increased to unknown levels, and in early 1986 protest against the Duvalier regime had become rampant even in Port au

PAGE 23

23 Prince. On 7 February 1986, the army forced President Duvalier out of office, and he was eventually exiled in Paris, France. he center of Haitian (Girard 201 0) It was the military between 1987 and 1991 that was the kingmaker. Without the backing of the military, no political actor became president, or maintained the power to remain president. After Duvalier left Haiti, the power was initially held by the N ational Government Council, and then in a short amount of time frequent governmental reshuffles, these years were marked by a societal status quo: the thugs inherited from th (Girard 2010) The military continued the political oppression and violence of the Duvalier era, and living conditions in Haiti did not change for the better. From the popular movement s (i.e., the organization populaires and the Ti Legiz ) a new movemen (Girard 2010) Lavalas translated means the flood, and it was a symbolic name for the masses unifying under leadership fighting against the oppression of the military regime. The popular leader of the Lavalas movement wa s Jean Bertrand Aristide. Aristide came from humble backgrounds with his mother being a simple merchant in Port au Prince. A Catholic priest took Aristide under his wings in his youth, and hence Aristide was able to get a top notch education gaining his M aster and PhD in Europe and the United States in psychology and theology respectively In 1982 Aristide returned to Haiti, after his studies were completed, where he eventually became a priest in Port au Prince. Aristide had publicly criticized the Duvali er government in the 1980s, and also defended the rights of the

PAGE 24

24 poor and provided for them. His theological and social views were influenced by liberation theology, which believed that Christians must actively work for the economic and social justice of al l people. Liberation theology believed in the preferential option for the poor, a notion that society and government must by all means protect the right s of the poor. While the political violence was a consistent force throughout the late 1980s, the Laval as movement remained alive and continued to oppose the military power in the country. (Girard 2010) b y the late 1980s, and continued domestic opposition in combination with international pressure forced a presidential election on December 16, 1990. The hope was that the election would finally bring stability to Haiti and its political system. There were m ultiple candidate s running in this election: Marc Bazin was a former World Bank economist, Victor Benoit was supported by left leaning political parties, Roger Lafontant was a former minister of the interior under the younger Duvalier, and Jean Bertrand Ar istide was the favorite of the masses and leader of the Lavalas movement. On the actual day of the election, it was clear to everyone in Haiti that Aristide would win the election, as long the election result would be determined in a fair manner. Aristide was remarkably different from the typical Haitian political candidates: Aristide was not a member of the mulattoos elite, from which the vast majority of Haitian leaders had come from previously, but aspired to be the leader of the masses defending their r ight s In 1990, the popular opinion about Aristide was that he truly protected the rights of the poor, and was in that sense different from all other politicians in Haiti. On Election Day, Aristide won the election by a significant margin winning 67 percen

PAGE 25

25 (Hallward 2007) The politica l and economic reforms of the first few months were certainly not as radical as one could have expected, due to the radicalism of his rhetoric before the (Hallward 2007) (Hallward 2007) Even though Aristide did not oppose the military and its hated section chiefs with elected officials and an apolitical police (Hallward 2007) Aristide also began implementing various social programs combating the poverty and destitute situation in Haiti. In September 1991, Aristide flew to New York to d eliver a speech at the United Nations. Once he landed in Haiti after the trip, he was informed that a coup had begun. Eventually Aristide was forced into exile. The new man in power was Raoul Cedras, a military leader. The reign of Cedras was marked by p olitical oppression and misery. (Girard 2010) The military harshly suppressed the grassroots movements, which had been so crucial to Aristide its long c ampaign against Lavala s In order to contain the popular mobilization, you (Hallward 20 07) also diminished during this time. The international community imposed a trade embargo

PAGE 26

26 on Haiti, and foreign aid decreased significantly as well. As a result of the dire situation in Haiti and the general disenfranchisement of the Haitian population, more and more Haitian s attempted to flee towards the United States using boats or other small vessels. Once Bill Clinton became the President of the United States in 1993, the politics of the United States towards Haiti changed; the previous administration had opposed the socialism of the Lavalas movement. Clinton was concerned about the humanitarian situation in Haiti and the rising stream of Haiti an illegal immigrants to the United St ates. Over the next two years, multiple UN resolutions were decided upon, and the United States as parts of an international coalition were preparing to invade Haiti. I n the end however, an agreement was negotiated that allowed Aristide to return to Haiti to fin ish his presidential term, and a UN peace keeping force under the leadership of th e United States entered Haiti. The eventual agreement between Aristide and the military government of Cedras which enabled Aristide return to Haiti, was a controver sial (Hallward 2007) including the adoption of a neo liberal economic policy due to the pre ssure of the (Hallward 2007) in his leadership approach Chavannes Jean poke for many travelers when he said in 2005 that Aristide completely changed in the US. He had become unrecognizable, a monster, obsessed with m oney (Hallward 2007) Such a statement is extreme, but also indi cates that the Aristide of post 1994 had changed his positions and approach. new president

PAGE 27

27 initiative s had been to disband the army on 31 December 1995 (Shamsie and Thompson 2006) The army in Haiti had intervened in the political arena frequently ousting scores of presidents in the History of Haiti, just as it had been instrumental in could be ensured that a un likely. Secondly, a new parliament was to be elected, and after several delays, a parliament was in place in October 1995. In the fall of 1995, Aristide began to rhetorically attack the presence of the United Nation mission in Haiti, and in particular the fact that United States military was stationed in Haiti. He focused more on maintaining and stabilizing his power basis and criticizing foreign powers, rather than on (Girard 2010) willingness t (Shamsie and Thompson 2006) ort was expended to bring perpetrators of past abuses to justice, while political violence and political abuses were (Shamsie and Thompson 2006) In 1996, Haiti experienced a peaceful transition of the presidential power. Aristide had finished out his presidential term, and the constitution prohibited him from running for a second term. The popular masses were still extremely supportive towards Aristide, and it was clear to people in Haiti that Aristide would be quasi appointing the new Rene Prval a Haitian who had been educat ed in Belgium and the United States, was a trained agronomi st, and who had known Aristide since the early 1980s when the two of them (Girard 2010) an organization with a mission

PAGE 28

28 of assisting the poor. Prval had also served as Aristide prime minister in 1991 and was one of the confidants of Aristide. Prval (Hallward 2007) while also not being directly part of a political party or association. Once in office, Prval loyalists and an increasingly anti (Hallward 2007) Prval Rosny Smarth, from the Organisation Politi que Lavalas. Smarth was a proponent of style (Hallward 2007) whereas Aristide was a left ward progressive wit h socialist preferences. In 1996, Aristide created a new political party, Fanmi Lavalas (FL) establish direct links between local branches of the Lavalas mobilization and its parliamentar y (Hallward 2007) k (Ha llward 2007) and the Organisation Politique Lavalas lost many seats (Hallward 2007) the el (Hallward 2007) of elections needed to finish the election process. From this point on, the legislative process came to dead halt with the Organisation Politique Lavalas blocking most of the legislative actions in Haiti. (Hallward 2007) was delayed until May 2000. In this time, Prval governed the country by presidential decree with no functioning parliament in place. In May 2000, elections were planned and held in Haiti. an overwhelming majority of the seats in both houses of parliament in 2000, thereby

PAGE 29

29 making it possible for him to govern without significant (Dupuy 2006) In November 2000, a presidential election took place, and Aristide han dedly won the election. After this election the political appointees of Aristide were surpr ising to many of his supporters and generally not supported by them. He included for one Marc Bazin, a former Wo rld Bank economist, and Stanley Theard, a minister under Bebe Doc, who allegedly was involved in a major corruption scandal. Right after the presidential election in 2000, a democratic opposition formed. Many of losers of the 2000 election regrouped as the Convergence Democratique (CD) The group primari (Shamsie and Thompson 2006) of the 2000 elections The Convergence Democratique was suppor ted by the United Aristide for ideological reasons and starved his regime of badly needed for eign (Shamsie and Thompson 2006) while simultaneously supporting the Convergence Democractique. The security situation in Haiti around 2000 was a complex situation with various agents and act ors involved. The Haiti National Police (HNP) was founded in 1995 to bring safety and security under civilian control, and the HNP replaced the military becoming the only and dominant security in Haiti. The presence of the HNP was particularly great in Po rt au Prince and much less so in the rural areas of Haiti. It is important to note the following two details about the HNP: for one, part of the HNP was recruited from the abolished military, indicating that at least part of the HNP was not supportive of the Aristide rule, and secondly the HNP was marred by corruption and cronyism.

PAGE 30

30 In addition to the HNP au Prince, Gonaive (Hallward 2007) These armed groups were frequently criminal groups focused on controlling certain areas of major cities, and gaining money and wealth from illicit activities. The gangs varied in strength and purpose. In ge neral the armed gangs, called (Hallward 2007) though i t is unclear and disputed how closely aligned the were to Arist ide and how much of their criminal activity was supported by Aristide. Alex Dup uy, a researcher and professors stated the following opinion: T he did Aristide's and the government's dirty work and, along with the police, attacked and killed membe rs of the opposition, violently disrupted their demonstrations, burned their residences and headquarters, intimidated members of the media critical of the government, and engaged in countless other human and civil rights violations. Some leaders among them also became a force in their own right by forming criminal gangs that acted autonomously, turned their neighborhoods into wards under their control, engaged in drug trafficking and other criminal activities, and even requisitioned the gove rnment itself (Dupuy 2006) To what extent this opinion is true cannot be conclusively determined, but the had a crucial part in the security situation in Haiti throughout the Prval and Aristide presidencies. 1990s saw a gradual and arguably premature reduc tion in the UN/OAS pr (Shamsie and Thompson 2006) The focus of the mission was to train the new HNP. The for ce was com prised of military personal of various countries. The actual size of the United Nations mission prohibited it to pacify and control the entire country even if it might have been necessary. When Aristide abolished the army he e pool of eligible and resentful ex m (Hallward 2007) These former army soldiers were well trained; most of them had been trained by the United State s. These former army

PAGE 31

31 soldiers represented a threat to the security situation in Haiti, as many, especially former officers, were politicized and the abolishment of the army in 1995 increased their economic un certainty. In the following years, former milita ry members played a crucial role in the history of Haiti. There are several interpretation s of what transpired between 2000 and 2004 in Haiti, and in particular the role of President Aristide and Fanmi Lavalas had throughout it. There are basically two c amps and interpretation s about President Aristide and his legislation during this time period (Hallward 2007) have been highly critical of President Ar istide and his legislation. A former supporter of President Aristide, Michele Montas stated : those ideals shared by Jean, including a generous but rigorous soci alism, respect for liberties within the framework of democracy, nationalist independence, based on a long history of resistance, those ideals that Jean used to call "Lavalas" are trampled every day in this balkanized State where weapons make right, and whe re hunger for power and money takes precedence over the general welfare, causing havoc on a party which, paradoxically, controls all the instit utional levers of the country (Dominique 2002) The critics of Aristide argue that Haiti in the period between 2000 and 2004 had be come more and more violent. These critics blame President Aristide for the increased violence arguing that he encoura ged and tolerated violent gangs and the ir t hroughout murde rous pro Lavalas gangs was deliberate government policy, that it was coordinated by local PNH boss Hermione Leonard and veteran Aristide loyalist Jean Claude Jean Baptiste along with other s in t (Hallward 2007) According to critics, the had been purposefully used, and were intended to be a

PAGE 32

32 counterweight to the opponents of Aristide, whether it was the democratic opposition or former military personal Aristide accepted the use of force and violence to protect his own power, and he himself had become a despot more fixated on maintaining his power than promoting development or social justice. In contrast to the critics of President Aristide and his regime, we can find people Jean Juste, John Jos eph Jorel, Jean Charlie Mose, Belizaire Print emps, Samba (Hallward 2007) who support a different view of the failure of the Aristide presidency. W ith its eve ntual overturn in 2004, these supporters interpret events as a successful act of sabotage, which removed a democratically elected government from office. Both internal and external actors had been discontent with the rule of President Aristide The internal actors were the rich elite of Haiti, which had been marginalized by the Lavalas movement, the former military leadership who had lost all of their power, and the Group 184 the democratic opposition to President Aristide. Additionally, it ha s been argued that the United States opposed President Aristide, while simultaneo usly supporting the Group 184. Former US ambassador to El Salvador Robert White said about Roger Noriega, Bush Assistant Secretary of State for Western Affairs, in February 20 (Hallward 2007) Arist ide was seen as a violent socialist and populist destabilizing both Haiti and the Caribbean region. Supporters of Aristide argued that US opposition to Aristide was founded on Aristide leaning populist. The internal actors were opposed to Aristid e,

PAGE 33

33 due to them lo Haiti. Hence, Aristide, according to his supporters, was brought down because of the internal and external resistance of powerful actors in Haiti. The proponents of Aristide argued f the rapidly expanding Lavalas (Hallward 2007) existed, whereas opponents of Aristide argued that he in particular had been a corrupt dictator along with his entire regime. Corruption and cronyism was a striking aspect of Haiti from 2000 and 2004, but how corrupt Aristide himself was has been a point of contention. The human rights condition was fragile in this time period. The supporters of Aristide argue that human rights violations, by the regime or regime supporters have been grossly overestimated and the human right s situation was not as tenuous as described by oppponents of Aristide. and Raboteau, as there was before Aristide and after Aristide. But if repor ts from Amnesty International can be trusted then from 2001 to 2004 perhaps thirty political killings can be attributed to the PNH (Hallward 2007) In 2004 B (Hallward 2007) The prop onents of Aristide argue that he and his administration did not approve of corruption, and political violence at least initiated by the administration, was nonexistent. However, the Group 18 4 and the United States used exactly these reasons to remove Aristide from office in 2004 proving that not facts where the reason for removing Aristide, but much more the opposition for what Aristide stood for, namely a populist and socialist approach advoc ating for social change.

PAGE 34

34 ive judgment must be withheld. Yet it is possible and necessary to outline the events that ultimately led to the ousting of Aristide in 2004 in order to provide a n adequate backdrop for the analysis conducted later in this thesis In February 2001, the inauguration of President Aristide took place. Aristide had named Jean ex (Hallward 2007) The CD wanted to establish a political alternative to the Lavalas movement, and consistently disputed the validity of 2000 election results. (Hallward 2007) and the fiscal and economic situation of the Haitian government was higher minimum wages, school construction, litera cy programs, higher taxes on the rich (Hallward 2007) The Haitian economy suffered in particular during this time from the receding internation al donor support, which limited the amount and extent of public programs. In July 2001, armed conflict was iniatiated by former army members, part of the Front National pour le Changement et la Democratie (FLRN), which was an anti Aristide rebel group led by Jodel Chamblain and Guy Phillipe. The rebel group attacked multiple targets throughout Haiti. At the end of 2001, another major attacked was staged by the FLRN. The FLRN (Hallward 2007) and seized the

PAGE 35

35 palace for several hours, before eventually failing in their effort to topple Aristide at this point. The political situation remained tenuous as well. The government and the CD had serious negotiations throughout 2001. The goal of the talks was to resolve the dispute about the election of 2000, but eventually the CD broke off all talks. Until 2004, no talks between the government and the CD to resolve their issues were ever succe ssful thus destab ilizing the political situation during this time period. In 2002, Jean in debilitating parliamentary squabbles with other ambitious members of FL like Prince Pierre Sonson o r Fo (Hallward 2007) The internal dispute in the Lavalas movement had a destabilizing impact upon Haiti throughout the second presidency of Aristide. In December 2002 another crucial opposition group was formed at a meeting in (Hallward 2007) met for a three day period discussing various strategies of opposing and political opposition to A (Hallward 2007) and to (Hal lward 2007) funds signaling the support of these opposition by the United States and its Western allies. While the security situation was bad throughout the reign of President Aristide, 003 and up to the resign (Faubert 2006) In 2003, a significant spike in violence occurred, and tension between rebel groups,

PAGE 36

36 the government, and HNP rose. The economy also performed poorly with negative growth rate of 3.4 percent while the annual growth rate of the populati on was (Faubert 2006) From the time they began in the summer of 2001 through to the middle of and run affairs. Then in the autumn of 2003, FLRN attacks become more regular and intense, spreadi ng from the Central Plateau to Petit d Cap Hatien (Hallward 2007) of the poorer neighborhoods of P ort au (Hallward 2007) by aligning themselves with gang s in the poorer areas of the capital. However, the dominance of the Lavalas gangs was too strong and too loyal to President Aristide for the G184 to gain control of these poorer neighborhoods. economic forc (Hallward 2007) leader of the Cannibal Army was murdered being shot multiple times. The brother of one of the new leaders of the Cannibal Army then, and he claimed that President Aristide had ordered the shooting of his (Hallward 2007) in the murder though, and claimed the murder was carried out (Hallward 2007) by attempting to we aken his position in Gonaves. The Cannibal Army turned against the government now joining the FLRN in their opposition. Violence increased now significantly throughout the country, and the Cannibal Army and FLRN assaulted important building s in Gonaves and throughout the surrounding area attacks on police station s by Fanmi Lavalas

PAGE 37

37 activist became r (Hallward 2007) with the government and allied gangs fighting back. O n February 7, after a sustained struggle for control, the Cannibal Army and on to take Hinche on 16 February and Cap Hatien on 22 February. By 27 February they appeared to control most of the northern half of the country, and parts of the south and south (Hall ward 2007) The international community, i.e. United States and France in this case, were not force to enter Haiti to stop the rebels and protect Aristide, therefor e, was a logical conclusion to a decision taken earlier by the three governm ents to remove him from (Dupuy 2006) The international community did not remove Aristide, but it did allow a rebel force to oust a democratically elected president from Haiti. On February 29th 2004 Aristide was ousted as the President of Haiti, due to the Aristide's departure, the United Nations authorized the deployment of a Multinational Interim Force (MIF) c omprised of troops from the United Stat es, France, Canada, and (Dupuy 2006) The international community had not been willing to protect former president Aristide, but was willing to intervene now. The new president, who was backed by the international community, was Gerard Latortue. The Multinational Interim Force (MIF), a UN peace keeping force also named MINUSTAH, came to Haiti to improve the security in Haiti. The new president, who was backed by the international community, was Gerard Latortue. Latortue headed a governm ent compromised of

PAGE 38

38 members from the CD and the G 184 The purpose of the Latortue government was to and the rebel forces of the defunct (Dupuy 2006) which was a paramilitary affiliated with the military. The government instead persecuted and killed supporters of the former President Aristide. Both the insurgents from the disbanded Haitian army, who precipitated pt their weapons. Under the MINUSTAH Latorture regime the country and especially Port au Prince, endured a climate of insecurity. Gang violence, be came widespread (Shamsie and Thompson 2006) The support for Aristide remained high, even after the ousting in February 2004, and it remained particularly stro au Prince. The Fanmi Lavalas movement was still existent and had p opular support throughout Haiti; it just experienced significant backlash. In late 2004, political violence spiked in the capital, in partic ular in the neighborhoods of the MINUSTAH on one side, and the gangs of the Port au Prince neighborhoods on the other side. Additionally, violence also occurred in other poorer are as of Port au Prince. The human causalities in Port au Prince were extremely high throughout this time. In 2004, Haiti suffered from numerous natural catastrophes and tropical storm s hitting Haiti, the usual high levels of corruption in government circles and the extremely high levels of violence and social unrest throughout the country. The situation was so dire that the United Nations mission (MINUSTAH) focused almost exclusively maintaining order in downtown Port au Prince, and doing little throughout the country.

PAGE 39

39 In early 2005 the United Nations and the United States were convinced that a certain level of progress in regards to stabilizing Haiti had been achieved. he All of the seams were coming apart, and there was no control anywhere. By the end of 2005, some parts of the country had been pushed to the brink of Port au Prince as virtually para lyzed by kidnappings, spreading panic among ri ch and poor alike (Hallward 2007) led to four postponements of general elections which were finally held on 7 February (Shamsie and Thompson 2006) The presidential election hoisted a broad field of candidates coming from various political factions i.e. G184, the CD, and former army leaders that had supported the overthro w of Aristide in 2004. Marc Bazin a former World Bank economist ran as well, in addition to an e vangelical candidate was ideologically part of the Lavalas movement, entered the presidential race fairly late in 2005. elected president with an overwhelming majority. Receiving more than 51% al by 40 points (Shamsie and Thompson 2006) ed Jacques Edouard Alexis as his prime minister. In September 2006, the UN mission launched a p rogram attempting to reduce violence and conflict by disarming gang members. The initial strategy by the UN was to e conomic terms (Hallward 2007) offering financial incentive s to give up guns. In January 2007, the UN launched a new offensive against gangs throughout Port au Prince. Over the next months, the UN carried out multiple raids attempting to decrease the number of arms in the hands of gangs reducing their

PAGE 40

40 potential for violence. Overall, the level of violence decreased in 2007 throughout Haiti. disarmame (Hallward 2007) in mid 2007. Throughout 2008 world food prices had risen. In April, riots throughout Haiti rose up protesting the high food prices. The government responded by subsidizing the pr ice of rice to halt the serious unrests. The UN also approve d additional food aid to improve food security in Haiti. Prime Ministe r Jacques Edouard Alexis resigned and is replaced by Michele Pierre Louis. In August 2008 a severe tropical storm hit Haiti k illing around 800 people. In April 2009, Haiti held a senate election. The election was very controversial, since a election commission, whose members were appointed by President Pr val, excluded the Fanmi Lavalas party. Voters responded by protesting the election, and only 10 percent of eligible voters turned out in this election. In January 2010, Haiti was hit by a particular ly devastating earthquake in Port au Prince and its surrounding region. The earthquake is estimated to have killed around 300,000 people The international community responded by increasing aid to Haiti. The rebuilding effort in Hai ti has been rather slow since. At the end of 2010, a presidential election was held in Haiti, which was cont roversial again with multiple candidates being held out from election. Fanmi Lavalas candidates were not allowed to run again. In March, 2011 Michel Martelly won the runoff election for the Haitian presidency. Martelly had previously stated that he was an ally or sympathizer of former President Latortue, and had opposed and been critical of former President Aristide.

PAGE 41

41 Haiti recent historical past has been characterized by political instability, and periods of high political conflict and violence. In the 19 80s the Lavalas movement, a socialist, grassroots movement, formed in Haiti. The goals of the Lavalas movement were socialistic and pro poor being opposed to the goals of the business and political elite, who traditionally held economic and social power in Haiti. The clash between the ruling elite and the Lavalas movement was the defining aspet to political and social life in Haiti, since the late 1980s. Armed Conflict Location and Event Dataset The ACLED records conflict in various developing countries, i ncluding Haiti. A dataset that collects reported information on internal political conflict disaggregated by (Raleigh et al. 2010) The ACLED dataset records the political violence inside of a particular country, and allows for a disaggregated appro ach to study government nsfers of military control from governments t b and violent events that are crucial to the dynamics of political violence (e.g. rallies, recruitment drives, peace talks, high (Raleigh, Linke, and Dowd 2012) The ACLED records political violence, and labels every single outbreak of political violence as an event. For every event, the dataset includes further information, such as location or time. Understand ing all of the incl uded information in the dataset helps researchers have a better comprehension of the dataset.

PAGE 42

42 event, and it is labeled as event date in the datasets. It records the time, month, and year, when the event occurred. As long as violent activities continue the ACLED on March 1st, 1999 and lasts until March 5th 1999 with violent activity reported on each day, is coded as five different events in ACLED with a different date fo (Raleigh, 2012). Additionally, the dataset also provides information about the precision of the time information. The time precision records t his information. If time precision is labeled as 1, then data sources provide an accurate, precise date of the violent. If time precision is labeled as 2, then data sources provided only the specific week of when the event occurred. If time precision is labeled as 3, then data sources provided only the month of when the event occurred and the midpoint is chosen in this case. The ACLED dataset also separates the event into eight types of events, in other words the ACLED dataset recognizes eight variou codes for eight types of events, both violent and non violent, that may occur during a no change of territory event type characterizes an event, w here two factions or groups, such as the rebels control location event type characterizes an event, wher a battle where a rebel group win s control government regains control (Raleigh, 2012). The headquarters or base establishment event type characterizes an

PAGE 43

43 event, when a rebel groups establishes a new headquarter or base camp. The non violent conflict / militia / governments participate that does not involve active fighting but is within the context of the war/dispute. For example recruitment drives For these events, the note section of the dataset gives further clarification of the type of activity that actually occurred. The rioting / protesting event type characterizes an event in which a group protests against the government or one of the governmen t institutions. R ioting and protesting is included, since it marks a non v iolent form of protest against the rules and institutions of a country. The violence against civilians event type characterizes an event, in which a rebel group, the government, or a militia engages in a violent activity towards a civilian group. The n o n violent transfer of even t type characterizes an event by which the possession of a location is transferred between two actors without the use of violence. The ACLED dataset also includes the actors or pa rticipant s in the recorded events. The conflict actors are primarily rebel groups, the government, militia, active political groups, or civilians in various countries. All of these groups struggle or fight over political control inside a country. Hence, the ACLED dataset gives information abo ut who was in a conflict with each other. contr Furthermore, changes in the government control can also be understood, and recognized from the ACLED d ataset. In the case of Haiti, the control of the government changed throughout the dataset. The police, as being the military arm of the government in Haiti, is labeled in various ways throughout the

PAGE 44

44 dataset (e.g. Police Forces of Haiti (2000 2004), Poli ce Forces of Haiti (2004 2006), Police Forces of Haiti (2006 )). The labeling allows for a better und erstanding of how the control over various institutional forces changes, and extenuates that institutional forces are temporally not uniform in form or function. Rebel groups are movements with the explicit goal to gain political or territorial often have predecessors and successors due to diverging goals within their memb Militias are paramilitary groups created by the government, military, or rebel group to work in conjunction with them helping to accomplish their overall political and military goals. Militias are o ften created for a specific purpose at a specific time. Overall perpetrates political vi olence; the default assumption in ACLED is that such groups can activities are coded a s riots if the spontaneous civilian actors become violent against rotester are non violent uprising s by groups of people. The ACLED dataset provides information about the location of the recorded event. It is a geo referenced data set providing the longitude and latitude of where the conflict event occurred. Additionally, it also provides the name of the country, administrative unit, and city. The dataset informs the user of the spatial precision of an

PAGE 45

45 event. If geoprecision coordinates are available for that town (Raleigh, 2012). If geoprecision is labeled as egion, and notes a general area, a town with georeferenced coordinates to represent that area is The dataset provides information about the number of fatalities in any e vent. The sourc e data frequently provides information about casualties for the different event Furthermore, the data set also provides eventnotes for every event. These notes provide explanation about the event, and assist the researcher in understanding the details of the event better. The ACLED uses press reports as the source of information about events in the dataset. T he reliability of sources, such as the government, and/or non governmental organizations, has frequently been questioned. These actors can have a vested interest to report incorrect numbers, either overestimating o r underestimating the amount of (Bocquier and Maupeu 2005) l evels of conflict and violence. Bocquier and Maupeu tested the reliability of newspapers to accurately measure levels of con flict in Kenya. The authors try to determine the ability of news paper source to accurately record trends in violent events and confl ict. They conclude: W e believe that, if properly analyzed, press reports can certainly serve as a surveillance system of violence, especially when it is collective. Press

PAGE 46

46 repor ts help mea suring the trend of collective violence over time and/or at a sub nat ional level. The method is particularly appropriat e to analyze low intensity con flicts, which persist over decades and are typical of violent political cultures (Bocquier and Maupeu 2005) Bocquier a nd Maupeu defend the use of news paper source s to analyze and measure conflict in a society, even if such sources include measurement errors and inaccuracies. Using newspaper sources, the ACLED measures and records conflict events in Haiti. It includes temporal and spatial information, which allow for a temporal and spatial analysis of conflict in Haiti. Country level Studies of Violence, Conflict, and Civil War Civil war and conflict have historically been studied in country level studies. The outbreak or continuation of civil war and conflict serves as the dependent variable upon which various correlates are tested. T he premise of these studies is the identification of robust correlates of conflict and violence. The objective of many country level studies is to find global or regional determinants of civil war. In this literature review, multiple, influential country level studies are included, which outline the significant finding s of country level studies their assumption s frequently used determinants, and overall methodological procedures. Paul Collier and Anke Hoeffler, from the University of Oxford, published the These authors note that t he political science discipl ine has contributed heavily towards the study of the economics of conflict, and the determinants of violence. It has in particular offered answers to the following two questions: Why does civil war break out? What contributes to prolonged civil wars and violence?

PAGE 47

47 For one, the political science literature explains violence and civi l war (Collier and Hoeffler 2004) Hence, conflict is seen as a reaction toward a particular social and/ or economic situation, which gives sufficient motivatio n to rebel against the status quo. Rebellion s or civil wars are seen as a path way toward institutional change and reform in a society. It occurs when the grievances have been significant enough to provide a motivation for violent protest. Economists offer another explanation. Civil war and rebellion can be explained (Collier and Hoeffler 2004) According to this explanation, civil war and rebellion must b e understood in terms of a cost benefit analysis. The benefits of rebellion mus t out weigh the cost of rebellion, and the explanation of civil war and rebellion becomes necessarily economic. opportunities for rebellion ag (Collier and Hoeffler 2004) Furthermore, the authors decide to treat the explanation s separately with this article focusing solely (Collier and Hoeffler 2004) The authors develop a model explaining the i ni t i ation of violence including multiple explanatory variables. Collier and Hoeffler use an aggregated approach to study determinants of civil war and violence. The authors use a data set developed by Small and Singer, which cover s 169 countries identifying 79 civil wars ranging from 1965 to 1999. The dependent variable in this case is the outbreak of civil war in a country in a five year period; the variable is coded as zero, if no civil war breaks out, and as one, if a civil war breaks out. Additionally th e regression an (Collier and Hoeffler

PAGE 48

48 2004) which means that the grievances and greed explanatory variables are tested in two separate models: one model tests the grievance hypothesis, and the other tests the greed hypothesis. The greed or opportunity variable explain s civil war or violence as a consequence of economic opportunity. The authors justified the inclusion of variables either by economic intuition or because the literature of the economi cs of conflict has determined them to be important variable s to consider. Looting or the extortion of resources can often be used as a source of finance for conflicts, and the variable significant opportuni ty costs for fighter s in a civil war; the greater the education level of a potential fighter the greater the opportunity cost for a fighter to join a rebellion, due to higher forgone wages Th is le d to an inclusion of the male secondary schooling variable which attempts to show the opportunity cost s for fight ers. Furthermore, the variable GDP growth was included to measure the infl uence of economic growth on the outbreak of civil war. The authors also included the on dispersion and soc (Collier and Hoeffler 2004) to measure possible military advantages enjoyed by opposition or rebel groups If a country is more mountainous, rebels or opposition groups could seek shelter and protection in the mou ntains, and such cover makes civil war and conflict more likely. The (Collier and Hoeffler 2004) model measure s the impact of grievances upon the likelihood of civil war and conflict It does not attempt to explain civil war as a conse quence of economic factors, but believes that grievances motivate

PAGE 49

49 ial (Collier and Hoeffler 2004) polarization of the society, and et hnic dominance of a group as explanatory variable s. As societies become more plural, and share power less equally the potential for civil war should rise. Furthermore, the authors include a the impact of democratic institutions upon conflict and civil war propensity The authors also include (Collier and Hoeffler 2004) as explanatory variable s. Inequality in asset ownership could be a motivating factor for people to initiate a civil war or rebellion. T he (Collier and Hoeffler 2004) whereas the perceived grievance model does not perform well. The authors highlight that primary exports, in particular oil, can be a significant so urce of finance that make civil war s and conflict s possible. The existence of a primary export commodity rises conflict and civil war propensity in a society. Additionally (Collier and Hoeffler 2004) impact the decision s by individuals to participate in civil war and rebellion. If individuals have low earning and economic potential the probability of participating in a civil war for these individuals increases. Hence, low earn ings and low economic potential raises the likelihood of conf lict and civil war. Military advantages are also important to consider in this discussion; in particular geographic dispersion of the population increase s the probability of civil war and rebellion. The objective grievance model suggests that ethnic domin ance by a particular ethnic group raise s conflict potential and democracy reduces the probability of civil war.

PAGE 50

50 The article concludes that economic opportunities enable civil wars and civil wars can be better explained through economic explanations The authors are not willing to dismiss objective grievances as a common cause of civil wars and conflict though but social concerns of inequality, political rights, and ethn ic or religious identity (Collier and Hoeffler 2004) and David Laitin in the American Political Sciences Review. The article explore s the reasons for civil war and violence in several countries between 1950 and 1999. In this time period, the casualties of intrastate conflict outnumbered the casualties of interstat e (Fearon and Laitin 2002) The authors dismiss the standard explanations of the political sciences literature in regard to the occurrence of conflict and civil war Civil wars have not been more common since the end of the cold war era, (Fearon and Laitin 2002) is not a favor of the dominant view that one can predict where a civil war will br eak out by looking for where ethnic or other broad political gr (Fearon and Laitin 2002) Fearon and Laitin responde d to the article by Collier and Hoeffler. Collier and Hoeffler argued that war and violence can be explained primarily by economic opportunities, and not by perceived grievances. Civil wars are caused and enabled by economic reasons, not by social or political reasons.

PAGE 51

51 theoretical interpret ation is more H (Fearon and Laitin 2002) The strength of institutions and of the state is the primary determinant of civil war and conflict encourage (Fearon and Laitin 2002) the occurrence of civil wars and conflict. ew scholars argued that the real source of rebellion w as of ten (Fearon and Laitin 2002) The authors t est this hypothesis using a data set of civil war outbreak s between 1945 to 1999. The analysis of the researchers variation in ethnic homogeneity (Fearon and Laitin 2002) Furthermore religious fractionalization does not increase the p robability of civil war either. (Fearon and Laitin 2002) is a very significant explanatory one year) is strongly significant in both a statistical and substantive sense. $1,000 less in per capita income is associated wit h 45% greater annual odds of civil war onset, on (Fearon and La itin 2002) M frequent in democracies after controlling for income, as shown by the positive and statistically insignificant coefficient for Democracy (Fearon and Laitin 2002) The authors suggest that per capita income is a statistically significant variable explaini ng civil war propensity in a society Income inequality on the other hand is not statistically significant, and does not provi d e an explanation for the outbreak of civil wars. Whether or not a country has a democratic political system does not impact civi l war propensity but anocratric countries, countries with a mixed political system containing autocratic and democratic elements, have a much greater probability of civil

PAGE 52

52 war breaking out The instability of such political systems increases the likelihood of a civil war being successful, which makes civil wars more feasible for opposition gr oups whic h is a huge effec (Fearon and Laitin 2002) The p olitical stabi lity of a country impacts conflict propensity. Fragile and new states have a greater propensity of conflict. Based on their study, t he authors reject many determinants of civil war, which other researchers in the discipline find significant and important Fearon and Laitin state: The conditions that favor insurgency and in particular state weakness marked by poverty, large size, and instability are better predictors of which countries are at risk for civil war than are indicators of ethnic and religi ous diversity, or measures of grievances such as economic inequality, lack of democracy or civil liberties, or state discrimination against minority religions or languages (Fearon and Laitin 2002) The authors claim that demographic and ethnic factors are not determinants of civil war and conflict They also point out that civil wars can break out, if a relatively small force encounters economic hardship s, and the possibility of hiding from the government or military is given Civil war must not be a large scale event initiated by a significant movement in the population, but rather a small pocket of the population period have structural roots, in the combination of a simple, robust military technology and decolonization, which created an international system numerically dominated by fragile states with limited administrative contro (Fearon and Laitin 2002) In 2002, t derstanding In this issue, the journal summarizes the current literature on civil war and

PAGE 53

53 violence in societies, and publishes ne w research findings. Paul Collier wrote the introduction to the special issue, and in it he explains his view on the current state of the discipline. Collier emphasizes that economics can and should contribute to this debate and research agenda. Conflict more obvious econ (Collier and Sambanis 2002) Civil wars can be caused or the initiation of them influenced by economic factors, which is an overlooked dimension of conflict. Furthermore, civil war and conflict have economic ramifications for a society. Collier states that because of these two important principle s the economic discipline has to continue to contribute to the research agenda of civil wars and conflict Both, political science and economics have an important role to play in furthering our understanding of the causal relations hips between determinants and civil wars Collier begins his discussion by stating possible explanations for the causes of conflict. In the political science grounded (Collier and Sambanis 2002) The situation individuals or groups find themselves in motivate s them to initiate civil war and violence to alter the ir situation. ation for rebellion is not grievance but greed (Collier and Sambanis 2002) Conflict and civil war are motivated by the opportunities they create or offer Conflict occurs when it is profitable for the involved (Collier and Sambanis 2002) The i mportant aspect here is that actors decide upon conflict based on their perceived opportunities and grievances. Information play s an important role in the initiation of

PAGE 54

54 conflicts, and misinformation influence s when and how civil war and conflicts play itse lf out. Collier then expounds on the reasons why conflicts prolongate and continue to out (Collier and Sambanis 2002) If the agents are uncertain who will win a rebellion or civil war, they will be motivated to continue the civil war or rebellion. Imperfect information can be a reason for the prolonged continuation of conflict s Furthermore, rebels are frequently not willing to disarm or stop their participation in a civil w ar because doing so would increase their potential to be harmed by the government. The government also has frequently no incentive to cease the conflict, since it could lead to a power sharing agreement with the rebels Conflict continues because the cess ation of conflict would make one or both of the parties worse off. Hence, once conflict has started little motivation exists to cease such conflict nd (Collier and Sambanis 2002) Such varying preferences would make it difficult for different agents to reach a settlement and agree on peace, if they were already engaged in a civil war. If motivation and preferences are strong enough and too antithetical, settlements of conflict bec ome more difficult to achieve. Thus, conflict often becomes a th (Collier and Sambanis 2002) Once civil war or violence occurs, stopping it becomes difficult and states of aggression are often self reinforcing. the Journal of Conflict Resolution by Marta Reynal Querol. The researcher begins the

PAGE 55

55 article by discussing the causes of civil war. In 1998, Collier and Hoeffler had (Reynal Querol 2002) Collier and Hoeffler had rejected the notion that ethnic fragmentation plays an important role in determining where civil wars and violence occur. Reynal Querol argues in this article that the conclusion reached by the researchers was incorrect, and argues that e thnic and religious fragmentation play a crucial role in understanding civil war and violence. Reynal Querol also argues that researchers should distinguish between different types of civil war: some civil wars are primarily ethnic and reli gious and others are caused by economic grievances The study reaches three major conclusions. eligious polarization and (Reynal Querol 2002) have explanatory significance in explaining the incidence of civil war and conflict. Reynal Querol used a different data source for religious fragmentation which distingu ishes between A nimistic, Christian, and Muslim religious affiliations more accurately and combines various data sources into one. Second the author determine s cleavage that can develop into civil war (Reynal Querol 2002) Religion has the potential to increase soci (Reynal Querol 2002) in nature separating people from each other Religion also implies that people understand the world in different ways. These differences in unders tanding can extenuate conflicts and make them at times even necessary. The res earchers also determined that religious differences are more important than linguistic differences in determining conflict and civil war. Linguistic differences do not polarize a society as strongly as religious differences do. Lastly, a

PAGE 56

56 emocracy is a political system that significantly reduces the incidence of ethnic civil war A consociational democracy provided guaranteed group representation, and attempts to share the power between various factions in society. The author remarks that democracy itself does not reduce the potential for conflict and violence, but that equal representation and power sharing reduces conflict potential The article Recruitment and Allegiance: The Microfoundations of Rebellion was published in 2002 in the Jo urnal of Conflict Resolution by Scott Gates. The article (Gates 2002) Rebel organizations exist outside of the legal structure of society and are by definition not under control of the government. It is the government that usually has the most influenc e on who defines and determines the rules and configurations of institutions, both formal and informal, which govern the interaction of various agents in a society. enf orcement is the root of recruitment and (Gates 2002) This study attempts to understand how rebel groups organize the mselves, how allegiances are formed, and how collective action is created. In other words the researcher attempts to shed further light on the organizational structure that allow s rebel groups to exist. Further clarifying the author does not attempt to understand what allows groups or people to form into a fully fledged rebellion, but w hat allows rebel groups to continue to exist and operate the theoretical foundations (Gates 2002) The author focus es (Gates 2002) role s that shape the organizational structure of a rebel force. The author distinguish es

PAGE 57

57 between two types of rewards that participant s rent (Gates 2002) (Gates 2002) Rent seeking award s are primarily financial benefits, such as looting, that participants of a reb ellion share in rewards (Gates 2002) Functional rewards refer to the utility received from the action itself, such as the utility received from fighting or conspiring, but also the utility received from working on the lofty goals of a rebe llion, such as overthrowing the government and establishing a democracy. Solidary reward s refer to the utility received from the (Gates 2002) Rebel groups differ from each other in the types of rewards they offer, and the amount of rewards they offer. The authors t hen propose that individuals decide to participate in a rebellion, if the offered rewards from participating are greater than the rewards offered by (Gates 2002) Gates define s participation in a rebel movement as a cost benefit analysis, where the benefits, in financial and non financial form, e xceed the cost of participation (Gates 2002) and compliance. Members of the rebel army often have an incentive not to comply or adhere to the commands of the leadership of the rebel group. Facing this problem, a rebel group must create enforcement mechanisms that diminish the likelihood of noncompliance. Punishment is a frequently employed strategy to deal with the problem of noncompliance. Gates proposes that geographic location plays an important role in the

PAGE 58

58 decreased ability to successfully punish defection. If agents are far away, they are going to be more difficult to punish (G ates 2002) Additionally, ideological or ethnically homogenous rebels groups will have greater amounts of (Gates 2002) at their disposal. Ethnic and ideological unity can help unify a rebel group increasing the benefit of rebellion to rebel fighters. Military altercations are d etermined (Gates 2002) of both t military effectiveness, distance as it relates to geography and a stochastic element that incorporates aspects of technology, strategy, and rand (Gates 2002) M ilitary strength is partly defined by a stochastic el ement introducing a random or variable term Additionally, the gov (Gates 2002) plays a crucial role in determining military success. Government groups are more likely to win battle s against the rebel groups the closer they are to the center of their territory. Geography also play s a role in allowing safe havens for the group. Sanctuary within a country or within a neighboring country play an instrumental role in giving the rebel movement a chance to develop and grow. Sanctuary implies a place to retreat away from governmental forces. Typically, such sanctuaries are in remote territories well away from the center of government (Gates 2002) (Gates 2002) T he benefits from rebellion can be high enough for a group to outweigh the proximity to the government and proximity is not an all determining factor of rebellion As long as significant rewards exist, civil wars and rebellions can rise up independent of location

PAGE 59

59 The study attempts to understand the underlying factors that allow rebel groups to organize. Geographic dispersion ca n be an important factor allowing rebel groups to form and to maintain organizational structure. Additionally, ideology and ethnicity serve as a crucial factor creating nonpecunitary rewards for participations of rebellion. The stronger the id eological and ethnic ties are the greater the nonpecunarity rewards can demonstrates that physical geography, ethnicity, and ideological distance play an important role in determ ining military success, deterring defection within a rebel group, and sh (Gates 2002) The article Third Party Interventions and the Duration of Intrastate Conflict was published in the Journal of Conflict Resolution in 2002 by Patrick M. Regan. The article focuses on the impact of third parties u pon the duration of civil wars. Third parties are understood to be foreign military forces or multi national peace keeping forces. These third party interventions are usually in the form of a military intervention, but can also be economic or social interve co (Regan 2002) The researchers begin with the assumption that the goal of intervention, from a third party, into a civil war is to control and manage the violence occurring in it. The goal of the third party is to influence the co sts and benefits of fighting, and changing them so that peace and the settlement of conflict will become the preferred outcome. A conflict can be understood as successive moves by various actors gaining an understanding of their likelihood to win the confl ict and their potential payouts from it. The conflict is settled when we reach equilibrium, where both parties are willing to settle the conflict.

PAGE 60

60 An intervention can influence the potential benefits and costs both to continued rebellion and to reaching continuing to fight until victory is a function of expectations of future victory and current (Regan 2002) If the intervention is on the side of the government, Regan argues that the power balance is shifted furt her towards the government, and the government can expect to behalf of the government sho (Regan 2002) Interventions on behalf of the rebel group or opposition would shift the power towards the r ebel group impacting costs and rewards of continuing to fight both for the government and the rebel group. The timing of the intervention is also important to consider. In the early stages of a rebellion, the rebel forces are often less organized and the ir military strength is often marginal As time goes on and the rebel movement survives, the organizational structures of the rebel group solidifies, and the rebel group usually gains in military strength in an armed rebellion should have a considerably greater impact on the (Regan 2002) When an intervention occurs matters to the outcome of the conflict and the impact the intervention can have itself. (Regan 2002) to test empirically the impact of intervention on the length of c ivil wars. The conclusion of whether military or economic do not act as effective tools (Regan 2002) The

PAGE 61

61 data and methodology used in this study appears to have little or no effect on the durat (Regan 2002) bi (Regan 2002) i.e., an intervention on b ehalf of one of the conflict sides, reduces the duration of conflict in comparison to a neutral intervention. The important conclusion, according to the authors, is that an intervention does little to reduce the length or duration of the conflict. Hence, the authors question the effectiveness of military interventions in conflict. Wars Have the Same Causes?: A Conflict Resolution by Nicholas Sambanis. According to theoretical and empirical research has recently helped to identify important economic and political determinants of civil war onsets and prevalence (Sambanis 2001) The theoretical work lumps all civil wars into a single homogenous category, and rese archers identify determinants of civil war using a single homogenous category However, Sambanis argues that there are different types of civil wars, and an analysis of civil wars must be done according to the specific causes of the civil war. There are tw (Sambanis 2001) To distinguish between the two causes of civil war, the definition of ethnicity and ethnic wars has to be established. Ethnic war s are episodes of violent conflict between governments and natio nal, ethnic, religious, or other communal minorities (ethnic challengers) in which the between rival communal groups is not coded as ethnic warfare unless it involves conflict or political power or government policy (Sambanis 2001) All othe r civil w ar are characterized as nonidentity wars.

PAGE 62

62 Collier and Hoeffler (Collier and Hoeffler 2004) and Fearon and Laitin (Fearon and Laitin 2002) argued for an economic explanation of civil wars and conflict. According to these researchers, ethnic and political fractionalization does not explain the causes of civil war in a country. Collier and Hoeffler even argue t hat as ethnic fractionalization rises (Sambanis 2001) Both Collier and Hoeffler and Fearon and Laitin argue d that civil war is primarily caused by economic factors. As long as the benefits of rebellion and civil war are greater than the cost of civil war and conflict, civil war will occur. Sambanis argues that Collier and Hoeffler and Fearon and Laitin arrived at this conclusi on due to faulty a do not consider if different war types have different causes, and their research design, which aggregates all civil wars in a single category, implicitly suggest that there are no (Sambanis 2001) Sambanis uses a data set, which distinguishes between ethnic and revolutionary civil wars. The underlying assumption here is that the causes of such wars are fundamentally different, and only a disaggregated approach will help us understand the determinants of civil wars for both fundamental cases. Sambanis uses a probit model to identify the determinants of ethnic civil wars. (Sambanis 2001) Political variables have previously been determined to be insignificant in analysis of c ivil war ; in this disaggregated approach political variables have become significant. Economic variable s are insignificant and non explanatory for ethnic

PAGE 63

63 variables and is positive ly correlated with the onset of ethnic war: as a country becomes (Sambanis 2001) Furthermore, whether or not a neighbor is in a civil war is an important factor as well If neighboring countries are at war, the probability of civil war increases significantly. The amount of time elapsed, since the last civil war, also i mpact s the probability of civil war breaking out The longer removed the last outbreak of a civil war is, the less the probability is that a civil war will br eak out i n a country. study has concluded that significant differences between ethnic and non ethnic civil wars exist, and determinants vary Ethnic civil wars have a causal relationship with the following variables: the level of democratization in country, the time elapsed since the last civil war, ethnic fractionalization, and neighbors being at war. N on ethnic civil wars have a causal relat ionship with the following variable s: the level of economic growth, overall GDP, and the population size in a country. Sambanis concludes: Ethnic heterogeneity is significantly and positively correlated with the onset of ethnic wars, whereas the economic literature on war initiation has suggested that ethnic heterogeneity either decreases the risk of war onset or has no significant association with the risk of war (Sambanis 2001) Sambanis argues that two types of civil wars exist, and both types of conflict have varying and unique determinants. The article The World E conomy journal by Halvor Mehlum, Karl Moehne, and Ragnar Torvik The economic develop (Mehlum, Moene, and Torvik 2006) Some countries benefit significantly from natural resources, and other countries do not benefit

PAGE 64

64 at all fr om the immense potential offered by the natural resources abundance in their (Mehlum, Moene, and Torvik 2006) between countries The researche r and roducer (Mehlum, Moene, and Torvik 2006) Grabber friendly institutions allow producer s entrepreneurs, or individuals to benefit from an activity or natural resources through the mechanism of rent seeking. Actors, participating in grabbing, profit (Mehlum, Moene, and Torvik 2006) A grabber friendly institution enables or to lerates rent seeking activities, and wealth is not a function of participating in productive activities. A producer (Mehlum, Moene, and Torvik 2006) Economic profit s only can be gained by entrepreneurs through productive and profitable economi c behavior, and rent seeking is not a feasible alternative As institutional quality in a country rises, it becomes less profitable to engage in rent seeking activities and incentives of entrepreneurial behavior rises. The researchers assess (Mehlum, Moene, and Torvik 2006) The result of this study is ality is the key to understanding the resource curse: when institutions are bad, resource abundance is a growth curse; when institutions are good resources a (Mehlum, Moene, and Torvik 2006) The somewhat (Mehlum, Moene, and

PAGE 65

65 Torvik 2006) with high institutional quality have the highest growth rates. The next highest growth rates are experienced by countries with high institutional quality and low resource abundance, followed by countries with low institutional quality and low resource abu ndance. The countries with the lowest growth rates have low institutional quality and high resource abundance. (Hegre and Sambanis 2006) in the (Hegre and Sambanis 2006) there is significant uncertainty about fac tor s impeding and influencing the occurrence of civil war and political conflict. The causal inferences are tenuous at best, a (Hegre and Sambanis 2006) about the significant deter (Hegre and Sambanis 2006) studies Hence, the authors employ a sensitivi ty analysis testing the robustness of models and important result s to small changes in the set of variables included in a regression. Second, we test how fragile our subst antive inferences are to small changes in the way we operationalize theoretically significant variables (which (Hegre and Sambanis 2006) To assess the robustness of the model and Sala i parameter estimates to determine the level of confidence in each of the explanatory (Hegre and Sambanis 2006) Sala i Martin uses the following model form:

PAGE 66

66 (Equation 2 1) (Hegre and Sambanis 2006) vector of (Hegre and Sambanis 2006) z is (Hegre and Sambanis 2006) ctor of up to three (Hegre and Sambanis 2006) taken from a pool of possible, explanatory variables. The distribution of estimated t values is then calculated from the various model specifications thus indicating whether certain variables are significant or insignificant. This al lows the researcher to identify if variable s are robustly significant or insignificant in explaining the onset of civil war. The researchers always include (Hegre and Sambanis 2006) in all of their models: The natural log of population (lnpop), the length of peacetime until the outbreak of a war (pt8, which we model as a decay function of time at peace), and the natural log of per capita gross domestic product (GDP) in the constant dollars (lngdp) (Hegre and Sambanis 2006) Furthermore, the dependent variable, in this case, is the onset of civi l war, with 1 indicating the outbreak of a civil war in a given time period and 0 indicating a peaceful time in a country in a given time period In his original model, Sala i variables for each z (Hegre and Sambanis 2006) Hegre and Samban is alter the Sala i (Hegre and Sambanis 2006) A concept variable is a category under which a variable falls, for example the authors define political systems, measure the same thing in the same model, we restrict the combination of x variables to those that measure thre (Hegre and Sambanis 2006) The

PAGE 67

67 authors include a broad variety of variables in this study. The variables, chosen for this study, have been included in other studies, which attempted to explain the onset of civil war. The included concept variab and (Hegre and Sambanis 2006) Overall, the researchers in (Hegre and Sambanis 2006) able the risk of civil war decreases by 65 percent (Hegre and Sambanis 2006) The population variable also has a significant result, an d as population increases in a country the risk of civil war rises as well. However, one of the core variable s is not as robust as the previous two variables. T he time elapsed since the previous conflict is not signficant in explain ing the onset of civil w ar. categories, and eighteen variables have average p values less than .05 under the least (Hegre and S ambanis 2006) relationships: civil wars are more likely to occur in countries with recent political instability and inconsistent democratic institutions; countries with small militaries and rough terrain; countries located i n war prone, undemocratic regions; and countries with (Hegre and Sambanis 2006) Many other variables do

PAGE 68

68 not explain the onset of civil war though, such as resource dependence. The problem and issue uncovered by this study is the frequently tenuous inference made by researchers in the discipline. The goal of this study was to identif y more causal factors, which robustly explain the onset of civil war. The authors encourage researchers to but differently measured variables in their regressions and add control variables to make sure the result on focus variab les are robust be fore (Hegre and Sambanis 2006) Disa ggregated Studies of Violence, Conflict, and Civil War Disaggregated studies of violence, conflict, and civil war have a differen t methodological approach than country level studies. Disaggregated studies examine conflict not at a national level, but at a small scale level. A country is divided into various regions or geographic areas most commonly a raster which then serve s as th e unit of analysis. A raster typically consists of a matrix of cells organized into rows and columns, where each cell is of an identical size. Conflict and civil war intensity is not homogenous throughout a country, but varies within it. Conflict intensit y can be high in one part of a country and low in another part of the country, and frequently civil war is limited to a certain region of the country. By design, c ountry level studies must assume homogeneity of conflict diffusion in a country, which can in troduce significant causal inference problems. Disaggregated studies account for conflict diffusion and can focus on local determinants of conflict Geography in 2005. The authors argue that historically it had been assumed that there is a strong link between geographical factors and the prevalence and occurrence of civil

PAGE 69

69 between the geographical distribution of physical and human factors in (Halvard Buhaug and Lujala 2005) though Using national or aggregate data sets, researchers have been unsuccessful in establishing a clear causal link between geographical factors and the prevalence of civil war. According to the authors, the use of disaggregated or small scale level data would be suitable to understand causal structures of civil war, and small scale and disaggregated data has two particular advantages. s are by definition subnational events, and the fighting rarely spans (Halvard Buhaug and Lujala 2005) Using aggregated data or country level data to assess civil war consequently is erroneo us, since such analysis would include areas not impacted by the civil war, which are not rel evant for the analysis. Second population distr ibution, and ethnic composition have subst antial va (Halvard Buhaug and Lujala 2005) Disaggregated studies can consider these variations, and include them in their analysis, which increase s control f or sub national variations since there may be huge deviations between nationa l level statistics and conflict specific (Halvard Buhaug and Lujala 2005) The researcher s point out that GIS (Geographical Information System) generated datasets allow for sub national analysis to take place. The researchers developed two different models to test, whether or not there are significant differences between country level analysis onflict (Halvard Buhaug and Lujala 2005) analysis, which is the small scale disaggregate approach. The model measure s the impact of various geographic variables upon the duration of c ivil

PAGE 70

70 (Halvard Buhaug and Lujala 2005) and employing (Halvard Buhaug and Lujal a 2005) For the country analysis, the researcher s include country size, population, territorial confl ict, initial density, mountainous regions forest covers rainy season, commodity exports/ GDP, gemstones, coca, cannabis, and opium as explanatory va riables. For the conflict level analysis, the same general explanatory variables are used The difference between the two models is that the country level model include level measu (Halvard Buhaug and Lujala 2005) whereas t he conflict level model includes disaggregated or small scale measures, which consider the variations inside of a country. stan dard errors and significance levels but even the substantive impact of some (Halvard Buhaug and Lujala 2005) The conflict level analysis indicates that significant impact upon the duration of conflicts, whereas the country level analysis rejects such correlation Additionally, the conflict level model shows no significant impact of mountainous terrain upon the duration of conflict, whereas t he country le vel analysis shows a significant impact. Another discrepancy between the models is the effect of forest cover upon the duration of conflict. T he country level analysis shows a significant effect upon conflict duration, and the conflict level analysis did n ot show such an effect. Importantly, the conflict level analysis included the distance to capital as an explanatory variables, which could not be included in the country specific model is the relati ve location. Civil wars occur at a distance from the capital the

PAGE 71

71 presumed center of state power are much more likely to turn into protracted contest t h an relativel (Halvard Buhaug an d Lujala 2005) The authors conclude that the use of disaggregated data, in conflict analysis, is (Halvard Buhaug and Luja la 2005) Civil wars are regional in nature and the use of GIS generated data improves the analysis of such conflicts. for the conference in 2005 by Halvard Buhaug and Jan Ketil Rd. The literature of civil war and conflict has (Halvard Buhaug and Rd 2005) The unit of analysis in the study of civil war and conflict traditionally was the individual (Halvard Buhaug and Rd 2005) consequently using the c ountry as the unit of analysis is incorrect, since it aggregates the determinants and characteristics of a country wihtout maintaining the variations and differences existing inside of a country. The author s mention three primary reason s why such methodol ogy is continued to be used: there is a lack of disaggregated dataset s explanatory variables are frequently measured at the country level and conflict need to include null and the country might appear to be the only suitable (Halvard Buhaug and Rd 2005) The authors argue that the disaggregated approach is superior to the aggregated approach and should be used to study civil wars and conflicts

PAGE 72

72 The rese ak out on a (Halvard Buhaug and Rd 2005) The country is disaggregated i nto (Halvard Buhaug and Rd 2005) which now serve as the unit of analysis every contestation between a state government and an organized opposition group that caused at least 25 battle (Halvard B uhaug and Rd 2005) In this (Halvard Buhaug and Rd 2005) The authors then test how covariate variation impact s the prevalence or outbreak o f violence in the dif ferent units of analysis. Their objective is to te st several hypotheses from the literature of civil war and conflict concerning the determinants of the prevalence of civil war. ed with the risk of (Halvard Buhaug and Rd 2005) Remote regions are more diffi cult to control for governments and can not easily be reached Additionally, regions with lower road density have typically lower economic development. Hence, as road density increase s, the probability of civil in each unit of observation. associat (Halvard Buhaug and Rd 2005) In the literature there has been an argument that rebels are, at least in part, motivated by economic gains, which often present themselves in the form of natural resources. Rebel groups loot natural resources, and frequently depend on the income of looting

PAGE 73

73 Additionally, the literature has also suggested that ethnic fractionalization impacts ominant (Halvard Buhaug and Rd 2005) Ethnic differences result in political, social, and economic difference s increasing the conflict potential between vari ous ethnic groups. movement by providing shelter out (Halvard Buhaug and Rd 2005) Additionally, geographic dispersion has fre quently been mentioned as an important factor. As region s of the country have a greater geographical dispersion from the capital or center of the country, these areas become more difficult to control, and it becomes more probable for rebellions to begin or is negatively associated with the risk of civil (Halvard Buhaug and Rd 2005) In other words, as population density rises the risk of civil war falls. The logic behind this is that rebel gro up s find it easier to organize and sustain themselves in the outlying, non densely populated regions. The researchers distinguish between two types of civil war in Africa: territorial war, in which the rebel group attempts to secede from the control of th e national government in order to gain territorial control over a region and governmental war s in which the rebel group wants to oust the government, and take political control over the country. Territorial and governmental wars have varying objectives a nd goals. If there is conflict in a grid, the likelihood of conflict in neighboring grids rises as at least at t he selected level (Halvard Buhaug and Rd 2005)

PAGE 74

74 (Halvard Buhaug and Rd 2005) Additionally, c po (Halvard Buhaug and Rd 2005) Surprisingly, (Halvard Buhaug and Rd 2005) an overall unexpected result and the author s propose th is finding to be a unique aspect to conflict in the African context. risk of (Halvard Buhaug and Rd 2005) Overall, insurgency is more li kely to occur in regions that rom the (Halvard Buhaug and Rd 2005) In th e governmental model, the significant explanatory variables differ from the ones determined in the (Halvard Buh aug and Rd 2005) Furthermore, the likelihood of civil wars and conflict rises in densely populated areas. prove invaluable as a supplement to conventio nal country level (Halvard Buhaug and Rd 2005) The researchers point out that a disaggregated research a pproach should be used in cases where sub national characteristics are o f importance. published in the Journal of Conflict Resolution in 2009 by Hvard Hegre, Gudrun stby, ts the location

PAGE 75

75 (Hegre, stby, and Raleigh 2009) In every country, there are areas of relative poverty and relative wealth, and wealth is not equally distributed in a country. uality was arguably an important motivat ion in the Liberian civil (Hegre, stby, and Raleigh 2009) and many fighters were motivated to partici pate in the civil war because of inequality. The research ers investigate in this article whether or not there is relationship between economic deprivations and the location of civil war. The authors first develop and explain the theoretical background to their model. egre, 2009) and (Hegre, stby, and Raleigh 2009) which a local population is likely to support either the rebel group or the government in a (Hegre, stby, and Raleigh 2009) Relatively rich areas of a country can be expected to support the government, whereas the deprived or poorer areas of country can be expected to support rebels. The richer areas pro tect and support the government since they are the benefactors of the current social and economic system. The poorer areas experience neglect from the government, hence these areas will support value refers to the extent to which a location (Hegre, stby, and Raleigh 2009) Certain geographical locations or ar eas in a country have a greater strategic value for the government or rebel groups. The strategic value of a location impacts the extent to which the various actors have a preference of control ling it and are willing to fight for it. In the case of Liberi a, (Hegre, stby, and Raleigh 2009) and rebel groups were militarily strong Since the rebel groups are

PAGE 76

76 militarily s trong and superior to the government, the conflict will occur in areas more supportive of the government The rebel groups will attempt to capture or gain control of the areas supporting the government, which were in theory the economically prosperous area s of Liberia value argument implies that the risk of conflict events increases with the level of (Hegre, stby, and Raleigh 2009) In the Liberian civil war, the (Hegre, stby, and Raleigh 2009) was important for rebel groups to maintain and secure a source of income. Areas with higher relative wealth, in particular diamond deposits, wer e more likely to be the target of attacks Liberia has a population of 2.5 million officially recognized ethnic gr oups living in (Hegre, stby, and Raleigh 2009) The ci vil war in Liberia can be which led to the surrounding of Monrovia, the killing of Samuel Doe, and finally the electoral process that made Taylor president in (Hegre, stby, and Raleigh 2009) unti l (Hegre, stby, and Raleigh 2009) The conflict in Liberia occurred between various ethnic factions; some ethnic groups supported the government, whereas other ethnic groups Liberian conflict followed a clear logic in which warlord pursuit of commerce has been (Hegre, stby, and Raleigh 2009) The civil war occurred along ethnic factions, but the motivation of the rebel g roups can be explained by greed and economic motives.

PAGE 77

77 The authors continue by explaining the aspects of working with a disaggregated data s et information on conflict events by politica ( Hegre, stby, and Raleigh 2009) etween (Hegre, stby, and Ra leigh 2009) reported A basic grid system for Liberia was created (Hegre, stby, and Raleigh 2009) which created 1,375 squares for Liberia events is associated with each grid square and ranges from zero events (in 1,312 squares) to thi rty nine events (Hegre, stby, and Raleigh 2009) referenced information on economic development is deriv ed from the 1986 Liberian (Hegre, stby, and Raleigh 2009) The DHS survey primarily focuses on collecting household data about the health of households, but does include other measures, such as socioeconomic indicators The researchers use the (H egre, stby, and Raleigh 2009) (Hegre, stby, and Raleigh 2009) 1,375 cells wer e covered (Hegre, stby, and Raleigh 2009) and the vast majority of cells do no t encompass a sample point. To o Inverse D points to estimate the values at surrounding points, using an inve rse distance weighting (Hegre, stby, and Raleigh 2009)

PAGE 78

78 with the location of civil w (Hegre, stby, and Raleigh 2009) The authors the third order (Hegre, stby, and Raleigh 2009) Additionally, the authors include the (Hegre, stby, and Raleigh 2009) as an explanatory variable (Hegre, stby, and Raleigh 2009) To include the spatial dependence, a dummy variable, indicating violence in neighboring squares, is included. Furthermore, the authors also include the populatio n density for the different grid s, and include a dummy variable (Hegre, stby, and Raleigh 2009 ) The authors continue by describing the statistical method employed in this study. grid ce (Hegre, stby, and Raleigh 2009) The occurrence of violence in a grid cell is not an independent event, but depends upon the violence in other grid cells, as well as previous violence in the grid cell itself. The authors employed inflated negative binomial model that allows for a large number of zero count observations and possible over dispersion within th e positive (Hegre, stby, and Raleigh 2009) (Hegre, stby, and Raleigh 2009) Diamond deposits do not seem to affect the occurrence of violence in locations (Heg re, stby, and Raleigh 2009) do not significantly

PAGE 79

79 explain the prevalence of conflict or violent events either. Furthermore, the amount of conflict in neighboring cells does not seem to have a significant impact either. conflict in (Hegre, stby, and Raleigh 2009) Conflict intensi ty rises as the wealth and relative wealth in an account (2005), which focuses on the wealth of Liberia as motivatio n for fighting during (Hegre, stby, and Raleigh 2009) explains the location of civil war events best. Civil war events occur primarily in locat ions that present economic opportunities dh Raleigh was published in 2000 level studies of civil war is that large countries more frequently have c (Raleigh and Hegre 2009) There is a strong correlation between country size and the onset of civil war. However, the reason or causal inference why larger countries experience more civil wars has not been clearly determined In this article, the researcher s employ a disaggregate d and sub state level approach to answer the question. related explanations of civil (Raleigh and Hegre 2009) in the conflict and civil war literature. The author s include these explanations and attempt to test them r (Raleigh and Hegre 20 09) Individuals have the same propensity or likelihood to engage in a conflict

PAGE 80

80 or civil war. As the population size increases in a location the like lihood of a conflict breaking rises in the given location A ability of conflict events at a location is likely to be dependent on where the location is situated relative to the capital, a rebel an (Raleigh and Hegre 2009) There and that the utility individuals derive from the public (Raleig h and Hegre 2009) Regions closer to the capital or political center of the country will have better provision of public goods, and regions farther away from the capital will have worse provision of public goods. Peripheral regions have a greater incen t ive to secede from the country and/or engage in civil war, since the provision of public goods is smaller there. Additionally, it also becomes more difficult for the government to military control require s better or ganization, supply lines become more vulnerable to guerilla attacks, and the (Raleigh and Hegr e 2009) (Raleigh and Hegre 2009) Rebel groups, in border regions, have the ability to flee into the neighb oring country and find shelter and protection there Hence, border regions should experience more conflict, since civil war or conflict becomes more advantageous for rebel groups in border regions decreases with the distance from the location to (Raleigh and Hegre 2009)

PAGE 81

81 the imm ediate geogra (Raleigh and Hegre 2009) As population density in an area increases, the location frequently becomes more valuable for both rebels and government gro ups. Hence, these areas naturally become the target of military attacks. more strongly in local population concentrations in locations distant from the capital of countrie (Raleigh and Hegre 2009) and population concentration are factors that are l (Raleigh and Hegre 2009) Civil war events are most likely in distant regions with greater varies with the size of the population of the cou ntry to which the location belongs, cont (Raleigh and Hegre 2009) The researchers employ the Armed Conflict Location Event Dataset (ACLED) to examine the links between population size and the onset of civil war in particular location s The dataset consists of 4,145 battle events for the 1960 2004 period in Africa. In the present analysis, we use 2,530 of these. The remai ning events were dropped either because they we re in countries not included in the analysis, or because information was missing for one of the k ey variables. (Raleigh and Hegre 2009) It is crucial to note that the author s do not distinguish or discriminate between the different events classified in the ACLED dataset. The researchers separate the er portions of 8.6km by (Raleigh and Hegre 2009) which serve as the unit of analysis.

PAGE 82

82 An important aspect in disaggregate d studies is the temporal and spatial dependence between t he various units of analysis. A civil war event in a unit of analysis can be both temporally as well as spatially dependent upon other civi l war events. The these data must handle the de (Raleigh and Hegre 2009) The res earcher can (Raleigh and Hegre 2009) which controls f or the dependency Since the unit of analysis is disaggregated, the control and independent variable must necessa rily be disaggregated as well. The res (Raleigh and Hegre 2009) appropriate for a disaggregated approach. The researchers created the following independent variable s to examine the possible relationship popu ion country variable (Raleigh and Hegre 2009) The researchers also included the following contr ol f event in neighboring (Raleigh and Hegre 2009) The result of the study o have frequencies in proportion to the size location (Raleigh and Hegre 2009) In o ther words, as the population areas rises, the potential for conflict rises there as well. Conflict occurs more frequently in populous areas of a country.

PAGE 83

83 predominately wher (Raleigh and Hegre 2009) Additionally, the researchers also determine that distance t o the capital is not as crucial cture of African internal conflict as prim (Raleigh and Hegre 2009) The result s still indicate that countries with populations that are largely concentrated around th e capital have few er internal conflict events than countries with populations that are spread out, or, even more strongly, are also concentrated in location far from the capital (Raleigh and Hegre 2 009) l in and others examined the determinants of violence and conflict for the North Caucasus in Russia. The authors report the following: the Chechen locational factor region, the Federal Caucasian Hig that in non and in a location are determinants of conflict and civil war in the North Caucasus oughin in several of his papers explores not only the det erminants of conflict, violence and civil war, but also identifies spatial and temporal dimensions of conflict. Conflict and civil war change in their temporal intensity, as well as in their spatial n highlights these changes, which provide for a greater understanding of the particular mechanism s of conflict and civil war. distributions scores for various time periods

PAGE 84

84 ell to analyze conflict patterns and diffusion of violence 2011) To determine 2011) employed a scan statistics imbedded in the SatScan he space time permutation scan statistic used here compares the observed number of events in a space time cylinder to the e xpected number of events within sp The space time permutation scan statistics identifies hot spots of violence and conflict. Furthermore, the I index score for different time period s index provide s information about spatial autocorrelation measuring the r 2011) in the data. The varys in different time periods indicating the altering spatial diffusion of conflict and violence throughout the observation period. Similarly, the mean centers and sta Loughlin allowed for an examination of spatio temporal diffusion. The mean center indicates the center of all events in the datase t, for a particular time period, calculated by averaging the longitudinal and latitudinal data. Hence, the mean center provides a geographic center of conflict and violence enabling a tracking of the spatial center over time. The standard deviational ellipses identify the diffusion of conflict measuring the standard deviation for all conflict events in a given time period. The greater the stan dard deviational ellipse, the greater the diffusion of conflict is, and vise versa. The standard deviational ellipses and mean centers facilitate a tracking of the spatial aspects of conflict for different temporal periods of conflict.

PAGE 85

85 Conflict and civil war data is not homogenous through time and space, but varies with time and through space Understanding spatial and temporal variations of conflict and civil war is necessary for a more comprehensive analysis of conflict and civil war Disaggregated studies of conflict have identified localized determinants of conflict. Determinants of conflict remain context and location specific with demographic determinants being some of the most robust determinants. Most disaggreagated studies of con flict have been undertaken in the African context, and our study will contribute to the disaggregated study of conflict by examining localized determinants of conflict in the n has highlighted the importance of assessing spatial and temporal structure and the spatial diffusion of conflict which improved the understanding of conflict dynamics

PAGE 86

86 Figure 2 1. Political Map of Haiti. Source: Nations Online Project (Nations Online Project)

PAGE 87

87 CHAPTER 3 TEMPORA L AND SPATIAL PATTER NS IN THE HAITI ACLED The ACLED records violent events including temporal and spatial information of the event. The temporal and spatial information of the event provides the researcher with information useful to understand patterns and causal relationship in the underlying been gathered over time. Observations on a variable that arranged temporally, or in a time seque (Burt, Barber, and Rigby 2009) On the other observations on a variable of interest with reference to their (Burt, Barber, and Rigby 2009) Spatial data provides the researchers with data on spatial locations. Both the temporal and spatial components of the data enable the researchers to further explore the caus al structure of conflict Hence, in thi s chapter we will explore the temporal and spatial patterns of conflict in Haiti. Temporal Patterns in the Haiti ACLED sometimes even control the underlying process generating th e observations in the time (Burt, Barber, and Rigby 2009) The tem poral aspect of the ACLED discover statistical regularities consistent with whatever physical or social processes (Burt, Barber, and Rigby 2009) The tem poral aspect of the data allows the researcher to test for significant trend s in the data, which must be considered in the analysis of the data. (Raleigh, Linke, and Dowd 2012) for every conflict event, which is labeled as event date in the datasets. It records

PAGE 88

88 the time, month, and year, when the event occurred. As lo ng as violent activities continue, the ACLED records an event for every single day. For example, if a military campaign in an area starts on March 1st, 1999 and lasts until March 5th 1999 with violent activity reported on each day, then it is coded as five different events in ACLED with a different date for each e ntry (Raleigh, Linke, and Dowd 2012) Sim ple Temporal Frequency Analysis The ACLED for Haiti includes 1055 events from 1997 to 2010. Table 3 1 and Figure 3 1 display the events for every individual year in the observation range. The ACLED events are clearly centered in 2003, 2004, and 2005. 17.82 % of events occur in 2003, 30.52% of events occur in 2004, and 8.72% of events occur in 2005. Hence, 57.02% of all events occurred throughout a three year time period in the dataset, and this simple temporal analysis point s out that political violence had increased significantly then, and underlying social patte rns must be considered in order to explain this pattern. Between 1997 and 2003, the A CLED event count per year range d from 18 to 83 events per year. For four of the six year s the event count per yea r ranged between 49 and 57 events per year. Between 2006 and 2010, the ACLED event count per year range from 8 to 42 eve nts per year, and in 2007 only 8 events were recorded, the lowest total event count per year. Carrol Faubert stated in 2006 in a United Nations Development Report about the security situation in Haiti : Since the departure of Jean Claude Duvalier in February 1986, Haiti has been engaged in a seemingly endless political transition punctuated by several military coups, outbursts of violence and foreign military interventions. The case of Haiti cannot be described as a conflict situation. There has been no recent situation of war with a neighboring country, nor has there been a civil war between opposing Haitian factions or communities. Haiti is a case of a lingering political and governance crisis accompanied by a severe degradation of the economy, of security and of

PAGE 89

89 livelihoods. The country has been trapped in an accelerating downward spiral that will be difficult to halt and reverse (Faubert 2006) Violence, conflict, and uprising occurred frequently between 1997 and 2000. Some years, such as 1998, might have seen less violence and political conflict, but the ACLED suggest that throughout the first presidency of Prval political violence and instabil ity existed. In November of 2000, Aristide was elected to become the new president of Haiti. The ACLED indicates a small spike of political violence and conflict. The electoral process, in a country with high political instability and a tenuous security si tuation, seemed to temporally destabilize Haiti. In 2001 and 2002, political violence and conflict was reduced to the levels prior 2000 again. It was in this time that the opposition against Aristide via the CD and G184 formed and gained in strength. Begin ning in the summer 2001 until the middle of 2003, former Haiti military member s (Hallward 2007) Car second Presidency (Faubert 2006) The ACLED event counts for the various year markedly different than the late 1990s. In 2003, the security situati on in Haiti rapidly deteriorated The ACLED observed 188 events for 2003. In 2003 the opposition against President Aristide had gained in force and resolve. operating mainly in Gonaves, or former members of the disbanded army and the had a free hand in many parts of the country, perpetuating the cycle of violence and impunity (Faubert 2006) In 2004, President Aristide had been ousted from office, and violence and conflict became

PAGE 90

90 epidemic throughout Haiti. The ACLED recorded 322 events for Haiti that year, which we re in part caused by the clashes between various military and paramilitary groups. In 2005, violence and conflict began to drop in Haiti. Another presidential election was held in 2006, and Rene Prval was elected for his second term as the Haitian president. The ACLED data suggest that poli tical violence and conflict noticeably decrease d since the election of Prval politically motivated violence to purely criminal activities,including kidnappings, drug (Faubert 2006) occurred. Since the ACLED focuses primarily on political conflict and violence, the increase in criminal activities might have not been reported. Overall, experts in Haiti agree that levels of political conflict and violence have decreased since 2006. Ex ponentially Weighted Moving Average Statistical Process Control Many of the statistical tools employed for temporal surveillance of events were (Rogerson 2009) Industrial processes need to be controlled to detect shifts in the mean outcome of a process, and are often monitored so that various process paramete rs stay within tolerable limits, and manufactured product (Rogerson 2009) Multiple s tatistical tools, i.e. control charts, have been developed to test for (Lucas and Saccucci 1990) Recently, these control charts and statistical tools have also bee n employed in other disciplines, su ch as epidemiology or geography, though t he ir purpose to use them has not changed To analyze the temporal structure of the data, an Exponentially Weighted Moving Average ( EWMA) scheme is employed for the monthly event co unt of Haiti ACLED. The EWMA control scheme detect s changes in the mean of a variable of int e rest.

PAGE 91

91 The EWMA control scheme can be applied to the ACLED. It allows a n examination of the temporal trend in the historical data. If the EWMA statistics for the AC LED event count per month exceeded the Upper Control Limit ( ) the mean event count per month temporally shifted above the his torical mean, which represent a period of greater political violence in Haiti. If the EWMA statistics for the ACLED event cou nt per month falls below the Lower Control Limit ( ) the mean event count per month temporally shifted below the historical mean, which repres ent a decreased period of political violence in Haiti. Hence, the EWMA will inform us of any sustained tempor al trend s in the data, while disregarding random outliers in the event count per month, and unsustained temporal trends of political violence. To perform the EWMA chart analysis, the qcc package (Root 201 2) for RStudio (RStudio 2012) was employed The weighting parameter was set to be 0.3, and k was set to be 3. The mean of the data is 6.478528, and the standard deviation is 4. 39979. Table 3 2 shows the ACLED event count for every month, as well as the and the EWMA statistics for every single month in the ACLED for Haiti. Since the weighting parameter, to 0.933683, and the UCL converges to 12.02337. If the EWMA statistic falls between 0.933683 and 12.02337, the process is in control. If it the EWMA statistics falls outside that range, the process is out of control. temporary shift of the mean occurred, and violence and conflict surpass ed the historical occurred, and violence and conflict are lower than the historical average.

PAGE 92

9 2 With the help of the qcc package (Root 2012) for RStudio (RStudio 2012) Figure 3 2, of the EWMA control scheme was created. 38 observations of 163 observations are outside the limits, or 23.31% of all months fall outside the In June a n d Jul y indicating a period of lower political conflict and violence. Haiti had just gone throu gh a political transition period; President Aristide had left office in 1997, and was succeed ed by his political all y, Rene Prval The political landscape was dominated by the Lavalas movement, which enjoyed high levels of support throughout Haiti, and in particular with the poor in urban areas Until the middle of 2003, no periods when the EWMA statistic exceeded the were observed. In 2000 from March to June of that year, slightly increased political violence and conflict occurred, while the presidential election process transpired with protest and violence against civilians taking p lace As noted earlier, Carrol Faubert state chaotic situation prevailed during (Faubert 2006) The ACLED does not confirm such an interpretation of the security situati on throughout Aristide presidency ; the levels of violence an d political conflict had remained similar to the levels experienced under President Prval From October 2003 t o June 2005, the EWMA statistic indicating a long sustained period of high political violence and conflict. Political violence and conflict actually started to spike up after Cannibal Army leader Metayer was killed in Gonaves After the killing of Metayer in September of 2003, the level of political violence and conflict skyrocketed in Haiti. The ACLED had recorded between 0 and 10

PAGE 93

93 per month prior to the killing, for the next couple of month s ACLED counts per month rose to be between 40 and 110. The G184 and CD were the political leadership of the movement, whereas former military members, rebel groups, a nd gangs, such as the Can nibal A rmy, led the military effort against Aristide, and his allied gangs and pro Aristide HNP forces. In February 2004, Aristide was ousted from his presidency, and left the country. February 2004 was also the most violent and conflict ridden month in th e entire observation period ; 11.09 percent of all ACLED events occurred in this particular month. Violence and political conflict rapidly fell after the ousting of Aristide from office but only for a short period of time. In the fall of 2004, violence esc in the poorest neighborhoods (Griffin 2004) of Haiti, where the most avid Aristide supporters were. There were numerous reports of g overnment s ponsored violence and extra judicial killings (Griffin 2004) supporters of the ousted co (Griffin 2004) The battle for political control over Haiti was won in Februar y 2004, but the fight continued until the middle of 2005. Carrol Faubert noted the following: The despatching of the Multinational Interim Force at the beginning of 2004 and the subsequent deployment of MINUSTAH helped restore a reasonable level of securit y in most regions of Haiti. The situation was quite different in Port au Prince, however, where armed groups continued their activities and took control of whole neighborhoods By the end of 2005, some neighbourhoods of Port au P rince were considered no go areas for both the de non over the population and engage in criminal activities from those havens. Cit Soleil, with i ts 300,000 people, is the most notorious of these enclaves. (Faubert 2006) The country had been stabilized by the end of 2005, except in the poorer neighborhoods of the big cities in Haiti, which now were in full control of rebel gangs. In

PAGE 94

94 February 2006, Haiti elected a new pre sident. Surprisingly, former president Rene Prval who belonged to the Lavalas movement, won the election, while candidates from t he Latortue government, G184, and CD were unsuccessful. The victory of Prval did not destabilize the country again. The ACLED event counts per month, indicating levels of political violence and conflict, remained low throughout the presidency of Prval between May 2007 and March 2008, and another time between November 2008 and December 2009. Under President Prval Haiti saw long periods of sustained low levels of conflict and violence. Prval has the reputation as a conciliator, and in particular the conflict data reported during his seco nd reign as president confirms this reputation. The EWMA control chart analysis indicates that two distinct periods of conflict and violence were observed in Haiti: from 1997 to September of 2003, and then again from July 2005 to 2010 persistent low inten sity political conflict occurred in Haiti with protest, violence, and clashes between armed groups taking place at a frequent basis. The persistent conflict indicates a violentization of the political process. From September 2003 to June 2005, the conflic t situation had changed. At first a full fledged rebellion similar to a civil war, broke out throughout Haiti, ending in the ousting of President Aristide. Violence and conflict continued afterward, due to the instability created by the ousting and resist ance by various groups opposed to the ousting. Hence, two distinct periods of violence and political conflict exist inside of the Haiti ACLED. Spatial Patterns in the ACLED Event Dataset s locational information as well as attribute (Rogerson 2009) S patial data are data that include geographic information.

PAGE 95

95 O bservations are recorded at different locations and these locations are coded as part (Rogerson 2009) Spatial data impr oves our ability to explore research problems and answer pertinent questions by employing it. The unique aspect of the ACLED is the provision of spatial conflict data. Every recorded event is provided with longitude and latitude information, which means c onflict can now be studied at a localized level. Analysis Structure Dissaggregated studies of conflict, violence, and civil war have historically employ ed a grid structure in their research design (Hegre, stby, and Raleig h 2009; Halvard Buhaug and Rd 2005; Raleigh 2007) The study area typically several adjacent countries, is divided into a grid of varying sizes. possible level, bot h spatially and temporally, leaving decisions regarding the appropriate unit of aggregation to the analyst (Raleigh et al. 2010) For this analysis of conflict violence and civil war in Haiti, we decided not to employ a grid structure for the analysis In 2005 order administrative enti (Halvard Buhaug and Rd 2005) Buhaug has reservations towards such an approach though, since first order administrative units are frequently time variant, and have different geographic sizes, which m akes analysis more challenging. Most of the disaggregated studies encompass unction and size of regions varying extens ively (Halvard Buhaug and Rd 2005) The goal of this study is to examine conflict, violence, and civil war in Haiti. We decided to employ first order administrative units as the unit of analysis. Since the focus

PAGE 96

96 of this study is only Haiti, the reservations about size and function variations in first order administrative units between different countries (Halvard Buhaug and Rd 2005) in a particular research context will not exist in our study design. Furthermore, the demographic variables are provided at the first order a dministrative unit, which makes the choice of first arrondisments, and 133 communes. The following analysis employs the 133 communes as the unit of analysis. Simple Spatial Analysis Using the geo coordinates of the events, we created several maps outlining the geographic patterns of violence and conflict in Haiti. Using ArcGIS, we counted the number of events in the various communes giving us a count of event s for every single commune We c reated several frequency maps for different temporal period s and for different types of violence and conflict, which are included in the ACLED. Geographic pattern of conflict and violence in Haiti between 1997 and 2010 Conflict and political violence in H aiti is clustered in certain communes throughout Haiti, as Figure 3 3 and Table 3 3 shows. The commune Port au Prince has 44.36%, Delmas has 12.42%, Gonaves has 9.86%, and Cap Hatien has 3.98% of all conflict events The communes encompass some of the l argest cities in Haiti. Hence, conflict events occur with great frequency in the large Haitian cities. Conflict, in particular, is clustered in the metropolitan area of Port au Prince, which includes the commune s of Port au Prince, Delma s Ption Ville an d Carrefour.

PAGE 97

97 In 70 communes zero conflict events were recorded and for 26 communes only one event was recorded between 1997 to 2010. Conflict event s are clustered in Haiti, and political violence and conflict can be extreme in the population centers of Hai ti. Geographic pattern of conflict and violence in Haiti for three distinct time periods The EWMA process control suggested that three distinct periods existed in Haiti between 1997 and 2010. Conflict and violence were particularly high between 2003 and 2005. From 2006 to 2010, the EWMA suggested that conflict was at times historically low, and Haiti observed relative political stability. From 1997 to 200 2, president Prval and Aristide, both leaders in the Lavalas movement, were in power and average lev els of conflict were observed. We decided to examine the geographic pattern of conflict and violence for these three distinct time period s The objective i s to examine, whether or not difference in spatial diffusion between the three time periods exist From 1997 to 2002 314 conflict events occurred, from 2003 to 2005 602 events occurred, and from 2006 to 2010 139 events occurred Conflict and violence, regardless of time period are clustered in the greater metropolitan area of Port au Prince For areas in Haiti difference in the spatial pattern for the different time period s can be observed. Between 1997 and 2002, as noted in Table 3 4 and Figure 3 4, the greater metro politan area of Port au Prince observed 66.24% of the event s Significant levels of conflict also occur in Gonaves, Saint Marc, Cap Hatien, and Petit Gove. Conflict is clustered in the population centers of Haiti in this time period. A vast majority of communes also experience zero events of conflict in this time period. Between 2003 and 2005 as noted in Table 3 5 and Figure 3 5, the intensity of conflict increased throughout Haiti, and 57.1 percent of all conflict events occur in thi s

PAGE 98

98 period Conflict continues to be clustered around the greater metropolitan area of Po r t au Prince. However, conflict intensity rises in the north of Haiti. In Gonaves 14.78% of all events occurred, and 5.65% events occurred in Cap at against Aristide in 2003 and 2004 began in those two communes, with rebel groups seizing control of both cities before the ousting of President Aristide. local phenomena, solely centered in one part of the country, but conflict existed in most parts of Haiti with many communes at least one zero conflict event Between 2006 and 2010, as noted in Table 3 6 and Figure 3 6, the spatial patterns of violence and political violence evolved Conflict intensity has historically been high in the greater metropolitan area of Port au Prince. I n those year s the spatial clustering in Port au Prince intensified with 78.41percent of all conflict events taking place there. A striking aspect of this time period is the pacification of both G onaves and Cap Hatien. B oth communes had been the epicenter of the rebellion against Aristide in 2003, but now returned to a state of extremely low violence. The rest of the country also returned to a state of political stability with low levels of conf lict. The geographic diffusion of conflict changed between the three time periods. Between 2003 and 2005, conflict and violence was more diffused in Haiti, a nd conflict intensity had also become high in population centers, such as Gonaves, Saint Marc, C ap Hatien, and Petit Gove. In the periods post and prior to 2003 to 2005, conflict has a more clustered pattern with political conflict remaining stable and relatively high in the greater metropolitan area of Port au Prince. Geographic pattern of confli ct and violence in Haiti for different event types currently codes for eight types of events, both violent and non violent, that may occur

PAGE 99

99 during a civil war, instability or state fai (Raleigh, Linke, and Dowd 2012) Tabl e 3 7 signifies that the ACLED records some of the event types more frequently than other s Battle No Change of Territory, Protest, and Violence Against Civilians combine to account for 92.51% of all recorded events. We created maps for the three most com mon event types to assess the spatial diffusion of them. The other event types had insufficient number of observation s to reliably assess their spatial diffusion. The spatial diffusion for the three different event s appears very comparable, as demonstrated in Figure 3 7, Figure 3 8, Figure 3 9. Conflict and violence are clustered in the major population centers of Haiti with the greater metropolitan area of Port au Prince, consisting of the communes of Port au Prince, Delmas, Carrefour, and P tion Ville, having the highest conflict intensity In the greater metropolitan area of Port au Prince 61.88% of all protest occurs, 88.54% of all violence against civilians occurs, and 67.20% of all battle with no change in territory occur s Interestingl y, the amount of violence against civilians is highly clustered in Port au Prince. The exact reason for such significant clustering is not known, but could indicate that violence against civilians is more frequently reported there, or armed groups, such as the police, rebel, or army groups, have a greater propensity for violence directed toward civilians. The spatial diffusion of conflict appear s similar throughout Haiti, and no particular distinct difference can be detected with a simple spatial comparison method. Spatial Autocorrelation space in which its values at a set of locations depends on values of the same variable at (Burt, Barber, and Rigby 2009) When spatial autocorrelation is (Burt, Barber,

PAGE 100

100 a nd Rigby 2009) and spatial clustering exists. If spatial autocorrelation is zero on the other hand the distribution of a variable is random, and no spatial patterns exist. If (Burt, Barber, and Rigby 2009) Positive autocorrelation indicates that there is a spatial pattern in the data, and the values of a variable are closely related in space to each other. Such spatial patterns are not random, but a process causes them to occur. etects spatial autocorrelation in data In the context of this research, we want to assess the spatial distribution of conflic t events in communes in Haiti, so we To calculate the M score for conflict in Haiti we employ ArcGIS (ESRI 2011a) T he number of conflict events in a given commune was the var iable of interest ArcG IS provides several ways to calculate a spatial weight matrix and the polygon contiguity method to calculate the spatial weight matrix was chosen With polygon contiguity the spatial weight matrix is solely based upon the values of the variables of the adjacent areas or polygons Additionally, we decided to utilize r ow standardization, which is preferred to no standardization, when using a contiguity method. I Results After initially determining that there is spatial clustering of conflict events in Haiti between 1997 and 2010, we tested for spatial autocorrelation using the G test. The G 80444 ( Table 3 11 ) which is significant at the one percent level. The data exhibits positive spatial

PAGE 101

101 autocorrelation; conflict events are clustered in Haiti, and conflict is more prevalent in certain regions of the country, especially in the greater metro politan area of Haiti. Furthermore, significant positive autocorrelation indicates that a spatial pattern exist in the data. Such spatial patterns are not random, but a process causes them to exist Hence, the spatial autocorrelation indicates that a proc ess influences the occurrence and pattern of conflict in Haiti, and the goal of this research is to uncover the underlying process causing the spatial pattern of violence. The Haiti ACLED records three event types frequently enough to have them u ndergo st atistical analysis by themselves whereas other events are recorded to o infrequently to be tested Battle No Change of Territory, Protest, and Violence Against Civilians combine to account for 92.51% of all recorded events. We tested the spatial autocorrel ation of all three event types. Table 3 12 records the result of the global for each of the event types The conflict events for all three types of events, namely protest, violence against civilians by an armed groups, and battles between tw o armed groups, are all significantly spatially clustered at the one percent 8.065148 for battles between various actors and 7.890661 for violence aga 3.715248 T he indicating that spatial clustering of protest is not as severe as the spatial clustering of the two other event types. Conflict diffusion for various event types diverges in Haiti. A possible explanation could lie in a particular distinction between the event types. Protest requires a group of

PAGE 102

102 people with an agenda and sufficient motivation to advocate for a change in social norm, social practice or in institutional framework Violence against civilians by an armed group or battle between two armed groups requires an armed group willing to engage in a battle or violent behavior. Protest can easily arise in any part of the country, but violent conflict only occurs when an armed group is present. The difference in spatial clustering might simply be a result of the fact that armed group s are either not present throughout all of Haiti, or find it not necessary to engage in violent conflict throughout Haiti. B enefit s of engaging in v io lent conflict might not exist in some parts of the country or may not be possible but protest can still occur there. If that is the case, spatial clustering of violent clustering necessarily must be higher We also calculated a rent time period s to allow for sample sizes (Bailey and Gatrelll 1995) 25 (Bailey and Gatrelll 1995) Due to low numbers of t otal events for several years, we decided to combine 1997 and 1998 with each other 2006, 2007, and 2008 with each other and also 2009 and 2010 with each other Figure 3 9 and Table 3 12 clustering of conflict events intensified in Haiti from an already significa nt level. The peak of spatial cluste ring occurred in 2000. T then fell from 10.87464 in 2000 to 5.45333 in 2002 which is primarily caused by conflict intensity rising throughout Haiti. Interestingly, from 2000 to 2002 opposition to President Aristide was strengthening gaining in organizational capacity and vigor in their effort to topple

PAGE 103

103 Aristide. In September 2003, the opposition toward Aristide broke out into a fully fledged rebel movement with paramilitary and gangs operating thr oughout Haiti, while gaining control of which is not significant at the five percent level. Conflict events in communes no longer exhibit ed a pattern of spatial autocorrelatio n in 2003. So why did conflict diffuse throughout the rebel movement against President Aristide? The capital often represents the center of power for the government. The power of the government frequently diminishes with distance from the capital, since it becomes more challenging to militar il y and /or politically control areas further away from the capital. Hence, a rebel movement has a greater probability of succeeding in regions in the country farther away from the capital. Conflict spread through Haiti m ost likely in 2003 because initiating a rebel movement had the greatest probability of success in areas other than Port au Prince. President Aristide was ousted from office in 2004, and spatial clustering o f conflict start ed to rise again. Conflict s have r emained spatially clustered since 2004. In general, conflict events in Haiti have tended to be significantly clustered. The rebellion against Aristide buck ed that trend. In 2003, conflict was diffused throughout Haiti. We employed ArcGis (ESRI 2011a) to detect local spatial cluster for conflict events in Haiti. ArcGis allows the weight matrix, in other words the conceptualization of the spatial relationships between areas to be determined in several different ways. We employed the inverse distance weight method with a maximum distance of 63 kilometers Only features of less kilometer than 63 kilome ter will be used to calculate the and nearby features will impact local

PAGE 104

104 Furthermore, we also employed row standardizat ion, as recommended by Anselin (Anselin 1995) The shown in Figure 3 11, indicate s that the greater metropolitan area of Port au Prince has increased levels of conflict and violence. The LISA has significant positive Z Score s for the communes of Port au Prince, Delmas, and Ption Ville ; all three communes are part of the greater metrop olitan area of Port au Prince Port au Prince is the economic, social and political center of Haiti, and most conflict is centered there as well. Furthermore, the LISA indicates that the commune of Gonaves is a spatial outlier. The communes surrounding Gonaves have a low conflict intensity, whereas in Gonaves 9.86 percent of total conflict occurred. The LISA points out that Gonaves is an area with high conflict in a region with low surrounding conflict, in that sense Gonaves is considered to be a spa tial outlier. identify Cap Hatien as a spatial outlier, a commune with a high event count surrounded by communes with relatively low conflict events. ts are centered in the greater metropolitan area of Port au Prince, with Gonaves having been a commune of greater conflict intensity as well. Mean Center and Standard Deviational Ellipse In combination, the mean and the standard deviational ellipses enab le an employed the use of mean center and standard deviational ellipse in the ir analysis of conflict in the N orth Caucasus and it will serve as an important tool to analyze conflict patterns in Haiti.

PAGE 105

105 E mploy in g ArcGIS, we calculated both the mean center of the data and the standard deviational el lipse for various time periods to assess va riations in the spatial diffusion of conflict data (Figure 3 12) C onflict and viole nce is centered and primarily located in and around Port au Prince in 1999, 2000, and 2001 In 2002, the mean center shifted north indicating that political violence and conflict intensity rose in the north of Haiti The mean conflict continued to shift northward in 2003; in the fall of that year an open rebellion against President Aristide broke out, and rebel forces were particularly active in Gonaves and Cap Hatien A t the end of 2003, rebel forces, gangs, and former military members controlled the north of Haiti, and started to move towards Port au Prince with the explicit goal of gaining political power and ousting President Aristide. In 2004, the trend of northward shifting political conflict and violent was reversed as the mean center of conflict shift to the south again Early in the year, the rebel movement had been so successful militarily that President Aristide fled Haiti, and relinquished his presidency. Confl ict intensity shifted southward again In 2005, the mean center of conflict and political violence was located in Port au Prince, just as it had been in years prior to 2002. O verall, the mean center of conflict has remained fairly stable over the year s be ing in close proximity to Port au Prince and thus indicating that conflict intensity is greatest there. However, a geographical diffusion of conflict occurred in 2002, and intensified in 2003. In 2004, conflict shifted southward then. A rebellion against a government frequently is initiated in parts of the country that are farther away from the capital. As the rebel movement gained in strength, the southward move toward the political power center of Port au Prince occurred The rebellion against the reign of Aristide shifted the spatial pattern of conflict in Haiti. The spatial pattern of this period is

PAGE 106

106 markedly different from the years prior and post. Since 2005 conflict was centered in and around Port au Prince again. From 1999 to 2001, t he standard deviational ellipses indicate that conflict intensity was limited in the north of Haiti; the northward extension of the standard deviational ellipses was small in this time period In 2000, political violence and conflict was slightly elevated in the south west ern part of Haiti i n particular in Anse d'Hainault stretching the standard deviational ellipses in the west east direction. C onflict were centered and located geographically in the south of Haiti. In 2002, conflict intensified in the no rth of Haiti, in cities such as Saint Marc, Gonaves and Cap Hatien Figure 3 13 illustrate s the change in the north south direction in kilometers over time. Conflict intensity remains high in Port au Prince over time, but intensified throughout Haiti a nd the overall spatial pattern of conflict changed Figure 3 13 illustrate the diffusion, and prevalence of conflict in the north, as distance rises conflict intensity increases in the north of Haiti. In 2000, the north south distance of standard deviatio nal ellipses was 29.44 kilometers, and increased in 2003 to 76.68 kilometers. The standard deviational ellipses clearly indicate the significant spread of violence and conflict toward the north of Haiti, as the opposition movement against Aristide intensif ied and rebel activity strengthened in resolve. Buhaugh had (H. Buhaug 2010) President Aristi de enjoyed large support in the capital of Haiti, where the Lavalas was strong and gang supported him. The political and military strength was high in Port au Prince for President Aristide whereas political and military

PAGE 107

107 strength throughout Haiti remained more tenuous and uncertain. The rebellion and civil war was initiated in Gonave and Cap Hatien which confirms the theory that outbreaks of conflict and civil war occur in more remote regions from the capital. The small geographical extent of the standard deviational ellipses in 2005 indicates that the pattern of spatial diffusion was altered in 2005. In 2005, the Minustah troops had been in Haiti for over a year, and pro Aristide groups had been demilitarized or defeate d throughout Haiti. Conflict intensity remained high throughout Port au Prince, but the standard deviational ellipse for 2005 indicates the significant pacification of the rest of Haiti. It is not surprising that conflict intensity remained high in Port au Prince, since popular support for President Aristide had been high there. Port au Prince represents the political power base for Aristide, and the pacification of Port au Prince was not as easily accomplished. The standard deviational ellipse for 2006 to 2008 had a substantial west east direction, since protest and conflict was high in Les Cayes, which is in the south western part of Haiti. Additionally, we also calculated the mean center and standard deviational ellipse for different types of events, nam ely for battles between two armed groups with no territorial change, protest of a group, and violence of an armed group against civilians. Figure 3 14 represents the spatial diffusion for the different types of conflict The mean centers for all the diffe rent conflict types are in close proximity to each other. The distance between the mean center for battle s no change of territory and violence against civilians is 6.67 kilometers, the distance between protest and violence against civilians is 5.30 kilomet ers, and the distance between violence against civilians is just 1.70 kilometer. The center of conflict for the various event types is similar.

PAGE 108

108 Furthermore, the standard deviational ellipses for the event types do not vary significantly from each other. T different from the other two event types, and was slightly more diffused. The standard deviational ellipses and mean center approach does not suggest that a significant difference between event types exists.

PAGE 109

109 Table 3 1 Simple temporal frequency analysis of Haiti ACLED Year Number of Events Percent of Total Events 1997 55 5.21% 1998 18 1.71% 1999 57 5.40% 2000 83 7.87% 2001 52 4.93% 2002 49 4.64% 2003 188 17.82% 2004 322 30.5 4 % 2005 92 8.72% 2006 23 2.18% 2007 8 0.76% 2008 27 2.56% 2009 39 3.70% 2010 42 3.98% Total 1055 100 .00 % Note: Percent of total events is calculated by dividing the number of events per year by the amount of total events in the observation period. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 110

110 Table 3 2 EWMA s tatistic for Haiti ACLED Month Event Count EWMA Statistic LCL UCL Jan 97 5 6.034969325 2.518717 10.43834 Feb 97 10 7.224478528 1.644962 11.31209 Mar 97 14 9.257134969 1.270059 11.687 Apr 97 7 8.579994479 1.09588 11.86118 May 97 3 6.905996135 1.012558 11.9445 Jun 97 2 5.434197294 0.972191 11.98486 Jul 97 3 4.703938106 0.952519 12.00454 Aug 97 4 4.492756674 0.942905 12.01415 Sep 97 2 3.744929672 0.9382 12.01886 Oct 97 1 2.92145077 0.935896 12.02116 Nov 97 2 2.645015539 0.934767 12.02229 Dec 97 2 2.451510877 0.934214 12.02284 Jan 98 2 2.316057614 0.933944 12.02311 Feb 98 1 1.92124033 0.933811 12.02324 Mar 98 3 2.244868231 0.933746 12.02331 Apr 98 1 1.871407762 0.933714 12.02334 May 98 0 1.309985433 0.933698 12.02336 Jun 98 0 0.916989803 0.933691 12.02336 Jul 98 0 0.641892862 0.933687 12.02337 Aug 98 4 1.649325004 0.933685 12.02337 Sep 98 2 1.754527503 0.933684 12.02337 Oct 98 1 1.528169252 0.933684 12.02337 Nov 98 2 1.669718476 0.933684 12.02337 Dec 98 2 1.768802933 0.933683 12.02337 Jan 99 4 2.438162053 0.933683 12.02337 Feb 99 3 2.606713437 0.933683 12.02337 Mar 99 7 3.924699406 0.933683 12.02337 Apr 99 10 5.747289584 0.933683 12.02337 May 99 11 7.323102709 0.933683 12.02337 Jun 99 8 7.526171896 0.933683 12.02337 Jul 99 1 5.568320327 0.933683 12.02337 Aug 99 3 4.797824229 0.933683 12.02337 Sep 99 5 4.85847696 0.933683 12.02337 Oct 99 3 4.300933872 0.933683 12.02337 Nov 99 1 3.310653711 0.933683 12.02337 Dec 99 1 2.617457597 0.933683 12.02337 Jan 00 3 2.732220318 0.933683 12.02337 Feb 00 6 3.712554223 0.933683 12.02337 Mar 00 11 5.898787956 0.933683 12.02337 Apr 00 8 6.529151569 0.933683 12.02337 May 00 15 9.070406098 0.933683 12.02337 Jun 00 11 9.649284269 0.933683 12.02337

PAGE 111

111 Table 3 2. C ontinued Month Event Count EWMA Statistic LCL UCL Jul 00 6 8.554498988 0.933683 12.02337 Aug 00 5 7.488149292 0.933683 12.02337 Sep 00 1 5.541704504 0.933683 12.02337 Oct 00 2 4.479193153 0.933683 12.02337 Nov 00 12 6.735435207 0.933683 12.02337 Dec 00 3 5.614804645 0.933683 12.02337 Jan 01 2 4.530363251 0.933683 12.02337 Feb 01 1 3.471254276 0.933683 12.02337 Mar 01 4 3.629877993 0.933683 12.02337 Apr 01 2 3.140914595 0.933683 12.02337 May 01 0 2.198640217 0.933683 12.02337 Jun 01 6 3.339048152 0.933683 12.02337 Jul 01 1 2.637333706 0.933683 12.02337 Aug 01 3 2.746133594 0.933683 12.02337 Sep 01 0 1.922293516 0.933683 12.02337 Oct 01 4 2.545605461 0.933683 12.02337 Nov 01 8 4.181923823 0.933683 12.02337 Dec 01 21 9.227346676 0.933683 12.02337 Jan 02 2 7.059142673 0.933683 12.02337 Feb 02 2 5.541399871 0.933683 12.02337 Mar 02 0 3.87897991 0.933683 12.02337 Apr 02 2 3.315285937 0.933683 12.02337 May 02 6 4.120700156 0.933683 12.02337 Jun 02 2 3.484490109 0.933683 12.02337 Jul 02 1 2.739143076 0.933683 12.02337 Aug 02 4 3.117400153 0.933683 12.02337 Sep 02 8 4.582180107 0.933683 12.02337 Oct 02 1 3.507526075 0.933683 12.02337 Nov 02 11 5.755268253 0.933683 12.02337 Dec 02 10 7.028687777 0.933683 12.02337 Jan 03 10 7.920081444 0.933683 12.02337 Feb 03 10 8.544057011 0.933683 12.02337 Mar 03 0 5.980839907 0.933683 12.02337 Apr 03 3 5.086587935 0.933683 12.02337 May 03 1 3.860611555 0.933683 12.02337 Jun 03 1 3.002428088 0.933683 12.02337 Jul 03 2 2.701699662 0.933683 12.02337 Aug 03 7 3.991189763 0.933683 12.02337 Sep 03 22 9.393832834 0.933683 12.02337 Oct 03 45 20.07568298 0.933683 12.02337 Nov 03 34 24.25297809 0.933683 12.02337 Dec 03 53 32.87708466 0.933683 12.02337

PAGE 112

112 Table 3 2. C ontinued Month Event Count EWMA Statistic LCL UCL Jan 04 70 44.01395926 0.933683 12.02337 Feb 04 117 65.90977148 0.933683 12.02337 Mar 04 31 55.43684004 0.933683 12.02337 Apr 04 3 39.70578803 0.933683 12.02337 May 04 8 30.19405162 0.933683 12.02337 Jun 04 1 21.43583613 0.933683 12.02337 Jul 04 6 16.80508529 0.933683 12.02337 Aug 04 4 12.96355971 0.933683 12.02337 Sep 04 15 13.57449179 0.933683 12.02337 Oct 04 33 19.40214426 0.933683 12.02337 Nov 04 14 17.78150098 0.933683 12.02337 Dec 04 20 18.44705069 0.933683 12.02337 Jan 05 4 14.11293548 0.933683 12.02337 Feb 05 14 14.07905484 0.933683 12.02337 Mar 05 16 14.65533839 0.933683 12.02337 Apr 05 10 13.25873687 0.933683 12.02337 May 05 10 12.28111581 0.933683 12.02337 Jun 05 12 12.19678107 0.933683 12.02337 Jul 05 8 10.93774675 0.933683 12.02337 Aug 05 10 10.65642272 0.933683 12.02337 Sep 05 0 7.459495906 0.933683 12.02337 Oct 05 3 6.121647134 0.933683 12.02337 Nov 05 0 4.285152994 0.933683 12.02337 Dec 05 5 4.499607096 0.933683 12.02337 Jan 06 2 3.749724967 0.933683 12.02337 Feb 06 3 3.524807477 0.933683 12.02337 Mar 06 0 2.467365234 0.933683 12.02337 Apr 06 3 2.627155664 0.933683 12.02337 May 06 0 1.839008965 0.933683 12.02337 Jun 06 2 1.887306275 0.933683 12.02337 Jul 06 6 3.121114393 0.933683 12.02337 Aug 06 0 2.184780075 0.933683 12.02337 Sep 06 0 1.529346052 0.933683 12.02337 Oct 06 4 2.270542237 0.933683 12.02337 Nov 06 1 1.889379566 0.933683 12.02337 Dec 06 3 2.222565696 0.933683 12.02337 Jan 07 3 2.455795987 0.933683 12.02337 Feb 07 1 2.019057191 0.933683 12.02337 Mar 07 0 1.413340034 0.933683 12.02337 Apr 07 0 0.989338024 0.933683 12.02337 May 07 0 0.692536617 0.933683 12.02337 Jun 07 1 0.784775632 0.933683 12.02337

PAGE 113

113 Table 3 2. C ontinued Month Event Count EWMA Statistic LCL UCL Jul 07 1 0.849342942 0.933683 12.02337 Aug 07 0 0.594540059 0.933683 12.02337 Sep 07 0 0.416178042 0.933683 12.02337 Oct 07 1 0.591324629 0.933683 12.02337 Nov 07 1 0.71392724 0.933683 12.02337 Dec 07 0 0.499749068 0.933683 12.02337 Jan 08 0 0.349824348 0.933683 12.02337 Feb 08 1 0.544877043 0.933683 12.02337 Mar 08 1 0.68141393 0.933683 12.02337 Apr 08 22 7.076989751 0.933683 12.02337 May 08 0 4.953892826 0.933683 12.02337 Jun 08 1 3.767724978 0.933683 12.02337 Jul 08 2 3.237407485 0.933683 12.02337 Aug 08 0 2.266185239 0.933683 12.02337 Sep 08 0 1.586329667 0.933683 12.02337 Oct 08 0 1.110430767 0.933683 12.02337 Nov 08 0 0.777301537 0.933683 12.02337 Dec 08 0 0.544111076 0.933683 12.02337 Jan 09 0 0.380877753 0.933683 12.02337 Feb 09 1 0.566614427 0.933683 12.02337 Mar 09 2 0.996630099 0.933683 12.02337 Apr 09 3 1.597641069 0.933683 12.02337 May 09 3 2.018348749 0.933683 12.02337 Jun 09 22 8.012844124 0.933683 12.02337 Jul 09 0 5.608990887 0.933683 12.02337 Aug 09 2 4.526293621 0.933683 12.02337 Sep 09 4 4.368405535 0.933683 12.02337 Oct 09 0 3.057883874 0.933683 12.02337 Nov 09 0 2.140518712 0.933683 12.02337 Dec 09 2 2.098363098 0.933683 12.02337 Jan 10 15 5.968854169 0.933683 12.02337 Feb 10 10 7.178197918 0.933683 12.02337 Mar 10 5 6.524738543 0.933683 12.02337 Apr 10 3 5.46731698 0.933683 12.02337 May 10 5 5.327121886 0.933683 12.02337 Jun 10 3 4.62898532 0.933683 12.02337 Jul 10 1 3.540289724 0.933683 12.02337 Note : Event count indicates the number of ACLED event per month. EWMA statistic represents the exponentially weighted moving average statistics for each month. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 114

114 Table 3 3. ACLED event count for communes in Haiti between 1997 and 2010 Commune Event Count Percent of Total Event Port au Prince 468 44.36% Delmas 131 12.42% Gonaves 104 9.86% Cap Hatien 42 3.98% Saint Marc 38 3.60% Ption Ville 36 3.41% Petit Gove 35 3.32% Cayes 18 1.71% Hinche 15 1.42% Carrefour 13 1.23% Belladre 11 1.04% Mirebalais 11 1.04% Ounaminthe 10 0.95% Jacmel 10 0.95% Jrmie 7 0.66% Croix des Bouquets 7 0.66% Grand Gove 7 0.66% Ennery 6 0.57% Miragone 6 0.57% Note: Communes with less than 0.50 percent of total events have been excluded in this table. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 115

115 Table 3 4. ACLED event count for communes in Haiti between 1997 and 2002 Commune Event Count Percent of Total Event Port au Prince 147 46.82% Delmas 42 13.38% Petit Gove 19 6.05% Ption Ville 19 6.05% Gonaves 15 4.78% Saint Marc 10 3.18% Cap Hatien 8 2.55% Carrefour 8 2.55% Belladre 6 1.91% Croix des Bouquets 4 1.27% Hinche 3 0.96% Anse d'Hainaul 3 0.96% Cabaret 3 0.96% Jacmel 3 0.96% Mirebalais 2 0.64% Jrmie 2 0.64% Ounaminthe 2 0.64% Grand Gove 2 0.64% Logne 2 0.64% Cayes 2 0.64% Note: Communes with less than 0.50 percent of total events have been excluded in this table. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 116

116 Table 3 5. ACLED event count for communes in Haiti between 2003 and 2005 Commune Event Count Percent of Total Event Port au Prince 244 40.53% Gonaves 89 14.78% Delmas 61 10.13% Cap Hatien 34 5.65% Saint Marc 28 4.65% Petit Gove 16 2.66% Ption Ville 15 2.49% Hinche 11 1.83% Mirebalais 8 1.33% Ounaminthe 7 1.16% Ennery 6 1.00% Miragone 6 1.00% Jacmel 6 1.00% Belladre 5 0.83% Grand Gove 5 0.83% Jrmie 4 0.66% Trou du Nord 4 0.66% Port de Paix 4 0.66% Dondon 4 0.66% Anse Galets 4 0.66% Massade 3 0.50% Croix des Bouquets 3 0.50% Ganthier 3 0.50% Carrefour 3 0.50% Note: Communes with less than 0.50 percent of total events have been excluded in this table. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 117

117 Table 3 6 ACLED event count for communes in Haiti between 2006 and 2010 Commune Event Count Percent of Total Event Port au Prince 77 55.40% Delmas 28 20.14% Cayes 14 10.07% Pestel 2 1.44% Gros Morne 2 1.44% Carrefour 2 1.44% Ption Ville 2 1.44% Hinche 1 0.72% Lascahobas 1 0.72% Mirebalais 1 0.72% Jrmie 1 0.72% Verettes 1 0.72% Baraderes 1 0.72% Ounaminthe 1 0.72% Acul du Nord 1 0.72% Limb 1 0.72% Saint Saint Raphal 1 0.72% Logne 1 0.72% Jacmel 1 0.72% Note: Communes with less than 0.50 percent of total events have been excluded in this table. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 118

118 Table 3 7 ACLED event type distribution Event Type Event Count Percent of Total Event Battle Government Regains Territory 7 0.66% Battle No Change of Territory 253 23.98% Battle Rebels Gain Territory 25 2.37% Headquarter / Basecamp 1 0.09% Nonviolent Rebel Activity 32 3.03% Nonviolent Transfer Territory 14 1.33% Protest 389 36.87% Violence Against Civilians 334 31.66% Source: User creation using Haiti ACLED (Raleigh et al. 2010) Table 3 8 Spatial diffusion of protest in Haiti Commune Event Count Percent of Total Event Port au Prince 202 51.93% Gonaves 47 12.08% Cap Hatien 19 4.88% Delmas 19 4.88% Saint Marc 16 4.11% Petit Gove 15 3.86% Ption Ville 14 3.60% Cayes 10 2.57% Jacmel 6 1.54% Hinche 5 1.29% Anse Galets 5 1.29% Grand Gove 4 1.03% Jrmie 3 0.77% Ganthier 3 0.77% Carrefour 3 0.77% Ennery 2 0.51% Miragone 2 0.51% Ounaminthe 2 0.51% Cabaret 2 0.51% Logne 2 0.51% Note: Communes with less than 0.50 percent of total events have been excluded in this table. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 119

119 Table 3 9 Spatial diffusion of violence against civilians in Haiti Commune Event Count Percent of Total Event Port au Prince 165 65.22% Delmas 38 15.02% Gonaves 27 10.67% Cap Hatien 15 5.93% Ption Ville 14 5.53% Petit Gove 10 3.95% Saint Marc 9 3.56% Carrefour 7 2.77% Hinche 5 1.98% Belladre 5 1.98% Croix des Bouquets 4 1.58% Jacmel 4 1.58% Cayes 4 1.58% Ounaminthe 3 1.19% Miragone 2 0.79% Jean Rabel 2 0.79% Port de Paix 2 0.79% Logne 2 0.79% Note: Communes with less than 0.50 percent of total events have been excluded in this table. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 120

120 Table 3 1 0 Spatial diffusion of battles with no change of territory in Haiti Commune Event Count Percent of Total Event Port au Prince 90 35.57% Delmas 70 27.67% Gonaves 25 9.88% Petit Gove 7 2.77% Ption Ville 7 2.77% Saint Marc 6 2.37% Cap Hatien 4 1.58% Hinche 3 1.19% Belladre 3 1.19% Lascahobas 3 1.19% Mirebalais 3 1.19% Croix des Bouquets 3 1.19% Carrefour 3 1.19% Cayes 3 1.19% Miragone 2 0.79% Note: Communes with less than 0.50 percent of total events have been excluded in this table. Source: User creation using Haiti ACLED (Raleigh et al. 2010) Table 3 1 1 conflict in Haiti Moran's I Score Expected Moran's I Z score P Value 0.203949 0.007576 6.980444 0.00000 Note: = Z score of 1.96 indicates significant clustering at the five percent confidence level. Source: User creation using Haiti ACLED (Raleigh et al. 2010) Table 3 1 2. types of conflict events in Haiti Event Type Moran's I Score Expected Moran's I Z score P Value Battle No Change of Territory 0.353833 0.007576 8.065148 0.00000 Protest 0.083154 0.007576 3.715248 0.00020 Violence Against Civilian 0.196605 0.007576 7.890661 0.00000 Note: = Z score of 1.96 indicates significant clustering at the five percent confidence level. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 121

121 Table 3 1 3. for various time periods Year Number of Events M oran's I Score Expected Moran's I Score Z score P Value 1997 1998 73 0.14440 0.0076 6.27276 0.00000 1999 57 0.12154 0.0076 9.25688 0.00000 2000 83 0.277577 0.0076 10.87464 0.00000 2001 52 0.32935 0.0076 6.44471 0.00000 2002 49 0.27651 0.0076 5.45333 0.00000 2003 188 0.04453 0.0076 1.11851 0.26335 2004 322 0.14651 0.0076 4.95860 0.00000 2005 92 0.25965 0.0076 7.69937 0.00000 2006 2008 58 0.28282 0.0076 6.03969 0.00000 2009 2010 81 0.09853 0.0076 8.48079 0.00000 Note: = Z score of 1.96 indicates significant clustering at the five percent confidence level. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 122

122 Figure 3 1 Temporal frequency chart for Haiti Armed Conflict Location and Event Dataset between 1997 and 2010 0 50 100 150 200 250 300 350 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Number of Conflict Events

PAGE 123

123 Figure 3 2 EWM A chart for monthly Haiti ACL ED event data. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 124

124 Figure 3 3 ACLED event count for t he 133 communes between 1997 and 2010. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 125

125 Figure 3 4 Percentage of total events for the 133 communes between 1997 and 2002 Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 126

126 Figure 3 5 Percentage of total events for the 133 communes between 2003 and 2005 Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 127

127 Figure 3 6 Percentage of total events for the 133 communes between 2006 and 2010 Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 128

128 Figure 3 7 Percentage of total protest events for the 133 communes between 1997 and 2010 Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 129

129 Figure 3 8 Percentage of total violence against civilians events for the 133 communes between 1997 and 2010 Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 130

130 Figure 3 9 Percentage of total battles no change of territory events for the 133 communes between 1997 and 2010 Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 131

131 Figure 3 10 evolution over time. Note: Z score of 1.96 indicates significant clustering at the five percent confidence level. Source: User creation using Haiti ACLED (Raleigh et al. 2010) 0.00000 2.00000 4.00000 6.00000 8.00000 10.00000 12.00000 Z-Score

PAGE 132

132 Fig ure 3 11 Note: High high clustering refers to a cluster of features with high values. Low low clustering refers to a cluster of features with low values. High low outlier refers to a feature with a high value surrounded by features with low values. Low high outlier refers to a feature with a low value surrounded by features with high values. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 133

133 Figure 3 12 Mean center and standard deviational ellipses for different time periods Note: The point shape represents the mean center, and the circle shape represents the standard deviational ellipse. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 134

134 Figure 3 1 3 Standard distance of north south direction in kilometers over time Source: User creation using Haiti ACLED (Raleigh et al. 2010) 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2008 2009 2010 Standard Distance of North South Direction in kilometers

PAGE 135

135 Figure 3 1 4 Mean center and standard deviational ellipses for different conflict events from 1997 to 2010 in Haiti. Source : User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 136

136 CHAPTER 4 ANALYZING LOCAL PATTERNS OF CONFLICT IN HAITI The study of conflict and civil war at a local, disaggregated level has become more frequent and common in recent years ( e.g. Buhaug and Rd 2005; Hegre, stby, 2011; Buhaug, Cederman, and Rd 2008) Prior to the last decade conflict and civil war were studied at a country level The goal of country level studies was to identify determinants of civil war, intrastate war, and conflict. These studies have deter diverse, and oil (Halvard Buhaug 2010) are more prone to civil war and conflict. At the same time, country level studies face challenges in fully explaining confli ct and civil war s Some of the empirical results in the civil war literature are frag ile, but others are not (Hegre and Sambanis 2006) Country level studies have robustly identified several determinants of c onflict but inference s have remained tenuous for others, such as religious and ethnic fractionalization. Furthermore, the result of country level studies most likely do not rudimentary transition of knowle dge from the state to the sub state level provides some (Halvard Buhaug 2010) Country level determinants of conflict and civil war must not be expected to be similar to sub state level determinants. Disaggregated studies of conflict and civil war have become more frequent in recent years. The use of Geographical Information S ystems (GIS) and disaggregated data sources have made studies of local conflict and civil war possible. Disaggregated studies have further expl ored conflict propensity and conflict characteristics (Halvard

PAGE 137

137 Buhaug and Lujala 2005) in localized contexts, and broadened the understanding of conflict mechanisms The geographical focus of disaggregated studies has primarily been in Africa so far primarily due to the high prevalence of conflict in Africa, and the prevalence of several dataset with sub state level information on conflict in Africa ( e.g. Buhaug and Rd 2005; Hegre, stby, and Raleigh 2009; Buhaug and Lujala 2005) Other res d focus ed on the Middle E ast or parts of the former Soviet Union in their research Little attention has been given to the Americas in disaggregated studies of conflict. Disaggregated Study of Conflict in Haiti Carol Faubert stated the following about Haiti in 2006: Haiti has been engaged in a seemingly endless political transition punctuated by several military coups, outbursts of violence and foreign military interventions. Haiti is a case of a lingering political a nd governance crisis accompanied by a severe degradation of the economy, of security and of livelihoods (Faubert 2006) The Haitian state has been described as a predatory state. Political power is primarily used to gain economic prosperity and wealth in Haiti and seldom has p olitical power been used to engage in long term economic growth and development. Mats Lundhal explained the predatory state in this way: Successful politics was a way to make the best of livings, pure and simple. A monotono usly repetitive pattern was es tab lished whereby a clique made its way into government via a revolution or a coup, emptied the treasury, sometimes indebting the state in the process, before it was toppled by another revolution or coup, which put the next clique into office, etc., ad nausea m (Lundahl 2008)

PAGE 138

138 Historically, control over the state was crucial in gaining economic prosperity a nd wealth in Haiti. The existence of a predatory state in Haiti encouraged the frequent use of violence to gain power and control of the stat e Haiti never had a stable and peaceful s ecurity situation, but rather political instability and low levels of con flict remained part of the Haitian identity throughout its history. history, most of the analysis of conflict has been purely descriptive in nature, and few systematic analyses h ave been undertaken. In 2006, Kolbe and Hutson used a random sample household survey to assess levels of conflict and violence in Port au Prince in (Kolbe and Hutson 2006) in the greater metropolitan area of Port au Prince and asked them about their exposure to violence since ime and systematic abuse of human rights were common in Port au P rince. Although c riminals were the most identifi ed perpetrators of violations, political actors and UN soldier s were also (Kolbe and Hutson 2006) Geographic patterns of v iolence in Haiti have not been systematically explored, nor have determinants of conflict and violence for Haiti been determined. The ACLED provide s geographic and temporal data of conflict and violence in Haiti. The dataset covers the period between 1997 and 2010, and allows for the disaggregated study of conflict and violence in Haiti. It enable s the exploration of multiple questions about conflict in Haiti. What determinants correlate with conflict and violence in Haiti? Are the in other contexts such as countries in Africa ?

PAGE 139

139 Determinants of Conflict in Haiti The determinants included in the analysis of conflict i n Haiti were primarily selecte d apriori the statistical analysis since those determinants had been shown to be correlates of violence in other countries or regions. Environmental, g eographic, demographic, economic, and institutional determinants have been used in prior disaggregated st udies of conflict. All of our variables were collected at the third order administrative level. In Haiti the third ord er administrative units are the communes. The choice of the commune as the level analysis enables us to study conflict determinants at a s mall scale level while main taining an internal coherence for the unit of analysis. Haiti is a n ethnic, religious, and cultural ly homogenous country. In most other studies of conflict and civil war, ethnic and religious fractionalization was considered in the analysis of conflict and violence. In Africa, in particular, multiple distinct ethnic groups exist inside of a country, and ethnic tension impacts conflict propensity. Furthermore, in countries, such as Nigeria, civil war and conflict has had a religi ous dimension. The unique, relative homogeneity of Haiti allows us to disregard ethnic and religious difference focusing our attention on economic, demographic, institutional, and geographic factors. Population Size in a Given Location T he U.S. Census Bu reau provides population and demographic data for Haiti The U.S. C ensus B resolution gridded population maps based ( U.S. Census Bureau 2010) to provide more information and help in times of natural disaster s which are a common occurrence in Haiti. The U.S. census bureau also provide d census data at the commune level, which were employed to analyze conflict in Haiti.

PAGE 140

140 The proposed relationship between population size and conflict propensity is that when popul ation size rises in a given location, the propensity for conflict increases as well. Hegre and Raleigh stated the following about the proposed relation ship : The simplest explanation of the national level relationship between population size and the risk an d extent of conflict is based on the If ther e is a given probability that a randomly picked individual starts or joins a reb ellion, then the risk of rebellion increases with populati on (Raleigh and Hegre 2009) All other things equal, the risk of conflict and violence at a given location must increase with population size rising at the given location. Figure 4 1 shows the popul ation size per commune in Haiti. The population size per commune varies significantly throughout Haiti. Population size is particularly high in the greater metropolitan area of Port au Prince, Gonaves Cap Hatien and Saint Marc Overall population size is relatively low in the south west ern part of Haiti, and in the north east ern part of Haiti. In the particular context of Haiti, the conflict propensity should be greater in the Port au Prince, Gonaves Cap Hatien and Saint Marc and lower in the south west and north east of Haiti. Urban Population Percentage in a Given Location The U.S. Census Bureau also provided the percentage of urban population in a given commune. Several researcher s have determined that a link between the percentage of urban population and conflict and violence in a location exists and Witmer 2011; Raleigh and Hegre 2009; Halvard Buhaug and Rd 2005) tiple studies find that urban and densely populated places experience far more conflict than peripheral, rural areas (Raleigh et al. 2010)

PAGE 141

141 In urban areas, the proximity of people to each other is smaller than in rural area s. P eople find it easier to organize themselves together to engage in protest, violent b ehavior, or engage in conflict, since proximity to each other is lower. Furthermore, an economic explanation for the tendency of urban areas to be more conflict ridden exists as well. Urban areas represent economic and political power center s and due to the ir importance urban areas become the target of conflict and violence. Additionally, in urban area s cluster s of wealth and poverty exist in close proximity to each o ther The overall wealth disparities of urban areas could motivate groups or individuals to engage in acts of conflict to possibly capture some of the wealth and negotiate a different distribution of the wealth Similarly, wealth disparities present diffe rence s in preferences and ideas, which can lead to greater conflict potential between the wealthy and the poor. The Lavalas movement advocated the rights, ideas and preference s of the poor, and wealthy Haitian s generally were opposed to these goals. Betwee n 1997 and 2010, the political ideas and preferences between the poorest Haitian s and wealthiest Haitian s did not overlap Figure 4 2 shows the urban population percentage in a given commune in Haiti. Not surprisingly, commune s with major population center s in them also have large urban population percentage s The interior of the country predominately has low urbanization rates indicating that most of the country is rural, instead of urban In general, urbanization is greater on the Haitian coastline. Lastl y, there are several communes with low population size number, but larger urbanization rate s A particular

PAGE 142

142 example of this is the north eastern commune of Fort Libert which has 26,679 inhabitants but an urbanization rate of 62.31 percent. Male Population Percentage in a Given Location The U.S. Census Bureau also provided the percentage of male population in a given commune. The male or female population percentage throughout a developing country is not homogenous, b ut significant variations in m ale and female population percentage s can occur throughout a country. The variations are partially caused by internal and international migration patterns. The economic opportunities for male and female citizen s vary geographically, and migration enables c itizens to benefit from the presented opportunities. Stefan Alscher a researcher at the Centre of Migration, Citizenships and Development, in Bielefeld Germany stated the following about the Dominican omen are looking for alternatives in do mestic service and are forced to migrate t o urban centers (Alscher 2011) Females have a greater propensity for urban migration than male citizen s in the Dominican Republic, and our a priori assumption is to expect a similar trend in Haiti. Figure 4 3 shows the male population perce ntage in various communes throughout Haiti. Male population percentage is high in the western part of Haiti. The male population percentage is generally low in the major population centers of Haiti, such as the greater metropolitan area of Port au Prince, Gonaves Cap Hatien and Saint Marc The possible impact of male population percentage on conflict propensity could be two fold For one, males might have a higher propensity of engaging in conflict or violent behavior. Armed groups, armies, and rebel groups recruit primarily males, and most armies and armed groups consist solely of males.

PAGE 143

143 Furthermore, females migrate to metropol itan areas searching for economic opportunities. Urban and metropolitan areas might present these economic opportunities but these economic opportunities could in turn enhance conflict propensity. If e conomic opportunities are scarce in migration destinat ion areas then in migration centers significant parts of the population might be economically grieved or dissatisfied. Hence, male population percentage might increase as well as decrease conflict propensity in a given location Adult Population Percentag e in a Given Location The U.S. Census Bureau also provided the percentage of adult population percentage in a given commune. The adult population percentage in a given location depends upon an array of factors. The birth rate in ru ral areas exceeds birth r ates in urban areas, and the number of adult population consequently should be higher in urban areas. Additionally, rural to urban migration could impact adult population percentage in a locatio n In general, young adult s have greater social mobility, and young rural adult s would be most like ly to migrate to urban center s The adult population percentage in a given commune is impacted by migration patterns, and the econ omic necessities behind the causes of migration, and most impo rtantly it represents the r ural urban divide. In general, adults have a great er conflict propensity than non adult s Conflict should be more frequent as the adult population percentage in a commune rises. Distance from Political Center In previous disaggregated studies of confl ict and civil wars (Halvard Buhaug, Cederman, and Rd 2008; Raleigh and Hegre 2009) the relative distance of a location in relation to the political center was shown to have an impact upon conflict propensity

PAGE 144

144 in a given location The political power of the government remains strongest in the (Halvard Buhaug, Cederman, and Rd 2008) ; the government is expected to find it easier to militarily and politically control areas in the capital or close to the capital than areas further away from the cap The capability of a country (a.k.a. its national strength) is largest at its home base and declines as the nation moves away. Capable states are relatively less impeded by distance and can therefore (H alvard Buhaug 2010) The Haitian state has historically been weak, but interestingly between 1997 and 2010 the state might have been temporarily stronger Robert Fatton described the massive corruption a nd state predation (Fatton 2006) existed. The Haitian state did not change much since the 1980s However, the Lavalas movement headed by Aristide, but also supportive of Prval was strong and aligned themselves with the state throughout their respective reigns. Hence, s ince both regimes enjoyed popular support the power of the state might have been higher than usual both during Prval s The distance between a given commune and Port au Prince is calculated by measuring the distance of the centroid o f the commune and the centroid of the Port au Prince commune. Elevation Data in Haiti American Geospatial Information Center (MAGIC) provides access to (MAGIC 2011) remote sensing programs with Texas based developments in leading edge information (MAGIC 2011) In response to the earthquake in Haiti in 2010, MAGIC developed geospatial elevation models from remote sensing data sources.

PAGE 145

145 Localized geographical factors can impact conflict propensity in a given location. Elevated areas can serve as a refuge area for rebel groups, and can be of strategic value to rebel groups, due to the cover and protection these areas offer. Furthermore, military actions in low lying areas should be preferable to military groups with high military strength. The probability of success of ambushes by military group s with limited military strength is lower in low elevation areas. Hence, the stronger military force in a conflict would prefer military actions to take place in low elevation areas while the weak military force would prefer action to take place in high elevation areas, which provide s shelter and protection. Using the ele vatio n data provided by MAGIC, we calculated the mean elevation per commune. Figure 4 5 shows the mean elevation per commune. Furthermore, we also calculated the standard deviation of elevation in a given commune. The standard deviation of elevation measures elevation differences in a commune. Commune with greater elevation differences could possibly make ambush attack s more likely. The likelihood of success of attack might be greater for an armed group if geographic retreat locations are given. Areas with lo w standard deviations of elevation represent fewer opportunities to retreat after a battle, and could reduce overall conflict propensity in that location. Figure 4 6 shows the standard deviation of elevation. Both the mean elevation per commune measurement and the standard deviation of elevation measurement attempt to encompass a localized geographic component into the analysis of conflict. The standard deviation of conflict measure s localized variation of elevation to a greater degree, whereas the mean ele vation is a more global measurement

PAGE 146

146 Border with the Dominican Republic In disaggregated studies of conflict, violence, and civil war a particular emphasis has been ebel groups may operate more easily in border areas since neighboring countries may provide (actively or tacitly allow) safe zones (Raleigh and Hegre 2009) The propensity of conflict in border regions can be higher, due to the prevalence of rebel groups operating in the bordering countries and crossing the border to engage in battles or armed conflict. In the particular context of Haiti, the G184 and CD held frequent strategic meetings throughout the Dominican Republic the only country to have an i nternational border with Haiti Furthermore, former Haitian military members resided in the Dominican Republic. The Dominican Republic was a safe haven for the anit Aristide opposition. To inc lude the border effect into the study of conflict propensity, we developed a dummy variable indicating whether or not a commune directly borders the Dominican Republic border or not. Departmental Capital in Haiti vided into three different levels: ten deparments, 41 arrondisments, and 133 communes. Every department also has a provincial capital. The provincial capital represents the regional center of political power. As a departmental capital, a city has important political infrastructure and serve s as a regional hub. The political power of the departmental capital also influence s conflict propensity. P rotest frequency can be expected to be higher in provincial capitals, since groups of people will advocate for cha nge more frequently in regional centers of political power. Furthermore, armed groups and rebel groups should have an incentive to attempt to

PAGE 147

147 gain control of a provincial capital, due to its strategic importance. Similarly, government forces have a greater motivation to defend the provincial capital, due to the importance of controlling the provincial capital Provincial capitals represent a significant strategic value to protester s armed groups, and government forces. Hence, conflict propensity should be greater in these cities. If a commune was a provincial capital, a dummy variable was created for it. Distance to Route Nationale is comprised of the eight Route Nationales The Route Nationale s connect the major population centers of Haiti, and most of the inter regional transport goes along the Route Nationale. Geocommons (GeoCommons 2010) provided a shapefile including all of the road s in Haiti; from the shapefile we extracted the eight Route Nationale highways Figure 4 7 shows the Route Nationale highway system in Haiti. The major highway system in Haiti runs thr ough many of the communes, and certainly all of the major population centers, but there are also areas not connected to the Route Nationale highway system, such as the south western part and north western part of Haiti. Conflict propensity should be greate r along the Route Nationale highway system Buhaugh stated that m ajor highways and strategically located airfields and military bases massively increase the mobility and speed (Halvard Buhaug 2010) of armed groups. Armed groups engage in conflict along major transportation networks, due to the strategic importance of the transportation network Furthermore, regions of the country, which are not connected to the major transportation network, should experience lower conflict propensity. In developing countries, travel to regions not connected to the major road network becom es arduous

PAGE 148

148 and time consuming. Armed groups should have few incentives to participate in armed conflict in remote regions. Regions not connected to the major highway sy stem are frequently less developed and less populous, which could reduce conflict propensity in those regions. DHS Wealth Index Score Sub national income and consumption surveys can be sparse in developing countries. In Hai ti, a couple of surveys measure income or consumption at the department level, which is the first order administrative level In this study, we attempt to examine conflict at the third order administrative level. The Demographic and Health Surveys (DHS) conduct surveys in many population, health, HIV, and nutrition through more than 3 (ICF International 2012) In addition to providing a vast array of information about health, education, and nutrition, the DHS provide s information about household wealth. The DHS surveys include information about A common problem with both hous ehold income and consumption expenditures is their volatility. Income is very changeable in less developed countries, on b (Rutstein and Johnson 2004) The DHS wealth index measures the amount of assets a household own s and composites them into a single index. Household assets are distributed different ly at different wealth leve ls, and this fact allows the composition of a wealth index. Assets included in DHS wealth surveys usually are type of flooring, water supply, sanitation facilities, electricity, radio, television, telephone, refrigerator, type of vehicle, persons per sleep ing room, ownership of agricultural land, domestic servant, and country specific items. The DHS wealth index is calculated in the following way:

PAGE 149

149 Each household is assigned a standardized score for each asset, where the score differs depending on whether o r not the household owned that asset (or, in the case of sleeping arrangements, the number of people per room). These scores are summed by household, and individuals are ranked according to the total score of the household in which they reside. The sample is then divided into population quintiles -five groups with the same number of individuals in each (ICF International 2012) In Haiti the DHS wealth index scores range from 159 743 to 390 349 with the higher score in dicating greater asset wealth. We then standardized the index score dividing it by 10,000, so that the index score ranged from 15.9743 (lowest) to 39.0349 (highest) for each household The DHS survey in 2005 included 339 survey sample locations, and 9998 household observations throughout Haiti. Six of the survey locations were not provided with geographic coordinates, indicating that the GIS experts working on the Haitian DHS survey did not trust the reliabili ty of the provided coordinates. Hence, we in turn discarded those six survey location s At every survey location, we calculated the average DHS wealth score index to have a composite wealth score for every survey location. The survey locations contain geo (ICF International 2012) The displacement is not expected to significantly impact our analysis, due to the chosen methodology. Figure 4 8 shows the approximate DHS survey locations in Haiti. The survey locations are distributed on a frequent basis representative at the third order administrative level, the communes, though. Hegre, stby, and R aleigh encountered a similar issue in their analysis of conflict in Liberia (Hegre, stby, and Raleigh 2009) 101 of the

PAGE 150

150 1,375 grid cells were covered by the Liberian DHS (Hegre, stby, and Raleigh 2009) The vast majority of their units of analysis included no survey sample locations. The researchers employed the Inverse Distance Weighted spatial interpolation methodology points to estimate the (Hegre, stby, and Raleigh 2009) Using this interpolation method, the researcher estimated th e DHS wealth index for every unit of analysis in their analysis of conflict in Haiti Simila r to Hegre, stby, and Raleigh we employed a spatial interpolation method to estimate continuous DHS wealth scores throughout Haiti. Instead of using the IDW, I employed the ordinary K riging method. The ordinary K based on statistical models that include autocorrelation that is, the statistical relationships among the measured points (ESRI 2011b) The following was stated about Kriging: Kriging assumes that the distance or direction between sample points reflects a spatial correlation that can be used to explain variation in the surface. The Kriging tool fits a mathematical function to a specified number of points, or all points within a specified radius, to determine the ou tput value for each location (ESRI 2011b) The Kriging meth od and IDW both use the value of the variable of interest in surrounding location s to estimate a value in a given location. The Kriging method not only the distance between the measured points and the prediction location but also on the overall spatial arrangement of the measured points (ESRI 2011b) Using ArcGIS, we estimated the prediction valu e for every location in Haiti. We decided to limit the number of sample points to perform the interpolation to six points,

PAGE 151

151 and we chose this relatively low number of points to ensure greater local variability i n the estimated wealth values Furthermore, we limited the search distance to 25 kilometers hence only sample points within a 25 kilometers radius were employed in the interpolation. For the vast majority of regions and points the maximum number of sample points were used to calculate the location estimate, but some locations, in particular in the rural areas of Haiti, used less than six sample locations. Figure 4 9 shows the wealth distribution estimated using the ordinary Kriging method in Haiti. To arrive at an average wealth index score per commune we multiplied the DHS score with the population size in a given location Figure 4 10 shows the population for every 0.1 kilometers by 0.1 kilometers location in Haiti and divide d it by the overall popu lation count in a commune: (Equation 4 1) where is the estimated population at a particular point location, is the estimated DHS wealth score at a particular point location, and is the overall p opulation in a given commune. The AWC is the estimated average wealth per person in a given commune. Figure 4 11 displays the average wealth per commune in Haiti. Generally, the coastal areas of Haiti appear to be wealthier than inner lying communes, whic h are primarily rural, agricultural areas. The major population centers in Haiti have greater wealth levels than communes with no major population centers. Statistical Analysis The dependent variable in our case is the number of conflict event s per commun e ; the conflict data set for Haiti is a count data set (Hubbard 200 6) The

PAGE 152

152 Poisson distribution assumes that the mean is equal to the variance. In our model, the mean of the data is 7.9323 events per com mune, but the variance is 1863.2; the high variance in the data suggests that a Poisson regression model is not suit ed to analyze conflict dynamics. T here is an extension to Poiss on regression, called negative binomial regression, which can account for greater than Poisson variation and is based on the negative binomial distribution (Hubbard 2006) The negative binomial regression model is explained in detail in Appendix E. Covariate Selection The dependent variable is the number of events in a commune in a given time period, or for a specific ACLED event type, such as protest. The covariate selection depended upon an a priori selection of pertinent determinants of conflict, and functionality inside of the specified model. Initially, we creat ed a covariance matrix Table 4 1, for the time invariant covariates and the dependent variable s event counts per commune between 1997 and 2010. The covariance analysis shows no significant correlations between any of the proposed covariates. However, we decided to exclude the percentage of urban population in a given commune, since urban population percentage and the sum of population in a given commune are too similar in what they attempt to measure The included covariates in our analysis are: Mean elev ation of the commune. A dummy variable indicating whether or not a commune borders the Dominican Republic. A dummy variable indicating whether or not a commune also has a city with a department capital in it.

PAGE 153

153 The total population in a given commune divid ed by 10,000. We use a scalar to simplify the interpretation of our results. Dis tance from the geographic center point of the commune to the capital, Port au Prince. The distance of the commune cemtroid to the Route Nationale highway sy s tem. Estimated average wealth level per commune. The Variance Inflation Factor (VIF) (National Institute of Standards and Technology 2002) Mulitcollinearity can result in numerically unstable estimates of the regression coefficients (National Institute of Standards and Technology 2002) The minimum VIF a covariate can have is one; in that case no correlation between the covariate of interest and all other covariates exists. Frequently, a value of greater than four for a covar iate has been mentioned as critical, and indicates that multicollinearity could be an issue in the data. The largest VIF for any covariate is 2. 2071 for the population measure in a commune For the vast majority of covariates, the VIF is actually less than 2, and no concern arises for these variables. In general, it is probably safe to assume that multicollinearity is not a significant issue for the included covariates. Model Validity The ACLED is both temporal and spatial. The temporal aspect will be considered to some extent in our analysis of conflict; the spatial aspect of the data will not be considered by the nega tive binomial regression model, which could introduce significant bias. Spatial autocorrelation impacts the estimates of the regression analysis, when the residual s are spatially correlated with each other To assess any introduced bias we tested the residuals of the global model, the negative binomial model including all

PAGE 154

154 events between 1997 and 2010. Table 4 3 shows the for the residuals of the negative binomial regression T 0.05958, which is not significant at the five percent confidence level. In our data, the residuals of the negative binomial regression are not spatially correlated with e ach other, and are independent of each other. Simon Jackman, part of the Stanford Political Science D epartment, published the pscl (Jackm an 2012) package for the R statistical program platform. The program contains a fu nested hypothesis test, which compares (Jackman 2012) In our analysis, we compared the negative binomial regression to a zero inflated negative binomial regression model. A zero inflated negative binomial regression model accounts for exces sive zero c ounts in the dependent variable in addition to having the same properties as the negative binomial regression model. The zero inflated negative binomial regression model assume s that a distinct process occurs, which causes the zero counts in th e data to be inflated A zero inflated negative binomial regression model has been used by Hvard Hegre, Gudrun stby, and Clionadh Raleigh in their analysis of conflict in Liberia (Hegre, stby, and Raleigh 2009) due to overdispersion, use of count data, and excessive zero count dependent variable s nested hypothesis test, we compared the model fit of the negative binomial regression model and the zero inflated negative binomial re gression nested hypothesis test established a test statistic of 2.3516 w hich correspond to a p value of 0.0093. The null hypothesis can be rejected that both

PAGE 155

155 models fit the data equally well, and it is 99 percent certain that the negative binomial model outperforms the zero inflated negative binomial model. Residual and deviance plots are an important model in checking statistical normal regression situation s such as logistic regression or log linear analysis, the residuals may be so far of f normality and from having equal variance as to (Dunn and Smyth 1996) The authors develop a randomization (Dunn and Smyth 1996) Figure 4 12 shows the randomized quantile residuals for the general conflict model, which enables us to check the residual distributition. The residual distribution does not indicate that the model is mispecified The assumption of normality is given in our model. Determinants of Conflict i n Haiti for Various Time Periods In Table 4 4 and Table 4 5 we report the estimates of the negative binomial regression for the Haiti conflict data. The table reports the coefficients, standard errors, and p values for the various determinants of conflict. The dependent variable in the negative binomial regression is the natural logarithm of the count data, and interpretation of the coefficient must be done in light of that. The negative binomial regression model is defined as: (Equation 5 2) where is the dependent variable, number of conflict events per commune. The number of conflict events depends upon the model specifications. is a vector of covariates, including the following covariates: Elev refers to the mean elevation of t he commune.

PAGE 156

156 DomR refers to a dummy variable indicating whether or not a commune borders the Dominican Republic. Dep refers to a dummy variable indicating whether or not a commune also has a city with a department capital in it. Pop refers to the total population in a given commune divided by 10,000 We use a scalar to simplify the interpretation of our results. DC refers to the d istance to the capital, Port au Prince. RN refers to t he distance of the commune centriod to the Route Nationale highway sy stem. Wea refers to the e stimated average wealth level per commune. Model 1 is the global model of conflict including all conflict events in Haiti between 1997 and 2010. Three determinants are significant at the one percent level in the model, and two de terminants are significant at the five percent level. If a commune includes a capital of the first order administrative units, (i.e. departments), conflict propensity in the first order administrative unit rises. The population size per commune is the stro ngest determinant of conflict. As population size rises in a commune, conflict (Raleigh and Hegre 2009) In Haiti, conflict propensity is greater in the population centers of the country, and significantly less in the rural, less populated areas of the country. The conflict literature has identifie (H. Buhaug 2010) of civil war and conflict, and our analysis confirms this. The distance to Port au Prince, the capital, is significant as well; the coefficient is negative. As distance to the capital rises, conflict propensity decreases. Conflict is more prevalent in Port au Prince than in areas further away from it. Civil wars primarily have two motivations and justifications. It is to seek autonomy/secession or incre ased governmental influence (H.

PAGE 157

157 Buhaug 2010) In the case of Haiti, conflict is not caused by considerations of autonomy or secession by individual or groups, but Haiti is marked by a continuous struggle for government influence and contro l. Distance to the capital matters, since political actors and agents desire to control Port au Prince and are willing to employ force and violence to accomplish their goals. Additionally, areas in close proximity to the Route Nationale highway system have a greater conflict propensity. Clashes between armed groups, whether militia, gang, or army, occur often along or in proximity to the major national highway system, probably due to the strategic importance of the highway system. Access to the highway tran sportation systems makes participation in conflict and civil war events possible. Furthermore, conflict becomes more prevalent in areas with lower mean elevation. Conflict is more prevalent in low lying areas of Haiti, which are primarily the coastal area s of the country. The average wealth level and border regions are not statistically significant. Surprisingly, there is no economic dimension to conflict in the global model of conflict. The average wealth determinant is not significant at the five percen t level. Civil war has been explained in terms of greed or grievance, economic deprivation or economic opportunity. Even though Haiti is one of the poorest and unequal countries in the Western Hemisphere, poverty and inequality seem to have no effect on c onflict propensity. A possible explanation might be the general disenfranchisement of the popular masses in Haiti. The popular masses have not been consistently involved in the political process over the last 200 years in Haiti, and the economic plight has been dismal for centuries now. Instead of being grieved by their economic plight, people in Haiti might have grown accustomed to the depravation, and are hence not motivated to

PAGE 158

158 participate in acts of political conflict. Furthermore, Haiti has no natural r esources, such as diamonds (Hegre, stby, and Raleigh 2009) or oil, that would be of strategic importance and induce a struggle to control them. Co nflict must be rather understood in terms of demographic, political, and military strategic concerns. Conflict propensity rises in locations of high population density, areas of political importance, low lying coastal areas, and in close proximity to Route Nationale highway system. Model 2, the civil war model, covers the time period between September 2003 and June 2005. In September 2003, the rebellion against President Aristide broke out in the north of Haiti, continuing until President Aristide was oust ed in February 2004. Afterwards, conflict intensity remained high because of clashes between pro Aristide gangs and armed groups on the side of the new government of Gerard Latortue. Conflict remains more prevalent in locations with higher population densi ty, areas of political importance, low lying coastal areas, and in close proximity to Route Nationale highway system. Political, demographic, and military strategic reasons remain the main causes of conflict during the civil war time period. The civil war model of conflict is similar to the global model of conflict. Both models only differ from each other in the significance of the coefficient estimates. The wealth determinant is almost significant at the five percent level, and is actually significant at the ten percent level; the coefficient for the wealth is negative. As average wealth level falls in a third order administrative unit, the conflict propensity in it rises. If (Hegre, stby, and Raleigh 2009) the location population can objectively be grieved by the inequality, and engage

PAGE 159

159 in conflict to accomplish a redistribution of wealth, or to change the institutions causing the economic inequality. While this is certainly a plausible explanation, another explanation should be offered. President Aristide was primarily supported by the poor and economically deprived. Conflict might have bee n greater in economically deprived areas, since these areas where most supportive of President Aristide, and hence these areas became the target of violence and conflict. The rebel groups, and later the new government of President Gerard Latortue, might ha ve purposefully targeted poorer areas, since these areas were in support of President Aristide. Additionally a civil war, in its nature, is a dynamic process with the location of conflict and battles constantly evolving. Highways are the main route of t ransportation, and facilitate the movement of armed groups, and present a strategic value as well. And findings show that conflict intensity is higher in locations in close proximity to the Route Nationale highways. Model 3 includes all events from 1997 to August 2003, and from July 2005 to crisis accompanied by a severe degradation of the economy, of security and of (Faubert 2006) Histo rically, Haiti has experienced low intensity continuous conflict. The use of force and violence to accomplish economic and political goals has been part of the Haitian social and political environment. The model employed here identifies local determinants of conflict, and offers a possible explanation of why low intensity constant political conflict occurs in Haiti. The results of Model 3 are surprising. The low intensity continuous conflict has a mass population effect, where conflict is greater in areas w ith higher population. No other determinant is significant at the five

PAGE 160

160 percent level. The low intensity conflict occurs with no particular political, economic, or military strategic motivation. Conflict propensity is solely greater in population centers of Haiti. In Table 4 5, we report the estimates of the negative binomial regression for the various types of conflict reported in the Haiti ACLED. We model protest, violence against civilians by armed groups, and battles between two armed groups, since only these types of events had a significant number of conflict events to test their particular determinants. The negative binomial regression provides similar estimates for the significance of the various coefficients for the three event types. Protest, vio lence against civilians by armed groups, and battles between two armed groups is greater in communes with a departmental capital in it, and also greater as total population in the commune rises. Regardless of the type of conflict, a political and demograph ic aspect to conflict propensity exists. Model 1 indicates that the distance to the capital is a significant determinant of protest. Protest propensity rises in close proximity to Port au Prince, the capital. Port au Prince is the political center of Hait i, and any agent willing to change institutions in Haiti is more inclined to advocate, such changes in the capital. Additionally, abuses of political power are most frequently observed and noticed in Port au Prince, which could also impact protest propensi ty in Port au Prince. The distance to the Route Nationale highway system is a significant determinant of battles between two armed groups at the one percent confidence levels. Battles between two armed groups occur more frequently in close proximity to t he Route

PAGE 161

161 Nationale highway system, but the distance to the Route Nationale highway system is not a significant determinant of conflict for the two other types of conflict tested here. and military (Halvard Buhaug 2010) of armed groups. Major highways play a crucial role in conflicts between two armed groups, due to the strategic and military importance of them. In ge neral, little variation between the determinants of conflict for the protest, violence against civilians by armed groups, and battles between two armed groups exists. Protest most commonly is a non violent activity, whereas the other two event types are vi olent in nature. Even though this systematic difference exists, there is little variation in the determinants of the different types of conflict. The analysis indicates that conflict has similar root causes, which do not depend on how violent the conflict event itself is. Different types of conflict are more similar than they are different from each other.

PAGE 162

162 Table 4 1 Correlation matrix of covariates and dependent variable Evts Elev DomR Dep Pop DC RN Wea Evts 1.0000 Elev 0.0740 1.0000 DomR 0.0510 0.2142 1.0000 Dep 0.3950 0.1170 0.1054 1.0000 Pop 0.8065 0.0212 0.0963 0.3975 1.0000 DC 0.2437 0.3343 0.2204 0.0204 0.4064 1.0000 RN 0.1549 0.0577 0.0173 0.2354 0.2259 0.2813 1.0000 Wea 0.4030 0.1611 0.1292 0.3287 0.6051 0.3397 0.3540 1.0000 Note: Evts refers to the number of ACLED events per commune between 1997 and 2000. Elev refers to the mean elevation of the commune. DomR is a dummy variable indicating, whether or not a commune borders the Dominican Republic. Pop refers to the totat popul ation in a given commune. LogPD refers to the logarithm of the population density in a given commune. UrbPer indicates the amount of urban population percentage in a given commune. DC refers to the distance to the capital, Port au Prince. RN refers to the distance of the commune to the Route Nationale highway sytem. Wea refers to the estimated average wealth level per commune. Source: User creation using Haiti ACLED (Raleigh et al. 2010) Table 4 2. Variance inflation factor VIF Elev 1.2575 DomR 1.1132 Dep 1.3229 Pop 2.2071 DC 1.7061 RN 1.2199 Wea 2.0624 Note: Elev refers to the mean elevation of the commune. DomR is a dummy variable indicating, whether or not a commune borders the Dominican Republic. LogPD refers to the logarithm of the population density in a given commune. UrbPer indicates the amount of urban population percentage in a given commune. DC refers to the distance to the capital, Port au Prince. RN refers to the distance of the commune to the Route Nationale highway sytem. Wea refers to the estimated average wealth leve l per commune Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 163

163 Table 4 3. Global Moran Moran's I Score Expected Moran's I Z score P Value 0.0 5958 0.0075 8 1.11785 0. 2636 3 Note: = Z score of 1.96 indicates significant clustering at the five percent confidence level. The distance threshold was set to 50 kilometers. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 164

164 Table 4 4. Determinants of conflict events in Haiti for various time periods Model 1 1997 2010 Model 2 Civil War Model Model 3 Low intensity Conf lict C 0.8513 0.6525 0.3154 ( 0.5501 ) ( 0.663 4 ) ( 0.6791 ) [ 0.121 8 ] [ 0.325 3 ] [ 0.6423 ] Elev 0.001 4 0.001 2 0.0010 ( 0.0006 ) (0.0008) ( 0.000 8 ) [ 0.033 5 ] [ 0.125 3 ] [ 0.1917 ] DomR 0.4782 0.0934 0.840 2 ( 0.4129 ) ( 0.51 50 ) ( 0.489 6 ) [ 0.246 8 ] [ 0.8560 ] [ 0.0861 ] Dep 1.216 6 ** 1.4601 ** 0.888 4 ( 0.4607 ) ( 0.49 50 ) ( 0.5573 ) [ 0.0082 ] [ 0.003 2 ] [ 0.1109 ] Pop 0.0139 ** 0.0128 ** 0.013 3 (0.0017) ( 0.0018 ) ( 0.00 20 ) [0.0000] [0.0000] [0.0000] DC 0.006 9 0.008 4 ** 0.0122 ( 0.0033 ) ( 0.004 1 ) ( 0.0041 ) [ 0.0402 ] [ 0.0385 ] [ 0.2615 ] RN 0.0380 0.0745 0.0186 ( 0.016 6 ) ( 0.0232 ) (0.0210) [ 0.021 8 ] [ 0.001 4 ] [ 0.3403 ] Wea 0.0449 0.0771 0.0059 ( 0.0357 ) ( 0.041 3 ) ( 0.0433 ) [ 0.208 6 ] [ 0.0615 ] [ 0.8916 ] N 133 133 133 Non zero 63 46 44 Theta 0.770 0.722 0.578 Log likelihood 437.240 320.321 322.922 AIC 455.24 338.32 340.92 **p<0.01. *p<0.05 (two sided tests). Table entries are coefficients, with standard errors in parentheses, and with p values in brackets. Note: C refers to the constant of the negative binomial regression model. Elev refers to the mean elevation of the commune. DomR is a dummy variable indicating, whether or not a commune borders the Dominican Republic. Pop refers to the total population in a commune divided by 10,000.DC refers to the distance to the capital, Port au Prince. RN refers to the distance of the commune to the Route Nationale highway sytem. Wea refers to the estimated average wealth level per commune. N is the number of areas, com munes. Theta is the dispersion factor. Log likelihood and AIC assess the model fit. Note: Model 1 includes all events from 1997 to 2010. Model 2 includes only events from September 2003 to June 2005. Model 3 includes only events from 1997 to August 2003, and from July 2005 to 2010. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 165

165 Table 4 5. Determinants of Conflict Events in Haiti for Various Time Periods ` Protest Battle No Change of Territory Violence Against Civilians C 0.05 20 0.2009 1.3855 ( 0.923 3 ) ( 0.6886 ) ( 0.779 2 ) [ 0.9551 ] [ 0.770 5 ] [ 0.0754 ] Elev 0.0014 0.0003 0.000 7 (0.0011) ( 0.0007 ) (0.0009) [ 0.203 8 ] [ 0.662 5 ] [ 0.4620 ] DomR 0.047 7 0.7262 0.718 9 ( 0.7142 ) ( 0.486 4 ) ( 0.553 2 ) [ 0.946 8 ] [ 0.1354 ] [ 0.1938 ] Dep 1.9608 0.9354 1.3289 ( 0.6897 ) ( 0.4639 ) ( 0.5480 ) [ 0.004 5 ] [ 0.043 8 ] [ 0.0153 ] Pop 0.014 2** 0.0090 ** 0.014 8 ** ( 0.0025 ) ( 0.0016 ) ( 0.0020 ) [0.0000] [0.0000] [0.0000] DC 0.0123 ** 0.0065 0.003 5 ( 0.0057 ) ( 0.0043 ) ( 0.0046 ) [ 0.031 2 ] [ 0.132 4 ] [ 0.4568 ] RN 0.030 8 0.0780** 0.0165 ( 0.029 1 ) ( 0.0271 ) ( 0.0226 ) [ 0.289 9 ] [ 0.003 2 ] [ 0.4646 ] Wea 0.050 6 0.018 8 0.0328 ( 0.0567 ) ( 0.0398 ) ( 0.047 2 ) [ 0.3725 ] [ 0.6376 ] [ 0.4865 ] N 133 133 133 Non zero 28 36 34 Theta 0.376 1.232 0.705 Log likelihood 240.525 223.677 243.396 AIC 258.52 241.68 261.4 **p<0.01. *p<0.05 (two sided tests). Table entries are coefficients, with standard errors in parentheses, and with p values in brackets. Note: C refers to the constant of the negative binomial regression model. Elev refers to the mean elevation of the commune. DomR is a dummy variable indicating, whether or not a commune borders the Dominican Republic. Pop refers to the total population in a commune divided by 10,000. DC refers to the distance to the capital, Port au Prince. RN refers to the distance of the commune to the Route Nationale highway sytem. Wea refers to the estimated average wealth level per commune. N is the number of areas, co mmunes. Theta is the dispersion factor. Log likelihood and AIC assess the model fit. Note: Model 1 includes all events from 1997 to 2010. Model 2 includes only events from September 2003 to June 2005. Model 3 includes only events from 1997 to August 2003, and from July 2005 to 2010. Source: User creation using Haiti ACLED (Raleigh et al. 2010)

PAGE 166

166 Figure 4 1 Population size per commune in Haiti in 2003. Source: User creation using U.S. Census demographic data ( U.S. Census Bureau 2010)

PAGE 167

167 Figure 4 2 Urban population percentage per commune in Haiti in 2003. Note: Urban population percentage per commune was derived by dividing the total number of urban population p er commune by the overall population per commune. Source: User creation using U.S. Census demographic data ( U.S. Census Bureau 2010)

PAGE 168

168 Figure 4 3 Male population percentage per commune in Haiti in 2003. Note: Male population percentage per commune was derived by dividing the total number of male population per commune by the overall population per commune. Source: User creation using U.S. Census demographic data ( U.S. Census Bureau 2010)

PAGE 169

169 Figure 4 4 Adult population percentage per commune in Haiti in 2003. Note: Adult population percentage per commune was derived by dividing the total number of adult population per commune by the overall population per commune. Source: User creation using U.S. Census demographic data ( U.S. Census Bureau 2010)

PAGE 170

170 Figure 4 5 Mean elevation per commune in Haiti in meters. Source: User creation using Mid American Geospatial Information Center (MAGIC) data (MAGIC 2011)

PAGE 171

171 Figure 4 6 Standard deviation of elevation per commune in Haiti in meters. Source: User creation using Mid American Geospatial Information Center (MAGIC) data (MAGIC 2011)

PAGE 172

172 Figure 4 7 Route Nationale highway system in Haiti and commune centroids Source: User creation using Geocommons (GeoCom mons 2010)

PAGE 173

173 Figure 4 8 Approximate DHS survey locations in Haiti Source: User creation using Demographic and Health Surveys (DHS) for Haiti (ICF International 2012)

PAGE 174

174 Figure 4 9 Estimated wealth distribution in Haiti using ordinary Kriging Source: User creation.

PAGE 175

175 Figure 4 10 U.S. census database population estimate for Haiti Note: Cell size was set at a height of 0.1 kilometers times width of 0.1 kilometers. Source: User creation using U.S. Census demographic data ( U.S. Census Bureau 2010)

PAGE 176

176 Figure 4 11. Estimated wealth value per commune in Haiti. Source: User creation.

PAGE 177

177 Figure 4 1 2 Randmo n ized Quantile Residual. Source: User creation.

PAGE 178

178 CHAPTER 5 D ISCUSSION AND CONCLU SION This research examines the spatial and temporal patterns of conflict and political violence in Haiti. It highlights and explains conflict dynamics in Haiti in an exhaustive manner, and identifies determinants of conflict and violence. Summary of Results T he temporal analysis of conflict identifies unique temporal patterns of conflict. Conflict intensity, in Haiti between 1997 and 2010, is incongruous Two distinct periods of conflict existed in Haiti during this time period. Between September 2003 and June reigning President in Haiti between 2000 and 2004, and conflict remained high after the period can be best characterized as a short, intense civil war. Apart from the short, intense civil war period, Haiti experienced a long period of low intensity continuo us conflict. The ACLED indicates that violence and political conflict remained a constant aspect to political life in Haiti. Carrol Faubert described Haiti as a a case of a lingering political and governance crisis accompanied by a severe degradation of t (Faubert 2006) Conflict and violence have been instrumentalized by political actors frequently to accomplish their agenda and goals of gaining personal economic prosperity and wealth. The temporal analysis of the ACLED also shines l ight on the security situation, (Collaborative Learning Projects 2009)

PAGE 179

179 However, the ACLED suggests that the levels of violence and political conflict had rema ined similar under President Artistide to the levels experienced under President Prval who served as President from 1997 to 2000 No significant sustained change in conflict intensity can be observed from 1997 to September 2003. The ACLED for Haiti could certainly underestimate conflict levels experienced under President Aristide, and consequently misinform the analysis of conflict in Haiti. However, the temporal analysis of conflict data casts serious doubt that the security situation worsened signi ficantly under President Aristide. The spatial analysis of conflict highlights that significant spatial clustering of events exists in Haiti. Most of the political conflict and violence occurs in the greater metropolitan area of Port au Prince, Gonaves, Saint Marc, Cap Hatien, and Petit Gove Conflict events are clustered in the major population centers of Haiti. An urban bias of conflict events exists, and conflict events are limited in rural areas of Haiti. The spatial diffusion of conflict changed o ver time. Throughout the civil war, and really beginning in 2002, the diffusion of conflict events increased. The civil war period experienced a greater dispersion of conflict through Haiti. A violentization of the north of Haiti began in 2002, and continu ed until 2004. Interestingly, since 200 5 the north of Haiti has experienced little political conflict and violence. The civil war in Haiti was a nationwide experience, whereas conflict has at times been relatively restricted to the greater metropolitan ar ea of Port au Prince in year s prior to and after the civil war. The spatial diffusion is similar for different types of conflict. The geographic diffusion for protest, violence against civilians by armed groups, and battles between two armed groups does n ot vary much in the case of Haiti.

PAGE 180

1 80 The deterministic analysis of conflict attempted to identify determinants of conflict and violence in Haiti. The general, global model including all conflict events from 1997 to 2010 indicates that c onflict propensity is greater in major population centers in communes with departmental capitals, in closer proximity to Port au Prince, the capital in closer proximity to the Route Nationale highway sytem, and in locations with lower elevation. Demographic, political and military strategic factor s impact conflict propensity in a given location. Surprisingly, average wealth levels and wealth distribution has no significant impact on conflict propensity in the general, global model of conflict in Haiti. The analysis of conflict for the civil war period, and the continuous low intensity conflict model show variations in the determinants of conflict for the two time periods The civil war model of conflict is similar t o the global model of conflict Political, demographic, and military strategic reasons remain the main causes of conflict during the civil war time period The wealth variable was almost significant at the five percent level; c onflict propensity was greater in areas of low average wealth levels. The low intensi ty continious conflict model was markedely different from the two civil war and global model of conflict. The only signicificant determinant of low intensity continiuos conflict is the population size in a given location Conflict propensity rises with pop ulation size. The determinants of conflict for protest, violence against civilians by armed groups, and battles between two armed groups are similar. Protest, violence against civilians by armed groups, and battles between two armed groups is greater in communes with a departmental capital, and in commune with greater total population Protest propensity rises with proximity to Port au Prince. The distance to the Route

PAGE 181

181 Nationale highway system is a significant determinant of battles between two armed grou ps. Battles between two armed groups occur more frequently in close proximity to the Route Nationale highway system, but the distance to the Route Nationale highway system is not a significant determinant of conflict for the two other ty pes of conflict tes ted. Protest most commonly is a non violent activity, whereas the other two event types are violent in nature. Even though this systematic difference exists, there is little variation in the determinants of the different types of conflict. The analysis indicates that conflict has similar root causes, which do not depend on how violent the conflict event itself is. Suggestions for Future Researc h Disaggregated studies of conflict offer vital insight s into the dynamics of conflict and violence. In our case s, the analysis of the ACLED, in combination with determinants, from various data sources, allowed an analysis of localized conflict in access make the modeling of viol challenging. Localized corruption, election, institutional, and in particular economic data could have enhanced the study of conflict in Haiti. Data access is the greatest challenge to studies of localized conflict. Most d isaggregated studies of conflict have studied localized conflict in Africa, with a limited number of studies focusing on Asia, Central Asia, and the Middle East The focus on Africa has been primarily caused by the large number of ci vil wars and conflicts in the region Since most studie s have focused on Africa n conflicts conclusions about localized civil war dynamics have been frequently gained from

PAGE 182

182 studies focusing on Africa The study design employed in the analysis of conflict in Haiti is similar to the study design used by Hegre, s tby, and Raleigh in their analysis of conflict and civil war in Liberia (Hegre, stby, and Raleigh 2009) Hegre, stby, and conflict events are more frequent in locations that were (Hegre, stby, and Raleig h 2009) In our analysis of conflict in Haiti, the wealth variable was only significant in the civil war model, and conflict propensity was greater in relatively poor areas of the country. In the general, global model and the continuous low intensity co nflict model the wealth variable was not significant. The localized study of conflict in Liberia and Haiti used a similar methodological approach, but arrive s at different conclusion s about the impact of wealth on conflict propensity. Conflict dynamics mu st not necessarily be similar in different geographic regions of the world, and can even vary significantly for two neighboring countries. Additionally, in Haiti the set of determinants for the civil war model was different from the set of determinants in the low intensity constant conflict model. General conclusion about all civil wars and conflict based on a single study or group of related studies must be avoided. Determinants of conflict can be context specific varying for different geographic regions and determinants can be time varying even for the same geographic area. It is unclear how time and location specific determinants of conflict are, and further research answering this basic question would be of immense value. Our research in Haiti indic ates a change in spatial diffusion of conflict, as conflict intensity rises. A possible link between conflict intensity and conflict diffusion could

PAGE 183

183 exis t A further exploration of a general model of conflict diffusion and conflict intensity would be of gr eat benefit to further understand conflict dynamics.

PAGE 184

184 APPENDIX A EWMA METHODOLOGY monitoring the process that averages the data in a way that gives less and less weight to data a The EWMA is mathematically defined in the following way: (A 1) where Y i j is the event count for each individual month, is which is referred to as the weighting factor or smoothing parameter, Z 0 v (Lucas and Saccucci 1990) or target value. The greater the value of the stronger more recent observations influence the EWMA statistics. A value of = 1 is usually set betwee (NIST/SEMATECH 2012) The Exponential Weighted Moving Ave rage control s cheme control (NIST/SEMATECH 2012) If the EWMA shifted, and is out of control at Z 0 + k*s ewma (A 2) Z 0 k*s ewma (A 3) Z 0 (Lucas and Saccucci 1990) or target value. S ewma is the standard deviation of the EWMA statistic (NIST/SEMATECH 2012) in most cases.

PAGE 185

185 The variance of the EWMA statistic is defined in the following way: s 2 (Zi) = [{1 (1 2i } ( /(2 )] s 2 (A 4) where the starting value soon dissipates and the variance quickly converges to its asymptotic (Lucas and Saccucci 1990 ) : s 2 (Zi) = { /(2 )} s 2 (A 5) The EWMA statistics measures the cumulative deviations over time from the process is out of control, and if the EWMA remains within the control limits the process in contr ol. The EWMA detects the mean value of a variable of interest changing.

PAGE 186

186 APPENDIX B TISTIC statistic detects spatial autocorrelation in a variable of the following way: (B 1) Where n is equal to the number of areas or polygons, is the mean of the variable of interest, is the value of the variable of interest in a given polygon or area, and is the value of interest in another polygon or area. is a spatial weight matrix of i relative to j. A spatial weight matrix measures the relative location of all points i and j, and determines the relationship and pattern between various areas and polygons. then compared to a random and expected spatial distribution to detect (Burt, Barber, and Rigby 2009) An index is created, which represent the expected distribution under randomness: (B 4) The variance is then defined as: (B 5) where, (B 6) (B 7) (B 8) The test statistic is then defined as being based:

PAGE 187

187 (B 9) The Global calculates a Z score that compare s the spatial distribution of a variable to a spatial distribution of the same var iable under randomness. If the observed is (Burt, Barber, and Rigby 2009) ,the data exhibits positive spatial than the expected value, the data exhibits negative spatial autocorrelation.

PAGE 188

188 APPENDIX C LOCAL INDICATOR OF S PATIAL ASSOCIATION E Local indicators of spatial association (LISA) identify (Anselin 1995) (Burt, Barber, and Rigby 2009) (C 1) where is the value of a variable of interest in a given area i. is the overall mean of the variable of interest. is the spatial matrix defining the spatial relationship between various areas. where n is the total number of areas included. Additionally, the Z (C 2) where (C 3) (C 4) The L G lobal L (Anselin 1995) occurs. If the local rea is negative, it indicates occurs (for example, a location with high values surrounde d by neighbors with low (Anselin 1995)

PAGE 189

189 APPENDIX D MEAN CENTER AND S T ANDARD DEVIATIONAL E LLIPSE S The mean center of concent (ESRI 2011c) It is derived by calculating the (ESRI 2011c) : (D 1) (D 2) The mean center identif ies changes or variations in the spatial distribution of data tracking the central or average location. A standard deviation of the x coordinates and y coordinates from the mean center to (ESRI 2011d) The standard deviational ellipses can assess th e spatial diffusion of data, and track the variation in geographical dispersion for different sets of data or over time The standard deviational ellipse is defined as: (D 3) (D 4) where and represent the coordinate of the feature i, and and are the mean center of the feature. The angle of rotation is calculated in the following way: (D 5) ) (D 6) (D 7) (D 8)

PAGE 190

190 w and are deviations of the xy (ESRI 2011d) The standard deviational ellipse creates an ellipse that shows the deviations of the observations from th (ESRI 2011d) The orientation of the ellipse indicates the direction and orientation of the data.

PAGE 191

191 APPENDIX E NEGATIVE BINOMIAL GE NERALIZED LINEAR MODEL The negati ve binomial regression model uses count data with ovedispersion; overdispersion in data occurs when the variance of the data exce eds the mean of the data. The negative binomial regression model has a flexible distribution to account for overdispersed data : (E 1) where is the mean or expected value parameter. The dispersion factor is estimated, and not a priori known. The log likelihood function for the negative binomial regression model is then defined as: (E 2) (E 3)

PAGE 192

192 LIST OF REFERENCES International Migration 49 (June 19): e164 e188. doi:10.1111/j.1468 2435.2010.00664.x. http://doi.wiley.com/10.1111/j.1468 2435.2010.00664.x. Geographical Analysis 27 (2) (September 3): 93 115. doi:10.1111/j.1538 4632.1995.tb00338.x. http://doi.wiley.com/10.1111/j.1538 4632.1995.tb00338.x. Arnson, Cynthia J., and I. William Zart man. 2005. Rethinking the Economics of War: The Intersection of Need, Creed, and Greed Bailey, Trevor, and Tony Gatrelll. 1995. Gatrell 7.4.5) The William and Mary Quarterly 63 (4): 643 674. Journal of Conflict Resolution 46 (1): 74 90. http://jcr.sagepub.com/content/46/1/74.short. Bocquier, Philippe, and Herv Maupeu. European Journal of Population 21 (2 3) (June): 321 345. doi:10.1007/s10680 005 6858 z. http://www.springerlink.com/index/10.1007/s10680 005 6858 z. Bolker, Ben. 2007. Ecological Models and Data in R Princeton and Oxford: Princeton University Press. national Analysis of Economi c and Institutional Determinants, 1971 European Sociological Review 26 (5) (July 14): 519 540. doi:10.1093/esr/jcp036. http://esr.oxfordjournals.org/cgi/doi/10.1093/esr/jcp036. The J ournal of Negro History 73 (1/4): 1 11. Brunsdon, C., S. Fotheringham, and M. Charlton. 1998. Geographically Weighted Regression Journal of the Royal Statistical Society: Series D (The Statistician) Vol. 47. doi:10.1111/1467 9884.00145. http://doi.wiley. com/10.1111/1467 9884.00145.

PAGE 193

193 Conflict Management and Peace Science 27 (2) (January 28): 107 128. doi:10.1177/0738894209343974. http://cmp.sagepub.c om/cgi/doi/10.1177/0738894209343974. Buhaug, Halvard, Lars Ethno International Organization 62 (3): 531 551. doi:10.1017/S0020818308080181. Political Geography 24 (4) (May): 399 418. doi:10.1016/j.polgeo.2005.01.006. http://linkinghub.elsevier.com/retrieve/pii/S0962629805 000181. 1970 Disaggregating the Study of Civil War and Transnational Violence 1 26. Burt, James E., Gerald M. Barber, and David L. Rigby. 2009. Elementary Sta tistics for Geographers https://www.cia.gov/library/publications/the world factbook/geos/ha.html. Cederman, Lars Models of Civil W First World Congress on Social Simulation 1 8. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.88.735&rep=rep1&type= pdf. https://www.cia.gov/library/publications/the world factbook /geos/ha.html. Population Research and Policy Review 27 (1) (September 27): 17 42. doi:10.1007/s11113 007 9051 8. http://www.springerlink.com/index/10.1007/s11113 007 9051 8. Childs, Colin. 2004. Interpolating Surfaces in ArcGIS Spatial Analyst Collaborative Learning Projects. 2009. A Brief Background to Conflict in Haiti Journal of Con flict Resolution 46 (1): 13 28. http://jcr.sagepub.com/content/46/1/13.short.

PAGE 194

194 Oxford Economic Papers 56 (4) (June 22): 563 595. doi:10.1093/oep/gpf064. http://oep.oupjournals.org/cgi/doi/10.1093/oep/gpf064. Co The Journal of Conflict Resolution 46 (1): 3 12. http://www.jstor.org/stable/10.2307/3176236. A Feb02/Montas Dominique.htm. Disaggregating the Study of Civil War and Transnational Violence 1 36. Dunn, Peter K, and Gordon K Smyth. 1996. Randomized Quantile Residual. From Jean Journal of Latin American Anthropo logy 10: 186 205. Prval Latin American Perspectives 33 (3) (May 1): 132 141. doi:10.1177/0094582X06288058. http://lap.sagepub.com/cgi/doi/10.1177/0094582X06288058. http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//0031000000320000 00.htm. http://webhelp.esri.com/arcgisdesktop/9.2/index.cfm?TopicName=Mean Center (Spatial Statistics). http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#/How_Directional_Dist ribution_Standard_Deviational_Ellipse_works/005p0000001q000000/. Research Institute. http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//005p0000000z0000 00.

PAGE 195

195 omparison of UCDP GED and ACLED Samfak.uu.se http://www.samfak.uu.se/digitalAssets/83/83575_CoCo__Eck__final_.111204.relea se_vers.pdf. Explaining the Pr val Journal of Conflict Resolution 46 (3): 307 334. Current Anthropology 45 (3): 305 325. The New England Journ al of Medicine 350 (15): 1483 1486. Fatton, Robert. 2005. Third World Quarterly 27 (1) (February): 115 133. doi :10.1080/01436590500369378. http://www.tandfonline.com/doi/abs/10.1080/01436590500369378. Faubert, Carrol. 2006. Evaluation of UNDP Assistance to Conflict Affected Countries Journal of the National Cancer Institute Vol. 104. doi:10.1093/jnci/djs060. http ://www.ncbi.nlm.nih.gov/pubmed/22872654. American Political Science Review of Forced Disp Society WGAPE 1 51. The Journal of Conflic t Resolution 46 (1): 111 130. Inconsistency and Political Instability: Polity Duration, 1800 American Journal of Political Science 50 (4): 893 908.

PAGE 196

196 Gaudart, Jean, B elco Poudiougou, Alassane Dicko, Stphane Ranque, Ousmane Toure, Issaka Sagara, Mouctar Diallo, et al. 2006. time Clustering of BMC Public Health 6 (286) (January): 1 13. doi:10.1186/1471 2458 6 286. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1684261&tool=pmcentre z&rendertype=abstract. Girard, Phillippe. 2010. Haiti: The Tumultuous History From Pearl of the Caribbean to Broken Nation Journal of Peace Research 44 (3): 293 309. http://jpr.sagepub.com/content/44/3/293.short. Griffin, Thomas M. 2004. Haiti Human Rights Investigation Miami, FL. Gros, Jean Third World Quarterly 17 (3): 455 471. Hallward, Peter. 2007. Damming the Floo d Haiti and the Politics of Containment Foreign Policy 91 (Summer): 105 122. Operations 2010 332. Hegre, Hvard, Mapping the Complexity of Civil Wars 1 25. on Journal of Conflict Resolution 50 (4) (August 1): 508 535. doi:10.1177/0022002706289303. http://jcr.sagepub.com/cgi/doi/10.1177/0022002706289303. Events: A Di Journal of Conflict Resolution 53 (4) (June 8): 598 623. doi:10.1177/0022002709336459. http://jcr.sagepub.com/cgi/doi/10.1177/0022002709336459.

PAGE 197

197 Heine, Jorge, and Andrew S. Thompson. 2011. Fixing Haiti: Minustah and Beyond Hubbard, Alan. 2006. Modelling Counts The Poisson and Negative Binomial Regression http://ehs.sph.berkeley.edu/hubbard/longdata/. University. Criminal Violations in Port au Lancet 368 (9538) (September 2): 864 73. doi:10.1016/S0140 6736(06)6 9211 8. http://www.ncbi.nlm.nih.gov/pubmed/16950364. Lauren, Taft CRS Report for Congress pdf.usaid.gov/pdf_docs/PCAAB593.pdf. Lucas, James M., American Statistical Association and American Society for Quality 32 (1): 1 12. Journal of Interamerican Studies and World Affairs 31 (1/2): 1 21. Scandia: Tidskrift Fr Historisk Forskning : 40 78. http://cts.lub.lu.se/ojs/index.php/scandia/article/download/907 /692. Comparative Economic Studies 40 (1): 43 71. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1663537. Lunde, Henriette. 2009. Haiti Youth Survey 200 9 Vol. II. http://magic.csr.utexas.edu/index.htm. The World Economy 29 (8): 1117 1131. doi:10.1111/j.146 7 9701.2006.00808.x. http://onlinelibrary.wiley.com/doi/10.1111/j.1467 9701.2006.00808.x/full.

PAGE 198

198 Journal of Conflict Resolution 46 (1): 91 110. http://jcr.sagepub.com /content/46/1/91.short. http://www.itl.nist.gov/div898/handbook/. time Cube: An Exploratory Data analysis Approach Using Space time Kernel Density Transactions in GIS 14 (3) (June 21): 223 239. doi:10.1111/j.1467 9671.2010.01194.x. http://doi.wiley.com/10.1111/j.1467 9671.2010.01194.x. http://www.itl.nist.gov/div898/software/dataplot/refman2/auxillar/vif.htm. http://www.nationsonline.org/oneworld/map/haiti_map.htm. Third World Quarterly 8 (4): 1239 1252. in the North Caucasus of Russia, 1999 Annals of the Association of American Geographers 101 (1): 178 201. War Logs, 2004 Eurasian Geography and Economics 51 (4) (July 1): 472 495. doi:10.2747/1539 7 216.51.4.472. http://bellwether.metapress.com/openurl.asp?genre=article&id=doi:10.2747/1539 7216.51.4.472. Pakistan Wars, 2008 2009: Micro geographies, Conflict Diffusion, and Cluster Eurasian Geography and Economics 51 (4) (July 1): 437 471. doi:10.2747/1539 7216.51.4.437. http://bellwether.metapress.com/openurl.asp?genre=article&id=doi:10.2747/1539 7216.51.4.437. Boston, MA. http://www.rstudio .org/.

PAGE 199

199 Raleigh, Clionadh, and Hvard Hegre. Political Geography 28 (4) (May): 224 238. doi:10.1016/j.polgeo.2009.05.007. http://linkinghub.elsevier.com/retrieve/pii/S0962629809000523. Raleigh, Clionadh, Andrew Linke, and Caitriona Dowd. 2012. Armed Conflict Location and Event Dataset (ACLED): Codebook Version 2 Raleigh, Clionadh, Andrew Linke, Hvard Hegre, and Joakim Karlsen. 2010. l Data Journal of Peace Research 47 (5) (September 28): 651 660. doi:10.1177/0022343310378914. http://jpr.sagepub.com/cgi/doi/10.1177/0022343310378914. party Interventions and the Duration of Intrastate Journal of Conflict Resolution 46 (1): 55 73. http://jcr.sagepub.com/content/46/1/55.short. Reynal Journal of Conflict Resolution 46 (1): 29 54. http://jcr.sagepub.com/content/46/1/29.sho rt. Ripley, Brian, Bill Venables, and Kurt Hornik. 2012. American Society for Quality 42 (1): 97 101. The SAGE Handbook of Spatial Analysis. p. 344 355. Statistics in Medicine 16 (18) (September 30): 2081 93. http://www.ncbi.nlm.nih.gov/pubmed/9308133. Ro Criminal Activities -A Cusum Nearest Towards Digital Earth Proceedings of the International Symposium on Digital Earth : 1 7. Rogerson, Peter a, Gaurav Si American Journal of Preventive Medicine 30 (2 Suppl) (February): S50 9. doi:10.1016/j.amepre.2005.09.006. http://www.ncbi.nlm.nih.gov/pubmed/16458790. Root, qualitytools.org. Rutstein, Shea Oscar, and Kiersten Johnson. 2004. DHS Comparative Reports 6 The DHS Wealth Index

PAGE 200

200 SAS Institute. 2011. Chapter 11 The COUNTREG Procedure Cary, NC: SAS Institute Inc. http: //support.sas.com/kb/26/addl/fusion26161_3_countreg.pdf. Journal of Conflict Resolution 45 (3): 259 282. http://jcr.sagepub.com/content/45/3/259.short. Sanitaire s, Ministere de la Sante Publicue et de Population Liste Des Institutions. 2010. http://sites.google.com/a/netspective.org/haiti health facilities/home. Shamsie, Yasmine, and Andrew S. Thompson. 2006. Haiti Ho pe for a Fragile State Comparative Politics 24 (4): 379 399. Sokoloff, Kenneth L., and Stanley L. Engerman. Endowments, and Paths of Development Journal of Economic Perspectives 14 (3): 217 232. http://siteresources.worldbank.org/INTWDRS/Resources/477365 1257315064764/3195_sokoloffb.pdf. World Bank Conference on Evaluation and Poverty Reduction 1 23. 2002. Horizontal Inequalities: a Neglected Dimension of Development Development WIDER. http://www3.qeh.ox.ac.uk/pdf/qehwp/qehwps81.pdf. http ://www.census.gov/population/international/data/mapping/demobase.html. United States Agency for International Development. 2005. Conflict, Poverty, Inequality, and Economic Growth Verner, Dorte. 2008. Making Poor Haitians Count Poverty in Rural and Urban Haiti Based on the First Household Survey for Haiti World Journal of Geographical Systems 11 (2) (February 28): 155 173. doi:10.1007/s10109 009 0080 1. http://www.springerlink.com/index/10.1007/s10109 009 0080 1.

PAGE 201

201 BIOGRAPHICAL SKETCH Jens Engelmann was born in Gronau, Germany in 1987. He graduated with a Bache lor of Science in f ood and r esource e conomics in 2010 from the University of Florida after which he pursued a Master of Science in f ood and f esource e conomics first at Texas A&M University, and then at the University of Florida, where he will graduate in the f all of 2012.