INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bieedthrough, substandard margins, and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps.
Photographs included in the original manuscript have been reproduced xerographicaliy in this copy. Higher quality 6” x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.
ProQuest Information and Learning 300 North Zeeb Road, Ann Aitor, Ml 48106-1346 USA 800-521-0600 UMI
CIVIL WAR, MACRO-SOCIAL CONTEXTS, AND INTERVENING MECHANISMS: IDENTIFYING STRUCTURAL LINKAGES
DISSERTATION
Presented in Partial Fulfillment of the Requirements for
the Degree Doctor of Philosophy in the Graduate
School of The Ohio State University
By
Matthew Christenson, M.Div., M.A.
*****
The Ohio State University
2001
Dissertation Committee: Approved by:
Edward M. Crenshaw, Adviser .dviser J. Craig Jenkins Department of Sociology Pamela Paxton UMI Number: 3031188
UMI'
UMI Microform3031188 Copyright 2002 by Bell & Howell Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.
Bell & Howell Infomiation and Leaming Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 ABSTRACT
Civil war is an especially threatening social phenomenon because of its catastrophic impact on human welfare. The gravity of this problem makes essential the development of frameworks that indicate where civil wars are likely to break out.
The research described here accomplishes this by developing an approach to the study of civil war that links existing perspectives on revolutions and other forms of collective violence with macro-social theories of societal development. This framework is then employed as a guide for multivariate statistical investigations. I use event-history analysis to study the onset of civil wars.
This investigation is novel in that it explores the ways that social grievance and institutional failure are contingent upon the larger social context from which they emerge and uses statistical techniques that permit the study of the temporal elements of civil war.
On the macro-social level, I confirm that mid-levels of both industrial development and techno-ecological Inheritance encourage civil war, while a strong service sector discourages civil war. In terms of intervening mechanisms, 1 find that mid-levels of inequality, as well as high levels of discrimination and minority grievances, are all associated with rapid movement toward civil war.
u Byaddingafocnyoir the^broadersociahcontexts-of crvtl^ war, tMy study expantfe the temporal perspective of existing early warning methodologies, making earlier detection possible and thereby improving the efficiency of potential interventions.
u i To My Wife Monica
IV ACKNOWLEDGMENTS
Numerous people have shaped and supported me in the time leading up to and
during the process of developing this dissertation. To begin, I wish to thank Ed
Crenshaw, my advisor, for his professional mentoring, his meticulous guidance at each
step of the dissertation process, and for setting a high standard of scholarship as an
excellent model to emulate. I also wish to acknowledge the members of my committee,
J. Craig Jenkins and Pamela Paxton, for their helpful comments and their flexibility at the
end of this process. Further, I wish to thank the professors and staff in the sociology
department for their support and collegiality, and the staff at the International Programs
Center of the U.S. Census Bureau for their encouragement as I sought to bring my formal education to closure.
On a more personal note, I thank the many friends I met in Latin America who encouraged me to embrace myself as a "first-worlder" and to struggle with the challenges of the world accordingly. My parents also deserve special recognition for providing me with a model of lives dedicated to issues that affect real people. Finally, I wish to thank my wife, Monica, to whom this work is dedicated, and my children, Kyle and Leah, for their sacrifice and dedication, but mostly for the relationships we share that consistently provide me with a sense of both vocational and personal afhrmation. VITA
June 18, 1964 Bom - Minneapolis, Minnesota
May, 1986 Bachelor of Arts, Summa cum Laude Concordia College, Moorhead, Minnesota
May, 1992 Master of Divinity Trinity Lutheran Seminary, Columbus, Ohio
1992-93 University Fellow The Ohio State University
August, 1994 Master of Arts The Ohio State University, Columbus, Ohio
1996 Admitted to Candidacy The Ohio State University, Columbus, Ohio
1997-present Statistician/Demographer International Programs Center, U.S. Census Bureau
Publications
1) Christenson, Matthew. IQOHforthcoming). “Global Population Growth.” Chapter 1 in Global Population Profile: 2000. International Programs Center, U.S. Census Bureau: Washington, D C.
2) Christenson, Matthew. 2Q0Ï (forthcoming). “Global Population Composition.” Chapter 2 in Global Population Profile: 2000. International Programs Center, U.S. Census Bureau: Washington, D.C.
3) Crenshaw, Edward, Matthew Christenson and Doyle Ray Oakey. 1999. “Demographic Transition in Ecological Focus.” American Sociological Review 65:371-391.
VI Christenson, Matthew; anchEdward Crenshaw. t997. '^Democracy's Handmaiden: Educational Constraints on Political and Economic Modernization," in N.F. McGinn and E.H. Epstein (eds.), The Role o fEducation in Democratization, Volume I: Transitional States and States o f Transition. New York: Peter Lang Publishers.
5) Crenshaw, Edward, Ansari Ameen and Matthew Christenson. 1997. "Population Dynamics and Economic Development: The Differential Effects of Age-Specific Population Growth Rates on Per Capita Economic Growth in Developing Countries, 1965 to 1990”. American Sociological Review 62:91
F ields o f St u d y
Major Field: Sociology
Topics of Interest: Demography, International Development, and International Conflict
vu TABLE OF CONTENTS
ABSTRACT...... ii
DEDICATION...... iv
ACKNOWLEDGMENTS...... v
VITA...... Vi
LIST OF FIGURES...... xi
LIST OF TABLES...... xiii
CHAPTER I: INTRODUCTION...... 1 Motivation for the Study of Civil Wa r ...... I Civil Wars , 1965-1992...... 4 The state of our knowledge : ...... 7 The Current Analysis ...... 13
CHAPTER 2: THEORIES OF CIVIL WAR...... 15 Conceptualizing Civil War ...... 15 The Immediate Conditions to Civil Wa r ...... 16 The Motives to Violent Collective Action ...... 18 Grievance and Violent Collective Action ...... 18 Greed and Violent Collective Action ...... 23 The Means to Violent Collective Action ...... 26 The Opportunities for Violent Collective Action ...... 29 The Macro -social Environment and civil war ...... 32 Structural Modernization Theory and Civil W ar ...... 33 Human Ecology Theory and Civil War ...... 37 Techno-Ecological hiheritance and Violent Collective Action: ...... 39 The Service Sector and Violent Collective Action: ...... 41 Ethnicity and Violent Collective Action: ...... 43 Political Economy Theory and Civil W ar ...... 46
viii CHAPTER ) : HATA ANEM«ÆETHODS ...... 49
The Scope of th e S tu d y ...... 49 The Modeling Framework ...... 50 The Dependent Variable ...... 54 Descriptive A nalysis ...... 56 Independent Variables and H ypotheses ...... 57 The Lagged Dependent Variable: Measures and Hypotheses ...... 58 The Macro-social Environment: Measures and Hypotheses ...... 59 Structural Modernization ...... 59 Techno-ecological Inheritance ...... 60 Service-sector Strength ...... 62 Ethnicity ...... 63 Political Economy ...... 66 Intervening Variables: Measures and Hypotheses ...... 68 Absolute Deprivation, Relative Deprivation, and Inequality ...... 68 Greed...... 72 Demographic Indicators ...... 73 Social Mobilization/Urbanization ...... 76 Government Characteristics ...... 77 Other Forms of Collective Action ...... 8 L Militarization ...... 83 Modeling Str.a.tegy .jiND A dditional H ypotheses ...... 83
CHAPTER 4: ANALYSIS...... 87 Bivariate Relationships ...... 88 Lagged Dependent Variables ...... 88 Macro-level Variables ...... 88 Indicators of Structural Modernization ...... 88 Indicators of Techno-ecological Inheritance ...... 89 Indicators of Service Sector Strength ...... 90 Indicators of Ethnicity ...... 90 Political Economy Indicators ...... 91 Intervening Variables ...... 93 Indicators of Absolute Deprivation, Relative Deprivation, and Inequality ...... 93 Indicators of “Greed” ...... 94 Demographic Indicators ...... 95 Indicators of Social Mobilization ...... 97 Indicators of Government Characteristics ...... 97 Indicators of Previous Experience with. Other Forms of Collective Action ...... 99 Indicators of Militarization ...... 99
IX MüITIVARKTERELATIONSfnPS-...... tOO^ Development of the Core Model ...... 100 Other Macro-social Variables and the Onset of Civil War ...... 106 Intervening Mechanisms and the Onset of Civil War ...... 107 CHAPTER 5: CONCLUSIONS...... 113
BIBLIOGRAPHY...... 118
APPENDIX A: CIVIL WARS BETWEEN 1965 AND 1992...... 134
APPENDIX B: RISK SET FOR THE 1965 ONSET ANALYSIS...... 136
APPENDIX C: RISK SET FOR THE 1970 ONSET ANALYSIS ...... 137
APPENDIX D: FIGURE 4...... 138
APPENDIX E: TABLE 2...... 139
APPENDIX F: CORE SECTION OF THE MAXIMUM MODEL DATA ANALYSIS
PROGRAM...... 154
APPENDIX G: TABLE 3 ...... 155
APPENDIX H: TABLE 4 ...... 156
APPENDIX I: TABLE 5...... 157
APPENDIX J: TABLE 6...... 159
APPENDIX K: TABLE 7 ...... 161
APPENDIX L: TABLES...... 164
APPENDIX M: HGURES5-16...... 167 LIST OF FIGURES
Figure I : Countries experiencing the onset of civil war between 1965 and 1992 ...... 5
Figure 2: Onsets of civil wars per year and number of ongoing civil wars with an onset between 1965 and 1992 ...... 6
Figure 3: Onsets of civil wars per year and number of ongoing civil wars with an onset between 1965 and 1999 ...... 115
Figure 4: Hazard function plots of the transition rate from peace to civil war from 1965- 1992 and 1970-1992...... 138
Figure 5: Independent effect of previous civil wars on civil war onset ...... 167
Figure 6: Independent effect of labor force in industry 1965 on civil war onset ...... 168
Figure 7: Independent effect of service sector dominance on civil war onset ...... 169
Figure 8: Independent effect of agricultural population density on civil war onset 170
Figure 9: Independent effect of ethnic homogeneity on civil war onset ...... 171
Figure 10: Independent effect of 20 percent ethnic dummy on civil war onset 172
Figure 11 : Independent effect of ethnic history o f lost autonomy * group size on civil war onset ...... 173
Figure 12: Independent effect of agricultural exports/total exports 1980 on civil war onset ...... 174
Figure 13: Independent effect of intensity of political discrimination 1975 on civil war onset ...... 175
Figure 14: Independent effect of etlinic grievance (all) * group population on civil war onset ...... 176
XI Figure 15: Independent effect of Gini coefficient of income inequality on civil war onset ...... 177
Figure 16: Independent effect of Gini coefficient of sectoral inequality 1970 on civil war onset ...... 178
xu LIST OF TABLES
Table I: Onsets of civil war and the countries experiencing them by region: 1965-1992...... 4
Table 2: Description of independent variables and bivariate statistics ...... 140
Table 3: Base model from the 1965 analysis ...... 155
Table 4: Base model from the 1970 analysis ...... 156
Table 5: Additional tests of macro-social variables: 1965 analysis ...... 157
Table 6: Additional tests of macro-social variables: 1970 analysis ...... 159
Table 7: Tests of intervening variables: 1965 analysis ...... 161
Table 8: Tests of intervening variables: 1970 analysis ...... 164
xm CHAPTER 1
INTRODUCTION
M o t iv a t io n for t h e St u d y o f C iv il W a r
The phenomenon of civil war has recently begun to attract significant attention in
both the scholarly and public policy communities. In the past several years, a burgeoning research literature has started to be seen in the academic journals (e.g., Licklider 1995,
Collier and Hoeffler 1998, Collier 1999, Ellingsen 2000, and Sambanis 2001). Some of this research has been supported by one of several major initiatives by institutions dedicated to the formation of public policy in response to civil war and other forms of state crises (e.g., the CIA’s State Failure Taskforce [Esty et al. 1998] and the World
Bank's Policy Research on the Causes o f Conflict in Developing Countries [World Bank,
2001]). Indeed, the heightened interaction o f academic scholars with public policy experts through joint taskforces, special conferences, and support for independent research points to the seriousness o f the search for answers to this pressing problem.
It is easy to make the case that such attention is warranted. Globally, civil war is currently the predominant form of large-scale violence (Wallensteen and Sollenberg
2000). In fact, of the 25 major armed conflicts going on in 2000, only two were intemational wars (Ethiopia-Eritrea and India-Pakistan in Kashmir - see SIPRI2001 ).
During the 1990s, it is estimated that civil wars were responsible for 90 percent of all
I war-rehrted deaths, and that more-thair 20-percent of thgpoptrhrtiotr in Snh-Saharag
Africa was living in countries directly affected by civil wars (World Bank 2001 ). Just
based on the prevalence of civil war, therefore, the need for an improved understanding is
warranted.
Perhaps the strongest case for more focused attention on civil war can be made by
considering the social costs. For individuals, civil war is horrific. In the 20'*' century
alone, it is estimated that as many as 134 million people were killed in domestic wars,
twice the number killed in intemational wars (Rummel 1994). In 1998, civil wars had
contributed to the creation of 13 million refugees and 38 million internally displaced
persons worldwide (World Bank 2001).
On an institutional level, civil war also does great damage. In the short run, much of the economic activity of a country is disrupted or diverted to military purposes (Collier
1999). This often includes stagnation in the consumer market and, consequently, a sharp reduction in health services. As a result, malnutrition and disease often rise sharply during war (Mohammed 1999).' Additionally, civil war generally entails the collapse of the rule of law (Collier 1999). This often leads to widespread human rights violations- and a climate of mistrust that makes resolving civil war difficult without external intervention (World Bank 2001).
'This situation is often hard to ameliorate even in the presence o f intemational assistance due to the difBculty of getting aid to affected populations. As a result, many o f the most dramatic humanitarian crises occur in countries experiencing civil war. “In fact, the State Failure Taskforce found that incidents of genocide and politicide since 19S5 could not be found outside of the context of civil or etlmic war, thus requiring tliat they all be studied as "consolidated cases" (unpublished report). 2 Inrthe^iong-nm; the damage^thatcmh wars caTisgofterr teaveg the societies iir which they occur severely handicapped. Civil war results in the destruction of infrastructure such as transportation networks, factories, and school buildings (Collier
1999). It also leads to the distortion of infrastructure, as in the conversion of manufacturing plants to military purposes and the proliferation of land mines on agricultural land (Mohammed 1999). Further, civil war reduces the supplies of financial capital by encouraging capital flight (Collier 1999) and of human capital through death, disability, lost education, and out-migration (Collier 1999, Mohammed 1999). Finally, civil war sets the precedent of using the military to solve political problems that can lead to ongoing social instability and even future wars.
On a broader scale, internal war may also become a threat to other countries.
Besides the possibility of the spread of the conflict across national borders, civil wars may endanger diplomatic and business relations, access to strategic resources, and even control over weapons of mass destruction. In addition, intemational organizations and individual countries are often called on to provide humanitarian assistance to counteract the effect of war on affected populations as well as mediation and peacekeeping initiatives to try to bring a halt to civil war. Indeed, since 1990 the number of peacekeeping operations being performed by the United Nations has increased from five to eighteen (SIPRI 2001). For all these reasons, research on the causes of civil war like the one described in this volume is of vital importance. This"study takey direct amratunderstandingtfae- dynamicy behindthe onset of
civil war during the period 1965-1992/ It draws upon a broad range of sociological
theory, a rich database, and a little-used statistical method to examine a fundamental
question; What causes some countries to fall into a state of civil war and not others? In
the end, the study suggests preliminary answers to this question and makes
recommendations for both domestic and international policy aimed at preventing and/or
curtailing the devastation brought on by civil war.
C ivil W ars. 1965-1992
According to Singer and Small (1993), there were 53 beginnings to civil wars in
37 countries between 1965 and 1992 (Table 1 )."* The region with the most civil wars was
Sub-Saharan Africa with 18 wars in 11 countries, followed by Asia with 13 wars in 8
Countries experiencing Region Onsets of civil wars civil war Sub-Saharan Africa 18 11 Asia 13 8 Latin America and the Caribbean 9 6 Eastern Europe 6 6 Near East/North Africa 6 5 Western Europe 1 1 Oceania 0 0 North America 0 0 Total 53 37
Table 1 : Onsets of civil war and the countries experiencing them by region: 1965-1992
^For a discussion of the reasons behind the choice o f years in this study, see chapter 2. ■*For a list of all civil wars with an onset between 1965 and 1992, see Appendix A. 4 “cotmtrieyandLatnrAmerica andthaCaribbearrwitirÇ^waranr6 countries. BotirthoNear
East/North Africa and Eastern Europe had six wars each, and there was only one war on
the very eastern edge of Western Europe (Turkey).
The map in Figure I gives a more graphic representation of the countries that
experienced the onset of civil war between 1965 and 1992. By far the majority of
countries experiencing civil war were in Sub-Saharan Africa and along a line stretching
from the southern part of Europe through the Middle East and South Asia (Figure 1 ).
The only other pocket of activity was in the Northwest part of South America and in
Central America.
In terms of the distribution of civil wars over time. Figure 2 presents the number
of wars that began each year between 1965 and 1992 as well as the number of ongoing
wars each year. As can be seen, there was an increase in both the number of onsets of
civil war and in the number of ongoing civil wars during the period. In the 1960s and
Figure 1 : Countries experiencing the onset of civil war between 1965 and 1992 5 25 25
( 3 Onsets of civil war -A-Ongoing civil wars (onset 1965 or later) 20 -"-Ongoing civil wars 20
15 ' 1 •3 ; I 10 10 n
5
0
Figure 2: Onsets of civil wars per year and number of ongoing civil wars with an onset between 1965 and 1992. first part of the 1970s, the number of onsets averaged just over one per year and the number of ongoing wars held steady at about six. After 1978, however, the scenario changed significantly. The number of onsets increased to an average of 2(6 per year and the number of ongoing wars steadily increased over the period to a high of 23 in 1992. THESTATE OFOUR KNOWLEDGE:
The cross-national study of civil violence has been going on for more than a generation. In 1962, Davies signaled the need for cross-national study of civil violence in his piece on “Toward a Theory of Revolution (p. 19).“ This was a signpost of the work that was to follow in the next several decades.
The first line o f cross-national work on violent collective action involved intense data collection (e.g.. Russet 1964a, Taylor and Hudson 1972, Taylor and Jodice 1983) and a string of cross-national studies that focused mostly on explaining deaths from political violence (e.g.. Russet 1964b, Gurr 1968, 1970, Hibbs 1973, Nagel 1974,
Sigelman and Simpson 1977, Hardy 1979, Weede 1981, Muller 1985, Muller and
Seligson 1987, Midlarsky 1988, Robinson and London 1991). Theoretically, the explanations of civil violence in this literature revolved around three constructs: grievance, resource mobilization/structural opportunity, and later, intemational dependency (see chapter 2 for a more in-depth discussion).
The initial studies were founded on the notion that the fundamental origin of civil violence is the grievances of the population regarding their economic and/or political circumstances (e.g.. Russet 1964b, Feierabend & Feierabend 1966, 1972, Gurr 1968,
1970). Empirically, grievances were measured either in terms of absolute or relative deprivation. The indicators of absolute deprivation consisted of measures such as general economic or political conditions and specific economic or political discrimination or setbacks (e.g. Gurr 1968, 1970), whereas the concept of relative deprivation was measured primarily with indicators of inequality (land inequality [Russet 1964b], sectoral inequality [Parvin 1973], income inequality [Sigelman and Simpson 1977]). 7 itt the 1970s andearijr t980s, theearfystudies that fbcnsed oit the reiationship
between grievance and violent collective action came under heavy criticism. On the one
hand, they were disparaged for failing to find a consistent relationship between grievance
and violent collective violence (Hibbs 1973, Sigelman and Simpson 1977, Hardy 1979,
Weede 1981). On a theoretical basis, they were criticized for neglecting the issue of the
organization of discontent - i.e., the extent to which dissident groups could acquire
control of the resources needed for strong and effective organization (resource
mobilization), the viability of the opposition to either repress or repel dissident groups,
and the ability of the system to channel dissident groups onto less destructive paths
(structural opportunity) (Hibbs 1973, Jenkins and Perrow 1977, Tilly 1978 - but see Gurr
1968, 1970 for an early exception).
In 1985, Edward Muller published an article that brought together and focused the
previous arguments and set the tone for the research on collective violence for the next
several years (Muller 1985). Muller’s dependent variable was deaths per 1 million population. His article was important primarily because he tested and found significant effects for measures representing both competing theoretical traditions. For grievance, he used indicators of income inequality (a new series from Taylor and Jodice 1983) and of historical separatism. For political opportunities, he used a measure of regime repressiveness.
In the years that followed, Muller’s article sparked several types of responses.
Some attacked his work directly by questioning the robustness of his results (see Weede
1986 and Hartman and Hsiao 1988, and Muller’s replies in 1986 and 1988 and his restatement of his findings with Seligson in 1987). Others scholars proposed alternative 8 approaches' to-hismodet hr terms of different specHîcations of the-dependent ■variaWe
(Boswell and Dixon 1990) and further indicators of “internal” variables (e.g., land inequality [Midlarsky 1988], government sanctions, and economic growth [Muller and
Weede 1991]).
The final response to Muller’s work involved the introduction of the third theoretical tradition into the debate: World System/Dependency Theory (see Timberlake and Williams 1987, London and Robinson 1989, Boswell and Dixon 1990, Robinson and
London 1991). Scholars working in this tradition tested variables that were argued to represent the “dependence” of less developed countries on more developed countries
(e.g., world system position and foreign capital penetration). In the end, however, the indicators they proposed failed to show consistent relationships with the dependent variable, and Muller’s results were upheld (Boswell and Dixon 1990).
In the 1990s, the focus o f research on violent collective action shifted in three directions. One was toward the study of ethnic violence. This included data collection efforts like Gurr’s ( 1993a) work on minority groups who have experienced discrimination and theoretical and empirical studies of the relationship between identity and conflict (Cannent 1993, 1994, Ellingsen 2000, Gurr and HarfF 1994, Horowitz 1985,
Jenkins and Rothchild 1998, Reynal-Querol 2001). Although some of these scholars would elevate the role o f identity as a direct predictor of collective violence to the level of either grievance, mobilization, or opportunity (e.g., Ellingsen 2000), the primary contribution was to underscore and map out the way that identity contributed to the development of grievance, mobilization, and opportunity. In addition, the data generated bythesesehola» wereuttique-^see-espeeiattyGtirr t993a); and they provided muclr helpful discussion about how the concept of identity should be specified in cross-national research.
A second focus of research on collective violence in the 1990s was in the direction of investigations directed at the policy-making community. Thus far, these efforts have been mostly data-gathering or data-producing exercises (e.g., Esty et al.
1998), or attempts at developing early warning systems with a focus on short-term predictions (e.g., Jenkins and Bond 2001). Among these studies, however, the work of the State Failure Taskforce deserves special note because of its potential relevance to this study (Esty et al. 1998). In addition to their extensive data gathering activities, the State
Failure Taskforce has defined a unique dependent variable and done some preliminary analyses of the correlates of this variable. Specifically, their dependent variable is the onset of “state failure," or a composite measure of the onset of revolutionary war, ethnic war, genocide/politicide, or adverse regime transition (movement toward autocracy).
Although they intend to study each of these variables separately, thus far they have only published results on the composite variable. Their preliminary findings suggest that three variables contribute to the failure of states: infant mortality (+), partial democracy (+), and trade openness (-). In an investigation of state failure in Africa, they confirmed these results as well as finding that GDP/capita (-), high urbanization with low GDP/capita
(+), low urbanization and with high GDP/capita (+), French colonial heritage (-), and ethnic discrimination (+) all contribute to the failure of states in the directions listed in parentheses.
10 The-finaffocusof researciroircoüectivaviolentactioirmthe- f99ôff had tad a with
the very recent study of civil war itself. This effort has been headed primarily by
economists at the World Bank (e.g.. Collier and Hoeffler 1998, 1999, 2001, Hegre et al.
1999, Elbadawi 2000). They have drawn primarily upon concepts from economic theory
and sophisticated econometric modeling techniques to study the onset and incidence* of
civil wars.
Theoretically, the main contribution o f this work has been to focus attention on
the possibility that, as an alternative to grievance, greed may be a motivating factor that
leads to the start of civil wars (Collier and Hoeffler 1998). The work identifies control of
taxes as the “loot” that may inspire such greed, and measures the amount loot in a country primarily in terms of the size of primary exports relative to GDP (Collier and
Hoeffler 1998, 1999,2001, Elbadawi 2000). In addition, this work parallels earlier analyses on deaths due to violence by suggesting that the ability of a government to repress, as measured by the level of democracy or autocracy in a country (as well as the size of the tax base), is a strong disincentive to starting a civil war (Collier and Hoeffler
1999,2001, Elbadawi 2000, Hegre et al. 1999). Lastly, this work has underscored the importance of ethnic, linguistic, and/or religious makeup of a country, which they discuss in terms o f the "transaction costs,” as an important predictor of civil war (Collier 2001,
EUingsen 2000, Hegre et al. 1999).
In terms of results, these studies have shown significant effects for the share of primary exports and its square, the level of repression (i.e., democracy) and its square, the
^Incidence is defined as the prevalence of civil war at a given point in time regardless of when the conflicts
11 tevetof devetopmenr (and soraetimes-itrsqaare [Hegre etah 1999J); andthe size of the
population (Collier and Hoeffler 1998, 1999,2001, Elbadawi 2000). Support has also
been found in some studies for the role of ethnicity, although in models that are under-
specified (Hegre et al. 1999, EUingsen 2000, Reynal-C^uerol 2001). In contrast to the
earUer studies of collective violence, however, little support has been found for the role
of inequality (CoUier and Hoeffler 2001, Elbadawi 2000).
As a final note, it is worth mentioning that World-System/Dependency theorists
are again on the fringe of this research. Gissinger and Gleditsch ( 1999), in a study of the
onset of civil war, investigated the effects of total trade (exports + imports/GDP, similar
to the State Failure Taskforce [Esty et al. 1998]), and a tnmcated dataset on foreign direct
investment ( 1980-93). TheoreticaUy, they argued that both of these variables reflect the
level of dependency of the country, and thus higher levels of trade and foreign direct
investment would lead to a higher probability of entering into civil war. In an under
specified model, they found support for the influence of both of these factors, although
the impact of trade was in the opposite direction than they had hypothesized.
In summary, over the past 40 years the different lines of cross-national research
on collective civil violence have had different strengths and weaknesses. The early
research hammered out a fairly sophisticated theoretical understanding of the immediate
causal mechanisms behind coUective civU violence while struggling with a paucity of
data on the dependent variables and a limited assortment of methodological tools with
started. The use of this variable has been harshly criticized (Beck et al. 1998, Beck 2001). For a more extended discussion of this issue, see chapter 3. 12 whiclr ta da analysis. Thcmore recent research hashadthe advantage of a macir larger set o f methodological tools and a broader range o f data but, until now, has either ignored or not yet taken advantage of the richness of the theory of the early work.
One final critique of this literature can be made based on some recent research on democratization. Although it is out of the scope o f this work to review the literature on democratization in depth, it is interesting to note that scholars have begun to intentionally incorporate another rich theoretical tradition into their analyses: theories of economic development (e.g., Crenshaw 1995). These theories have been used to tie the relationships between immediate predictors and democracy to the broader social context in which they interact. In so doing, they add a depth to the analysis that would be well worth adding to the analysis of the onset of civil war.
T he C u r r en t A n a l y sis
The research presented in this volume is an attempt to “bring it all together.” It does this by, first, developing a framework for studying civil war that links the rich theoretical heritage on collective violence with macro-social theories of societal development (chapter 2). Next, it describes the method used in the analysis, proportional hazards regression (Cox analysis), which is a relative newcomer to the study of collective civil violence, as well as describes the data that are employed as both dependent and independent variables (chapter 3). Then, the results of the analysis are presented (chapter
4). Finally, the work concludes by reflecting on the implications of the results of this study and speculating on the future o f civil war across the globe (chapter 5).
13 IrMhe- endv P corrfiim that itia important to-thinir of civif war not just as an act of agency carried out by motivated and equipped individuals who take advantage of an opening in the opportunity structure. Rather, it is helpful to conceptualize civil war as an emergent property that is highly dependent on the macro-social environment. In these terms, civil wars occur more readily in countries with mid-levels of techno-ecological inheritance and that have experienced only moderate levels of industrialization. In addition, civil wars are encouraged by higher levels of past experience with civil war, whereas they are discouraged by a strong service sector. The effect of ethnicity is less clear, with civil war either being encouraged by higher levels of homogeneity or by a situation where the size of the 2"** largest group is approximately 20 percent of the size of the largest group.
Related to this last finding, 1 also find that civil war occurs more readily in countries with ethnic groups that have a previous history of lost autonomy and, similarly, in countries with ethnic groups that have economic, political, and social grievances.
Finally, I find that general political discrimination, higher proportions of agricultural exports per GDP, and mid-levels of inequality all encourage civil war.
14 CHAPTER 2
THEORIES OF CIVIL WAR
C onceptualizing C ivil W a r
Civil war as a phenomenon for analysis presents some interesting conceptual
challenges. On the one hand, the destructiveness of civil war is a powerful incentive to
focus directly on war itself and the immediate conditions that lead it to break out. On the
other hand, civil war is arguably uncommon enough to make scholars question how this
occasional phenomenon is related to broader patterns of social development. An adequate synopsis o f the theories of civil war must take into account, therefore, the challenges posed by both of these perspectives.
In practice, the specification of the association between the inunediate conditions that lead to civil violence and the macro-social environment that encourages civil violence depends not only on an appreciation of these relationships in isolation, but also on an understanding of the way the determinants from the two levels are interrelated.
This task is complicated for two reasons. First, there are competing notions within both perspectives about the specific immediate conditions or macro-social contexts that are important for explaining violent collective action. For instance, some have argued that grievance in a population is a necessary immediate precursor to civil violence (Gurr
1970), while others have argued that grievance is either ubiquitous or uimecessary
15 f McCarthy arid Zald \9T}, Tilly I97d>. Second; there^art disparate views ort the
character of the relationships between the proximate conditions of civil war and the
macro-level environment. For example, modernization theorists have suggested that the
transition to industrialization predisposes a country to civil war by engendering
grievances in the population due to the inability of underdeveloped political institutions to channel the increasing mobilization of the population (Huntington 1968). Elite theorists, in contrast, view civil war as an outgrowth of a conflict between elites that is divorced from the underlying economic conditions (Higley and Burton 1989). For these two reasons, the most helpful theoretical treatment of the causes of civil violence will pay close attention to the immediate conditions of civil war, the macro-social context of collective violence, and the ways in which these two are related.
In the pages that follow, I divide the discussion of theories of civil war into two sections. First, I review current thought on immediate conditions that lead to civil war.
Second, 1 discuss civil war in its macro-social context with a special focus on the relationship of this context to the immediate conditions of civil war.
T he Im m e d ia t e C o n d it io n s t o C iv il W a r
The literature on the immediate conditions of social conflict lists a broad array of factors that could lead to a civil war. For the analyst wishing to summarize these conditions, it is helpful to look to another discipline where there has been a concentrated and scrutinized effort to formulate a comprehensive scheme linking actorfs) and a drastic
16 event ükeeivit war. I suggest that the^most appropriate realnr ftonr whieh te draw sueir a
scheme is the investigation of violent crime. Specifically, I propose that we examine the
immediate conditions of civil war in terms o f motive, means, and opportunity.
Drawing this conceptual scheme from the realm of the investigation of violent
crime is appropriate for several reasons. First, civil war and violent crime are similar
phenomena. They both involve two parties who tend to be identified with the language
of “victim” and “perpetrator.”* In addition, they are both drastic events for which the
consequences of engagement are potentially severe enough for both the actors themselves
and the larger community to require more than just passing thought into why actors
would pursue such courses of action.
Second, treatments of the theories of violent conflict already incorporate some of
the language of motive, means, and opportunity. In the competing hypotheses of Greed
vs. Grievance (see discussion below), the central contention of the debate is the specific
motives o f the actors involved. Likewise, discussions by Resource Mobilization theorists
revolve primarily around the different means that are mobilized for engagement in violent
collective action. Possibly the best example o f this is Structural Opportunity Theory,
whose relationship to the conceptual scheme is obvious.'
* It could be argued that the victim of a violent crime is not really “involved” in any meaningful way because he or she is simply on the receiving end of the violence, whereas both parties of a civil war actively engage each other. However, sometimes the intended victims of violent crime turn the tables on would-be perpetrators in an active way. In these cases the language of “self defense” comes into play, which also tends to be mirrored in war-time rhetoric. ^ In feet, 1 would argue that the category “opportunity” may even be more appropriate for organizing the immediate conditions of civil war than for violent crime. For. in violent crime, the issue of opportunity comes down to whether a person has an alibi for the time during which the crime was committed. In the case o f civil war, however, this construct points to a much richer set of possible incentives or deterrents (see discussion below). 17 Finatiy; there haa beeir a tong^ancF intense history of debate that hay gone into
generating this conceptual scheme. This can be seen in the fact that motive, means, and
opportunity has been incorporated into introductory textbooks on the science of criminal
investigation (e.g., Weston and Wells 1986) as well as applied as a general standard of
guilt in case law (for example, see Marshall vs. Commonwealth o f Virginia). Together,
the similarity between the phenomena of civil war and violent crime, the parallels
between the extant theories of collective civil violence and violent crime, and the well-
established character of the motives, means, and opportunity framework make it a good
tool for organizing the discussion of the immediate conditions of civil war.
Using this scheme, I divide the discussion of the immediate conditions that lead to
civil war into three sections. First, 1 describe the motives that exist in a population that might prompt actors to rise up against others. Second, I review the means that may be accessed and mobilized to mount a violent campaign. Finally, 1 discuss the opportunities that lend themselves to a civil war.
The Motives to Violent Collective Action
Probably the most extensively developed part of the literature regarding the immediate conditions to civil conflict has to do with the motives behind those who adopt violent strategies. Two motives have been suggested: grievance and greed. The following discussion summarizes the literature on both of these topics.
Grievance and Violent Collective Action
One of the most intuitive hypotheses about civil conflict is that grievance, or something that causes discontent within people, must be present within a society before 18 civihconftrcrwiU^brealcotir(Davres^t962;Fereraljend‘an(f&Feieraben(f t966, t972, Gurr
1968, 1970, Russett 1964b). The actors with whom grievance is most commonly associated are the masses in society (Komhauser 1959, Huntington 1968), especially when looking for relationships between grievance and large-scale violent conflict, but some have also suggested that elite grievance can lead to violence (see Gurr 1970 for a discussion on elite grievance and other forms o f violent conflict).
It has been argued that the strains that cause grievance result from two psychological conditions: absolute deprivation and relative deprivation (Gurr 1970). The absolute deprivation hypothesis posits that the experience of grievance always occurs in humans when a set of definable circumstances exists. This experience itself drives people to engage in violent acts. This idea has prompted scholars to test whether measures of the “general standard of living" are associated with civil conflict (Jenkins &
Schock 1992).
In contrast, the relative deprivation hypothesis suggests that civil unrest is motivated by a discrepancy between what people have attained and what people believe they should have attained. In short, the question is whether people feel they are getting their “just deserts” (Gurr 1970). It is argued that the experience of relative deprivation occurs in response to both economic and political conditions.
In terms of economic conditions, one type of comparison involves short-term setbacks in the economic situation. Severe recession, rapid inflation, and balance of payment crises have all been used to explain such setbacks (Gurr 1968, 1969, 1970, Gurr
& Duvall 1973, Gurr & Lichbach 1979, 1986). Civil war, however, is such a costly undertaking that it is hard to envision such setbacks by themselves leading to such large- 19 scaiemobüizatian; Instead; it may be^bettertaconceptuaüze-regressive-changes m people's conditions as “triggers” that might set off a civil war in the presence of other aggravating factors. This is analogous to what Jenkins and Schock ( 1992) entitle the
“sophisticated” approach to modeling the relationship between interests and conflict, where short-term events alter the calculations of rewards within a context of structurally fixed interests to create civil conflict.
A likely candidate for a more fundamental economic factor leading to the experience of relative deprivation is economic inequality. Part of the literature that draws upon this perspective treats general inequality as a potential producer of feelings of deprivation (see Lichbach 1989 for a review of this literature), while others focus on inequality between specific groups (e.g., ethnic groups — see Williams 1994 for a review of this literature). Regardless of the reference group, there are four different specifications of the relationship between inequality and the experience of deprivation
(Jenkins and Schock 1992):
1 ) A simple linear model that emphasizes the envy of the have-nots and the greediness of the haves (Russet 1964, Sigelman & Simpson 1977, Park 1986, Muller & Seligson 1987);
2) A positive, curvilinear effect that stipulates that people tend to compare themselves to others that have been or are presently situated similarly. Hence, the experience of deprivation is at its height at mid-ranges of inequality (Nagel 1974);
3) A “positively accelerated effect” that suggests that inequality will be insufficient to cause violent mobilization until it reaches very high levels (Muller 1985);
20 4> Hirschmatt'»fl97SK‘tunneteffect’” thesis‘thatposits^people‘experience deprivation as a result of personal economic stagnation or deterioration in the context of other people who are advancing. This model has been specified in two ways: As an inverted v-curve suggesting that grievance is most intense at higher and lower levels of inequality, or as a time lag before inequality creates grievance.
An analysis of civil war that claims to be thorough must test each of these specifications of the economic deprivation hypothesis.
Relative deprivation resulting fix>m circumstances in the political arena has also been suggested to encourage grievance and, consequently, civil conflict. This notion is expressed in the literature on the “strains of modernization” or “system imbalances”
(Feierabend & Feierabend 1966, 1977, Feierabendet al. 1969, 1970, Huntington 1968).
Here, an increase in the standard of living or in the level of social mobilization (e.g., urbanization, education, or media exposure) without a concomitant increase in the political opportunities in society (see below) is suggested to lead to an experience of deprivation and, consequently, to an increase in the likelihood of social conflict.
A notion similar to the strains of modernization hypothesis is Tilly's polity model
(1978). Without specifying the conditions in the broader context, he divides actors into those with routine, low-cost access to political power and those who are excluded from such influence. Unruliness, he argues, may flow from the excluded because of their lack of access when compared to those that have such access. Although he does not limit the connection between exclusion and disruptive acts to grievance alone, this is surely a part of his thinking. Tilly argues that those with access, in contrast, will only resort to unruliness if threatened with loss of access. In these ways grievance may flow from political as well as economic comparisons.
21 The Relative Deprivadoitperspeetive-has^ some sigoiiieant logieatliabilitieSi For
example, the leap from the existence of inequality to the experience of deprivation is at
best indirect and at worst can be faulted as an ecological fallacy or psychological
reductionism (Jenkins & Schock 1992). Likewise, there are a host of cognitive states
other than grievance such as cognitive liberation (McAdam 1982) or alienation (Paige
1975, Midlarsky 1988), each of which could explain the connection between the
deprivation and conflictual action. Despite these criticisms, it would be unwise to
abandon relative deprivation carte blanche, for the criticisms do not condemn the
perspective as much as push scholars to strive for validity in measurement, rigor in
testing, and the acknowledgment of the caveats of the perspective.
The relationship between grievance as a whole and violent collective action is also vulnerable to critique on logical grounds. Olson ( 1965) argued that social action is only rational when it gratifies the divisible interests of the actors who participate; i.e., it produces “private goods.” When social action primarily produces “public goods,” or when the number of interested potential beneficiaries rises above a certain relatively low number, it becomes irrational for any individual to participate. In short, people become
“free riders.” Consequently, he suggested that large-scale violent social actions only occur when participants are motivated by personal gain distinct from the indivisible collective gains that are likely to be produced.
There are several responses to the ideas that Olson elaborated. Smelser (1963), for example, hypothesized that irrational behaviors are at the root of violent collective action. He held that people are apt to act on irrational impulses when the equilibrium of society is upset, which includes engaging in a variety of types of violence including 22 value-OFientedsK>vements^(e.g., revc^uticms); OtbeFs^seholars^have attemptedto-dispel
the notion that public goods and private goods are always distinguishable, suggesting that
the two become fused through individual processes of value integration (Parsons 1978),
or via forces external to the individual (Jenkins 1982).
A final response to the logical difficulties of theories of grievance is to set the role
of grievance aside altogether. This has been done by scholars advocating that greed is the
real motivation to collective violence on the level of civil war, as well as by those who
argue that grievance is ubiquitous and so is a less important construct than the
mobilization of resources or structural opportunity. I discuss the former approach in the
next section, while the latter approaches follow below.
Greed and Violent Collective Action
An alternative hypothesis to the notion that grievance is a necessary precondition of civil war is that insurgents are motivated by greed, or by the financial resources over which they will gain control if they are successful in winning the war (Azam 1995,
Collier and Hoeffler 1998, 1999, 2001, Grossman 1995). This motivation, it is argued, avoids Olson's free rider problem because it frames the gains that motivate people to engage in war as^ private goods (Collier and Hoeffler 1999).
The greed hypothesis has been much less well developed than the grievance hypothesis. Economists have been on the forefront of this work, although their efforts began only recently (Collier and Hoeffler 1998). Sociologists and political scientists, in contrast, have tended to focus on mass movements or movements of the left, both of
23 whieb temtte^emphaske the Fole o^gnevanees^fMeCarthy and Zald 1977>. One
exception to this, however, is the literature on the role of elites in violent collective
action. I discuss both of these approaches below.
Economists came to emphasize the hypothesis that greed is a primary motivator to
violent collective action by applying a cost-benefit model to the issue of civil war.
Engaging in civil war, they suggest, implies an attempt to acquire resources either
through extortion of particularly profitable enterprises or by gaining control of the tax
base of the state (Collier and Hoeffler 1999). Presumably, these benefits accrue not only
to the leaders of the insurgency but also to their armed forces and the other allies who
assist in the war effort. On the other side of the equation, economists view the costs of
civil war primarily in terms of current lost earnings due to the switch away from direct
economic activity and future lost earnings due to the destructiveness of civil war
(Hoeffler 1998)*. The more the potential benefits of war outweigh the costs, therefore, the more likely potential insurgents are to start a war.
Elite theory from sociology and political science is a good complement to the ideas put forth by the economists. This theory has its origin not in the question of why people engage in violent collective action but, rather, in the question of who is primarily responsible for such action. Their answer is that we must look to the elites of society and their conflict with one another (Mosca [1896] 1939, Pareto [1901] 1968, Lachman 1990).
In this literature, theorists follow Weber ([1921] 1978) in defining elites as members of society who have the capacity to appropriate resources from non-elites
24 througbtheÎF autheritattve-positieny at the-headr of targg bureaucracies (Higtey and
Burton 1989, Lachman 1990, Mosca [1896] 1939, Pareto [1901] 1968). Elite conflict, in
turn, is deflned as the attempt by one group o f elites to subordinate part or all of the
organizational apparatus o f another group of elites within its own (Lachman 1990). Civil
war, then, is a special case of elite conflict that can occur if the organizational apparatus
at stake is the state (Higley and Burton 1989).
Although elite theorists do not specifically list the motivations for this action, they
can be inferred through the distinction that is made between elite conflict and class
conflict (Lachman 1990). By definition, elites are in control of significant resources by
virtue of their authoritative position at the head of large bureaucracies. Consequently, the
claim that they are motivated to engage in collective violence due to experiences of either
absolute or relative deprivation is a stretch, especially when compared with the
conditions of non-elites. Rather, the language of elite theorists is more often in terms of
the “interests” of elites, a meaning much closer to our definition of greed than of
grievance.
In the present analysis, I use these theories on the motives to civil war to identify
indicators of both greed and grievance that are, in turn, used in multivariate models to predict the onset of civil war. If the grievance hypothesis is true, indicators representing the general standard of living (absolute deprivation) or inequalities in the economic or
* When economists detail the calculation of the costs of civil war, they write in terms of the resource balance between potential insurgents and the state. In this work, this line o f thought is discussed under the heading of the means to violent collective action (below). 25 poKticat sphere^ (relative deprivation)^'with correlate witirthronser of civil war. If greed is
the more helpful construct, then indicators of the amount of potential loot in a society will
be related to the onset of civil war.
It is possible, however, that indicators from neither theoretical framework will aid
in this task. If this turns out to be true, it will have been foreseen by theorists espousing
the next set of theories on the means and opportunities to violent collective action.
The Means to Violent Collective Action
The literature on the means to violent collective action arose, in part, in reaction
to the extensive debate over the motives of violent collective action. Some argued that to
posit a motive is a necessary but insufficient condition for explaining violent action (Gurr
1970). Others suggested that the motives to collective violence may be ubiquitous, and thus attention needed to be focused elsewhere to explain violent collective action (Jenkins and Perrow 1977, McCarthy & Zald 1977, Tilly 1978). In either case, scholars agreed that the story about what causes violent collective action was incomplete when told strictly in terms of motivation.
One o f the missing pieces to explaining violent collective action that scholars have suggested is a discussion about the means to such behavior. Whatever the motive, h is argued, people must be able to mobilize resources in order to carry out an act of civil unrest (Gamson 1975, Jenkins 1981, 1983, McCarthy and Zald 1973, 1977, Oberschall
1973, Tilly 1978). This is especially true in the case of civil war.
Extensive reviews of the Resource Mobilization literature are available elsewhere
(Jenkins 1983, McCarthy and Zald 1978). For this reason, in the following section 1 give
26 a brief review" of thetypes of resources^ that have been suggestedare mvoivecT m violent
collective action. I organize these types into three categories of resources: people,
money/capital, and organization.
In terms of people, three general classes are commonly mentioned. First are those
with some specialized skill that can contribute to the organization and direction of the
movement (e.g.. Freeman 1979). These will often include people with leadership skills,
the ability to organize, and technical expertise in the methods that are used to carry out
the goals of the movement organization. In the case of a civil war, these will especially
include military leaders, which tend to be the most cohesive elite groups in less-
developed societies (Kennedy 1974, Jenkins and Kposowa 1990, Kposowa and Jenkins
1993).
A second category of individuals is a general cadre of followers, or what
McCarthy and Zald (1977) call constituents. These people typically act as flexible,
unspecialized labor or provide other resources to the organization. Again in the case of a
civil war, the most important category would be the soldiers themselves.
Finally, a third class of people is supporters from outside of the movement. These
often include special benefactors who are willing to lend aid in the form of money or capital without directly becoming involved (McCarthy and Zald 1977). In the case of civil war, supporters who are willing to finance the war effort are an especially important group. Additionally, scholars have pointed to importance o f support from what
McCarthy and Zald (1977) have called adherents, or members of the general populous who believe in the general aims of the movement and so are willing to put up with the disturbances and inconveniences associated with the movement. 27 The second class of resources that are commonly referred to in the literature is money and/or capital. Although little needs to be said about the content of this resource, the literature does include a discussion about the sources of money and capital of which a movement organization may make use. One traditional assumption about resources is that they come primarily from those who will be direct beneficiaries of the outcomes of the movement (McCarthy and Zald 1977). However, more recent scholars have noted the importance of outside contributors of money and capital as well as the strategy of co opting the resources of institutionalized actors (McCarthy and Zald 1973, 1977).
Regardless of the source of money/capital, there is little disagreement about its importance to the success of social movement organizations.
The final category of resource that is often listed in the literature is organization.
Analysts such as Oberschall (1973) and Tilly ( 1978) have stressed the importance of the degree of preexisting group organization for the mobilizing potential of a movement organization. Movement organizers who are able to draw on existing groups with strong, distinctive identities and dense interpersonal networks exclusive to group members are more likely to be successful in recruiting participants and gathering money and capital.
In less developed societies,, two sources of such, organization have been mentioned: existing ethnic groups (Williams 1994), and the military (Jenkins and Kposowa 1990,
Kposowa and Jenkins 1993).
Before leaving the topic of the means to violent collective action, it is worth noting that the relationship between the amount of resources available to a social movement organization may not be directly related to the likelihood o f that organization
28 carrying out movement activities. Rather, as Gamson, Fireman & Rytina (1982) point
out, the relationship may be curvilinear, with movement organizations waiting until they
pass a certain threshold in the mobilization of resources before action is taken.
In the present analysis, the discussion of the means to violent collective action is
used to identify indicators of the types of people, money/capital, and organization that are
important in civil war. If they truly are associated with civil war, we should find that the more resources are available to potential insurgents, the higher the likelihood of the onset of civil war, although this relationship may be subject to a threshold. Conversely, situations in which resources are scarce should be less conducive to the onset of civil war.
The Opportunities for Violent Collective Action
The concept "structural opportunity” suggests that neither motive nor means may be either necessary or sufficient conditions to bring about violent collective action.
Rather, motivated and equipped actors may also be dependent on the opportunities afforded by the broader social context in order to initiate an act of collective violence.
One of the most straightforward structural factors that has been suggested to affect the opportunities to conflict is the power of potential opponents. This has been specified as the ability of elites to suppress potential threats, whether by placation or oppression (Dix 1984, Hibbs 1973, Lichbach 1987, Muller & Weede 1990, Zimmerman
29 1983)’. In addition, the coherence of the elite class has been suggested as a condition that could affect the cost-benefit calculation of engaging in civil conflict (Piven & Cloward
1977, Skocpol 1979).
Another set o f ideas about the structural opportunity for civil conflict is to consider other actors on the scene. One set of such actors is potential allies with whom an insurgent group could join, whether they are grass-roots organizations or foreign states
(Jenkins and Schock 1992). Another “actor” that has been more recently stressed is the state and its institutions. Scholars such as Skocpol (1979) have argued successfully that the state is sometimes a relatively independent actor within society and thus must be considered in any equation predicting political outcomes. The state itself has a stake in the continuation of the status quo and so may use its resources to repress. Alternatively, its interests may lead it to side with an insurgent group that seems to promise improvements in the lot of the state bureaucracy.
A final type of opportunity structure that has been discussed in the literature is related to broader social conditions. On the one hand, the basic economic organization of society has been mentioned. As Paige (1975) describes in his exploration of peasant revoltv situations where economic interaction is essentially a zero-sum-game may leave no alternative for aggrieved groups to address their grievance other than violence. The same may be true o f the political realm (Tilly 1978, Muller 1985, Muller & Seligson
’Two concepts, the capacity to repress political opponents and the autonomy of the rebel group (Wolf 1969, Skocpol 1979, Jenkins 1983), have been treated as distinct but are actually very similar to one another. The former has been considered a part of the opportunity structure while the latter is conceptualized as a resource to be mobilized. However, autonomy actually signifies the absence of something, in this case the influence of the opposition. This does not lend the construct well to translation into measurable indicators. For this reason, this work employs the concept “ability to oppress” instead. 30 1987, Gastil 1989, Gurr 1989, Boswell & Dixen 1990, Muller & Weede 1990). If civil rights (i.e., the ability to mobilize free of government control) or political rights (i.e., the ability to participate in the political system) are suppressed, little recourse may be left other than civil unrest. Alternatively, very closed and very open polities have been hypothesized to inhibit civil conflict, whereas those that are partially open or those undergoing change in the level of openness may allow such action (e.g., Muller 1985,
Muller and Seligson 1987, Esty et al. 1998). Each of these more intractable circumstances helps to define the political opportunity structure that, in turn, influences the potential for civil violence to break out.
In the present analysis, this discussion of the opportunities to violent collective action is used to identify indicators of the opportunities that may arise that make civil war more likely. If this perspective holds true, we should find that the onset of civil war is more likely when potential opponents are weak, potential allies are numerous, and the available alternatives to civil war are limited.
As a whole, all three of the concepts discussed above share one main strength:
They are rich sources of potential variables that can be tested for their relationship to civil war. As was argued above, however, conducting analysis based on the theory that can be developed from these concepts alone would provide an insufficient framework for adequately explaining civil war. For this reason, the following discussion shifts our attention to the level of macro-social theories.
31 T h e M a c r o - s o c ia l E n v ir o n m e n t a n d c iv il w a r
As logical as the previous accounts of political violence may seem, empirically verifying the extent of grievances, greed, means, or structural opportunity at micro- and meso-levels is very difficult. There are no scientific surveys of peasants to discern their levels of grievance, nor any comparative studies of elites or groups friendly to insurrection across the globe. Because of this dearth of proximal information, determining the degree of grievance or structural opportunity within a society is generally reconstructed after the fact on a case-by-case basis using documentary evidence and verbal argumentation (i.e., qualitative comparative research). While such thick description' can be quite useful as history and as a theory-building device, it lacks the rigor and generalizability of quantitative analysis (e.g., exacting variable construction, use of deductive theory).
Fortunately, the research problem can be recast in terms of macro-social theories.
In essence, the notion here is that grievances, opportunities for profit, means to war, and structural opportunities are sharply constrained by large-scale (i.e., macrosocial) social structures. Moreover, these motives, means, and opportunities are not sui generis, but in fact constitute emergent properties generated by the larger social environment.
If the notion of "bounded rationality" is correct (i.e., that rational action can only be defined by referencing the social environment within which it occurs), then it is clear that individual and group motivations and behaviors are generally epiphenomena of structural/environmental conditions. Given this, we can assume that all the world's people routinely employ rationality (i.e., self-interested behavior wedded to empirical assessments of possible success/failure within a given environment). \Miat differs are the 32 structural conditions that shape the rationality of action, conditions that are measurable
(unlike arguments about social psychology) and provide a reasonable hope of quantitative
comparisons. When applied to specific phenomena, macro-social theories are at their
best when they suggest the most relevant ways that the environment varies under
different circumstances (e.g., in different countries) and connect the environment to other
theories that suggest how people are likely to behave under these circumstances (e.g.,
rational actor theory).
Three macro-social theories are considered in this work: Structural
Modernization Theory, Human Ecology Theory, and Political Economy Theory. As
argued above, the most helpful discussion of macro-social theories and civil war will pay
close attention to the ways in which each is related to the immediate conditions of civil
war. For this reason, the discussion of each of these theories includes both a short
introduction to the most prominent features of the theory and a focused discussion on the
ways that the fundamental mechanisms highlighted by each theory will affect the
immediate context of civil war - i.e., the motives, means, and opportunities.
Structural Modernization Theory and Civil War
Structural modernization theory focuses primarily on the implications of industrial technology for society. Under this theory, all societies can be placed along a single continuum of economic development according to the extent to which they have incorporated industrial technology (Deutch 1961, Levy 1966, Parsons 1964, Rostow
1960). The central notion is that a society dominated by either a “modem” (i.e., industrial) sector or a “traditional” (i.e., non-industrial) sector, will conform to a logic
33 associated with that sector (Kerr et al. I960). Societies undergoing the transition from traditional to modem, in contrast, will be more heavily influenced by the upheaval caused
by that change (Olson 1963, Feierabend et al. 1969).
In general, modernization theory suggests that societies under the sway of non industrial technology (i.e., traditional societies) should experience limited violent collective action. This happens despite the fact that they exhibit characteristics encouraging grievance such as subsistence economies, poverty, social inequality, and authoritarian regimes (Moaddel 1994). Rather, the peace in these societies results from the institutional stability, social continuity, and effective social and political integration of groups and classes within the social order that is built around agricultural technology
(Feierabend et al. 1969, Huntington 1968). In short, modernization theory associates traditionalism with a social equilibrium that discourages violent conflict on the level of civil war.
In a similar way, societies under the sway of industrial technology should also reach a state of equilibrium that discourages civil war. In this case, however, the underlying conditions are more favorable to individuals and families. They include mass affluence, low inequality, and political democracy (Huntington 1968, Kerr et al. 1960).
In addition, mature industrialized societies should also exhibit institutional stability, social continuity, and effective social and political integration of groups and classes, only this time built around a dominant industrial sector (Moaddel 1994). Consequently, the balance of the effect of mature industriahsm on the prospects for civil war should also be negative.
34 Accordmg to structuraT modeinizatidn, it Is wfien à society is maldhg tEe transition from traditional to modem that it is most likely to experience civil war. During this time, the motivations to war reach the most intense levels, the means to war are most plentiful, and the opportunities to war become most conducive. This occurs due to a series of factors, including heightened inequality, rapid population growth and urbanization, and increased mobilization with decreased effectiveness of political institutions.
First, the process of modernization causes inequality to increase between the various sectors of the economy as capital is concentrated in the dynamic industrial sector
(Kerr et al. 1960, Moore 1966). This leads to both heightened means and motives to civil violence. In terms of means, the concentration of large amounts of capital in the hands of a few makes it easier for those who would wish to start a war to commandeer what they might need. In terms of motives, the increased productivity of workers in industrializing sectors means higher incomes for them as compared to others in the labor force as well as some level of worker dislocation as more productive workers force others out of traditional occupations (Olson 1963, Feierabend et al. 1969). As a result, modernization theorists look for income inequality to increase during industrialization (Adelman and
Morris 1973, Chenery and Syrquin 1975, Crenshaw 1992, 1993, 1994, Kuznits 1963).
Some economists view rising incomes and the increase in inequality in a slightly different fashion. For them, the essential aspect is the expansion of the tax base (Collier and Hoeffler 1998, 1999, 2001 ). This is specified both in terms of the general tax base of society and well as the specific tax base of especially profitable industries. At middle
35 levels of development, income an(f, consequently, the tax Base, reach high enough levels
to be a temptation for potential “looters” while not yet high enough to fortify the ability
of those who control the state to protect themselves.
The second factor associated with the modernization process that should heighten
the likelihood of civil war is rapid population growth. Very early on, modernization
theorists observed that industrialization is associated with rapid mortality decline without
an immediate, equivalent response in fertility (e.g., Notestein 1945). As a result,
societies undergoing modernization tend to experience rapid population growth (Kirk
1996, Raulet 1970, Crenshaw et al. 2000). As Goldstone (1997) has argued, population
growth tends to distribute resources away from those who labor to those who employ
labor. In this way, the high rates of population growth in modernizing societies increase
the level of grievance over that which is seen in either traditional or mature industrial
societies. Moreover, the younger part of the population increases faster than the older
part due to the simple mechanics of population growth (Christenson 2001 ). As a result,
the proportion of the population that could potentially become soldiers will increase
during industrialization (Goldstone 1997).
Population growth during industrialization contributes to an associated
phenomena that modernization theorists view as an intrinsic part of the industrialization
process and that may contribute to civil wan rapid urbanization. During modernization,
population growth is typically coupled with massive rural-to-turban migration due to both push and pull factors. This leads a society’s urban centers to expand rapidly (Alonso
1980, Hawley 1981 ), often swelling beyond the existing capacity of the city to provide
in&astructure, housing, social services, and employment (Kasarda and Crenshaw 1991). 36 In this way, a mismatch between tbè needs of people and the available resources is created that contributes to the level of grievance in the population. In addition, the increased proximity of those with interests in fomenting a civil war will increase the likelihood of them coming into contact, again raising the likelihood of civil war.
The final connection that modernization theorists suggest to exist between industrialization and civil war results from an interaction between an individual level phenomenon and an institutional level dynamic. Modernization theorists posit that the concentration of people that is attendant on industrialization (e.g., through urbanization) exposes people to literacy, mass media, and interest groups (Lipset [1960] 1981). These experiences, it is argued, "mobilize" the population, or heightens the motivation of people to participate in civic life (Deutch 1961). At the same time, modernization theory posits that the state will not be immune to the transformations that occur during industrialization, and will therefore experience a temporarily loss of effectiveness (e.g.,
Huntington 1968). Together, the inability of the political institutions of society to channel the rising mobilization of the population leads to a form of deprivation and, consequently, an increased likelihood of entering into civil war (Bollen 1983, Crenshaw
1995, Huntington 1968).
Human Ecolosv Theory and Civil War
Modernization theory has been criticized on many fronts. Scholars working out of the human ecological tradition have provided one source of such criticism. Their primary critique is that, in focusing on industrial technology. Modernization Theory
37 ignores the offier organizatîonar hases o f society tharcan also have a profound influence
on society. Specifically, they have suggested the perspective taken by modernization
theorists is limited in three important ways.
First, modernization theory tends to ignore the effect of pre-industrial
development, or techno-ecological inheritance, on a country’s future development
(Crenshaw 1995,2001, Lenski 1966, Lenski and Nolan 1984, Nolan and Lenski 1985).
Scholars who work from this perspective posit that historic differences in the physical
environment (e.g., climate, disease regimes, and topography), as well as the social environment (e.g., the level and type of contact with techno-ecologically rich of neighbors) prior to contact with industrialization are critically important. Such histories result in different configurations of subsistence technology, institutional development, and sociospacial efficiencies within countries that greatly influence the development trajectory of a country (Crenshaw 2001). Consequently, scholars who emphasize techno- ecological inheritance believe modernization theory is limited in its ability to explain the social and economic development of societies.
The second criticism that comes from scholars working from a human ecological perspective is that the emphasis of modernization theory on industrialism ignores other sectors of the economy that may exercise dominance during and after modernization
(Crenshaw et al. 2000). In particular, the service sector has been highlighted since it is generally larger and employs more people than the industrial sector (Sundrum 1990). If the dominance of this sector is as influential as scholars working firom this tradition suggest it can be, modernization theory will have missed an important aspect o f the macro-social environment that can help to explain the development of many societies. 38 The thinh criticism of modernization theory from homair ecotogicaf theory is that, due to its tendency to view social relationships strictly in economic terms, it tends to ignore the influence of the non-economic bases of relating that play an influential role in society. In particular, the role o f ethnicity has been highlighted (Crenshaw 2001,
Crenshaw et al. 2000, Williams 1994). Consequently, the characteristics of the industrialization process that modernization theorists have put forth may be distorted by the presence of ethnic solidarities in society.
In sum, the human ecological perspective suggests that modernization theory is limited in its perspective on the pertinent macro-social factors that influence society.
Looking at techno-ecological inheritance, alternative key functions, and the ethnic makeup of society, therefore, can give us a broader perspective on the origins of motives, the abundance of means, and the emergence of opportunities to violent collective action.
Techno-Ecological Inheritance and Violent Collective Action:
The core idea behind the impact of techno-ecological inheritance is relatively straightforward: the sociospacial organization of society is heavily influenced by past population density (Crenshaw 1995). In small, thinly populated societies, social relationships tend to be characterized by high levels of interdependence within very small groups (e.g., clans, tribes, etc.) and much looser ties elsewhere. As population density grows, however, competitive pressures build between individuals and groups as they begin to vie for scarce resources. In the long run, population density and the competition it elicits eventually encourage people to innovate and differentiate (Spencer [1852] 1972,
Durkheim [1933] 1984), thereby laying the foundation for complementary relationships
39 based oir interdependency (he., protomodemizatioir); hr the short term, however, the
competitive climate created by growing population density will cause strains similar to
those experienced by societies undergoing the transition to modernization.
In terms of civil war, the direct effect o f techno-ecological inheritance should
operate via the level of complexity and functional interdependency it has encouraged in
society. Countries with poor techno-ecological inheritances should not move toward civil
war due to the low level of competitive interactions in society. Likewise, those with rich
techno-ecological inheritances should not move toward civil war, but, in this case, due to
the high level of interdependence that has been generated. Countries with moderate
inheritances, in contrast, should be the most likely to experience civil war due to the
prevalence of competitive interactions without the accompanying interdependence to
counteract its more deleterious effects.
In addition to these direct effects, techno-ecological inheritance has been linked to
several other factors that could have an impact on the likelihood of a country to enter into
civil war. To begin, research has shown that countries with higher levels of techno-
ecological inheritance tend to modernize more quickly than those with lower levels
(Crenshaw 2001 ). For this reason, the benefits that accrue to a country as a result of modernization should also apply to a country that has a rich techno-ecological inheritance.
Moreover, techno-ecological inheritance has been independently correlated (net of the level of modernization) with many of the factors that have been hypothesized to suppress the likelihood of entering into civil war. First, countries with rich techno- ecological inheritance have been shown to exhibit lower levels of inequality (Crenshaw 40 t992, Noteir and tenskr 1985). This^shotdddiscouragethe^experience of relative
deprivation within society, thereby suppressing grievances, and should make it more
difficult to commandeer the resources needed to start a war.
Second, with respect to population growth, recent research suggests that fertility
decline is quickened by a rich techno-ecological inheritance (Crenshaw et al. 2000).
Similarly, the rate of urbanization has also been shown to be lower in societies with a rich
techno-ecological inheritance (Crenshaw and Oakey 1998). Moreover, it has been argued
that societies with rich inheritances experience more rural-urban articulation (Boserup
1981, Crenshaw 2001, Glover and Simon 1975). Hence, the strains associated with
population growth (Goldstone 1997) and urbanization (Evans and Timberlake 1980,
Bradshaw 1987) should be less in a country with a rich inheritance than in countries with
a poor inheritance. Likewise, the resources created by rapid population growth (e.g., a
large proportion of young people) and rapid urbanization should be reduced due to a rich
techno-ecological inheritance.
Finally, it has been suggested that political development, and democracy in
particular, is encouraged by a rich techno-ecological inheritance prior to industrialization
(Crenshaw 1995,2001). If this is true as well, then the likelihood of civil war should be
less in techno-ecologically richer societies than in poorer ones.
The Service Sector and Violent Collective Action:
The theory about the relationship between service sector dominance and the development of motives, means, and opportunities to violent collective action is somewhat ambiguous. On the one hand, the service sector has often been belittled as the
41 *^‘uniorpartner*^ofindastriatrsra and assocratedwitli dead-end, informât activities beings performed by an uneducated, impoverished labor force. As such, a large service sector has been cited as a symptom of the economic stagnation that comes from an anemic or distorted industrial sector (Bradshaw 1987, Evans and Timberlake 1980). For this reason, it would be easy to think that a large service sector would encourage the conditions that are associated with a higher likelihood of collective civil violence.
Upon closer examination, however, a case can be made to suggest that this portrayal is overly simplistic. For, although it is true that the service sector typically provides lower wages for its employees than does the industrial sector, this does not necessarily mean that the underlying economic conditions are dysfunctional. Rather, it may be that the service sector promotes a unique “logic” - something comparable to logic associated with the industrial sector by modernization theorists (Crenshaw et al. 2000).
Although this logic may not be as advantageous for individuals and families as is the logic of mature industrialism, a dominant service sector may still have other implications that nuance its effect on the conditions that promote collective civil violence.
I suggest that a vital aspect of the service sector that determines the relationship between it and the conditions that lead to collective violence is the high amount of functional differentiation and interdependence it encourages. These conditions are created because service sector firms are more heavily dependent on labor, tend to be smaller than industrial firms, and are more dependent on the external business environment (Granovetter 1984). As a result, the service sector should, in fact, reduce the conditions that encourage violent collective action.
42 As^ evidence fortins, empiricafresearcir has^ beeff shown that the reiationship
between the service sector and other phenomena associated with violent collective action
is negative. For example, although initial studies suggested that growth in the service
sector is associated with increasing inequality (Evans and Timberlake 1980), more recent
analyses have shown that the size of the service sector may actually contribute to reduced
levels of inequality (Fiala 1983). As a result, a dominant service sector may reduce grievance, limit the amount of capital that can be commandeered for war efforts, and decrease the appearance of “lootable” resources in society. As a result, a large service sector would decrease the likelihood of violent collective action.
Likewise, other research has shown that service-sector dominance is associated with lower levels of population growth (Crenshaw et al. 2000). Thus, the strains associated with population growth and the resources it can make available should be lower under an economy dominated by the service sector.
Moreover, one of the largest employers in many developing economies where the service sector is dominant is the state bureaucracy (Skocpol 1979). In most circumstances, the state bureaucracy should have an interest in seeing the existing polity preserved so as to protect their jobs. As a result, a large service sector should weaken mass support for a civil war against the state. In these ways, therefore, the service sector should, in fact, decrease the likelihood that a country will enter into a civil war.
Ethnicity and Violent Collective Action:
Ethnicity is a somewhat different factor from the other two phenomena promoted by Human Ecology theory due to its intrinsically non-economic character. Because of
43 this; there arena cleariogicat reasons^that woui(MBad"oneto connect the ethnic composition of society by itself \.o the production of motive, means, or opportunities for violent collective action. Rather, from a Human Ecology perspective, ethnicity is more aptly considered a fundamental axes within societies (see Crenshaw 2001 ) that can either heighten or diminish the influence of other, more direct causes of the motives, means, and opportunity to violent collective action. Thus, to identify the role of ethnicity in promoting civil war, it is important to ask a modified version of the question posed by
Jenkins and Kposowa ( 1990), “What politicizes ethnic diversity so that it heightens the risk of civil war?”
The exploration of this question has consisted of two lines o f thought. First, attempts have been made to identify the specific structure of ethnic composition that is most relevant. In this respect, there have been two hypotheses.'® On the one hand is ethnic competition theory (Bates 1974, 1983, Jenkins and Kposowa 1990, Kposowa and
Jenkins 1993, Melson and Wolpe 1970, Olzak 1983, Olzak and Nagel 1986, Young
1976). In this line of thinking, the closer the size of the ethnic groups, the more the potential for competition between groups over economic, political, and social resources.
On the other hand is ethnic dominance theory (Jackman 1978, Jenkins and Kposowa
1990, Kposowa and Jenkins 1993). This perspective views the existence of one large group and one or more smaller groups as the prime circumstances for contests over resources.
'® Previous discussions of these theories have mixed the issues of the relative size of ethnic groups and the extent to which different groups dominate the economic, political, and social resources of society (e.g., Jenkins and Kposowa 1990). It is important, however, to keep the two issues distinct in order to specify
44 The second Kneof thought is overthe spectficwayrthar ethnicity interacts with
other social processes to create the motives, means, and opportunities to violent collective
action. For example, some early modernization theorists identified ethnicity with the
solidarities prevalent in traditional society and hypothesized that the dismptiveness of
modernization might initially encourage people to take refuge in ethnicity until the
process of assimilation could take hold (Deutch 1969, Smelser 1968). In this way, the
grievances associated with the process of modernization are intensified in the presence of
highly salient ethnic identities. Similarly, scholars working out of the Human Ecology
tradition have suggested that salient ethnic identities may be an enduring characteristic in
society which can significantly impair the processes of social differentiation and the
development of interdependency (Crenshaw 2001 ). As a result, ethnic composition may
cause competition to intensify, again exacerbating the experience of grievance.
Overall, the general impetus of human ecology theory for explaining the onset of civil war is to encourage more sophisticated thinking than is present in structural modernization theory. In terms of our modeling effort, there are reasons to believe that mid-levels of techno-ecological inheritance and politicized ethnic divisions will encourage civil war. The role of the service sector, however, should be to decrease the likelihood of the onset of civil war.
clearly the way ethnicity interacts with the phenomena that directly produce the motives, means, and opportunities to civil war. 45 Politieat Economy^ Theory and Civil- War
The other major critique of modernization theory has come from scholars using political economy theory. The central criticism of this line of thought is that modernization theory fails to take into consideration the international context of the developmental process. As such, it promotes an incorrect understanding of the macro social environment in which civil wars occur.
For these scholars, the most influential aspect of the macro-social environment is the international hierarchy of asymmetrical exchange relationships on which the current world order is based. Under this system, the continued development of the dominant
“core” is flnanced through the appropriation of surplus from the dependent “periphery”
(Wallerstein 1979). This system has its roots in the legacy of colonialism, but is maintained through the unequal exchange of primary products for processed goods on the world market and/or via direct private foreign investment (Bomschier and Chase-Dunn
1985, Galtung 1971). Although peripheralized countries may experience temporary advances as a result of this relationship (i.e., dependent development), the long run consequences are always detrimental (Evans 1979). In the process, the entire macro social environment of less developed countries is conditioned by the structural disarticulation imposed by the external dictates of the world market (Amin 1976,
Wallerstein 1979).
According to scholars working from the Political Economy perspective, the connection between dependency/peripheralization and the likelihood of violent collective action should be positive and strong. To begin, the unequal exchange relationship between the core and periphery is thought to extract much o f a dependent society’s 46 surplus; Gonsequentiy, themaforityofsocietyis-left^atjrsubsistencc-levetofearmngs;
In this way, low standards o f living are associated with dependency (London and
Williams 1988, Stokes and Anderson 1990) and thus should increase absolute deprivation.
In a similar manner, dependency is attributed with increasing inequality in society
(Bomschier and Chase Dunn 1985, Evans and Timberlake 1980). For there are small parts of a dependent or peripheralized society that will be able to rise above the subsistence level of productivity. These few are able to do so as a result of their close involvement with the international interests, e.g., those employed in the firms in which foreign capital has been invested or in companies involved in the exportation of primary goods. Hence, dependency leads to an unequal distribution of resources within society and, subsequently, should increase relative deprivation. In addition, the unequal distribution o f resources should increase the amount of resources that could potentially be commandeered for use in violent collective action, and should increase the appearance of
“loot.” Each o f these factors, in turn, should increase the propensity of the society to experience violent collective action.
Finally, dependence has been connected with over-urbanization and the expansion of the informal labor market (Evans and Timberlake 1980, Bradshaw 1987). Dependency theorists believe that a dependent relationship with the core will lead to massive rural-to- urban migration. Such migration results firom the subsistence conditions associated with most agricultural production in the country side and the false appearance of opportunity in the city due to the small sector closely tied to international interests. These conditions lead a society's urban centers to expand rapidly, swelling them well beyond the capacity 47 of the city ta provide infrastructure, honsing^ aiKfsociat services. In addition, people are pushed into dead-end, iow-wage employment, mostly in informal service sector. In this way, the urbanization associated with dependency is linked to higher levels of grievance in the population and, consequently, a higher inclination to move toward civil war.
There is reason to believe, however, that the reiationship between dependency and violent collective action should be more ambiguous than at first glance. The main reason for this is that the state should be stronger with increasing dependency. Under conditions of dependency, it is argued that the state becomes a tool of the part of society most closely allied with international interests. In order to guarantee the benefits of this relationship, the state becomes more willing to use repression in order to guarantee the status quo (Timberlake and Williams 1984, O’Dormell 1973). Moreover, the international interests themselves have an interest in maintaining the status quo, and, therefore, provide support to the state to maintain stability (Moaddel 1994).
Consequently, higher levels of dependency should raise the expected costs associated with a civil war and should lower the prospects for victory, thus discouraging potential insurgents from moving in this direction. For these reasons, the relationship between dependency and violent collective action may be indistinct.
48 CHAPTERS
DATA AND METHODS
T h e Sc o p e o f t h e S tu dy
This study spans the time period from 1965 to 1992. These dates were set taking both pragmatic and theoretical considerations into account. Pragmatically, the decision to focus on wars beginning after January 1, 1965 was made in an attempt to include in the analysis as many years of information for as many independent countries as possible.
The year 1965 is the earliest date where data for the independent variables are available in sufficient quantities to allow reasonable testing of the proposed hypotheses. Even so, numerous African countries achieved their independence after 1965, and data availability is often circumscribed before the year 1970, so the analysis was also conducted using both dates as beginning points.
The ending date for the analyses, December 31, 1992, is the last date for which data were collected by the authors of the dataset (Singer and Small 1994). I stayed with these dates to maintain the integrity of the original dataset and to encourage comparability o f results with other analyses using these data. In addition, 1992 was the last year for which data for some of the crucial independent variables in the analysis were available (especially the variables from Summers and Heston 1991).
49 Theoretically, this study seekato-produee-resulta ttett are-poHcy-relevant and that
inform attempts to predict future outbreaks of civil war. The decision to start the analysis
after the end World War II was therefore made in an attempt to assure that the results
would reflect current and future conditions as closely as possible. In terms of the ending
date of the analysis, it has been argued that the end of the cold war created a period of
instability that was unique, having “lifted the veil” off previously suppressed conflicts
(Collier and Hoeffler 2001, Kaplan 2000). Thus, the decision to exclude post cold-war
wars from the analysis allows us to examine this unique period in history. For these
reasons, in the analyses below I consider civil wars that occurred between 1965 and 1992
(see Appendix A for a list of civil wars during this period).
T he M o d e l in g F r a m e w o r k
Recently, the issue of which statistical technique is most appropriate for the study
of dichotomous outcomes such as the onset of civil war has received considerable
scholarly attention (Beck et al. 1998, King and Zeng 2000, Beck 2001, Beck et al. 2001).
Several approaches have been used to tackle this type of problem.
One approach, used mainly in the Political Science literature, has been to apply
logit or probit models to- study the “incidence” or “prevalence” of conflict as a dependent variable (e.g.. Collier and Hoeffler 1998). In these studies, the unit of analysis is the country-year, and every country-year for which there are data on both the dependent and independent variables is included in the analysis. This technique has been soundly
50 entiesed, however, for faiUng to^eootrot for autoeoFreiatioi^betweeo subsequent country-
years of conflict, and so has been largely discredited as a valid approach (Beck et al.
1998, Beck 2001).
As an alternative, scholars have used the same statistical techniques (logit and
probit analysis) while censoring (excluding from the analysis) subsequent country-years
of the same conflict from the analysis (e.g.. Collier and Hoeffler 2001, Elbadawi 2000,
Gissinger and Gleditsch 1999, Sambanis 2001). This approach solves the problem of
autocorrelation between multiple country-years of war. However, it has been criticized
for having the opposite problem - autocorrelation between multiple country-years of
peace. In addition, it has been criticized for assuming a constant baseline probability of
entering into civil war - i.e., that there are no system-level effects which change the
overall likelihood of entering into civil war in a given year (Hegre et al. 1999). Finally, it
runs the risk of obtaining biased estimates due to the problems associated with estimating
rare events using logit or probit models (King and Zeng 2000). Nevertheless, since it
avoids the gross errors of the previous technique, it has been more readily accepted in the
literature.
In an attempt to further deal with the types of problems inherent in the above techniques, two other approaches have been used in the analysis dichotomous outcomes.
The first continues with either logit or probit models but uses a case-control sampling technique to determine the country-years that will be included in the analysis (e.g., Esty et al. 1998). This technique has investigators select several “control” country-years for every “problem” country-year of conflict. It has the advantage of avoiding the problems of autocorrelation (as long as the selection process is rigorously controlled) and the bias 51 associated w itt the estimatiott of rare-event^. If can-be criticized, however, for throwing
away potentially useful information as a result of the selection process. In addition, it
still assumes a constant baseline hazard.
The final approach to these types of problems, and the one used in this study, is
the application of an event history technique. The use of this type of technique is
equivalent to taking a snapshot of all the cases that are at risk of experiencing a transition
(i.e., those not already experiencing the event in question) any time that a unit in the
analysis experiences the transition of interest. The compilation of these snapshots is then
used to estimate the hazard rate of making a transition. Additionally, this hazard rate is
decomposed into two parts: 1 ) a parametric function of the risk factors thought to be
associated with a transition, and 2) a non-parametric function representing the effect of
time itself (i.e., the baseline hazard). In this way, event history techniques avoid the
problems associated with auto-correlation between country-years of the same conflict, use all of the available information in testing the model, and control for the possible effect of calendar time on the likelihood of experiencing a transition." When the analysis reduces to a country-year format, however, they still can be criticized for not taking into account autocorrelation between subsequent country-years of peace.
Mathematically, the hazard rate that is estimated in event history techniques is expressed as follows:
y \ A(f) = Ao(f)exp
" Despite an extensive search in the literature, at this point it is not clear to the author whether the issue of biased estimates due to the analysis of rare events is an issue that needs be considered when using this method. 52 vHssKrhoft/ is the baseline bazard-rate; X / is erisk factor; is the corresponding
regression coefficient, p is the number of risk factors, and t is calendar time. The
baseline hazard can be handled in one of two ways. First, a functional form (e.g.,
WeibuU) can be used to specify the shape o f the baseline. In most cases in the social
sciences, however, there is not enough past research on the phenomenon in question to
give an adequate indication o f what shape it might take, nor sound theoretical bases with
which to hypothesize about the shape of the baseline hazard.
Fortunately, the second method for dealing with the baseline hazard does not
require the analyst to specify its form. In this analysis, the hazard rates of sub-samples as
determined by the values of the independent variables are expressed as a proportion of
one another. In this way, the underlying hazard function is eliminated and the remaining
quantity is used to estimate the effect of the independent variables. In considering the results, therefore, one is examining whether the hazard rate functions are nested below one another.
Hence, in the models below, the coefficient pertaining to a specific independent variable should be interpreted relative to a reference category in the case of a dummy variable or with reference to a case with a one-unit difference when the independent variable is continuous. For our purposes, the important thing to notice is the sign and the significance of the coefficient. A positive and significant coefficient indicates that countries with higher levels o f the independent variable (or those in the comparison group) are moving faster toward civil war. In addition, the risk ratio can be interpreted.
It is calculated by taking the exponent of the regression coefficient. It indicates the approximate proportional increase (or decrease) in the rate of transition to civil war due 53 to-a^one^umt increase-ht the^exponentofthe^independent variable (or as-compared whir
the reference group). Values above I indicate increases in the hazard rate, while values
below 1 indicate decreases.
It is important to note that this method is based upon the assumption that the
underlying hazards of separate countries are proportional no matter how they might be
stratified. In addition, it is assumed that each unit's history during the time period of the
investigation is independent of the histories of the other units under investigation.'■
Further, it is assumed that each event occurs at a distinct time (i.e., no ties). Finally, the
assumption of non-collinearity among variables was tested and confirmed using beta
correlations.
T he D ependent V ariable
The dependent variable under consideration is the instantaneous hazard rate of entering into civil war. In this study, I follow Singer and Small (1993) in defining civil war as open, armed conflict perpetrated by citizens of a country that has the goal of overthrowing or transforming the structure of that state and in which: (a) military action was involved (centrally organized fighters and fighting), (b) the national government at the time of the conflict was actively involved, (c> effective resistance (as measured by the ratio of fatalities of the weaker to the stronger forces) occurred on both sides and (d) at least 1,000 battle deaths resulted during the war. Data on civil war are drawn fiom the
In Act. this assumption could be converted into a hypothesis and tested. In particular, the issue of the diffusion o f civil war across national borders is a potentially relevant phenomenon. This type of relationship falls into the category of external influences to civil war. which is treated only briefly in this study (under the heading of Political Economy). See the conclusions (below) for a discussion o f the need for additional research in this area. 54 correiateffof warprqject(Singerand‘Sraaü"t995); Although some o f the civiT wars in
this dataset include became internationalized, in this analysis they are not counted as civil
wars for the intervening countries. The actual civil wars that occurred during the period
under consideration are listed in Appendix A.
The criterion used to determine what constitutes a transition from peace to civil
war was based on the start of the first major battle of the war. In some cases, data on the
specific day that the first major battle broke out was unavailable. However, since the
dependent variables are available on a yearly basis, error in measuring the exact timing of
the onset of civil war should not significantly affect the results. With these caveats, no
data on the dependent variable in this analysis were missing.
Two other aspects of the analysis in this study are important to note. First, due to
the strictures of the SAS procedure used to perform this analysis (Proc PHREG), only
countries that were eligible to experience a civil war during the entire time period under
consideration were able to be included in the analyses. Thus, a number of cases, most
notably from Sub-Saharan Africa and from the breakup of the Soviet Union, were absent
because they had not yet reached independence at the beginning of the spell. In addition, a few other countries were excluded because they were already in the midst of a civil war
(i.e., they were left censored)." This limitation in the SAS procedure was the primary impetus behind the decision to complete two sets of analyses - one starting in 1965 and one starting in 1970. The final risk sets for each respective analysis are shown in
Appendices B and C.
55 Second; again dnetn the confines of thrSA S procednre described above, each
country was only allowed to enter once into the analysis. This restriction only affected
those countries that experienced the onset of more than one civil war during the time
period under investigation. The result was that wars subsequent to the first onset in a
coimtry were excluded from the analysis. However, although this limitation may have
affected the results, its impact should have been limited to that of weakening the
relationships that were uncovered since the conditions in a given country do not change
very rapidly. Hence, this limitation may actually have provided a more robust check on
the strength of the significant relationships.
D escriptive A nalysis
The figure in Appendix D displays plots of the hazard function of the transition rate from "peace" to civil war from 1965-1992 and 1970-1992, respectively. The point estimates of the hazard give a graphic representation of the rate of the onset of civil war over time. An upward or downward slope of the curve indicates that the rate of onset of civil war is rising or declining in that time period.
The plots clearly shows that the transition rate from "peace" to civil war fluctuates up and down over this time period, mostly due to the relatively small number of onsets.
Overall, the hazard rates from the two different starting years parallel one another fairly well despite the addition of cases in the 1970 analysis.
‘^At the beginning of 1965, civil wars were going on in Laos, Sudan, Vietnam, the Yemen Arab Republic, and Zaire. In 1970, civil wars were going on in Burma, Guatemala. Laos, Nigeria, and the Sudan. 56 Atf of the perspectives- discassed abovemay help-ta expiant thechangmg'ratc'of
the onset of civil war over time. The central concern of the analyses below is to connect
the changing rate o f onset of civil war to specific factors indicated by these theories.
In d epend en t V ariables a n d H ypotheses
The independent variables used in the analysis were chosen in order to represent as many of the arguments from the theoretical approaches to violent collective action as possible (see chapter 2). Each variable was constructed from publicly available data sets.
In all, 202 specifications were tested (see Appendix E for a listing of the variables and the sources from which they were drawn). The resulting sample sizes were typical of cross national research - 116-126 cases depending on data availability for the independent variables. See appendices B and C, respectively, for the risk sets for the 1965 and 1970 analyses.
Event history analysis techniques allow variables to be entered into the model either statically, i.e., as a single value for the entire time period, or dynamically, i.e., with values that change over the time period. In this analysis, variables representing historical characteristics or relatively permanent qualities of societies (e.g., techno-ecological inheritance, ethnic homogeneity and world system position^ were modeled statically.
Alternatively, variables that could be expected to change over the time period of the analysis (e.g., infant mortality, economic growth, or level of democracy) were modeled dynamically where possible. In addition, where theoretically appropriate, dynamic variables were also measured in terms o f the annual average percentage change in the indicator so as to capture the effect not only of differing levels of the indicator but also of
57 the eceuneace o f change itself. Finally, except wherenoted, alt dynamic variables were entered into the models with a two-year lag in accordance with other research (e.g., Esty et al. 1998).
The specific measures of the independent variables and the hypotheses about how they should be related to the onset of civil war are detailed below. One peculiarity about these hypotheses is that they are phrased in terms of “quickness of movement toward civil war.” This is done to be faithful to the event history method that is employed in the analysis. However, this method produces very similar results to time-series logistic regression when the underlying assumptions of both methods are not violated.'■* For this reason, it may be helpful for pragmatic purposes to think about these hypotheses in terms of the probability of entering into a civil war. However, the technically correct form of the hypotheses is maintained for the sake of precision.
The independent variables used in the analysis are organized into three groups: lagged dependent variables, variables representing the macro-social conditions of society, and variables representing conditions that may intervene between the macro-social conditions and civil war.
ITie Lapsed Dependent Variable: Measures and Hypotheses
In this study, the level of previous experience that a country has had with civil war is measured with three indicators: the number of previous civil wars experienced
(since 1800), the time since a country's last civil war, and a dummy variable indicating whether a country had any experienced with civil war (since 1800). If a country had no
58 experience-wittcivit war, thevariablemeasnring-time-since-the last war was set ta the time since independence. Data on involvement in past civil wars is taken from the same source as the dependent variable
Theoretically, the past history of civil war in a society should indicate a propensity toward civil war either because o f general instability in the elites of a society or because a society has adopted war as one of its strategies for dealing with social conflict. These data are therefore used to test the following hypothesis:
HI : All else constant, societies with more experience/more recent experience with civil war will move more quickly toward civil war than those with less experience/less recent experience.
The Macro-social Environment: Measures and Hypotheses
Structural Modernization
In this study, the extent to which a society has undergone structural modernization is operationalized with multiple indicators. Two general indicators of the level of development are used. The first is real gross domestic product (RGDP) per capita. This variable is standardized in 1985 international dollars, adjusted for the actual buying power of national currencies, and excludes factor income from abroad. The second is en er^ consumption per capita. Conforming toother studies, both indicators were logged to correct for skewness.
In addition to these variables, four indicators directly related to the size of the industrial sector are tested. They include:
" In 6ct, preliminary analyses showed this to be the case (results not shown). 59 ty The percent ofthetotattaborforc& iff indastry; 2) The percent of GDP due to the industrial sector; 3 ) Industrial sector dominance (# I + #2); 4) Real industrial GDP/capita (logged);
Data for all of these variables were available every five years between I960 and 1990.
Hence, the variables were tested both statically (assuming development to be a relatively
enduring characteristic) and dynamically (to test whether the different levels of
development over time had a different effect take advantage of the time series).
Additionally, in order to test for the curvilinear relationships, the square of each of these
variables was calculated and entered into the model.
In accordance with the theoretical treatment given in the previous chapter, these
variables were used to test the following hypothesis:
H2: All else constant, the societies at mid-levels of development should move toward civil war more quickly than societies at either high or low levels of development.
Techno-ecological Inheritance
There are two categories of variables that are used in the literature to represent the techno-ecological inheritance of a country. The first is a measure of the techno-economic inheritance of a society. Developed by Lenski and Nolan (1984), this variable is intended to measure the technological organization of the economy at the time of its exposure to industriaiism. It is used to construct three dummy variables, one indicating a horticultural inheritance, a second indicating a new agrarian inheritance, and a final one indicating an old agrarian inheritance. These indicators represent a historical condition of the country and so are modeled statically.
60 Gountrier witlt a horticultnrahinheritance arethoughtto have devetopedonty a
rudimentary level of social organization, while countries with an old-agrarian inheritance
will have developed a quite diverse and interdependent social organization. In
accordance with the theoretical discussion above, therefore, countries with either of these
histories should demonstrate a low propensity toward civil war. Those with a new-
agrarian history, however, will have inherited the middle level of pre-industrial social
organization that has been hypothesized to encourage violent collective action. Thus, these variables are used to test the following hypothesis:
H3: All else constant, societies with an new-agrarian inheritance will move toward civil war more quickly than those with either a horticultural or an old-agrarian inheritance.
The second set of measures of a society's techno-ecological inheritance is constructed using measures o f population size and density. Given the virtual impossibility o f measuring centuries of institutional development, several indicators serve as proxies for demographic inheritance, including:
1) Population density (population/arable land); 2) Rural population density (rural labor force/arable land); 3) Total population; 4) Rural population density * total population.
These variables are intended to capture pre-industrial social development, and are used to test the following hypothesis:
H4: All else constant, societies with a history of moderate population density will move toward civil war more quickly than those with a history of lower or higher population density.
The final indicator of the inheritance of societies refers to their colonial heritage.
This indicator is incorporated into the models because some have argued that the process
61 of decolonizatioir as condncted by thcBritislr was more orderly thair that conducted b y
other colonizers and that it helped to instill political democracy (Emerson 1960, Smith
1978). Alternatively, French colonial heritage may also have discouraged civil war since
the French tended to stay more involved in their former colonies after independence than
did other former colonial powers. Both of these indicators are specified with a dummy
variable and used to test the following hypotheses:
H5 : All else constant, societies with a history of British colonialism will move toward civil war less quickly than those without such a history.
H6: All else constant, societies with a history of French colonialism will move toward civil war less quickly than those without such a history.
Service-sector Strength
Comparable to the measures of the strength of the industrial sector (see section on
structural modernization above), four indicators were used to represent the relative
influence of the service sector. Specifically, the following measures were created:
1 ) Percent of the total labor force in the service sector; 2) Percent of GDP due to the service sector; 3) Service Sector dominance (#1 + #2); 4) Real GDP/capita firom the service sector (logged).
Data for each of these were available every five years between 1960 and 1990, so the indicators were tested dynamically.
The theoretical literature on the service sector suggests two conflicting hypotheses regarding the effect of the service sector on the onset of civil war. On the one hand, the view that the service sector consists of dead-end, informal activities being performed by an uneducated, impoverished labor force will tend to see a strong service sector encouraging civil war. Conversely, those who view the service sector as a key function 62 that encourages diverstficatiou amhinterdependency with toofr for the opposite effect.
Hence, testing the following two hypotheses will provide us with a critical test of these two theories:
H7: All else constant, societies that are under the sway of a stronger service sector will move toward civil war less quickly than those under the sway of a less dominant service sector.
H8: All else constant, societies that are under the sway of a stronger service sector will move toward civil war more quickly than those under the sway of a less dominant service sector.
Ethnicity
In this study, data on ethnicity was taken from four sources.By far the most extensive source of data came from the World Christian Encyclopedia (Barret 1982).
This source provides data on the ethnic makeup of all countries of the world circa 1970.
From this data, various specifications of the ethic composition of society were created, including 1) the largest group as a percent of the total population, 2) the ratio of the second largest group to the largest group, and 3) a dummy variable fixing a curvilinear specification of #2 coded as 1 when the size of the second largest group is between 10 and 30 percent of the size of the largest group and 0 otherwise.
The three other sources of data provided only a single measure of the ethnic composition of society. From Vanhannen (1991) came a measure of ethnic homogeneity.
Some analysts (e.g., the State Failure Taskforce) have used data that comes horn the same source as the civil war data used in this analysis. This data is called the Correlates of War Cultural Dataset and contains data on the size of the seven largest ethnic, linguistic, and religious groups of all countries. However, the authors of this data have warned against its use (personal communiqué) because of their own doubts about the reliability of the data. In an unpublished study, the International Programs Center of the U.S. Census Bureau conducted a review of this dataset and confirmed the authors' contention. For this reason, the COW cultural dataset was not used in this investigation. 63 or the largest ethnic gronp^g percentage of tfaepopntatiott. Second, TaytorandHudson
( 1972) published a measure of ethno-iinguistic fractionalization, specified as the
probability that two citizens drawn at random will be firom different ethno-linguistic
groups. Finally, in 1999, Vanhannen published a second measure on ethnicity, this time
labeled ethnic heterogeneity. This indicator is actually a composite of three measures:
the percent of the largest ethnic group based on racial differences, the percent of the
largest group based on linguistic, national, or tribal differences, and the percent of the
largest group based on stabilized old religious communities. These three measures are
summed to produce the composite ethnic heterogeneity measure.
There is no current agreement on the shape of the ethnic distributions that pose the most serious problems for society. Accordingly, 1 test several specifications that have been proposed in the literature (see theory section above for a more extended treatment):
H9: All else constant, the more dominant the largest ethnic group, the more quickly a society will move toward civil war.
H 10: All else constant, the closer in size the two largest ethnic groups, the faster that a society should move toward civil war.
H 11 : All else constant, societies in which the size of the second largest ethnic group approaches 20 percent of the size of the largest ethnic group will move toward civil war more readily than societies with other ethnic compositions.
A second type measure associated with the ethnic composition of society that was available for testing is whether a society has groups within it with a history of separatism or actual autonomy. Having such a history, I suggest, indicates the presence of social cleavages that are likely to have an ongoing affect on society. Specifically, they may directly contribute to the overall level of grievance, and they may act as kindling to intensify any other grievances that may exist. 64 The-data^oirseparatist/autonomotiyMstorywere-taketrfrom twasoarces. First,
Taylor and Jodice (1983) published two measures. The first is a five point scale of the
intensity of separatism in the country, ranging from 0 (no separatism) to 4 (intense
separatism). The second is an estimate o f the percent of the population that is associated
with the separatist/autonomous sentiments in a country. Both of these variables were available for the years 1960 and 1975. Although these variables will have experienced change over the time period of the analysis, they were entered as time-invariant variables for lack of an alternative. Hence, the results of the tests of these indicators should be considered exploratory.
Second, Gurr ( 1993) published indicators of both separatist potential and autonomous history for minority groups across the globe in the 1980s. His separatism index is a four-category variable that indicates whether a minority group exhibits separatist sentiments that are latent ( 1 ), historical (2), active (3), or other (4). His index reflecting lost autonomy was created by assigning a score reflecting the extent to which each minority group in his dataset had been autonomous (from 0 to 6) and then weighting this score by the length of time since their autonomy was lost. Data on both of these indicators was converted to country-level indicators by summing the values for each ethnic group within a given country. Additionally, a second score was derived for both measures by multiplying the resulting value by the sum of the population size of the minority groups expressed as a proportion of the total population.
These indicators of a history of separatism or autonomy were used to test the following hypothesis:
65 Ift2r Attelseconstant, societieywitir separatist movementg or g group that way once autonomous will move toward civil war more quickly than those without such groups in their population.
Political Economy
World systems/dependency theorists argue that the experience of dependence or peripheralization by a country will increase the propensity of that country to enter into civil war. In this study, variables representing five different specifications of the level of dependency/peripheralization are tested.
First, two dummy variables, one indicating semi-peripheral status and the other indicating peripheral status, represent world system position (core status being the reference category). World-system position is a relatively enduring quality and so is modeled statically. It is argued that those in the periphery or in the semi-periphery are more likely to experience tense class relations than are those in the core.
Foreign capital penetration, in turn, was constructed following conventional practice. One indicator was the book value of the total foreign-owned stock in the year in question divided by the total stock. A second was the book value o f the total foreign- owned stock divided by the total population. Finally, a third was the book value of the total foreign-owned stock divided by the square root of the value of total stock times the total population (Ballmer-Cao & Scheidegger, 1979). These variables were logged to correct for skewness. Values for these indicators were only available for a few years, so there was no choice but to model them statically. For this reason, the results of these tests should be taken as exploratory.
66 Thenext indicaror, export dependènce, is specîfîecTih several ways. Ffrst, totaT export dependence was measured as the value of total exports as a proportion of GDP.
Values for this indicator were available every five years fi-om 1960 to 1995, so this variable was modeled dynamically. In addition, an indicator intended to capture the effect of change in this variable was created by calculating the annualized yearly average increase in exports/GDP over the five years prior to the year in question.
The other two indicators were more circumscribed in their coverage. The first was agricultural exports as a proportion of total exports. The second was commodity concentration, or the value o f the top three exported commodities relative to total exports.
Since coverage on these variables across time is limited, their effect was modeled statically.
The next set of variables purports to measure the level of debt-dependence. There are four variables that are used to represent this construct: total debt as a proportion of
GDP, long-term debt as a proportion of GDP, short-term debt as a proportion of GDP, and debt service as a proportion of GDP. Values for each of these variables were available every five years from 1965 to 1995, so they were modeled dynamically. In addition, indicators intended to capture the effect of change in the level of debt- dependence were created by calculating the atmualized yearly average increase each indicator over the five years prior to the year in question. Again, a higher level o f debt- dependence is argued to be associated with a higher propensity toward civil war.
The final dependency measure is military dependency. It is specified both as the amount of military aid received firom all external sources relative to GDP and as the amount of aid from specific external sources (e.g., the former Soviet Union or the United 67 StafesXperGDP: These indicators were modefed statically and specIfiecT strictfy in terms
of their ievel due to data availabiiity. Miiitary aid is thought to suppress civil war since it
would strengthen the existing regime and, thus, lower the expectations of potential
insurgents for a positive outcome. One exception to this is aid received from the Soviet
Union, which was thought to more often to be a prop to regimes that were already in trouble.
All of the variables discussed above were used to test the following hypotheses:
HI 3: All else constant, societies in the periphery of the world system will move toward civil war more quickly than those in the semi-periphery or in the core.
HI 4: All else constant, societies in the semi-periphery of the world system will move toward civil war more quickly than those in the periphery or in the core.
HIS: All else constant, societies with higher levels of direct foreign investment will move toward civil war more quickly than those with lower levels
Hi 6: All else constant, societies with higher levels of export dependence will move toward civil war more quickly than those with lower levels.
HI 7: All else constant, societies with higher levels of debt dependence will move toward civil war more quickly than those with lower levels.
HI8: All else constant, societies that are less heavily dependent on other countries for miiitary aid, regardless of the source, will move more quickly toward civil war than those that are more dependent.
HI9: All else constant, societies that were more dependent on the former Soviet Union for military aid will move more quickly toward civil war than those who were less dependent.
Intervenins Variables: Measures and Hypotheses
Absolute DenrivatioiL Relative Deprivation, and Inequality
In this study, absolute deprivation within a society is measured with four variables: the infant mortality rate, life expectancy at birth, caloric intake per capita, and 68 protenr intake-per capita^(see-Appendix E): Datapoints- for these variables- were avariable every five years over the time period of the study. Values for the interim years were obtained through linear interpolation. Additionally, growth rates for these indicators were constructed to reflect the change in the previous 5 years. These four variables give an indication of the general standard of living within a populace and therefore of the general level of absolute deprivation being experienced.
Relative deprivation, in turn, is measured with a series o f variables taken &om
Taylor and Jodice (1983) and Gurr’s Minorities at Risk (MAR) dataset ( 1993). Taylor and Jodice developed indicators of general economic and political discrimination
(Appendix E). These measures were specified in terms of intensity of the discrimination, measured on a 5 point scale with 0 being no discrimination and 4 being intense discrimination, and an estimate of the percentage o f the population that is experiencing discrimination. Data points for these indicators were only available for the years 1960 and 1975. Although these variables will have experienced change over the time period of the analysis, they were entered as time-invariant variables for lack of an alternative. For this reason, the results of the tests of these indicators should be considered exploratory.
The indicators in the MAR dataset are slightly different in that they are intended to measure deprivation in minority groups that experience discrimination. Gurr and his colleagues formulated measures of economic and political discrimination, economic, political and social grievance, and demographic distress for the 1980s for individual minority groups across the globe (1993). Data on these indicators was converted to country-level indicators by summing the values for each ethnic group within a given country. Again, although these variables will have experienced change over the time 69 periodoftheanalysî^ there-wasno-akemative-exceptto-keep-theni constant. Fortins reason, the results of the tests of these indicators should be considered exploratory.
In the MAR data, the two discrimination indicators are ordinal measures with four categories:
1 ) Neglect o f the group with remedial policy, 2) Neglect of the group with no remedial policy, 3) Social ostracism with a neutral policy, and 4) Social exclusion with a repressive policy.
The three grievance indicators, in turn, are actually indices that were constructed by summing the values from four to six trichotomized variables. The underlying variables measure whether the following situations are highly salient (value= 1 ), of lesser salience
(=2), or an insignificant issue (=3)'®:
Economic Discrimination 1) Diffiise economic grievances 2) Seek greater public funds or services 3 ) Seek greater economic opportunities 4) Seek improved working conditions 5) Seek protection of land or resources 6) Other economic grievances
Political Discrimination 1 ) Differential access to power positions, national or regional 2) Differential access to civil service 3> Differential recruitment to- police or military service 4) Differential voting rights 5) Differential right to organized political activities 6) Differential legal protection
It should be noted that the scale of these variables would lead one to expect the opposite relationship to what one might expect A positive relationship would mean that grievances slow the approach of civil war. whereas a negative relationship would indicate that grievances quicken the approach. 70 I ) Seek freedom of religious practice 2) Seek recognition of language or culture 3 ) Seek protection from other groups 4) Other social or cultural grievances
Third, the demographic distress variable is an index constructed from 11 other
trichotomized variables. The underlying variables measure whether the following
situations are a minor condition (value=l ), a moderate condition (=2), or a serious
condition (=3):
1) High birth rate 2) Youthful population 3) Poor public health conditions 4) High rural-to-urban migration 5) High emigration abroad 6) High immigration from abroad 7) Competition for vacant land 8) Dispossession from land 9) Forced internal resettlement 10) Other demographic trait 1 II ) Other demographic trait 2
Finally, a second set of variables was created by multiplying each of the above variables by an estimate of the population size of the group. See Appendix E for the complete list of the variables fr-om this dataset.
The next set of variables consists of measures of inequality. For this, two different types of measures of income inequality and a measure of land inequality are used (see Appendix E). The indicators of income inequality are a Gini coefficient (4 formulations) and the percent o f income controlled by the top 20% of the population. For land inequality, a Gini coefficient weighted by the labor force in agriculture was used to capture concentration of agricultural resources. All measures were only available for the
71 one point in time, so, altbongb they wilt havechangect during^the time period, they were
entered into the modeis as static variables. For this reason, the results associated with
these variables should be considered exploratory.
Drawing on the theory regarding the supposed relationship between deprivation
and violent collective action, four different deprivation hypotheses were tested, each
reflecting a different functional form of the potential relationship;
H20: All else constant, the greater the deprivation in society, the more quickly it will move toward civil war (linear).
H2I: All else constant, societies with a very high level of deprivation will move quickly toward civil war than those without very high levels (exponential curve).
H22: All else constant, societies with moderate levels of inequality/deprivation will move more quickly toward civil war than those with low or high levels (inverted U curve).
H23: All else constant, societies with either low or high levels of inequality/deprivation will move more quickly toward civil war than those with moderate levels (U curve).
Greed
The economists working on the topic of civil war have highlighted the importance the quest for “loot” in encouraging civil war. More specifically, they view the earnings that come from the export of primary products as particularly enticing (Collier 1998).
In this study, four measures were constructed to operationalize this construct.
Data on the export of agricultural products, fuels, and minerals and ores, as well as a composite of all three, was taken from the World Bank's (2001) World Development
Indicators and normed by the GDP of the country. This data was highly circumscribed
72 before t980; and-the-coverageacrosecorartriey way scant Stiif, yearly data was
available, so the variables were entered dynamically into the models. These results were
then used to evaluate the following hypothesis:
H24: All else constant, the more “lootable” resources a society appears to have, the more quickly it should move toward civil war.
Demographic Indicators
A rich source of potential independent variables for this study came from demographic data sources, especially the United Nations World Population Prospects:
1998. In particular, this source was used to construct variables on population growth and the age structure of population that have been associated with violent collective action.
With respect to population growth, Goldstone ( 1997) posited that it could lead to violent collective action by placing greater demands on the resources of society, thus increasing the level o f social strain. In this study, I borrow and expand on this idea.
First, I test the effect of simple population growth with a measure of the annualized yearly average increase in population over the five years prior to the year in question.
Data for this variable was available every five years from 1960-1995, and indicators were constructed for the interim years through linear interpolation. This variable was then tested dynamically in the models.
I also tested several other indicators of population growth. Recently, research has shown that the age structure o f the population is important for determining the economic and political relevance of population growth (e.g., Crenshaw 2001, Crenshaw et al.
1997). Following this line of thinking, I suggest that growth in the adult population (ages
15-64) should have a far greater impact on the quickness with which countries move 73 towardcm h warthait growth iirthgcfaitd-popniatioir^agerO^t#); The reason forthijis the greater relevance of the social issues that would lead to violent collective action for this segment of the population.*’ These measures were constructed and tested in the same way as those for overall population growth.
The above measures population growth were used to test the following hypotheses:
H25: All else constant, societies with higher rates of population growth will move toward civil war more quickly than those with lower rates of growth.
H26: All else constant, societies with a more rapidly growing population under the age of 15 will move toward civil war less quickly than those with a population under age 15 that is growing more slowly.
H27: All else constant, societies with a more rapidly growing population between the ages of 15 and 64 will move toward civil war more quickly than those with a population ages 15 to 64 that is growing more slowly.
With respect to the age structure of the population, it has been suggested that a high proportion of youth within a society may have a destabilizing effect (Goldstone
1997). This is because young people tend to be less well integrated into the institutions of society. As a result, they may be more able and willing to make themselves available to serve as participants of violent collective action. Conversely, other analysts have argued that a large- proportion of children in a society may have the opposite effect.
Children require substantial investments in subsistence activity. A large proportion of children in society may therefore restrict the rest of the population's disposition to participate in movements bent on violent collective social action.
*’ Here again, this measure may also have relevance for the means to violent collective action by affecting the numbers of people available for participation in armed conflict 74 hr this study, t fbttow previous work by spectfyrngtheprevatence of youth iira
population in ways: the number of people ages 15 and 29 as a proportion of the total
population, and the nmnber of people 15 to 29 per person between the ages of 30 and 64.
In addition, since males are generally more likely to serve in wars than women, 1 also
developed and tested similar indicators for the male population. These variables were all
tested dynamically in the models.
Using these indicators as a basis, I also developed measures representing change
in the proportion of youth in society in the form of annualized yearly average increases in
each ratio over five years periods. Finally, I tested an indicator of the annualized yearly
average change in absolute size of the youth population over five year periods. Each of
these indicators was available every five years from 1960-1995 and so was tested dynamically in the models.
To test the effect of the child population on a society, I used two different
indicators: the child dependency ratio (the ratio of the population 0-14 to the population
15-64) and children (0-14) as a proportion of the total population. I also tested the effect of change in these indicators by constmcting variables of their annual average change over five year periods. Finally, I modeled the annualized yearly average change in the absolute size of the child population over five year periods. Each of these indicators was entered dynamically into the models.
With these indicators, I tested the following hypotheses:
H28; M l else constant, societies in which the proportion of youth or young men is larger will move toward civil war more quickly than those in which it is smaller.
75 H29r Att else constant, socictieynrwinclrthe^yotmgpopniation^ortheyotmgmale population is growing faster will move toward civil war more quickly than those in which it is growing more slowly.
H30: All else constant, societies with a large proportion of the population under the age of 15 will move toward civil war less quickly than those with a smaller proportion.
H31: All else constant, societies in which the child population is growing faster will move toward civil war less quickly than those in which the child population is growing more slowly.
Social Mobilization/Urbanization
The relationship between the mobilization of the population and civil war has
been depicted as follows. Once a population begins to be mobilized either socially or
geographically, expectations rise - i.e., people begin to expect action in response to their
desires, and they will be more willing to make themselves available as a resource in the pursuit of their desires. In addition, mobilization tends to increase interactions within society, thereby increasing the likelihood that like-minded people who are similarly disposed to engage in violent collective action will come into contact with each other.
In this study, mobilization is measured with three variables: the proportion of the secondary-school-age population that is enrolled in secondary school, the amount of education in years that the average person in a population has, and the percent of the population living in urban areas. In addition, average annual percentage change variables were constructed and tested for each of these indicators. Finally, the average annual percentage change in the absolute size of the urban population was tested. All of these variables were available in five-year panels between 1965 and 1995 and, consequently, are modeled as dynamic variables
76 These*measures are* usedto^tesfthe^fbUowing-hypotheses:
H32: All else constant, societies that are more highly mobilized will move toward civil war faster than those that are less highly mobilized.
H33: All else constant, societies that are mobilizing more quickly will move toward civil war faster than those that are mobilizing more slowly.
Government Characteristics
Since the state is, by definition, one of the two combatants in a civil war, it makes
sense that the characteristics of the state would influence the likelihood that a civil war
will begin. There are two notions about the basic character o f the state that drive the
thinking about its contribution to violent collective action. The first is that the state is an
arena where the members of society come together to compete for public goods (e.g.,
Huntington 1968). The second is that the state is a relatively independent actor with the potential to use the power at its disposal to influence the rest of society (e.g., Skocpol
1979). The perspective that one takes with respect to the state has significant consequences for the ideas about how it is related to the onset of civil war.
Modernization theorists were the primary source of notions about the state as an arena for competition over public goods. From this perspective, a democratic system was the most conducive way that the state could be organized te channel a mobilizing population, thus avoiding the types o f grievances that could lead to violent domestic conflict (Huntington 1968). Autocracies, conversely, while severely limiting the participation of the populace in political processes, also served to suppress dissent
77 throngh tight controh overpotiticah activity: Partiahdemocracies or autocracies, however;
would be the most vulnerable due to their limited ability to either channel or suppress the
political aspirations in society.
In this study, democracy is operationalized with data from several sources. First,
the Polity HI dataset provides measures of democracy and autocracy (Jaggers and Gurr
1995). Both are 10-point composite indices meant to tap the basic institutional structure
of the state. Additionally, a third indicator was created by subtracting the autocracy score
from the democracy score, thereby creating an indicator of the level of democracy/
autocracy with values from -10 to +10. Finally, a fourth indicator from the Polity II
dataset that was meant to measure the coherence of political institutions was used. It is
constructed as a dummy variable indicating that a country scored 7 or more on either the
autocracy or democracy variable. All o f these data are available for the entire period of
the study and thus are modeled dynamically.
The other source of data on this construct is the Freedom House indicators of civil
liberties and political rights. Both of these indicators are 7 point scales that range from 1
(most free) to 7 (least free). Following past research, these variables were also combined
into a composite score of “repression” by adding them together (e.g., Muller 1985). Data on these variables is available for the period 1972-1992 and so they are tested dynamically.
These indicators o f the openness of the political system to the participation of the populous are used to test the following hypothesis;
H34: All else constant, societies that are neither highly democratic nor highly autocratic will move toward civil war more quickly than those with consolidated polities.
78 The other perspective on the state is that if fs less an arena for competftfon and
more an actor that, due to its monopoly on the legitimate means o f coercion, can act
relatively independently of the rest of society. As such, characteristics concerning its modus operandi should influence the propensity of the society to move towards civil war.
To begin, the scope of political authority may be important. This is the extent to which all levels of government combined - national, regional, and local - attempt to regulate and organize the economic and social life of the citizens and subjects of the state.
On the one hand, an interventionist state may be resented by the populace and thus provoke rebellion. On the other hand, the extent to which a state is involved in the lives of its citizens may indicate an ability to suppress potential rebellions. Hence, it may actually be that rebellion is most likely at moderate levels of state scope.
In this study, the construct reflecting the scope of political authority is measured with a nine category index from the Polity II dataset (Gurr 1990). This variable ranges from minimal involvement ( 1 ) to totalitarian (9). Values are available from before the beginning year of this analysis up through 1990. Hence, it is modeled as a dynamic variable.
A second characteristic of the way the state operates is the centralization of political authority. This is the extent to which authority is vested in the federal government as opposed to regional or local governments. The level of centralization of the state should be an indication of its capacity to mobilize resources and act in the face of a crisis. For this reason, centralization should decrease the likelihood of a civil war.
For our purposes, the measure of this construct is a three category ordinal variable ranging from centralized (3) to decentralized ( 1 ) and is again taken from Polity U. As 79 witlr the scope varia We, vahies areforthiff indrcatorareavailaWeftont before the
begiiming years of this analysis up through 1990, so it is modeled as a dynamic variable.
The final characteristic of the state that may influence the propensity of a society
for civil war is the willingness of the state to engage in repression. In this study,
variables reflecting this aspect of the state were constructed using measures of government sanctions per capita firom Taylor and Jodice ( 1983). Unfortunately, sanctions data were only available for the early portions of the period ( 1960, 1965, 1970, and 1975), so the indicators were modeled statically. For this reason, the results based on these indicators should be viewed as exploratory.
The indicators just described were used to test the following hypotheses:
H35: All else constant, societies with govenunents that have a broad scope of power will be less likely to move toward civil war than those with governments that have a more narrow scope of power.
H36: All else constant, societies with governments that are centrist in structure will move toward civil war less quickly than those with governments where power is more diversified.
H37: All else constant, societies should move toward civil war less quickly when they have a government that is actively imposing sanctions against its populous than when not.
One final phenomenon related to the state that may influence the movement toward civil war is the level of stability. Specifically, when the state undergoes a change, or when changes in the state are more firequent, the opportunities to confront successfully the state apparatus may be more attractive than when the state is stable. In this study, three measures are used to reflect the stability of the state, each taken from Gurr ( 1990) or Jaggers and Gurr (1995). The first two are variables available from Polity U (Gurr
1990): 1) the occurrence of an abrupt change to an entire polity, and 2) transitions in 80 executive- leadership-. The- third- variable-was-created-ftonrthr democracy and autocracy variables in Polity HI (Jaggers and Gurr 1995). In these data, there are times when values could not be assigned to a country due to the transitional nature of political arrangements.
1 use the flags placed in the data when these situations occurred to create a dummy variable of regime transition.
These variables on political instability are used to test the following hypothesis:
H38: All else constant, societies should move toward civil war less quickly when they have a government that stable than when it is not.
Other Forms of Collective Action
Other forms of collective action in a society may either encourage or hinder the prospects of civil war. First, depending on the level of conflict, a past history of collective action may indicate either a propensity toward civil war or a buffer against it.
On the one hand, involvement in international wars may indicate an overall instability o f the elites in society and thus a propensity toward civil war. International war may also act as a destabilizing force, increasing the odds that potential insurgents will seize the opportunity to make a grab for the state. On the other hand, the occurrence of prior conflict at less-intense levels may mean that a society has adopted strategies that allow it to diffuse situations of conflict without going to the extreme of a civil war.
In this study, past collective action is measured on a variety o f levels. On the broadest level, data on involvement in past international wars is taken 6om the same source as the dependent variable. Measures are constructed in terms o f whether a society
81 bas-beeit involved nt intemationai war (since^hSOO); thenumberof warrthe society has- experienced (since 1800), and the time since a society’s involvement in the most recent international war.
On a smaller scale, several indicators of protest by minority groups were taken from the Minorities at Risk dataset (Gurr 1993). These included the number of non violent protests, the number of violent protests, and the number of ethnic rebellions since
1945. These data were collected for many different minority groups across the globe, and were converted to country-level indicators by summing the values for each ethnic group within a given country. Only one value could be constructed for each country, so the variables were modeled statically.
The final measures of collective action other than civil war concern the prevalence of coup d’etats. Two measures of coups d’etats were constructed from Polity
II data (Gurr 1990). The first was a measure of the whether the country experienced a coup d’etats in the previous 2 years. The second was a sum o f the number of coup d’etats in the previous five years. As with the other low-intensity forms of collective action, the occurrence of coup d’etats should decrease the likelihood of civil war.
Each of these measures of prior collective action are used to test the following two hypotheses:
H39: All else constant, societies with more past experience international war will move more quickly toward civil war than those with less experience.
H40: All else constant, societies with more past experience of lower-intensity violent conflict will move less quickly toward civil war than those with less such experience.
82 MiKtarizaticm
The militarization of society may help predict the onset of to civil war by giving
an indication of the availability of the technical means needed to wage war (arms) as well
as of the expertise to engage in war. In addition, the professional military itself may give
rise to intact groups that splinter off to wage war. The level of militarization of a society
may provide insight into its propensity to civil war.
The militarization of society is measured with two indicators:
1 ) The number of military personnel per 1000 population; 2) The proportion of GDP spent on the military;
Both military personnel per capita and military expenditures per GDP were modeled dynamically and specified in terms of both the level and the annual average percentage change in the variable.
With these data, I test the following hypothesis:
H41: All else constant, societies that are more militarized in terms of soldiers or military expenditures will move more quickly toward civil war than those that are less militarized.
M o d e l in g St r a t e g y a n d A d d itio n a l H y po t h e se s
One of the principal challenges of this research project was managing the large number o f variables that were specified, keeping track o f the plethora of hypotheses that were generated by linking these variables to existing theory, and trying to assure that these hypotheses were tested rigorously. This challenge was complicated by several concerns not yet discussed.
The first is the possible interactions between variables. For example, the discussion of macro-level theory (above) indicated that a confluence of motives, means, 83 andopportunitieg way neededta bring jsociety-mtacivth wan hr terms of modeKng; this
could be specified as interactions between any number o f indicators of these three
constructs. Additionally, other theoretical perspectives promoted interactions between
things such as mobilization and characteristics of the state (Huntington 1968), or level of
development and urbanization (Bradshaw 1987). In fact, the number of theoretically
défendable interactions is large enough that it is not feasible to list them all in this work.
Another potential problem concerning the specification of the models pertains to
issue of causal order. Specifically, most present day analyses do much to assure that the
spurious effects in a given model are uncovered. They do this by making indicators with
significant effects compete with many other indicators and combinations of indicators for
explanatory license. A more difficult prospect, however, is to uncover the presence of
suppressor effects. To do this, all insignificant effects would need to be treated with the
same suspicion as the significant effects. Consequently, they would be tested in
combination with just as many indicators and combinations of indicators as are used to
test for spuriousness. To do this, however, would tax any research using traditional
modeling strategies.
In response to this challenge, I developed a set of SAS macros that provide a
partial solution to this problem. These macros perform what has come to be called “data-
mining.” Stated somewhat differently, they perform a routine similar to the stepwise procedure available in most statistical analysis packages. As such, they perform tests of every possible combination of the variables that are given to the program within a set of defined parameters. They do so, however, in a way that makes the results superior to those obtained through the stepwise procedure. 84 The macros approaebthe-modeling-task by reqmrmg^the-anaiysttaspecify'tho constructs to be tested and then the specific indicators of each construct. This helps to maintain contact with the theory and the hypotheses that are to be tested. In addition, the program does not test the same indicators of a given construct to be tested in a given model, thus preserving the integrity of the logic of causal order.
2) The macros do not delete all cases with missing data on any indicator at the beginning of the procedure, but instead deletes cases on a model-by-model basis.
3) The macros do not output only the “best” model, but rather provide access to every model tested and an interface that permits for sorting and examination of the models according to the analyst’s wishes.
4) The macros model curvilinear terms correctly, only testing a quadratic or interaction terms if the base term(s) are in the model.
The code that makes up this program consists of thousands of lines of commands and so cannot be reproduced in full in this volume. Nevertheless, an example of the core of the program is given in Appendix F. An examination of this code will show the logic by which the basic models for testing are constructed.
In practice, the promise of this program turned out to be brighter than its impact.
The basic problem was a lack of computing power. For example, a test of just the macro level variables in a five-variable model was allowed to run for over a week before it was stopped manually. Upon investigation, it still had not completed even 5 percent of the runs it had been programmed to explore even-though it had run hundreds of thousands of models. Thus, it was not feasible to use this program to its full extent for this work.
The program was not without benefit, however. For although it did not allow me to rule out the presence all suppressed relationships, it did permit me to systematize the
85 searctfor spurious effects and interactionsbetween variabtes. Hence, ahhough only a few of the actual models run are discussed in this work, these models were tested in a very rigorous, methodical way.
86 CHAPTER 4
ANALYSIS
The results o f the analysis of the onset o f civil war can be found in the appendices
at the end of this volume. Appendix E presents the bivariate relationships between the
onset of civil war and each of the 204 variables considered as predictors (obtained
through a regression of the dependent variable on each predictor individually).
Appendices G through L, in turn, contain the results of the multivariate analyses. This
latter presentation is divided into three parts. First, appendices G and H trace the
development of the core model of macro-level variables that were found to best predict
the onset o f civil war. Next, appendices 1 and J show selected results of the models that
contain other, theoretically meaningful macro-level variables. Finally, appendices K and
L contain selected results of models specifying potential intervening variables between the indicators of the core model and the onset of civil war.'* The discussion that follows
is organized to parallel the structure o f the appendix tables.
'* V/hen considering the results of the multivariate analyses, it should be noted that I have tried to represent all of the significant relationships that I found. Thus, if the effect of a specific variable of interest is not represented in the tables, it can be assumed that its effect was insignificant. 87 BfVARtATgREfcATIONSHtPS
Lasted Dependent Variables
The first hypothesis states that societies with more experience or more recent
experience with civil war will move toward another bout with civil war more quickly
than those with less experience or less recent experience. The bivariate findings for the
three variables used to test these relationships provide general support for this idea (see
Appendix E). Both the number of previous wars and the fact that a country had
experienced any prior civil wars were positively and significantly related to the onset of
civil war. Similarly, although statistically insignificant, the negative coefficient on the
years since the last civil war variable suggests that shorter periods since the last
occurrence of civil war cause countries to move toward other civil wars more quickly.
Macro-level Variables
Indicators of Structural Modernization
The hypothesis concerning the effect of structural modernization on civil war says that countries at mid-levels of industrial development should move toward civil war more quickly than countries at either high or low levels of development. Since this relationship is curvilinear, it is not possible te test it by looking at simple bivariate relationships.
Despite this fact, the relationship between the variables used to measure the level of development and the onset of civil war is consistently negative and almost always significant (Appendix E). Only the level of industrial dominance fixed in 1965 does not reach significance at the .10 level. This finding suggests the higher the level of industrial
88 development, themoreslowlyacotmtrywütmove^towantcmtwar The^qnestioirof
whether there is an upper limit to this relationship after which the effect levels out or
turns negative will have to be left until the multivariate analysis.
Indicators of Techno-ecological Inheritance
I proposed four hypotheses relating the techno-ecological inheritance of a country and the onset of civil war. The first says that countries with a new-agrarian techno- economic inheritance will move toward civil war more quickly than those with either horticultural or old-agrarian inheritances. Although the coefficient for the dummy variable representing new agrarian inheritance is in the appropriate direction (positive), it fails to reach statistical significance (Appendix E). Nor do the dummy variables representing either horticultural or old agrarian inheritances achieve significance. In terms of their signs, however, the results suggest that a horticultural inheritance would promote civil war, while an old-agrarian inheritance would discourage it.
The second hypothesis states that countries with a history of moderate population density will move toward civil war more quickly than those with a history of either low or high population density. As with structural modernization hypothesis, the curvilinear character of this relationship makes it impossible to test by looking at simple bivariate relationships. Consistent with this, there is no stability in the signs coefficients of the twelve density variables tested against the onset of civil war (Appendix E). In addition, only three of the variables reach statistical significance, and two o f these only at the .10 level. As with the modernization indicators, therefore, we will have to wait until the multivariate analysis to judge the merits of this hypothesis.
89 Thetirirctandfourth hypothra^esrconcenr the effect of British and French cofoniaT
history on the onset of civil war. Consistent with the hypothesis concerning French
colonial heritage, the coefficient associated with this variable is negative, although
insignificant. Contrary to what was expected, however, the coefficient associated with
the British colonial heritage variable is positive, although again insignificant (Appendix
E).
Indicators of Service Sector Strength
Two competing hypotheses concerning service sector strength and the onset of civil war were provided for analysis. The first was that a strong service sector would encourage civil war, and the second was that a strong service sector would discourage civil war. The results show clear support for the latter hypothesis (Appendix E). Each of the coefficients associated with the variables representing the strength o f the service sector is negative and significant, with the strongest predictors being percent GDP from services, service dominance modeled dynamically, and the log of service sector real
GDP/capita.
Indicators of Ethnicitv
Three hypotheses concerning the ethnic composition of society and one about separatist or autonomous history have been proposed. The first suggests that the more dominant the size of the largest ethnic group, the more quickly a country will move toward civil war. The evidence for this hypothesis is mixed. Although insignificant, the coefficients associated with the variables measuring the size of the largest ethnic group
90 (Barret t982)^anctethnic homogeneity(Vanhanneit t99‘t>arenegative. Conversely, the
sign of the coefficient associated with Vanhannen’s {1999) (misnamed) ethnic
heterogeneity measure is positive and very significant.
The other two hypotheses concerning the ethnic composition of society suggest
we look for either relative parity or an 80/20 split in the size of the two largest ethnic
groups. Although in the correct direction (positive), the coefficient associated with the
variable comparing second largest group to the first largest group is insignificant. Thus, I
find no support for the relative parity hypothesis. In contrast, the coefficient of the 20
percent composition variable is positive and significant, and so this hypothesis is
supported.
With respect to the hypothesis concerning separatism and lost autonomy, I find
consistent and strong support. The coefficients for each of the seven variables measuring
these constructs are all positive, and six of the seven are highly significant. Only the
transformed measure of Gurr’s ethnic separatist history multiplied by the population size
o f the minority groups at risk in the population fails to reach significance (p=. 187).
Political Economv Indicators
In all, seven hypotheses were formulated regardingthe relationship between variables proposed by Political Economy Theory and the onset of civil war. In general, these hypotheses posit that higher levels of dependency or peripheralization will be associated with faster movement toward civil war. The two exceptions to this are the hypotheses that semi-peripheral world system position wül encourage movement toward civil war and overall military dependence should depress the movement toward civil war.
91 The-results of the bivariate-analysts show very mixed support for these hypotheses. On the positive side, both peripheral and semi-peripheral world system position have regression coefficients in accordance with their hypotheses (positive), but only semi-peripheral position is significant. Similarly, core world system position is negatively related to the onset of civil war, but again this coefficient is insignificant. In like fashion, the relationship between the onset of civil war and the variables measuring the amount of agricultural exports and the extent to which exports are concentrated in a few commodities are positive and significant. These few variables give support to the hypotheses proposed from political economy theory.
On the other side of the coin, many of the relationships between the dependency variables and the onset o f civil war are in the opposite direction of the hypotheses. For example, all of the variables measuring foreign capital penetration have negative coefficients, although only foreign stock/capita in 1967 is significantly related to the onset of civil war. Similarly, the measures of the proportion of GDP due to all exports and of the openness of the economy to trade also received negative coefficients, but these times they were all significant. Likewise, each of the measures of debt dependence had negative regression coefficients, although none of them achieved statistical significance.
Finally, none of the measures of military dependence achieved statistical significance, although the coefficients associated with the variables measuring soviet, western, and
U.S. military aid were in the appropriate direction.
92 Intervenin^Variable»
Indicators of Absolute Deprivation. Relative Deprivation, and Inequality
Three of the four hypotheses regarding the effects of relative deprivation, absolute deprivation, and inequality on the onset of civil war called for testing of curvilinear relationships. Hence, the results of the bivariate tests can give only preliminary indications about the helpfulness of these constructs in predicting the onset of civil war.
First, tests of the indicators of absolute deprivation gave quite strong support for a positive relationship between poor standards of living and the onset of civil war. In line with this hypothesis, the coefficients of both infant mortality and growth in infant mortality were positive and significant. Similarly, the effects of life expectancy at birth, caloric intake per capita, and protein intake per capita were all negative and significant.
When specified as growth rates, however, the implications of the results using these last three variables were less clear. The relationship between both growth in life expectancy at birth and growth in caloric intake per capita and the onset of civil war is positive, although only the coefficient associated with life expectancy is significant. Conversely, growth in protein intake per capita is negatively related to civil war, although this variable is far from reaching statistical significance.
The indicators of relative deprivation provided about the same level of confirmation for a linear relationship with the onset o f civil war as did the absolute deprivation variables. All of the indicators o f economic and political discrimination published by Taylor and Jodice (1983) were shown to encourage civil war, although only the intensity of political discrimination in 1960 and the proportion of the population experiencing discrimination in 1975 reached significance at the .10 level. Similarly, all 93 ofthevariaWes^ott minority discriminatroii^ grievances and demographic distress as developed by Gurr (1993) showed a positive relationship with the onset of civil war.
Among these variables, the direct indicators of discrimination, grievance, and demographic distress among minority groups were the most significant, with those variables that were weighted by the population of the minority groups receiving less support.
Finally, the results of the bivariate tests gave only weak support for a linear relationship between inequality and the onset of civil war. The coefficients associated with all four Gini coefficients of inequality and the three variables measuring the share of income earned by the top 20 percent of the population were positive but not significant.
Similarly, the relationship between the Gini coefficient o f land inequality fi*om Taylor and Hudson ( 1972) was positive but insignificant.
Indicators of “Greed”
In response to the work of the economists on civil war, a hypothesis was formulated linking the appearance of “lootable” resources in society with movement toward a civil war. Four variables tapping the extent to which the export of primary products was important to the economy were constructed to test this hypothesis. The results of the bivariate tests of these variables provided no support for this hypothesis.
The coefficient of each variable was negative and non-significant. However, the severely limited number of cases available to test this relationship makes these findings highly suspect.
94 E)emograDhic Indicators^
Drawing on the wealth of available demographic data, seven hypotheses were
proposed regarding the effects of population growth and population structure on the
quickness of movement toward civil war. With respect to population growth, I suggested
that higher rates o f population growth would lead a society to enter civil war more
quickly than lower rates of growth. However, I also posited that growth in the potential
labor force (ages 15-64) would be a more powerful predictor of movement toward civil
war than simple population growth. Conversely, I argued that growth in the child
population (ages 0-14) might actually inhibit countries from moving toward civil war.
The results of the bivariate analysis provide general support for the first two of these hypotheses. Higher rates of overall population growth and growth the potential
labor force were positively and significantly associated with movement toward civil war, with the growth o f the potential labor force being slightly more significant (p=.002 vs. p=.004 for overall population growth). Additionally, the relationship between faster rates of growth in the child population and movement toward civil war was also positive and significant (p=.009). However, this was to be expected due to its high correlation with overall population growth. The true test of the rate of growth in the child population will be net of the rate of growth of the potential labor force.
With respect to the age-sex structure of populations and movement toward civil war, the first set o f hypotheses deal with the effect of a large number of youth (ages 15-
29) in a population. Specifically, a high proportion of youth, as well as high growth rates
95 in^the-populatioirat these agea, wae hypothesized-te encotn-age a society ta move toward
civil war. Moreover, high proportions and high growth rates among males of this age
was suggested to be especially disruptive.
The bivariate results show fairly strong support for the hypotheses associating a
high proportion of youth in society with quickness of movement toward civil war. The
regression coefficients on both variables used to measure this construct were positive and
significant. Likewise, the coefficients for the two variables measuring the relative size of
the young male population were also positive and slightly more significant than the
measures that included the population of both sexes (p=.001 vs. p=.003 and p=.032 vs.
p=.035). As for the indicators of growth in the proportion of young people, the
coefficients of all four had the correct sign (positive) but were highly insignificant.
Finally, the indicators of average annual percentage change in the absolute size of the
youth and young male populations were positively and significantly related to the
quickness of movement toward civil war. However, it remains for these to be tested net
of total population growth to see if this growth truly has the hypothesized effect.
The other set of hypotheses that consider the age structure of population focus on the size and growth of the dependent population (ages 0-14 and 65+). Larger dependent populations were posited to be associated with slower movement toward civil war. The results of the bivariate analyses showed the exact opposite relationship. Both the overall dependency ratio and the child dependency ratio were positively related to the speed of movement toward civil war. Moreover, these relationships were highly significant.
96 Indicatory of Sociah MobtKzatiotr
I proposed two hypotheses relating the mobilization of society to the onset of civil
war. First, I suggested that more highly mobilized populations would move toward civil
war more quickly than those that are less highly mobilized. Second, I posited that
populations that were mobilizing more rapidly would move toward civil war faster than
those mobilizing more slowly.
To measure mobilization, I used indicators of both the level of education and the
level of urbanization. The bivariate results o f testing these indicators against the onset of civil war suggest the exact opposite of what was hypothesized. The relationships of secondary educational enrollment, mean years of education, and percent urban to the pace of movement toward civil war were negative and significant. Likewise, the coefficient pertaining to the measure of average annual percentage change in secondary school enrollment was negative, although insignificant. The other two variables measuring annual average percent change in the urban population and the percent urban had coefficients of the appropriate sign (positive). However, the effect of the latter was insignificant, and the effect of the former could be attributed to a high correlation with overall population growth. We will therefore have to wait for the multivariate results to sort out these issues of causal order.
Indicators of Government Characteristics
I proposed five hypotlieses concerning the characteristics of the state and the speed of movement toward civil war. Specifically, a more rapid movement toward civil
97 warwas^posited to be Knkedto partiab-deraocracy/partiat-aatocracy, a liraited scope of
governance, a decentralized structure, the infrequent application of sanctions, and
political instability.
The bivariate results (Appendix E) give mixed support for these hypotheses. The
democracy/autocracy effect cannot be adequately evaluated in this context since the
hypothesis calls for a curvilinear relationship to be specified. However, the relationship
between government coherence, which is the variable that is closest to specifying a
curvilinear relationship, and the speed of movement toward civil war was negative and
highly significant. As for the other variables, the regression coefficients for the
democracy indicators were negative but insignificant, while the coefficients for the
autocracy and repression indicators were positive but insignificant. If confirmed, these
relationships would suggest that democracy helps to stave off civil war, while autocracy
and repression encourage it.
As for the other hypothesized relationships, the exact opposite relationship from
that which was hypothesized was found for the scope of governance, the extent of
centralization, and the imposition of sanctions. Specifically, the bivariate results indicate
that the relationship between the breadth of the scope of government and quickness of
movement toward civil war is positive and significant. Likewise, the relationship
between the extent of centralization and the speed of movement toward civil war is
positive, although insignificant. Finally, the imposition of government sanctions is positively, although insignificantly, related to the propensity to move toward civil war.
The final hypothesis concerning the characteristics of the government and civil war suggests that instability will promote movement toward war. The bivariate results of 98 a test of the* three indicators of government stability providepmtiaf support for this
hypothesis. All three indicators (changes in polity, changes in executive, and regime
transitions) show a positive relationship to the rapidity of movement toward civil war, but
the regime-transitions variable is the only one that reaches significance.
Indicators of Previous Experience with Other Forms of Collective Action
I proposed two hypotheses that relate to the previous experience of a society with
forms of collective action other than civil war. On the one hand, I suggested that involvement in international wars would increase the speed at which societies move toward civil war. On the other hand, I posited that lower intensity forms of conflict would discourage societies from moving toward civil war.
The analysis of bivariate relationships provides very mixed support for these hypotheses. The measures of the number of prior international wars and the dummy tapping any past experience with international wars had coefficients in the expected direction (positive) but were insignificant. Conversely, the measure of time since involvement in the last international war was in the opposite direction hypothesized
(positive) but was also insignificant. As for the indicators of lower-intensity forms of conflict, each of the variables tapping ethnic collective action were in the opposite direction that was hypothesized (positive) and were highly significant. The variables measuring previous coup d’etats also had positive coefficients but were insignificant.
Indicators of Militarization
The final hypothesis that was offered in this study states that societies with higher levels of militarization as measured by soldiers per capita or defense expenditures/GDP 99 wütmevfrmore-quieldy toward eivit war. Thobivariatoanafysia of these indicatora
provides conflicting information about this hypothesis. Soldiers/capita is negatively
related to the speed at which societies move toward civil war, although the corresponding
coefficient is insignificant. Conversely, defense expenditures/capita is positively related
to the quickness of movement toward civil war, and this coefficient is significant.
M ultivariate R elationships
Turning to the multivariate analysis, the results are presented in Appendices G
through L. In the discussion of these results, I will concentrate on the findings of the
analyses that began in 1965 since they contained the largest number of civil war onsets.
However, I will also highlight the significant deviations in the results from the 1965
analyses from those obtained in the analyses starting in 1970. In reality, the cases of
these deviations are quite few, in part because of the general similarity of the results, but
also because I used the 1970 analyses as a means to verify the results of the 1965
analyses and rejected the latter if it could not be confirmed by the former. Only when the discrepancies were substantively interesting do I present the results.
Development o f the Core Model
In Appendix G, I present the development of the core macro-social model. In
Model I, a lagged dependent variable in the form of the number o f prior civil wars since
1800 was entered as a starting point. This variable was significant at the . 10 level in the
1965 analysis but insignificant in the 1970 analysis. Consistent with hypothesis ( 1 ), the sign of the coefficient is positive, signifying that more prior civil wars would increase the rate of transition to future civil wars. 100 hr Mode} 2, f enter the percent of tabor force emp}oye±in indastry-in t965
and the square of this term as representatives of the level of industrial development for
each country.*’ This specification contributes significantly to the explanatory power of
the model as evidenced by a jump in the pseudo R-squared fiom .01 to .09.-’ In addition,
the signs of the coefficients are consistent with the prediction of hypothesis (2),
suggesting that countries with a moderate amount of industrial development will move
toward civil war more quickly than countries with either low or high levels of
development. Finally, the introduction of this variable does not change the general effect
of the lagged dependent variable.
In Model 3 ,1 add a time-variant indicator of the dominance of the service sector
to the previous specification. Again, this addition contributes significantly to the
explanatory power of the model, causing the pseudo R-squared to rise from .09 to .15.
However, the addition of this variable carries the cost of reducing the number of onsets of civil war considered in the analysis to 19 from the original 29. With respect to the
independent effect of the indicator itself, its coefficient is negative and highly significant.
This results thus disconfirms the hypothesis based on Political Economy notions that a large service sector would encourage civil war because it is simply a symptom of poor social articulation (hypothesis 8). Rather, it provides support for the notion based on
Human Ecology theory that the service sector creates social conditions such as
*’ Most of the other indicators of industrial development that were tested could have replaced this variable in this and subsequent models without substantially altering the results. This variable was chosen because it was the most significant variable in the final base model. Unlike a regular R-squared, the pseudo R-squared cannot be interpreted as an indicator of the explained variance. Rather, its primary function is to allow the investigator to judge the relative strengths of various models. 101 heightenerfinterdependency thatwür discourage cîvît war (hypothesis 7). Finally, the
addition of this variable does not significantly change the effects of the previous
indicators in the model.
In Model 4 ,1 add an indicator of agricultural population density in 1960 and its
square to the previous specification. Again, the addition of this variable contributes to a
significant increase in the pseudo R-squared (from .15 to .22) while maintaining the
number of civil war onsets at 19. The coefficients are both significant and the signs
consistent with the hypothesis that countries with mid-levels of techno-ecological
inheritance should move more quickly toward civil war than those with either rich or
poor inheritances (hypothesis 4). As with the previous models, the addition of this
variable does not significantly alter the effects of the previous indicators in the model.
In models 5 through 7 ,1 show the effects o f two different indicators of ethnic
composition. In model 5, Vanhannen’s (1991) ethnic homogeneity variable is included in the model. In the analysis beginning in 1965, it raises the pseudo R-squared modestly over the previous model (=.25 vs. .22) and is only significant at the . 10 level. In the analysis beginning in 1970, in contrast, the increase in the pseudo R-squared is more marked (.29 vs. .24) and the variable is significant at the .05 level. The coefficient is positive, suggesting support for the hypothesis that higher levels of ethnic dominance encourage the onset of civil war (hypothesis 9). The rest of the variables in the models with both years as starting points retain their substantive effects.
In model 6 ,1 enter a dummy variable created from Barret ( 1982) which flags countries in which the second largest ethnic group comprises between 10 and 30 percent of the size of the largest group. In the 1965 analysis, this variable raises the pseudo R- 102 squared sKgfattymore than did etfani(rhomogeneity (=^ 5>an level. In the analysis beginning in 1970, in contrast, the increase in the pseudo R-squared is more restrained (.27 vs. .29), but the indicator is still significant at the .05 level. If this result stands, it would indicate support for the idea that competition between mismatched ethnic groups will lead societies toward civil war (hypothesis 11). As with ethnic homogeneity, the rest of the variables in the models with both years as starting points retain their substantive effects. In model 7,1 try to tease out the most important specification of ethnic composition by making the two indicators of ethnicity compete against one another for explanatory power. Unfortunately, the results of this test are inconclusive. In the 1965 analysis, the 20 percent dummy variable washes out the effect of ethnic homogeneity, while in the 1970 analysis the results are reversed. As a result, this study is not able to make a firm conclusion about which specification of ethnicity is the most important for predicting the speed at which countries move toward civil war. In order to verify the robustness of the results of the core model, I conducted a bootstrap analysis of the 1965 analysis (results not shown). Specifically, I drew a sample of 116 cases from the original 1965 dataset using a "with replacement" strategy. The resulting problem set included 70 of the original 116 countries, with some of the original units being included in the sample as many as five times. In addition, the number of wars in the resulting analysis increased from 19 to 24. Upon re-testing the core model, 1 confirmed the results of the original analysis. All of the variables in the core model obtained significance at the .05 level except for ethnic homogeneity (p=. 144). Thus, I conclude that the results of the core model are robust. 103 Having completed the-discussion of the-development of the core macro-levet model, it is instructive to consider the substantive effects of the variables that make up this model. To aid in this task, I have provided graphs of the independent effects of each of the core variables in Appendix M. Previous civil wars: The hazard ratio associated with the number of previous civil wars variable in the 1965 model indicates that each additional instance of war leads to approximately a 37 percent increase in the rate of transition. Thus, all else constant, a country with three prior civil wars moves toward civil war approximately 2.5 times as fast as a country with no prior civil wars. Labor force in industrv: Since the labor force variable is specified in with curvilinear term, the interpretation of the hazard ratio is more complex. However, the graph in Appendix M suggests that the rate of transition to civil war peaks when a country has approximately 17 percent of its labor force working in the industrial sector, all else constant. At this point, a country moves toward civil war over twice as fast as if it had either half that proportion in the industrial labor force (8.5%) or half again as much in the industrial labor force (25.5%). Service sector dominance: The hazard ratio of .896 associated with the service sector dominance variable indicates that, all else constant each additional unit increase in the level of service dominance creates approximately a 10 percent decrease in the rate of transition. This creates quite a steep downward slope at the early stages of service sector dominance, with countries that have a score of 30 on this variable (or an average of 15 104 percent of the tabor force employed in services" and percent of GDPdne to services^ moving toward civil war over five and a half times more slowly than countries with a score of 15 on this variable. Agricultural population density: As with the industrial development variable, the interpretation of the hazard ratio for the agricultural population density variable is complicated by the curvilinear relationship. However, the graph associated with this variable in Appendix M indicates that, all else constant, the rate of transition toward civil war peaks when there are approximately 25 people per hectare in rural areas. A country with this level of rural density will move toward civil war over twice as fast as if it had half that level of density ( 12.5 people/hectare) and 10 times as fast as a country with twice that level of rural density (50 people/hectare). Ethnic homogeneity: The hazard ratio of 1.03 associated with the ethic homogeneity variable indicates that each additional percent added to the size of the largest population group leads to a 3 percent increase in the rate of transition, all else constant. Thus, a country where the largest group comprises 50 percent of the population will move toward civil war approximately 1.5 times as fast as a country where the largest group makes up only 25 percent of the population. Similarly, a country with almost complete ethnic homogeneity (i.e., almost all of the population comes from the same ethnic group) will move toward civil war 2.5 times as fast as a country in which the largest ethnic group makes up just 50 percent of the total population.-' *' This last implication of the finding for ethnic homogeneity is the clearest indication that more research is needed on the relationship between ethnicity and the onset of civil war. 105 2Q^ercenr ethnie dnmmvr Finafly; the hazard ratio o f 4.9^ associated with the 20^ percent ethnic dummy variable indicates that, ail else constant, countries in which the size of the second largest ethnic group is between 10 and 30 percent of the size of the first ethic group move toward civil war almost 5 times as fast as those with other ratios of second to largest groups. Other Macro-social Variables and the Onset o f Civil War Appendices I and J present the results obtained when other macro-social variables were added to the core model. This presentation is not meant to be an exhaustive summary o f all of the analyses performed, but rather an illustrative summary of the effect of general categories of variables. Each variable is added separately to the core models to preserve parsimony. The presentation of these results serves two purposes. First, I have tried to include the most common variables that have been tested in the literature to date. Second, 1 have included several other variables that achieved statistical significance in the context of the core model. In the discussion that follows, however, I will limit my commentary to the two models with consistently significant results. The first variable that achieves statistical significance in the presence of the core model is the variable adapted from Gurr 1993 on lost autonomy multiplied by the proportion of the population associated with this loss. The addition of this variable reduces the number of onsets of civil war in the analysis to 18 and raises the pseudo R- 106 sqaaredta 35 ay compared t(r.23forthgcomnodeh Ayyresult of entering-tWy variably into the model, the service dominance variable becomes insignificant in the 1965 analysis (p=.158) but not in the 1970 analysis (p=.078). With respect to the independent effect of this indicator, its coefficient is positive and highly significant. This result supports the hypothesis that countries with groups that have been autonomous previously will move more quickly toward civil war (hypothesis 12). The fact that the service sector effect becomes less significant in the presence of this indicator suggests that it may counteract the positive affects of service sector dominance. The other variable that achieves significance in the presence of the core model is agricultural exports/total exports in 1980. The addition of this variable to the model does not affect the number of onsets of civil war in the analysis, nor the significance of any of the core variables. However, it only raises the pseudo R-squared to .266 as compared to .253 for the core model. The coefficient associated with this variable is positive and significant. This result provides support for the hypothesis that countries who are more dependent on the export sector will move toward civil war more quickly than those who are less dependent (hypothesis 16). Intervenins Mechanisms and the Onset o f Civil War The final stage of the analysis consisted of testing the plethora o f variables that might mediate between the macro-social variables and the onset of civil war (Appendices K and L). These variables were tested as both linear effects and, where appropriate, as curvilinear or interactive effects. Again, the presentation of these results is not meant to 107 be air exhaustive recount of att of thganatyses performed; but rather ihustrative of the effect of general categories of variables. In practice these variables were tested both alone, in combination with other variables, and, where theoretically appropriate, in different functional forms (e.g., quadratic and simple interactions). The effect of each variable is shown in a separate model so as to preserve parsimony. One of the surprising findings of this study was the lack of many consistently significant relationships between potential intervening factors and the onset of civil war. Variables representing constructs such as absolute deprivation and “greed”, as well as indicators of population growth, the age composition of society, characteristics of the state, previous collective action, and the militarization of society all failed to consistently achieve statistical significance in the presence of the core model. However, several variables proported to represent relative deprivation did achieve significance consistently. They were the intensity of political discrimination, grievances by minorities at risk, income inequality, and sectoral inequality. First, the intensity of political discrimination was found to be positively related to the speed at which countries moved toward civil war. The addition of this variable raises the pseudo R-squared to .35 as compared to .25 for the core model, but the number of civil wars in the analysis is reduced to 15. Partly because of this, both the lagged dependent variable and the indicators of ethnic composition failed to retain their significance in the 1965 analysis. However, all variables retained their significance in the 1970 analysis. With respect to the independent effect o f this indicator, its coefficient is positive and marginally significant in the 1965 analysis (p=.067) but more significant in the 1970 108 analysis^(p=;002); The hazard" ratrcr fronrthe t965 analysis that is associated with this variable indicates that each additional increase to this index will lead to approximately a 72 percent increase in the rate of transition. The hazard ratio from the 1970 analysis, however, has the rate of transition increasing more than five fold for each step of the index (see Appendix M for a graph o f these relationships). This result supports the hypothesis that countries with people experiencing highly intense discrimination will move toward civil war faster than those with less intense discrimination (hypothesis 20). Second, the presence of social, political, and economic grievances in society was also found to be positively related to the propensity of a country to move toward civil war. Moreover, the significance of this variable is very high. The addition of this variable raises the pseudo R-squared to .31 as compared to .25 for the core model, and the number of civil wars in the analysis is maintained at 18. Finally, all of the variables in the core model retain their significance when this variable is added. The independent effect of this indicator is positive and very significant. The hazard ratio suggests that a one-unit increase in this index will translate into a 46 percent jump in the rate of transition (see Appendix M for a graph of this variable). As with the intensity of political discrimination indicator, these results support the hypothesis that countries with minority groups that have more economic, political, and social complaints will move toward civil war faster than those with less (hypothesis 20). The final two indicators that achieved statistical significance are Gini coefficients of inequality - one of income inequality (Deininger 1996) and the other of sectoral inequality. The addition of these variables raised the pseudo R-squared to .35 and .30, 109 respectively, anddroppedthe rannberof wars considered in the analysis to i6 and X7. However, all of the variables in the core model maintained their significance in the presence of these variables. The independent relationship of both of these indicators to the dependent variable was curvilinear, with higher rates of transition to civil war occurring at mid ranges of inequality (see Appendix M for a graph of these relationships). Therefore, these findings provided support for hypothesis 22. Of the intervening variables that did not achieve consistent statistical significance, several that were significant in the 1965 model but not in the 1970 model deserve mention. These are infant mortality, several specifications of population growth, and secondary school enrollment. For each of these variables, I was able to determine by an examination of beta correlations that the failure to reach statistical significance in the 1970 model was not due to multicoUinearity. Nor was their failure to reach significance due to the presence of a different set of wars, as determined by other analyses that excluded from an analysis o f the 1965 problem set those wars not included the 1970 problem set (results not shown). I can only speculate that these results are due to influential cases. However, due to the lack of adequate regression diagnostics with the PHREG procedure in S AS, this fact cannot be determined conclusively. If we were to assume that these relationships were valid, their effects would be interpreted as follows. The addition of infant mortality and its square to the core model results in one less war in the analysis and only a modest increase to the pseudo R-squared (.287 as compared to .276 in the core model). Predictably, since the level of development in a country is typically highly correlated with the general standard of living, entering this 110 variable resuit» le the effect of the levetof^developraent^tebeeome insignifieant^pef square term =.141). With respect to the independent effect of this indicator, the base term is positive and significant at the .10 level, and the quadratic term is negative and significant at the .10 level. Thus, its effect is curvilinear and supports hypothesis 22, with countries at mid-levels of infant mortality moving toward civil war more quickly than countries with either low or high levels. Next, three variables that reflect population growth were all significant when added to the core model in the 1965 analysis but not in the 1970 analysis. The average annual changes in the total population, the population 15 to 64, and the population 15-29 were all significant at the p<.05 level, with the change in the population 15 to 64 being the most significant. When forced to compete against one another in pairs of two, none of the variables retained significance due to high levels of multicoUinearity (results not shown). However, growth in the population 15 to 64 consistently came closer to achieving significance than the other two. Therefore, 1 assume that its effect is the primary mechanism at work. As such, the addition of the annual average percent change to the population 15 to 64 to the core model resulted in one less war in the analysis and a modest increase to the pseudo R-squared (.301 as compared to .276 in the core model). In addition, each of the variables in the core model retained significance. With respect to the independent effect of this indicator, each additional percent added to the rate of change of this population resulted in approximately a 73 percent increase in the rate of transition to civil war. It therefore provides support for hypothesis 27, which states that a society with a more 111 rapiëly^growinçpopulatiott betweei^theages-of 1-S amt64^ witt move toward civd war more quickly than one with a population ages 15 to 64 that is growing more slowly. The final variable to be described is the secondary educational enrollment ratio. The addition of it and its square to the core model results in two less wars in the analysis but a relatively high increase to the pseudo R-squared (.349 as compared to .276 in the core model) and the rest of the core model retaining statistical significance. With respect to the independent effect of this indicator, the base term is positive and significant at the .10 level, and its quadratic term is negative and significant at the .05 level. Thus, its effect is curvilinear, with countries with mid-levels of secondary educational enrollment moving toward civil war more quickly than countries with either low or high levels. This effect does not support either o f the hypotheses regarding mobilization and the onset of civil war (hypotheses 32 and 33), which both posited a linear relationship. In summary then, 1 found solid support for the notions from both structural modernization and human ecology theory. Countries move toward civil war more quickly when they are at mid-ranges of development, have a moderate techno-ecological inheritance, and exhibit an anemic service sector. Additionally, the ethnic composition of society seems to be important for the onset of civil war, although the exact specification of this relationship could not be determined in this study. Conversely, I found little support for Political Economy Theory, with only one measure of export dependence reaching statistical significance. Finally, I found little support for possible intervening mechanisms between the macro-social context and the speed o f movement toward civil war except for a positive effect of relative deprivation. 112 CHAPTERS CONCLUSIONS This study has provided an in-depth look at the factors that contribute to the onset of civil war. It has been heavily informed by previous research on the causes of violent collective action as well as more recent research on civil war itself. It has also harnessed the theoretical insights from two distinct traditions, social movements and international development, to construct a broad theoretical framework regarding both the macro-social context and the immediate conditions of civil war. Further, it has drawn upon a broad range of data sources from which to construct variables, and utilized event history analysis, a relatively uncommon but powerful statistical technique, to explore both the static and dynamic aspects that can lead to civil war. Finally, a proprietary data-mining tool was developed during the course of this study that, although it could not be used fully due to the physical limitations of current technology, shows much promise for use in future research. In the end, the results of the analysis showed that civil war may be best thought of in terms of an emergent property as opposed to an agency driven phenomenon. By far, the hypotheses concerning the macro-social conditions related to the onset of civil war received much more support than did the hypotheses concerning the immediate conditions of war. Specifically, 1 showed that mid-levels o f both industrial development 113 andtechno-ecotogicat inheritanccencourage civibwai; while a strong'service sector discourages civil war. In addition, I found support for several different aspects of the relationship between ethnicity and civil war, including the importance of ethnic homogeneity, ethnic competition, and separatist or autonomous history. With respect to intervening mechanisms, it was surprising to find that a plethora of indicators representing constructs as diverse as absolute deprivation, “greed,” rapid population growth, the age composition of society, characteristics of the state, previous collective action, and the militarization of society all failed to show a consistently significant relationship to the onset of civil war. However, several variables associated with relative deprivation did make the cut. Specifically, 1 found that discrimination, the grievances o f minorities, and mid ranges of both income and sectoral inequality were all independently associated with rapid movement toward civil war. There are several shortcomings of this study that highlight the need for future research. First, the study was limited to the years 1965 to 1992. Almost a decade has past since the end point of this study, and it will be important to incorporate the experience o f the 1990s into future analyses as a test o f the generalizabiiity of these results to the non-cold war era. In this vein, Gleditsch (2001) has begun a formal extension of the Correlates of War dataset. In a paper he presented this summer at a conference sponsored by the World Bank, Gleditsch released the preliminary results of this work, the implications of which for civil war are plotted in Figure 4. If his work is correct, the decrease in both the incidence of civil war across the globe and the new onsets of civil war in the last half of the 1990s are striking. The number o f ongoing wars drops from a high of 25 in 1992 114 30 ^ 3 Onsets of civil war 25 Ongoing civil wars (onset 1965 or later) -«-Ongoing civil wars ______ 20 15 f 10 5 0 Figure 3: Onsets of civil wars per year and number of ongoing civil wars with an onset between 1965 and 1999. back to only 10 wars in 1999. Similarly, in contrast to the 16 civil wars that began between 1990 and 1994, only one war began between 1995 and 1999 (Uganda). It will be therefore important in the near future to model the civil wars of the 1990s in conjunction with the rest to see if anything can be inferred about this dramatic decrease in the global prevalence and initiation of civil war. A second avenue for future research encouraged by this study is the exploration of the effect of the ethnic composition of society on civil war. The implication of the finding reported here that movement toward civil war is at its height in situations of almost complete homogeneity chafes at intuitive notions. In addition, the finding that societies in which the second largest ethnic group comprises between 10 and 30 percent 115 of the-size-of the largest ethirio groop^tnove more quickly toward civil war supports this need. However, the focus of such a line of research will need to be not only on the specification of ethnic composition variables but also on the source and quality of the data. A third line of future research that this study has highlighted is the effect of other states on the movement of societies toward civil war. In this study, the problem first arose in the need for the assumption that each unit's history with civil war was independent of the histories of the other units under investigation. In fact, civil war may diffuse across national borders, and this possibility needs to be tested. In addition, the relatively simple specifications by Political Economy theorists of other types of influence from the international system on movement toward civil war calls for redress using the Human Ecological principles of interdependence, dominance, and isomorphism. Moreover, the possibilities for testing such effects has been greatly enhanced recently with the advent o f multivariate analysis using information gleaned from Geographic Information Systems (CIS). Work of these types would help to elucidate the specific effects of the external environment on the phenomenon of civil war. The results of this study may not be very heartening for the policy community. If it is true that civil war is more closely related to relatively intransigent social conditions than it is to the desires and decisions o f individuals, then the ability to formulate policy aimed at preventing wars before they start and at bringing wars to an end may be highly constrained. It may be that the international community will need to adopt a policy of 116 “triage’*^ with respect ta civih wars; preparingay wett arircair hr advance for the tumult that is created by civil war in vulnerable countries and readying itself to provide both humanitarian aid and peacekeeping services when civil war breaks out. 117 BIBLIOGRAPHY Adelman, Irma, and Cynthia Taft Morris. 1973. Economic Growth and Social Equity in Developing Countries. San Francisco: Stanford University Press. Alonso, William. 1980. "Five Bell Shapes in Development." Papers o f the Regional Science Association 45:5-16. Amin, Samir. 1976. Unequal Development. New York: Monthly Review Press. Auvinen, Juha. 1997. "Political Conflict in Less Developed Countries, 1981-89." Journal o f Peace Research 34:177-195. Azam, J.P. 1995. “How to Pay for the Peace? A Theoretical Framework with Reference to Afiican Countries.” Public Choice 83:173-84. Barret, D.B. (ed.). 1982. World Christian Encyclopedia. Oxford: Oxford University Press. Barro, Robert J., and Jong-Wha Lee. 1996. “International Measures of Schooling Years and Schooling Quality.” American Economic Review, Papers and Proceedings 86:218-23. Bates, Robert H. 1983. “Modernization, Ethnic Competition, and the Rationality of Politics in Contemporary Afnca.” Pp. 152-71 in D. Rothchild and V.A. Olorundsola (eds.). State Versus Ethnic Claims. Boulder, CO: Westview. ------. 1 9 7 4 . “Ethnic Competition and Modernization in Contemporary Africa.” Comparative Political Studies 6:457-84. Beck, Nathaniel. 2001. "Time-Series-Cross-Section Data: What Have We Learned in the Past Few Years?" Annual Review o f Political Science 4:271-93. Beck, Nathaniel, David Epstein, Simon Jackman, and Sharyn OTIalloran. 2001. "Alternative Models of Dynamics in Binary Time-Series-Cross-Section Methods: The Example of State Failure." Paper prepared for delivery at the 2001 Annual Meeting o f the Society for Political Methodology, Emory University. 118 Beek, Nathaniet, Jonathan N. Katev and-RiehardTuekef. "Taking Time SeFieusly: Time-Series-Cross-Section Analysis with a Binary Dependent Variable." American Journal o fPolitical Science 42:1260-1288. Berry, RA., and W. Cline. 1979. Agrarian Structure and Productivity in Developing Countries. Baltimore, MD: Johns Hopkins University Press. Blau, Peter M. 1977. Inequality and Heterogeneity: A Primitive Theory of Social Structure. New York: Free Press. Bollen, Kermeth A. 1983. "World System Position, Dependency, and Democracy: The Cross-national Evidence." American Sociological Review 48:468-73. Bomschier, Volker. 1980. “Multinational Corporations and Economic Growth: a Cross- National Test of the decapitalization Thesis.” Journal o f Development Economics 7:191-210. Bomschier, Volker, and Christopher Chase-Dunn. 1985. Transnational Corporations and Underdevelopment. New York: Praeger. Bosemp, Esther. 1965. The Conditions of Agricultural Growth: The Economics of Agrarian Change Under Population Pressure. Chicago: Aldine. Boswell, Terry, and William J. Dixon. 1990. "Dependency and Rebellion: A Cross national Analysis." American Sociological Review 55:540-559. Boyle, K.E. and T.B. Starr. 1985. "Survival models for fertility evaluation." Journal o f the American Statistical Association 80:823-827. Bradshaw, York. 1987. “Urbanization and Underdevelopment: A Global Study of Modernization, Urban Bias, and Economic Development.” American Sociological Review 52: 224-39. Brown, Michael E. 1996. "Introduction," in Michael E. Brown (ed.>. The International Dimensions o f Internal Conflict. Cambridge, MA: MIT Press. Cannent, D.B. 1993. "The International Dimensions of Ethnic Conflict: Concepts, Indicators, and Theory." Journal o f Peace Research 30:137-50. Chenery, Hollis, and Moises Syrquin. 1975. Patterns o f Development 1950-1970. London: Oxford University Press. Christenson, Matthew. 2001 (forthconung). "Global Population Composition." Ch. 2 in U.S. Census Bureau, Global Population Proflle: 2000. Washington, D C.: Government Printing Office. 119 CoUier, Paut. 2060; “RebeüionrayffQaasi^CrimniahActivity;"' Revolution 44:839-53. 1999. "On the Economic Consequences of Civil War." Oxford Economic Papers 5 1:168-83. Collier, Paul, and Anke Hoeffler. 2001. "Greed and Grievance in Civil War." World Bank Policy Research Working Paper 2355, World Bank. 1999. "Justice-Seeking and Loot-Seeking in Civil War." Paper presented at the World Bank's Development Economic Research Group conference on Civil Conflicts, Crime, and Violence, Washington, D C., February 22-23, 1999. 1998. "On the Economic Causes of Civil War." Oxford Economic Papers 50:563-73. Coser, Lewis A. 1956. The Functions o f Social Conflict. Glencoe, IL: Free Press. Crenshaw, Edward. 2001. “The Demography o f Development: An Evolutionary Approach.” Unpublished manuscript. ------. 1995. "Democracy and Demographic Inheritance: The Influence of Modernity and Proto-Modemity on Political and Civil Rights, 1965-1980." American Sociological Review 60:702-718. ------. 1993. "Polity, Economy, and Technology: Alternative Explanations for Income Inequality." Social Forces 71:807-816. ------. 1992. "Cross-national Determinants of Income Inequality: A Replication and Extension Using Ecological-Evolutionary Theory." Social Forces 71:339-363. Crenshaw, Edward, and Ansart Ameen. 1994. “The Distribution of Income Across National Populations: Testing Multiple Paradigms." Social Science Research 23:1-22. Crenshaw, Edward, and Doyle Ray Oakey. 1998. "Jump-Starting Development: Hyperurbanization as a Long-Term Economic Investment." Sociological Focus 23:1-22. Davies, James C. 1962. "Toward a Theory of Revolution." American Sociological Review 27:5-19. 120 Deinmger, K., and L. Squire. Î996: **A New Data Set Weasnrùig hcome Inequality. World Bank Economic Review 10:565-91. Delacroix, J., and Charles Ragin. 1981. "Structural Blockage: A Cross-National Study of Economic Dependency, State Efficacy, and Underdevelopment." American Sociological Review 86:1311-1347. Deutch, Karl W. 1961. "Social Mobilization and Political Development." American Political Science Review 55:493-514. Dix, R.H. 1984. “Why Revolutions Succeed and Fail.” /*o//rv 26:423-46. Durkheim, Emile. [1933] 1984. The Division of Labor in Society. New York: Free Press. Eckstein, H. 1980. "Theoretical Approaches to Explaining collective Political Violence." Pp. 135-166 in T.R. Gurr (ed.). Handbook o fPolitical Conflict. New York: Free Press. Elbadawi, Ibrahim A. 2000. "How Much Civil War Will We See? Estimating the Incidence of Civil War in 161 Countries." Working Paper, World Bank. Ellingsen, Tanja. 2000. "Colorful Community or Ethnic Witches' Brew? Multiethnicity and Domestic Conflict During and After the Cold War." Journal o f Conflict Resolution 44:228-249. Esty, Daniel C., Jack A. Goldstone, Ted Robert Gurr, Barbara Harff, Marc Levy, Geoffrey D. Dabelko. Pamela T. Surko, and Alan N. Unger. 1998. State Failure Task Force Report: Phase II Findings. McLean, VA: Science Applications International Corporation. Evans, Peter. 1979. Dependent Development. Princeton: Princeton University Press. Evans, Peter, and Michael Timberlake. 1980. “Dependence, Inequality, and the Growth of the Tertiary: A Comparative Analysis of Less Developed Countries.” American Sociological Review 45:531-51. Feierabend, Ivo K., & Rosalind L. Feierabend. 1966. “Aggressive Behaviors Within Polities, 1948-1962: A Cross-national Study.” Journal o f Conflict Resolution 10:249-71. —. 1972. “Systemic Conditions of Political Aggression: An Application of Frustration-Aggression Theory.” Pp. 136-83 in I.K. Feierabend, R.L. Feierabend & TJl. Gurr (eds.). Anger, Violence, and Politics. Englewood Cliffs: Prentice-Hall. 121 Feierabend; IvaK,RosafmdL. Feierabend; frB:A.Nesvold; i969; “SociaFGhangeand Political Violence: Cross-national Patterns.” Pp. 632-87 in HJ). Graham & T.R. Gurr (eds.). Violence in America. New York: Âneger. Feierabend, Ivo K., B.A. Nesvold, & Rosalind L. Feierabend. 1970. “Political Coerciveness and Turmoil.” Law Sociological Review 5:93-118. Fiala, Robert. 1983. “Inequality and the Service Sector in Less Developed Countries: A Reanalysis and Respecification.” American Sociological Review 48: 421-428. Gamson, William A. 1975. The Strategy o fSocial Protest. Homewood, BL: Dorsey. Galtung, Johan. 1971. “A Structural Theory of Imperialism.” Jbz/mn/ q/" fence Research 8:81-117. Gamson, William A., Bruce Fireman, and Steven Rytina. 1982. Encounters With Unjust Authority. Homewood, IL: Dorsey Press. Gastil, Raymond D. 1989. Freedom in the World. New York: McKay. Gerlach, Luther P., and Virginia H. Hine. 1970. People, Power, and Change. New York: Bobbs-Merrill. Gissinger, Ranveig, and Nils Petter Gleditsch. 1999. "Globalization and Conflict: Welfare, Distribution, and Political Unrest." Journal o f World-Systems Research 2:237-365. Gleditsch, Kristian S. 2001. "A User's Guide to Omitted Wars and Participants in the COW War Data." Paper presented to the conference. Identifying Wars: Systematic Conflict Research and its Utility in Conflict Resolution and Prevention, Uppsala University, June 8-10, 2001. Glover, Donald R., and Julian L. Simon. 1975. “The Effect of Population Density on fefrastructtner The Case of Roadbuilding.” Economic Development and Cultural Change 23:453-468. Goldstone, Jack A. 1997. "Population Growth and Revolutionary Crises." Pp. 102-120 in John Foran (ed.), 77ieonr/«g^ Revo/i/nb/i^. New York: Routledge. ------. 1991. Revolution and Rebellion in the Earlv Modem World. Berkeley: University of California Press. -. 19S6. Revolutions: Theoretical, Comparative, and Historical Studies. San Diego: Harcourt Brace Jovanovich. 122 GoWst€>ne, Jack A., Teé Robert GuTFiand-FanrokltMoshirh Revolutions o f the Late Twentieth Century. Boulden Westview. Granovetter, Mark. 1984. “Small is Bountiful: Labor Markets and Establishment Size.” American Sociological Review 49:323-34. Grossman, H.I. 1995. “Insurrections.” In K. Hartley and T. Sandler (eds. ), Handbook of Defense Economics, Vo\. 1. Amsterdam: Elsevier Science B.V. Gurr, Ted Robert. 1993a. Minorities at Risk: A Global View o fEthnopolitical Conflict. Washington, DC: United States Institute of Peace Press. —. 1993b. "Why Minorities Rebel: A Global Analysis of Communal Mobilization and Conflict Since 1945." International Political Science Reviev.’ 14:161-201. ------. 1990. Polity II: Political Structures and Regime Change, 1800-1986. Ann Arbor: ICPSR data set #9263. ------1989. “Protest and Rebellion in the 1960s: The United States in World Perspective.” Pp. 101-30 in T.R. Gurr (ed.). Violence in America, Vol. 2. Newbury Park: Sage. -. 1970. Why Men Rebel. Princeton: Princeton University Press. -. 1969. “A Comparative Study of Civil Strife.” Pp. 572-632 in H.D. Graham «& T.R. Gurr (eds.), Fito/encem.<4menca. New York: Praeger. -. 1968. “A Causal Model of Civil Strife: A Comparative Analysis Using New Indices.” American Political Science Review 62:1104-24. Gurr, Ted R., and Robert D. Duvall. 1973. “Civil Conflict in the 1960s: A Reciprocal Theoretical System with Parameter Estimates.” Comparative Political Studies 6:135-169. Gurr, Ted R., and Michael I. Lichbach. 1979. “A Forecasting Model for Political Conflict within Nations.” Pp. 153-93 in J.D. Singer & M.D. Wallace (eds.). To Auger Well. Beverly Hills: Sage. Hartman, John, and Wey Hsiao. 1988. "Inequality and Violence: Issues o f Theory and Measurement." American Sociological Review 53:794-800. Hawley, Amos. 1981. Urban Society: An Ecologfcal Approach, 2'^^ Edxiion. New York: Ronald. 123 ------. 196S. "Ecology:'^ ^p;3^%-yyT‘ ’va:The~lnremati(mat Encyclopedia o f the Social Sciences, Vol. 4. New York: MacMillan. Hechter, Michael. 1987. Principles o fSocial Solidarity. Berkeley, CA: University of California Press. Hegre, Havard, Tanya Ellingsen, Nils Petter Gleditsch, and Scott Gates. 1999. "Towards a Democratic Civil Peace? Opportunity, Grievance, and Civil War, 1816-1992. Paper presented to the DECGR Launch Workshop Civil Conflicts, Crime, and Violence in Developing Countries, February 22-23, 1999, Washington, D C. Hibbs, Douglas A. 1973a. Mass Political Violence. New York: Wiley. ------. 1973 b. “Industrial Conflict in Advanced Industrial Societies.” American Political Science Review 70:1033-58. Higley, John, and Michael G. Burton. 1989. “The Elite Variable in Demographic Transitions and Breakdowns.” American Sociological Review 54:17-32. Hirschmann, A.O. 1973. “The Changing Tolerance for Income Inequality in the Course of Economic Development.” Quarterly Journal o fEconomics 87:544-66. . 1991. The Third Wave. Norman, OK: University of Oklahoma Press. Horowitz, Donald. 1985. Ethnic Groups in Conflict. Berkeley: University of California Press. Huntington, Samuel. 1968. Political Order in Changing Societies. New Haven: Yale University Press. Jackman, Robert. 1978. “The Predictability of Coups D’Etat.” American Political Science Review 72:1262-75. Jaggers, Keith, and Ted Robert Gurr. 1995. "Tracking Democracy's Third Wave with the Polity HI Data." Journal o f Peace Research 32:469-482. Jenkins. J. Craig. 1983a. "Resource Mobilization Theory and the Study of Social Movements." Annual Review o fSociology 9:527-53. ------. 1983b. “Why Do Peasants Rebel? Structural and Historical Theories of Modem Peasant Rebellions.” American Journal o f Sociology 88:487-514. Jenkins, J. Craig, and Doug Bond. 2001. “Conflict-Carrying Capacity, Political Crisis, and Reconstruction: A Framework for the Early Warning of Political System Vulnerability.” Journal o f Conflict Resolution 45:3-31. 124 Jenkins; J; Craig; and^AugnstineJ; Kposowa. t990: '^Expiaming^MiHtary Conps EXEtatr Black Africa 1957-1984.” American Sociological Review 55:861-75. Jenkins, J. Craig, and Charles Perrow. 1977. "Insurgency of the Powerless: Farm Worker Movements 1946-1972." American Sociological Review 42:249-68. Jenkins, J. Craig, and Kurt Schock. 1992. "Global Structures and Political Processes in the Study of Domestic Political Conflict." Annual Review o f Sociology 18:161- 185. Kaplan, Robert D. 2000. The Coming Anarchy. New York: Random House. Kasarda, John D., and Edward M. Crenshaw. 1991. "Third World Urbanization: Dimensions, Theories and Determinants." Annual Review of Sociology 17:467- 501. Kerr, Clark, et al. 1960. Chs. 1 & 3 in Industrial and Industrial Man: The Problems of Labor and Management in Economic Growth. New York: Oxford University Press. Kick, Edward L. 1980. "World System Properties and Mass Political Conflict Within Nations: Theoretical Framework." Journal o fPolitical and Military Sociology 8:175-90. King, Gary, and L. Zeng. 2000. "Logistic Regression in Rare Events Data." http:. ■gking.har\'ard.edu preprints.html. Kirk, Dudley. 1996. "Demographic Transition Theory." Population Studies 50:36l-3S7. Komhauser, William. 1959. The Politics o f Mass Society. Glencoe, IL: Free Press. Kposowa, Augustine J., and J. Craig Jenkins. 1993. "The Structural Sources of Military Coups in Postcolonial Africa, 1957-1984.” American Journal o fSociology 99:126-63. Lachmann, Richard. 1990. “Class Formation Without Class Struggle: An Elite Conflict Theory of the Transition to Capitalism.” American Sociological Review 55:398- 414. Leitenberg, Milton. 2001. "Deaths in Wars and Conflicts Between 1945 and 2000." Paper prepared for the Conference on Data Collection in Armed Conflict., Uppsala, Sweden, June 8-9,2001. Lenski, Gerhard. 1966. Power and Privilege: A Theory of Social Stratification. New York: McGraw-Hill. 125 Lenskr, Gerhard; and Patrick Notan: t984r '*Tngectories“ofDevelopmentr A Test of Ecological-Evolutionary Theory. " Social Forces 63:1-23. Lemer, Daniel. 1958. The Passing of Traditional Society: Modernizing the Middle East. New York: Free Press. Levy, Marion J., Jr. 1966. Modernization and the Structure o fSociety. Princeton, NJ: Princeton University Press. Licklider, Roy. 1995a. Stopping the Killing. New York: New York University Press. ------. 1995. The Consequences of Negotiated Settlements in Civil Wars, 1945-1993. American Political Science Review 89:681-90. Lichbach, Mark Irving. 1989. "An Evaluation of Does Economic Inequality Breed Political Conflict?' Studies." World Politics 41:431 -70. Lipset, Seymour M. [1960] 1981. Political Man: The Social Basis fo r Politics., expanded edition. Baltimore, MD: Johns Hopkins University Press. Little, Daniel. 1991. Ch. 4-6 in Varieties of Social Explanation An Introduction to the Philosophy of Social Science. Boulden Westview Press. London, Bruce, and Thomas Robinson. 1989. "The Effect of International Dependence on Income Inequality and Political Violence." American Sociological Review 54:305-308. London, Bruce, and Bruce A. Williams. 1988. “Multinational Corporation Penetration, Protest, and Basic Needs in Non-Core Countries: A Cross-National Analysis.” Social Forces 66:747-73. McAdam, Doug. 1982. Political Process and the Development o fBlack Insurgency. Chicago: University o f Chicago Press. McCarthy, John D., and Mayer N. Zald. 1977. "Resource Mobilization and Social Movements: A Partial Theory." American Journal o fSociology 82:1212-41. 1973. The Trend o f Social Movements. Morristown, NJ: General Learning. Melson, R., and H. Wolpe. 1970. “Modernization and the Politics of Communalism.' American Political Science Review 64:1112-30. Midlarsky, Manus. 1988. "Rulers and Ruled: Patterned Inequality and the Onset of Mass Political Violence." American Political Science Review 82:587-95. 126 “The-RevototionaryTransfonnatroirof Foretgit PoKcyr Agrarianism and Its International Impact.” In Charles W. Kegley and Pat McGowan (eds.). The Political Economy o fForeign Policy Behavior. Beverly Hills, CA: Sage. Moaddel, Mansoor. 1994. "Political Conflict in the World Economy: A Cross-National Analysis of Modernization and World-System Theories." American Sociological Review 59:276-303. Mohammed, Nadir A.L. 1999. "Civil Wars and Military Expenditures: A Note." Paper presented at the World Bank conference on Civil Conflicts, Crime, and Violence, Washington, D.C., February 22-23, 1999. Moore, Barrington. 1966. The Social Origins o f Dictatorship and Democracy. Boston: Beacon Press. Mosca, Gaetano. [1896] 1939. The Ruling Class. New York: McGraw Hill. ------. 1986. "Income Inequality and Political Violence: The Effects of Influential Cases." American Sociological Review 51:441-45. Muller, Edward. 1985. "Income Inequality, Regime Repressiveness, and Political Violence." American Sociological Review 50:47-61. Muller, Edward, and Mitchell A. Seligson. 1987. "Inequality and Insurgency." American Sociological Review 82:425-451. Muller, Edward, and Erich Weede. 1990. "Cross-national Variation in Political Violence: A Rational Action Approach." Journal o f Conflict Resolution 34:624- 651. Nagel, Jack. 1976. “Erratum” World Politics 2%:315. -. 1974. “Inequality and Discontent: A non-linear hypothesis.” World Politics 26:453-72. Nas, Tevfik F., Albert C. Price, and Charles T. Weber. 1986. "A Policy-Oriented Theory of Corruption." American Political Science Review 80:107-19. Nolan, Patrick, and Gerhard Lenski. 1985. "Technoeconomic Heritage, Patterns of Development, and the Advantage of Backwardness." Social Forces 64:341-58. Notestein, Frank W. 1945. "Population — The Long View." Pp. 36-57 in T.W. Schultz (ed.). Food for the World. Chicago, IL: University of Chicago Press. 127 Oberschatfc Anthony: t973: SocfatCanfticranctSociatMavements^. Engtewood Cliffs, NJ: Prentice-Hall. O’Donnell, Guillermo A. 1973. Modernization andBureaucratic-Authoritarianism: Studies in Latin American Politics. Berkeley, CA: Institute for International Studies, University of California. Olson, Mancur. 1965. The Logic o f Collective Action. Cambridge, MA: Harvard University Press. ------. 1963. "Rapid Growth as a Destabilizing Force." Journal o f Economic History 23:529-52. Olzak, Susan. 1983. “Contemporary Ethnic Mobilization.” Annual Review o f Sociolog}’ 9:355-74. Olzak, Susan, and Joane Nagel. 1986. “Competitive Ethic Relations.” Pp. 1-14 in S. Olzak and J. Nagel (eds.), Competitive Ethnic Relations. New York: Academic Press. Paige, Jeffrey M. 1975. Agrarian Revolutions. New York: Free Press. Pareto, Vilfredo. [1901] 1968. The Rise and Fall o f the Elites. Totowa, NJ: Bedminister Press. Park, Kang H. 1986. “Reexamination of the Linkage Between Income Inequality and Political Violence.” Journal o f Political and Military Sociology 14:185-97. Parsons, Talcott. 1978. Action Theory and the Human Condition. New York: Free Press. ------. 1964. "Evolutionary Universal."/Imencan 5bao/og/cct/Rev/evv 29:339-57. Parvin, M. 1973. “Economic Determinants of Political Unrest: An Econometric Approach.” Journal o f Conflict Resolution 17: 271-96. Piven, Frances F., and Richard A. Cloward. 1977. Poor People's Movements. New York: Pantheon. Raulet, Harry M. "Family Planning and Population Control in Developing Countries." Demography 7:211-234. 128 Reynat-Queroh, Marta. 20frh “Ethnicity; PoKticahSystems; and CivihWars;" Paper prepared for the conference. Identifying Wars: Systematic Conflict Research and its Utility in Conflict Resolution and Prevention, Uppsala University, June 8-10, 2001. Robinson, Thomas, and Bruce London. 1991. "Dependency, Inequality, and Political Violence: A Cross-national Analysis." Journal o f Political and Military Sociology 19: 305-308. Rostow, W.W. 1962. The Stages o fEconomic Growth: A Non-Communist Manifesto. Forge Village, MA: Cambridge University Press. Russet, Bruce. 1964a. World Handbook o fPolitical and Social Indicators. New Haven, CT: Yale University Press. ------. 1964b. "Inequality and Instability: The Relation of Land Tenure to Politics." World Politics 15:442-54. Rummel, R.J. 1994. Death by Government. New Brunswick, NJ: Transaction Publishers. Sambanis, Nicholas. 2001. "Do Ethnic and Nonethnic Civil Wars Have the Same Causes: A Theoretical and Empirical Inquiry (Part 1)." Journal o f Conflict Resolution 45:259-282. Sigelman, L., and M. Simpson. 1977. " A Cross-National Test of the Linkage Between Economic Inequality and Political Violence." Journal o f Conflict Resolution 21:105-28. Singer, J.D., and Melvin Small. 1994. Correlates o f War Project: International and Cm/fParDara, 6-/992 (Computer file). Ann Arbor, MI: Inter-University Consortium for Political and Social Research. ------. 1982. Resort to Arms. Beverly HtUsr Sage Publications. SIPRI. Various Years. SIPRI Yearbook. Oxford: Oxford University Press. Skocpol, Theda. 1994. Social Revolutions in the Modem World. New York: Cambridge University Press. ------. 1979. States and Social Revolutions. Cambridge: Harvard University Press. Smelser, Neil. 1968. Essavs in Sociological Explanation. Englewood Cliffs: Prentice- Hall. 129 19€3-. TheoryofColleetive Behavior New York: Free Press. Snyder, D. and Kick, E. 1979. "Structural Position in the World System and Economic Growth, 1955-1970: A MuWple-Network Analysis of Transnational Interactions." American Journal Sociology 84:1096-1126. So, Alvin. 1990. Social Change and Development: Modernization, Dependency, and World Systems Theory. Newbury Park, CA: Sage. Sollenberg, Margareta, Peter Wallensteen, and Andrés Jato. 1999. “Major Armed Conflicts.” Pp. 15-75 in SIPRI (ed.), SIPRI Yearbook 1999: Armaments, Disarmament, and International Security. New York: Oxford University Press. Soysa, Indra de. 2001. "Greed and Grievance: Data, Proxies, and Alternative Specifications." Paper prepared for the Conference on Data Collection in Armed Conflict,, Uppsala, Sweden, June 8-9, 2001. Spencer, Herbert. [1852] 1972. “Population and Progress.” Chapter 5 in J.D.Y. Peel (ed.), Herbert Spencer: On Social Evolution. Chicago, IL: University of Chicago Press. Stokes, Randall G., and Andy B. Anderson. 1990. “Disarticulation and Human Welfare in Less Developed Countries.” American Sociological Review 55:846-60. Summers, Robert, and Alan Heston. 1991. "The Penn World Tables (Mark 5): An Expanded Set of International Comparisons, 1950-1988." Quarterly Journal o f Economics (May):327-368. Supplemented to 1992 with data online directly from authors. Sundrum, R.M. 1990. Economic Growth in Theory and Practice. London: Macmillan. Taylor, Charles L., and Michael C. Hudson. 1972. World Handbook o f Political and Social Indicators, 2"** edition. New Haven: Yale University Press. Taylor, Charles L., and David A. Jodice. 1983. World Handbook o f Political and Social Indicators, 3"* edition. Vols. I and 2. New Haven: Yale University Press. Tilly, Charles. 1993. European revolutions, 1492-1992. Cambridge, MA: Blackwell. ------. 1978. From Mobiltation to Revolution. Reading: Addison-Wesley. Timberlake, Michael, and Kirk R. Williams. 1987. "Structural Position in the World- System, Inequality, and Political Violence." Journal o f Political and Military Sociology 15:1-15. 130 1984; “Depend^ieevPotitieatExeittsionr and Government Repression: Some Cross-National Evidence.” American Sociological Review 49:141-46. U.S. Arms Control and Disarmament Agency. Various years. World Military Expenditures and Arms Transfers. Washington, D.C.: U.S. Arms Control and Disarmament Agency. Vanhaiuien, Tatu. 1999. "Domestic Ethnic Conflict and Ethnic Nepotism: A Comparative Analysis." Journal o fPeace Research 36:55-73. ------. 1991. The Process of Democratization: A Comparative Study of 147 States, 1980-1988. New York: Crane Russak. Wallensteen, Peter, and Margarete Sollenberg. 2000. "Armed Conflicts, 1989-1999." Journal of Peace Research 37:635-649. Wallerstein, Immanuel. 1974. The Modem World System. New York: Academic Press. Walton, John, and Charles Ragin. 1990. "Global and National Sources of Political Protest: Third World Responses to the Debt Crisis." American Sociological Review 55:876-90. Weede, Erich. 1987. "Some New Evidence on Correlates of Political Violence: Income Inequality, Regime Repressiveness, and Economic Development." European Sociological Rev/ew 3:97-108. 1986. "Income Inequality and Political Violence Reconsidered." American Sociological Review 51:438-41. -. 1981. "Income Inequality, Average Income, and Domestic Violence." Journal o f Conflict Resolution 25:639-54. Weede, Erich, and Edward Mueller. 199R. "Rebellion, Violence, and Revolution: A Rational Choice Perspective." Journal o f Peace Research 35:43-59. Weston, Paul B., and Kenneth M. Wells. 1986. Criminal Investigations: Basic Perspectives. Fourth Edition. Englewood Cliffs, NJ: Regents/Prentice Hall. Williams, Robin M. 1994. “The Sociology of Ethnic Conflicts: Comparative International Perspectives.” Annual Review o fSociology 20:49-79. World Bank. 2001. "The Economics of Civil Wars, Crime, and Violence: Motivation." Description of World Bank program on the study of civil war. See http://v\Avvv.vvorldbank.org.^research/conlfict motivation.htm. 131 World Bank. Annual. Social Indicators o f Development. Baltimore. MD: Johns Hopkins University Press. Young, Crawford. 1976. The Politics o f Cultural Pluralism. Madison: University of Wisconsin Press. Zald, Mayer N., and R. Ash. 1966. “Social Movement Organizations.” Social Forces 44:327-41. Zimmerman, Ekkart. 1983. Political Violence, Crises, and Revolutions: Theories and Research. Cambridge, MA: Schenkman. 132 APPENDICES 133 APPENDIX A CIVIL WARS BETWEEN 1965 AND 1992- Dominican Republic (1965) Chad (1980-1988) Uganda (1966) ♦♦Nigeria (1980-1981) ^Guatemala (1966-1972) ♦Uganda (1980-1988) China (1967-1968) ♦♦Iran (1981-1982) #Nigeria (1967-1970) Peru (1982-Ongoing) #Burma (1968-1980) ♦♦Nicaragua (1982-1990) Kampuchea (Cambodia) (1970-1975) Somalia (1982-Ongoing) Jordan (1970) ♦♦Burma ( 1983-Ongoing) ♦♦Guatemala ( 1970-1971) ♦♦Sri Lanka Tamil Rebellion Paidstan (1971) (1983-Ongoing) Sri Lanka (Ceylon) (1971) t t Sudan (1983-Ongoing) Burundi (1972) ♦♦Nigeria (1984) Philippines New People's Army Colombia (1984-Ongoing) (1972-Ongoing) Iraq Kurdish/Shiite Rebellion tfZimbabwe (Rhodesia) (1972-1979) (1985-Ongoing) ♦♦Pakistan ( 1973-1977) India (1985-Ongoing) Lebanon (1975-1990) ♦South Yemen (1986) t t Angola ( 1975-1991 ) ♦♦Sri Lanka JVP ( 1987-1989) **Guatemala (1978-1984) ♦♦Burundi (1988) Afghanistan (1978-Ongoing) Liberia (1989-1990) Iran (1978-1979) Romania (1989) Nicaragua (1978^1979) Rwanda (1990-Ongoing) **Cambodia (1979-1991) ♦♦Burundi (1991-Ongoing) El Salvador (1979-1992) ttOeorgia ( 1991-Ongoing) f f Mozambique (1979-Ongoing) (Continued) ~ Bold = Wars available for analysis t Not independent in 1965. f t Not Independent in 1970. X Left censored in 1965 analysis Left Censored in 1970 analysis • Subsequent war in 1965 analysis ** Subsequent war in both the 1965 and 1970 analyses 134 Soviet Union: Nagorao-Karabakh (1991-Ongoing) Turkey Kurdish Rebellion (1991-Ongoing) Yugoslavia Croatian Independence (1991-1992) ttBosnia-Serbia Rebellion (1992-Ongoing) tfTajikistan (1992-Ongoing) **Liberia (1992-Ongoing) f t Angola ( 1992-Ongoing) 135 APPENDIX B RISK SET FOR THE 1965 ONSET ANALYSIS (N=l 16) Afghanistan Egypt Libya Sri Lanka Albania El Salvador Luxembourg Sweden Algeria Ethiopia Madagascar Switzerland Argentina Finland Malawi Syria Australia France Malaysia Taiwan Austria Gabon Mali Tanzania Belgium German Democratic Malta Thailand Benin Republic Mauritania Togo Bolivia Federal Republic of Mexico Trinidad and Tobago Brazil Germany Mongolia Tunisia Bulgaria Ghana Morocco Turkey Burkina Faso (Upper Greece Myanmar (Burma) Uganda Volta) Guatemala Nepal United Kingdom Burundi Guinea Netherlands United States Of Cambodia Haiti New Zealand America (Kampuchea) Honduras Nicaragua Uruguay Cameroon Hungary Niger USSR Canada Iceland Nigeria Venezuela Central African India Norway Vietnam Republic Indonesia Pakistan Yugoslavia Chad Iran Panama Zambia Chile Iraq Paraguay Chma Ireland Peru Colombia Israel Philippines Congo Italy Poland Costa Rica Jamaica Portugal Cote d’Ivoire (Ivory Japan Romania Coast) Jordan Rwanda Cuba Kenya Saudi Arabia Cyprus North Korea Senegal Czechoslovakia South Korea Sierra Leone Denmark Kuwait Somalia Dominican Republic Lebanon South Africa Ecuador Liberia Spain 136 APPENDIX C RISK SET FOR THE 1970 ONSET ANALYSIS (N=126) Afghanistan Dominican Republic South Korea Senegal Albania Ecuador Kuwait Sierra Leone Algeria Egypt Lebanon Singapore Argentina El Salvador Lesotho Somalia Australia Equatorial Guinea Liberia South Africa Austria Ethiopia Libya Spain Barbados Finland Luxembourg Sri Lanka Belgium France Madagascar Sudan Benin Gabon Malawi Swaziland Bolivia Gambia Malaysia Sweden Botswana German Democratic Maldives Syria Brazil Republic Mali Taiwan Bulgaria Federal Republic of Malta Tanzania Burkina Faso (Upper Germany Mauritania Thailand Volta) Ghana Mauritius Togo Burundi Greece Mexico Trinidad and Tobago Cambodia Guinea Mongolia Tunisia (Kampuchea) Guyana Morocco Turkey Cameroon Haiti Nepal Uganda Canada Honduras Netherlands United Kingdom Central Afidcan Hungary New Zealand United States Of Republic Iceland Nicaragua America Chad India Niger Uruguay Chile Indonesia Norway USSR China Iran Pakistan Venezuela Colombia Iraq Panama Vietnam Congo Ireland Paraguay South Yemen Costa Rica Israel Peru North Yemen Cote d’Ivoire (Ivory Italy Philippines Yugoslavia Coast) Jamaica Poland Zaire Cuba Japan Portugal Zambia Cyprus Jordan Romania Czechoslovakia Kenya Rwanda Denmark North Korea Saudi Arabia 137 APPENDIX D FIGURE 4 0.035 0.030 0.025 0.020 0.015 0.010 0.005 0.000 1965 Analysts 1970 AnafysB Figure 4: Hazard function plots of the transition rate from peace to civil war from 1965- 1992 and 1970-1992 138 APPENDIX E TABLE 2 139 1965 Analysis 1970 Analysis Description Source // SE p N Events SIg. SE p N Events Sig. LAGGED DEPENDENT VARIABLES ------1 Number of previous civil wars 1800- 1992 Singer and Si)i 0.163 0.084 0.053 116 29 * 0.061 0.093 0.511 126 25 ns Years since last civil war Singer and Small i «3 -0.046 0.036 0.206 116 29 ns -0.030 0.035 0.389 126 25 ns Previous civil wars? (yes/no) Singer and Small 1993 0.826 0.372 0.026 116 29 ** 0.872 0.401 0.030 126 25 ** INDICATORS OF STRUCTURAL MODERNIZATION United Nations, various Log energy consumption/capita 1965 years -0.303 0.102 0.003 111 28 *** -0.116 0.052 0.027 120 24 ** United Nations, various Log energy consumption/capita years -0.235 0.075 0.002 116 28 *** -0.227 0.081 0.005 126 24 *** Sunamers and Heston Log GDP/capita 1965 1991 -0.606 0.221 0.006 102 26 ♦ ♦♦ -0.475 0.232 0.041 112 21 4>« Summers and Heston Log GDP/capita 1991 -0.557 0.206 0.007 116 25 *** -0.444 0.224 0.048 126 20 ** Percent labor force in industiy 1965 ILO1986 -0.044 0.016 0.007 112 28 *** -0.038 0.017 0.027 120 23 ** ' Percent labor force in industry ILO1986 -0.041 0.016 0.011 116 28 ** -0.036 0.018 0.041 126 23 V* ' Percent GDP from industiy 1965 World Bank 1994 -0.038 0.020 0.049 72 18 ** -0.036 0.020 0.078 87 17 * ' Percent GDP from industiy World Bank 1994 -0.015 0.018 0.422 116 19 ns -0.010 0.019 0.612 126 17 ns ILO 1986, World Bank Industrial dominance 1965 1994 -0.042 0.022 0.060 72 18 * -0.047 0.023 0.039 87 17 ** ILO 1986, World Bank Industrial dominance 1994 -0.024 0.020 0.244 116 19 ns -0.022 0.022 0.321 126 17 ns World Bank 1994, Summers & Heston Industrial sector real GDP/capita 1965 1991 -0.000 0.000 0.040 69 17 ** 0.000 0.000 0.054 84 16 * Table 2: Description of independent variables and bivariate statistics (Continued) Table 2: Continued 1965 Analysis 1970 Analysis Description Source /y SE p N Events Sig. // SE p N Events Sig. STRUCTURAL MODERNIZATION (Cont.t World Bank 1994, Summers & Heston Industrial sector real GDP/capita 1991 -0.000 0.000 0.026 116 18 ** 0.000 0.000 0.048 126 16 ** World Bank 1994, Log industrial sector real Summers & Heston RGDP/capita 1965 1991 -0.803 0.360 0.026 69 17 «* -0.403 0.243 0.098 84 16 * World Bank 1994, Log industrial sector real Summers & Heston RGDP/capita 1991 -0.899 0.323 0.005 116 18 *** -0.691 0.267 0.010 126 16 ♦♦♦ Annual average percentage change in Summers and Heston GDP/capita 1991 -0.110 0.066 0.097 116 25 * -0.129 0.072 0.073 126 20 * INDICATORS OF TECHNO-ECOLÇGICAL INHERITANCE Horticultural techno-ecological inheritance? (yes/no) Lenski and Nolan 1984 0.247 0.458 0.589 116 29 ns 0.146 0.500 0.770 126 25 ns New agrarian techno-ecological inheritance? (yes/no) Lenski and Nolan 1984 0.203 0.434 0.640 116 29 ns -0.024 0.500 0.961 126 25 ns Old Agrarian techno-ecological inheritance? (yes/no) Lenski and Nolan 1984 -0.064 0.401 0.874 116 29 ns -0.090 0.445 0.839 126 25 ns Agricultural population densi^ I960 Taylor and Hudson 1972 -0.006 0.008 0.438 110 29 ns -0.009 0.009 0.322 117 25 ns Log agricultural population density I960 Taylor and Hudson 1972 0.114 0.137 0.407 110 29 ns -0.010 0.131 0.942 117 25 ns Crenshaw, Unpublished Agricultural population density I960 data 0.002 0.004 0.647 90 24 ns 0.000 0.004 0.922 94 20 ns Log agricultural population density Crenshaw, Unpublished I960 data 0.322 0.153 0.035 90 24 ** 0.213 0.155 0.169 94 20 ns (Continue*)) Table 2: Continued 1965 Analysis 1970 Analvsis Description Source n SE p N Events Sig. /y SE p N Events Sig. TECHNO-ECOLOGICAL INHERITANCE (Conl.t ' United Nations 1998, Total population density I960 FAO 1995 -0.067 0.205 0.744 110 27 ns -0.116 0.197 0.556 118 22 ns United Nations 1998, ' Total population density FAO 1995 0.022 0.146 0.881 116 27 ns 0.045 0.107 0.672 126 22 ns United Nations 1998, Log total population density I960 FAO 1995 0.157 0.132 0.234 110 27 ns 0.077 0.132 0.562 118 22 ns United Nations 1998, Log total population density FAO 1995 0.201 0.187 0.155 116 27 ns 0.130 0.137 0.343 126 22 ns United Nations 1998, ' Ruralpopulation densi^ I960 fAO 1995 0.056 0.1 II 0.616 105 25 ns -0.026 0.061 0.668 113 22 ns United Nations 1998, Rurol^opuiation density FAO 1995 0.074 0.103 0.473 116 25 ns -0.017 0.047 0.714 126 21 ns United Nations 1998, Log rural density I960 FAO 1995 0.239 0.142 0.093 105 25 * 0.133 0.132 0.313 113 21 ns United Nations 1998, ’ Log rural density FAO 1995 0.281 0.147 0.055 116 25 ♦ 0.153 0.128 0.232 126 21 ns Taylor and Hudson Population density * total population 1972, World Bank 1986 0.000 0.000 0.147 90 24 ns 0.000 0.000 0.147 94 20 ns Created from Taylor and Techno-ecological inheritance Hudson 1972 0.057 0.035 0.102 105 25 ns 0.066 0.037 0.073 113 21 * Rural population pressure United Nations 1998 0.080 0.038 0.035 116 25 ** 0.015 0.022 0.483 126 21 ns Total population I960 United Nations 1998 0.004 0.002 0.004 110 27 *** 0.001 0.002 0.586 118 22 ns Total population United Nations 1998 0.004 0.001 0.002 116 27 ♦♦♦ 0.001 0.001 0.430 126 22 ns Log total population I960 United Nations 1998 0.229 0.129 0.075 110 27 * 0.179 0.126 0.155 118 22 ns Log total population United Nations 1998 0.273 0.130 0.035 116 27 ** 0.219 0.126 0.083 126 22 * Crenshaw, Unpublished British colonial heritage (yes/no) data 0.553 0.434 0.203 116 29 ns 0.058 0.500 0.908 126 25 ns (Continued) Table 2: Continued 1965 Analysis 1970 Analvsis Description Source /f SE p N Events Sig. fl SE p N Events Sig. TECHNO-ECOLOGICAL INHERITANCE (Cont.) Crenshaw, Unpublished French colonial heritage (yes/no) data -0.346 0.610 0.570 116 29 ns -0.010 0.615 0.987 126 25 ns Years since independence CIA 1994 -0.041 0.025 0.108 116 29 ns -0.035 0.027 0.189 126 25 ns INDICATORS OF SERVICE SECTf^R STRENGTH Percent labor force in services 1965 ILO 1986 -0.035 0.014 0.011 112 28 ** -0.027 0.014 0.053 120 23 * Percent labor force in services ILO 1986 -0.028 0.012 0.015 116 28 ♦♦ -0.025 0.013 0.046 126 23 Percent GDP from services 1965 World Bank 1994 -0.036 0.021 0.087 72 18 * -0.069 0.022 0.002 87 17 *** Percent GDP from services World Bank 1994 -0.056 0.020 0.005 116 19 -0.073 0.021 0.001 126 17 ILO 1986, World Bank Service dominance 1965 1994 -0.040 0.021 0.054 72 18 * -0.061 0.022 0.005 87 17 *** w ILO 1986, World Bank Service dominance 1994 -0.049 0.019 0.008 116 19 *** -0.060 0.021 0.003 126 17 ***; World Bank 1994, Summers & Heston Service sector real GDP/capita 1991 -0.000 0.000 0.019 116 18 ** 0.000 0.000 0.026 126 16 ** World Bank 1994, Log service sector real RGDP/capita Summers & Heston 1965 1991 -0.746 0.421 0.077 69 17 • -0.936 0.427 0.028 84 16 ** World Bank 1994, Summers & Heston Log service sector real RGDP/capita 1991 -1.054 0.387 0.006 116 18 *** -1.081 0.405 0.008 126 16 *** INDICATORS OF ETHNICITY World Christian Size largest ethnic group Encyclopedia 1982 -0.011 0.007 0.129 115 29 ns -0.009 0.008 0.214 125 25 ns World Christian Size 2nd largest ethnic jgroup Encyclopedia 1982 0.021 0.017 0.208 115 29 ns 0.025 0.018 0.173 125 25 ns (Continued) Table 2: Continued 1965 Analysis 1970 Analvsis Description Source A SE p N Events Sig. ft SE p N Events Sig. ETHNICITY rConl.t Proportion 2nd largest ethnic World Christian group/largest ethnic group Encyclopedia 1982 0.006 0.006 0.278 115 29 ns 0.008 0.006 0.247 125 25 ns Is 2nd largest ethnic group = 10-30% World Christian of largest ethnic group? (yes/no) Encyclopedia 1982 0.866 0.402 0.031 116 29 ** 0.148 0.500 0.768 125 25 ns Etiinic homogeneity Vanhannen 1991 -0.007 0.007 0.328 114 29 ns -0.005 0.008 0.546 123 25 ns Ethno-linguistic fmclionalization Taylor and Hudson 1972 0.002 0.006 0.696 92 28 ns 0.001 0.007 0.838 100 24 ns Etluiic heterogenei^ Vanhannen 1999 0.015 0.005 0.005 110 27 *** 0.014 0.006 0.014 118 22 ** Separatist intensity I960 Taylor and Jodice 1983 0.446 0.138 0.001 79 22 *** 0.521 0.155 0.001 78 18 ♦iV ’ Separatist intensity 1975 Taylor and Jodice 1983 0.528 0.140 0.000 78 21 *** 0.550 0.154 0.000 77 17 Separatism * population size I960 Taylor and Jodice 1983 0.032 0.009 0.000 79 22 *** 0.035 0.010 0.001 78 18 Separatism * population size 1975 Taylor and Jodice 1983 0.031 0.009 0.001 78 21 *** 0.028 0.010 0.008 77 17 Ethnic separatist history * group size Qurr 1993 0.386 0.293 0.187 78 27 ns 0.436 0.315 0.167 78 23 ns Ethnic histoiy of lost autonomy Qurr 1993 0.104 0.034 0.002 78 27 *** 0.082 0.042 0.051 78 23 * ' Ethnic history of lost autonomy * group size Quit 1993 1.255 0.533 0.019 78 27 1.032 0.592 0.082 78 23 * POLITICAL ECONOMY INDICATORS Snyder and Kick 1979, Core world system position Dollen 1983 -1.657 1.018 0.104 116 29 ns -1.368 1.021 0.180 126 25 ns Snyder and Kick 1979, Semi-periphery world system position Qollen 1983 1.127 0.374 0.003 116 29 *** 1.481 0.401 0.000 126 25 *** Snyder and Kick 1979, Periphery world system position Bollen 1983 0.327 0.372 0.379 116 29 ns 0.177 0.403 0.661 126 25 ns Ballmer-Cao and Foreign owned stock/total stock 1967 Scheidegger 1979 -1.320 2.197 0.548 89 24 ns -1.595 2.493 0.522 87 20 ns Ballmer-Cao and Foreign owned stock/total stock 1973 Scheidegger 1979 -1.684 2.230 0.450 86 23 ns -2.395 2.741 0.382 84 19 ns (Continued) Table 2: Continued 1965 Analvsis 1970 Analvsis Description Source A SE p N Events Sin. A SE p N Events Sig. POLITICAL ECONOMY (Conl.) Ballmer-Cao and Foreign owned stock/capita 1967 Scheidegger 1979 -12.649 6.273 0.044 92 25 ♦ ♦ -12.602 6.786 0.063 91 21 * Ballmer-Cao and Foreign owned stock/capita 1973 Scheidegger 1979 -6.906 4.777 0.148 73 24 ns -8.227 5.571 0.140 72 20 ns Foreign owned stock/capita 1978 UNCTNC -2413 2127 0.256 70 22 ns -2645 2191 0.228 76 19 ns Foreign owned stock/GDP * total Ballmer-Cao and population 1967 Scheidegger 1979 -5.608 4.124 0.174 89 24 ns -5.846 4.641 0.208 87 20 ns Foreign owned stock/GDP * total Ballmer-Cao and population 1973 Scheidegger 1979 -5.989 4.447 0.178 67 22 ns -7.698 5.637 0.172 65 18 ns Ballmer-Cao and Foreign owned stock/GDP 1967 Scheidegger 1979 -814833 1350961 0.546 80 19 ns -830359 1534912 0.589 78 15 ns t Ballmer-Cao and Foreign owned stock/GDP 1973 Scheidegger 1979 -531227 12595II 0.673 78 18 ns -598951 1436475 0.677 78 15 ns Foreign owned stock/GDP 1978 UNCTNC -209219 976439 0.830 61 19 ns -63778 1023862 0.950 67 16 ns Exports/GDP 1965 World Bank 1994 -0.034 0.018 0.057 88 21 * -0.042 0.019 0.027 96 18 ♦♦ Exports/GDP World Bank 1994 -0.033 0.017 0.049 116 21 ♦♦ -0.041 0.018 0.024 126 18 ** Annual average percentage change in exjports/GDP 1965 World Bank 1994 0.027 0.021 0.187 81 20 ns -0.092 0.064 0.147 94 18 ns Annual average percentage change in exports/GDP World Bank 1994 0.040 0.029 0.160 116 17 ns 0.036 0.031 0.244 126 17 ns Agricultural exports/total exports 1965 Taylor and Jodice 1983 0.013 0.006 0.024 116 29 0.0! I 0.006 0.070 126 25 * Agricultural exports/total exports 1980 Taylor and Jodice 1983 0.012 0.006 0.047 109 29 ** 0.014 0.007 0.036 118 25 ** Commodity concentration 1965 Taylor and Jodice 1983 0.016 0.006 0.010 116 29 ** 0.016 0.007 0.015 126 25 ** Commodity concentration 1980 Taylor and Jodice 1983 0.010 0.006 0.071 103 29 * 0.010 0.006 0.120 110 24 ns Summers and Heston Trade Openness 1991 -0.023 0.009 0.014 116 25 ** -0.021 0.009 0.021 126 20 ** (Continued) Table 2: Continued 1965 Analysis 1970 Analvsis Description Source A SE p N Events Sig. A SE p N Events Sig, POLITICAL ECONOMY tConl.t Total debt/GDP World Bank 1994 -1.761 1.184 0.137 116 10 ns -1.792 1.213 0.140 126 10 ns Annual average percentage change in total debt/GDP World Bank 1994 -0.003 0.017 0.872 116 9 ns -0.004 0.017 0.800 126 9 ns Long-term debt/GDP World Bank 1994 -2.017 1.373 0.142 116 10 ns -2.069 1.404 0.141 126 10 ns Annual average percentage change in long-term debt/GDP World Bank 1994 -0.003 0.019 0.869 116 9 ns -0.004 0.019 0.828 126 9 ns Short-term debt/GDP World Bank 1994 -4.830 5.394 0.371 116 10 ns -4.267 4.985 0.392 126 10 ns" Annual average percentage change in short-term debt/GDP World Bank 1994 -0.000 0.000 0.873 116 9 ns 0.000 0.000 0.889 126 9 ns Debt service/GDP World Bank 1994 -6.338 8.236 0.442 116 10 ns -6.104 8.369 0.466 126 10 ns^ Annual average percentage change in debt service/GDP World Bank 1994 -0.004 0.012 0.731 116 9 ns -0.005 0.012 0.692 126 9 ns U.S. Arms Control and Pisarmament Agency, Total military dependence 1970-1985 various years 0.000 0.000 0.609 77 24 ns 0.000 0.000 0.312 84 21 ns U.S. Arms Control and Soviet military dependence 1970- Disarmament Agency, 1985 various years 0.000 0.000 0.611 78 24 ns 0.000 0.000 0.295 85 20 ns (J.S. Arms Control and Disarmament Agency, US military dependence 1970-1985 various years -0.016 0.023 0.496 80 25 ns -0.009 0.022 0.679 87 21 ns U.S. Arms Control and Western military dependence 1970- Disannament Agency, 1985 various years -0.000 0.000 0.475 80 25 ns 0.000 0.000 0.688 87 21 ns INDICATORS OF ABSOLUTE DEPRIVATION Infant mortality rate United Nations 1998 0.010 0.004 0.008 116 27 *** 0.010 0.004 0.014 126 22 ** Life expectancy at birth United Nations 1998 -0.046 0.017 0.007 116 27 *** -0.042 0.019 0.024 126 22 ** (Continued) Table 2: Continued 1965 Analysis 1970 Analysis Description Source SE p NEvents Sig. P SE p N Events Sig. ABSOLUTE DEPRIVATION (Cont.) Caloric intake per capita World Bank SID 1992 -o.Vdi 0.041 0.013 116 26 ** -0.086 0.043 0.046 126 24 ** Protein intake per capita World Bank SID 1992 -0.252 0.108 0.019 116 25 ** -0.195 0.108 0.071 126 24 * ' Growth in infant mortality rate United Nations 1998 0.306 0.149 0.040 116 27 ** 0.434 0.172 0.012 126 22 ** Growth in life expectancy at birth United Nations 1998 0.992 0.470 0.035 116 27 #* 0.172 0.485 0.724 126 22 ns Annual average percentage change in caloric intake per capita World Bank SID 1992 0.137 0.160 0.393 116 19 ns 0.089 0.164 0.587 126 20 ns Annual average percentage change in protein intake per capita World Bank SID 1992 -0.002 0.053 0.967 116 19 ns -0.006 0.057 0.913 126 20 ns INDICATORS OF RELATIVE DEPRIVATION Intensity of economic discrimination I960 Taylor and Jodice 1983 0.148 0.177 0.699 79 22 ns 0.141 0.197 0.473 78 18 ns ' Intensity of economic discrimination 1975 Taylor and Jodice 1983 0.031 0.187 0.868 78 21 ns -0.005 0.213 0.982 77 17 ns Economic discrimination/population I960 Taylor and Jodice 1983 0.005 0.008 0.573 79 22 ns -0.002 0.010 0.812 78 18 ns Economic discrimination/population 1975 Taylor and Jodice 1983 0.007 0.010 0.494 78 21 ns -0.005 0.013 0.676 77 17 ns Intensity of political discrimination I960 Taylor and Jodice 1983 0.313 0.189 0.097 79 22 * 0.363 0.208 0.081 78 18 * Intensity of political discrimination 1975 Taylor and Jodice 1983 0.211 0.164 0.199 78 21 ns 0.311 0.176 0.078 77 17 * Proportion of political discrimination I960 Taylor and Jodice 1983 O.OII 0.008 0.168 79 22 ns 0.001 0.010 0.928 78 18 ns Proportion of political discrimination 1975 Taylor and Jodice 1983 0.013 0.008 0.081 78 21 * 0.007 0.009 0.425 77 17 ns Ethnic economic discrimination Gurr 1993 0.064 0.042 0.131 78 27 ns 0.090 0.044 0.042 78 23 ** Ethnic political discrimination Gurr 1993 0.069 0.032 0.031 77 27 ♦ * 0.076 0.034 0.028 77 23 ** (Continued) Table 2: Continued 1965 Analysis 1970 Analvsis Description Source SE p N Events Sig. A SE p N Events Sig.... , RELATIVE DEPRIVATION (ContA Ethnie economic discrimination * g-oup population Ourr 1993 0.218 0.249 0.381 78 27 ns 0.270 0.268 0.313 78 23 ns Ethnie political discrimination * group population Gurr 1993 0.100 0.234 0.671 77 27 ns 0.156 0.250 0.532 78 23 ns Ethnie economic j^evance Ourr 1993 0.058 0.032 0.071 78 27 ♦ 0.063 0.034 0.064 78 23 * Ethnicpolitical grievance Gurr 1993 0.048 0.015 0.001 78 27 *** Ô.Ô49 0.016 0.002 78 23 *** Ethnic social grievance Ourr 1993 0.061 0.018 0.001 78 27 *** 0.064 0.019 0.001 78 23 *** Ethnic grievance (all) Ourr 1993 0.023 0.007 0.001 78 27 *** 0.024 0.008 0.002 78 23 *** Ethnic economic grievance • group population Ourr 1993 0.240 0.179 0.179 78 27 ns 0.294 0.189 0.119 78 23 ns Ethnic political grievance * group population Ourr 1993 0.159 0.109 0.146 78 27 ns 0.193 0.115 0.092 78 23 * Ethnic social grievance * group population Ourr 1993 0.777 0.204 0.000 78 27 *** 0.984 0.236 0.000 78 23 *** Ethnic grievance (all) * group population Gurr 1993 0.123 0.058 0.033 78 27 ** 0.148 0.062 0.016 78 23 Ethnic demographic distress Ourr 1993 0.030 0.014 0.035 77 27 ** 0.038 0.015 0.012 77 23 ** Ethnic demographic distress * group population Ourr 1993 0.123 0.073 0.092 77 27 * 0.151 0.078 0.052 77 23 * INDICATORS OF INEOIIALITV Gini income inequality 1970 Hoover 1989 0.007 0.033 0.835 54 10 ns 0.023 0.036 0.524 55 9 ns Muller 1988, Simpson Gini income inequality 1970 1990 0.040 0.033 0.223 60 12 ns 0.040 0.034 0.233 60 11 ns Gini sectoral inequality 1965 Taylor and Hudson 1972 0.010 0.012 0.402 98 25 ns 0.008 0.013 0.530 104 21 ns Gini sectoral inequality 1970 Taylor and Hudson 1972 0.002 0.012 0.846 98 23 ns 0.000 0.013 0.996 105 19 ns Income share of bottom 20% 1970 Hoover 1989 -0.005 0.152 0.972 54 10 ns -0.088 0.171 0.605 55 9 ns Income share of top 20% 1970 Hoover 1989 0.007 0.032 0.833 54 10 ns 0.021 0.034 0.537 55 9 ns (Continuée)) Table 2; Continued 1965 Analysis 1970 Analvsis Description Source SE p N Events Sig. SE p N Eveitts Sig, INEQUALITY IConl.» Income share of tojp 20% 1970 Simpson 1990 0.048 0.032 0.138 60 12 ns 0.044 0.034 0.186 60 II ns Income share of fop 20% 1970 Taylor and Jodice 1983 0.009 0.023 0.688 73 17 ns 0.019 0.026 0.468 74 14 ns Gini land inequality Taylor and Hudson 1972 0.014 0.013 0.264 81 17 ns 0.009 0.013 0.501 86 16 ns Peininger and Squire Gini income Inequality 1996 0.007 0.022 0.764 79 20 ns -0.006 0.025 0.800 83 16 ns INDICATORS OF "GREED" Exports of Raw Agricultural Products/GDP WDI2000 -38.089 49.321 0.440 116 5 ns -35.147 47.113 0.456 126 5 ns Exports of Fuel/GDP WDI2000 -13.384 16.388 0.414 116 5 ns -12.383 15.675 0.430 126 5 ns Exports of Raw Ores and Minerals/GDP WDI2000 -3.366 11.233 0.764 116 5 ns -3.686 11.166 0.741 126 5 ns Exports of Primary Products/GDP WDI2000 -10.094 10.316 0.328 116 5 ns -8.897 9.391 0.343 126 5 ns DEMOGRAPHIC INDICATORS Annual average percentage change of total population United Nations 1998 0.233 0.081 0.004 116 27 *** 0.280 0.102 0.006 126 22 Annual average percentage change in population 0-14/total population United Nations 1998 0.166 0.064 0.009 116 27 *** 0.195 0.074 0.009 126 22 Annual average percentage change in population 15-64/total population United Nations 1998 0.272 0.089 0.002 116 27 *** 0.317 0.120 0.009 126 22 Population 15-29/population 30-64 United Nations 1998 0.032 O.OII 0.003 116 27 ♦ ♦♦ 0.029 O.OII O.OII 126 22 ♦* Male population IS-29/male population 30-64 United Nations 1998 0.062 0.019 0.001 116 27 *** 0.064 0.023 0.006 126 22 Annual average percentage change in population 15-29/population 30-64 United Nations 1998 0.087 0.156 0.575 116 27 ns 0.138 0.176 0.434 126 22 ns (Continued) Table 2: Continued 196S Analvsis 1970 Analvsis Description Source A SE p N Events Sig. A SE p N Events Sig. DEMOGRAPHIC INDICATORS (Cont.) Annual average percentage change in male population 15-29/male population 30-64 United Nations 1998 0.057 0.153 0.710 116 27 ns 0.102 0.171 0.549 126 22 ns Population 15-29/total population United Nations 1998 0.190 0.090 0.035 116 27 ♦ ♦ 0.124 0.095 0.192 126 22 ns Male population 15-29/total population United Nations 1998 0.248 0.116 0.032 116 27 ** 0.241 0.168 0.152 126 22 ns Annual average percentage change in population 15-29/total population United Notions 1998 0.104 0.223 0.642 116 27 ns 0.069 0.245 0.779 126 22 ns Annual average percentage change in male population 15-29/male population 30-64 United Nations 1998 0.039 0.209 0.851 116 27 ns 0.007 0.228 0.975 126 22 ns o Annual average percentage change in population 15-29 United Nations 1998 0.269 0.091 0.003 116 27 *** 0.335 0.133 0.012 126 22 ♦ ♦ Annual average percentage change in male population 15-29 United Nations 1998 0.269 0.097 0.006 116 27 *** 0.295 0.126 0.019 126 22 Dependency ratio (population 0-14 and 65+/population 15-64) United Nations 1998 0.036 0.013 0.005 116 27 ♦♦♦ 0.036 0.014 O.OII 126 22 ** Child Dependency ratio (population 0-I4^opulation 15-64) United Notions 1998 0.035 0.012 0.003 116 27 *** 0.034 0.013 0.008 126 22 Dependency ratio (population 0-14 and 65+/population 15-64 - calculated) United Nations 1998 0.004 0.001 0.005 116 27 *** 0.004 0.001 O.OII 126 22 ♦♦ INDICATORS OF SOCIAL MOBILIZATION Secondary educational enrollment World Bank 1994 -0.020 0.009 0.024 116 22 ** -0.017 0.009 0.064 126 18 * Annual average percentage change in secondary educational enrollment World Bank 1994 -0.008 0.030 0.799 116 22 ns -0.001 0.039 0.985 126 18 ns Mean years of education Barro and Lee 1996 -0.227 0.089 O.OII 116 22 ** -0.183 0.093 0.049 126 19 ** (ContinuetJ) Table 2: Continued 1965 Analysis 1970 Analysis Description Source n SE p N Events Sig. ff SE p N Events Sig. SOCIAL MOBILIZATION (Cont.) Mean years male education Barro and Lee 1996 -0.219 0.090 0.015 116 22 ** -0.163 0.093 0.081 126 19 * Percent urban United Nations 1991 -0.022 0.008 0.010 116 29 *** -0.016 0.009 0.069 126 25 * Annual average percentage change of urban population United Nations 1991 0.134 0.055 0.015 116 27 ** 0.066 0.050 0.187 126 22 ns Annual average percentage change in percent urban United Nations 1991 0.116 0.085 0.175 116 29 ns 0.008 0.090 0.927 126 25 ns INDICATORS OF GOVERNMENT CHARACTERISTICS Gurr 1990, Jaggers and Democracy Gurr 1995 -0.048 0.048 0.311 116 27 ns -0.065 0.052 0.214 126 24 ns Gurr 1990, Jaggers and Ul Autocracy Gurr 1995 0.005 0.053 0.923 116 27 ns 0.030 0.057 0.606 126 24 ns Gurr 1990, Jaggers and Democracy-Autocracj' Gurr 1995 -0.014 0.024 0.577 116 29 ns -0.024 0.027 0.370 126 24 ns Civil liberties index Gastil200l 0.162 0.124 0.194 116 17 ns 0.188 0.122 0.121 126 19 ns Political Rights index Gastil 2001 0.025 0.108 0.820 116 16 ns 0.054 0.105 0.608 126 18 ns Repression Index Gastil 2001 0.082 0.125 0.513 116 16 ns 0.115 0.122 0.346 126 18 ns Scope Qurr 1990 0.205 0.115 0.074 116 20 # 0.336 0.132 O.OII 126 17 ** Centrism Gurr 1990 0.275 0.286 0.336 116 20 ns -0.026 0.399 0.948 126 17 ns Gurr 1990, Jaggers and Coherence Ourr 1995 -1.074 0.380 0.005 116 29 *** -0.874 0.409 0.033 126 25 ** Government sanctions I960 Taylor and Jodice 1983 9481 10319 0.358 III 27 ns 12428 11543 0.282 120 22 ns Log Government sanctions 1960 Taylor and Jodice 1983 9481 10319 0.358 III 27 ns 12428 11543 0.282 120 22 ns Government sanctions 1965 Taylor and Jodice 1983 8687 27054 0.748 111 27 ns -12822 27514 0.641 120 22 ns Log Government sanctions 1965 Taylor and Jodice 1983 8687 27054 0.748 111 27 ns -12822 27514 0.641 120 22 ns Government sanctions 1970 Taylor and Jodice 1983 22836 36815 0.535 III 27 ns 44669 38212 0.242 120 22 ns Log Government sanctions 1970 Taylor and Jodice 1983 22836 36815 0.535 III 27 ns 44669 38212 0.242 120 22 ns Government sanctions 1975 Taylor and Jodice 1983 -27301 39490 0.489 111 27 ns -22324 40204 0.579 120 22 ns (Continued) Table 2: Continued 1965 Analysis 1970 Analysis Description Source P SE p N Events Sig. P SE p N Events Sig. GOVERNMENT CHARACTERISTICS (ContA Log Government sanctions 1975 Taylor and Jodice 1983 -27301 39490 0.489 111 27 ns -22324 40204 0.579 120 22 ns Change in polity Ourr 1990 0.236 0.209 0.258 116 22 ns 0.024 0.314 0.940 126 19 ns Changes in executive Gurr 1990 0.557 0.355 0.116 116 23 ns 0.611 0.429 0.154 126 19 ns Regime transition Gurr 1990 2.343 0.561 0.000 116 24 *** 0.232 1.028 0.822 126 19 ns PREVIOUS EXPERIENCES WITH COLLECTIVE ACTION Number of previous international wars 1800-1992 Singer and Small 1993 0.007 0.022 0.746 116 29 ns 0.010 0.024 0.676 126 25 ns Number of previous wars (all) Singer and Small 1993 0.013 0.019 0.504 116 29 ns 0.011 0.021 0.600 126 25 ns Years since last international war Singer and Small 1993 0.033 0.082 0.689 116 29 ns 0.000 0.085 0.995 126 25 ns Years since last war (any) Singer and Small 1993 -0.051 0.106 0.629 116 29 ns -0.027 0.096 0.778 126 18 ns Previous international war? (yes/no) Singer and Small 1993 0.339 0.373 0.365 116 29 ns 0.718 0.408 0.079 126 25 * Previous wars (any)? (yes/no) Singer and Small 1993 0.436 0.391 0.265 116 29 ns 0.844 0.446 0.058 126 25 * " Non-violent ethnic protests (in year X)...... Gurr 1993 0.064 0.020 0.001 116 24 *** 0.064 0.020 0.002 126 20 *** Violent ethnicprotests (in year X) Gurr 1993 0.075 0.025 0.002 116 26 *** 0.075 0.025 0.003 126 22 *** Ethnic rebellions (in year X) Gurr 1993 0.146 0.038 0.000 116 26 *** 0.132 0.041 0.001 126 22 *** All ethnic conflict (in year X) Gurr 1993 0.036 0.012 0.003 116 24 *** 0.036 0.012 0.003 126 20 *** Coup d'etats (in year X) Gurr 1990 0.902 0.698 0.196 116 23 ns 0.760 1.037 0.463 126 19 ns Cumulative coup d'etats (in year X) since 1945 Gurr 1990 0.482 0.301 0.110 116 23 ns 0.363 0.386 0.347 126 19 ns INDICATORS OF MILITARIZATION U.S. Arms Control and Disarmament Agency, Soldiers/capita various years -0.008 0.025 0.744 116 28 ns 0.008 0.024 0.734 126 23 ns (Continued) Table 2: Continued 1965 Analysis 1970 Analysis Description Source P SE p N Events Sis. /y SE p N Events Sis^ MILITARIZATION (Cont.) U.S. Arms Control and Average annual percent change in pisarmament Agency, soldiers/capita various years -0.001 0.003 0.813 116 27 ns -0.005 0.016 0.763 126 22 ns U.S. Arms Control and Pisarmament Agency, Defense expenditures/GPP various years 0.059 0.029 0.044 116 29 ** 0.063 0.030 0.037 126 24 U.S. Arms Control and Average annual percent change in Pisarmament Agency, defense expenditures/GPP various years -0.003 0.014 0.840 116 27 ns -0.008 0.016 0.646 126 23 ns %; APPENDIX F: CORE SECTION OF THE MAXIMUM MODEL DATA ANALYSIS PROGRAM %IF &FORCE=0 %THEN %D0; %DO i=l %TO &CPTS; ?oDO ii=l %TO &&INDSC&i; %LET CPT&i=&&C&i.INDⅈ %IF &SIZE=1 «îbTHEN %D0: %DO iii=I %TO &CPTS: %T0 &CPTS; %IF &&YC&iii.C&iij=YES %THEN %D0; “oIF %EVAL(&iii=&i) %THEN %D0: "oIF %EVAL(&J=&i) “/oTHEN %D0; %LET MULT=YES; "oMPLUSREG; DATA WORK.A; SET &DATASET; %IF &METHOD=REG %THEN %D0; C&iü.C&üj=&&CPT&iii*&&CPT&jüj; «ÔEND; »'oMETHOD: MODEL &DV=&&CPT&i C&iii.C&jyj; %MPLUS: 'oFF &METHOD=PHREG %THEN %DO: C&iii.C&jjj=&&CPT&iii*&&CPT&jjj; %END; %DSOUT; %MERGE; %END; %END: SEND; SEND; SEND; SLET MULT=NO; SMPLUSREG; DATA WORK.A; SET&DATASET; SMETHOD; MODEL &DV=&&CPT&i; SMPLUS; SDSOUT; SMERGE; RUN; SEND; SEND; END; 154 APPENDIX G TABLES Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 P 0.1635 0.1890 0.1906 0.2442 0.2528 0.311 0.300 s.e. 0.0844 0.0897 0.1512 0.1447 0.1564 0.144 0.152 Number of previous civil wars P 0.053t 0.035* 0.208 0.092t 0.106 0.031* 0.049* 1800-1992 h.r. 1.063 1.208 1.21 1.2770 1.2880 1.365 1.350 P 0.1525 0.4086 0.3633 0.3918 0.356 0.349 s.e. 0.0812 0.1320 0.1298 0.1244 0.135 0.127 Percent labor force in industry P 0.060t 0.002* 0.005* 0.002* 0.008* 0.006* 1965 h.r. 1.165 1.505 1.4380 1.4800 1.427 1.417 P -0.0062 -0.01 -0.0093 -0.0094 -0.010 -0.009 s.e. 0.0026 0.0037 0.0037 0.0035 0.004 0.004 Percent labor force in industry P 0.016* 0.007* 0.013* 0.007* 0.009* 0.016* 1965 Squared h.r. 0.994 0.99 0.9910 0.9910 0.990 0.991 P -0.137 -0.1152 -0.1398 -0.110 -0.131 s.e. 0.0420 0.0375 0.0399 0.036 0.038 P 0.001* 0.002* 0.001* 0.002* 0.001* Service dominance h.r. 0.872 0.8910 0.8700 0.896 0.877 P 0.1770 0.1957 0.200 0.220 s.e. 0.0617 0.0610 0.064 0.065 Agricultural population density P 0.004* 0.001* 0.002* 0.001* 1960 h.r. 1.1940 1.2160 1.222 1.247 P -0.0031 -0.0034 -0.004 -0.004 s.e. 0.0013 0.0013 0.001 0.001 Agricultural population density P 0.015* 0.007* 0.005* 0.002* 1960 Squared h.r. 0.9970 0.9970 0.996 0.996 P 0.0195 0.014 s.e. 0.0110 0.012 P 0.O76t 0.250 Ethnic homojgeneity hj". 1.0200 1.014 P 1.591 1.500 Is 2nd largest ethnic group = s.e. 0.665 0.663 10-30% of largest ethnic group? P 0.017* 0.024* (yes/no) h.r. 4.9070 4.4810 -2 log likelihood without covariates 268.052 256.854 160.751 159.297 159.031 159297 159.031 -2 log likelihood with covariates 265.107 234.381 136.793 124.058 119.678 118.964 115.13 Pseudo R-Squared 0.011 0.087 0.149 0.221 0J247 0.253 0276 N 116 112 116 116 116 116 116 Events 29 28 19 19 19 19 19 ' = p<.05 t = P<-10 Table 3: Base model from the 1965 analysis 155 APPENDIX H TABLE 4 Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 P 0.0613 0.0404 0.31826 0.3126 0.3723 0.349 0.390 s.e. 0.0932 0.0944 0.1468 0.1426 0.1572 0.146 0.159 Number of previous civil wars P 0.511 0.669 0.030* 0.028* 0.018* 0.017* 0.014* 1800-1992 h.r. 1.178 1.041 1J75 1.3670 1.4510 1.417 1.477 P 0.1254 0.4287 0.4019 0.4747 0.421 0.458 s.e. 0.0793 0.1298 0.1257 0.1387 0.129 0.141 Percent labor force in industry P 0.114 0.001* 0.001* 0.001* 0.001* 0.001* 1970 h.r. 1.134 1.535 1.4950 1.6080 1.524 1.580 P -0.0045 -0.0090 -0.0087 -0.0103 -0.010 -0.010 s.e. 0.0022 0.0031 0.0031 0.0034 0.003 0.004 Percent labor force in industry P 0.042* 0.004* 0.005* 0.003* 0.002* 0.005* 1970 Squared h.r. 0.996 0.991 0.9910 0.9900 0.990 0.990 P -0.1922 -0.1621 -0.2023 -0.166 -0.200 s.e. 0.04985 0.0433 0.0495 0.042 0.048 P 0.000* 0.000* 0.000* 0.000* 0.000* Service dominance h.r. 0.825 0.8500 0.8170 0.847 0.819 P 0.1491 0.1931 0.161 0204 s.e. 0.0617 0.0644 0.061 0.066 Agricultural population density P 0.016* 0.003* 0.008* 0.002* 1960 h.r. 1.1610 1.2130 1.175 1.226 P -0.0027 -0.0032 -0.003 -0.004 s.e. 0.0013 0.0013 0.001 0.001 Agricultural population density P 0.036* 0.010* 0.017* 0.006* 1960 Squared h.r. 0.9970 0.9970 0.997 0.996 P 0.0312 0.028 s.e. 0.0126 0.013 P 0.013* 0.040* Ethnic homojgeneity h.r. 1.0320 1.028 P 1.391 1.068 Is 2nd largest ethnic group = s.e. 0.700 0.715 10-30% of largest ethnic group? P 0.047* 0.135 (yes/no) h.r. 4.019 2.910 -2 log likelihood without covariates 236.710 215.717 148.793 146.995 146.762 146.995 146.762 -2 log likelihood with covariates 236.333 204.216 120.519 111.212 103.783 107.786 101.789 Pseudo R-Squared 0.002 0.053 0.190 0243 0293 0267 0.306 N 126 120 126 126 126 126 126 Events 25 23 17 17 17 17 17 • = p<.05 f = p<.10 Table 4: Base model from the 1970 analysis 156 APPENDIX I TABLES Variable Model 8 Model 9^ Model 10 Model II Model 12--* P -0.1189 s.e. 0.0799 Annual average percentage change in P 0.137 GDP/capita h.r. 0.888 P I.I9I4 s.e. 0.8862 New agrarian techno-ecological P 0.179 inheritance? (yes/no) h.r. 3.292 P -0.0187 s.e. 0.1692 P 0.912 Log total population I960 h.r. 0.981 P -0.0093 s.e. 0.0358 P 0.794 Years since independence h.r. 0.991 P 0.1748 s.e. 0.2490 P 0.483 Separatist intensity 1975 h.r. 1.191 -2 log likelihood without covariates 156.992 159.297 150.3 159.297 114.384 -2 log likelihood with covariates 114.149 117.147 111.844 118.892 76.883 Pseudo R-Squared 0.273 0.265 0.256 0.254 0.328 N 116 116 116 116 116 Events 19 19 19 19 15 * = p<.05 f = p<.10 Base model = lagged dependent variable, percent labor force in industry 1965 and square, service dominance, agricultural population density I960 and square, and 20 percent ethnic dummy. All variables in the base model are significant at the p = .10 level except where noted. Table 5: Additional tests of macro-social variables: 1965 analysis (Continued) ^ For lagged dependent variable, p = .161. For 20 percent ethnic dummy, p = 214. 157 Table 5: Continued Variable Model 13“ Model 14 Model 15“ Model 16“ Model 17 P 3.506 s.e. 1.162 Ethnic history of lost autonomy * P 0.003* group size h.r. 33.306 P 0.79845 s.e. 0.5636 P 0.157 Semi-periphery world %stem position h.r. 2.222 P 0.01808 s.e. 0.01031 P o.osot Agricultural exports/total exports 1980 h.r. 1.018 P -6239768 s.e. 4109646 P 0.129 Forei^ owned stock/GDP 1973 h.r. 0.000 P 0.000744 s.e. 0.01062 P 0.944 Trade Opeimess h.r. 1.001 -2 log likelihood without covariates 136.904 159.297 157.398 122.777 158.454 -2 log likelihood with covariates 92.1 117.076 115.593 85.599 118.1 Pseudo R-Squared 0.327 0.265 0.266 0.303 0.255 N 116 116 116 116 116 Events 18 19 19 15 19 • = p<.05 t = p<-lO Base model = lagged dependent variable, percent labor force in industry 1965 and square, service dominance, agricultural population density 1960 and square, and 20 percent ethnic diunmy. All variables in the base model are significant at the p = . 10 level except where noted. For service dominance, p = .158. For lagged dependent variable, p =. 157. ■' For service dominance, p = .686. 158 APPENDIX J TABLE 6 Variable Model 8 Model 9 Model 10 Model 11 Model 12 P -0.1438 s.e. 0.08853 Annual average percentage change in P 0.1043 GDP/capita h.r. 0.866 P 0.796 s.e. 0.93584 New agrarian techno-ecological P 0.395 inheritance? (yes/no) h.r. 2.217 P 0.0194 s.e. 0.18655 P 0.9172 Log total population 1960 h.r. 1.02 P -0.00406 s.e. 0.0443 P 0.927 Years since independence h.r. 0.996 P 0.15409 s.e. 0.26978 P 0.568 Separatist intensity 1975 h.r. 1.167 -2 log likelihood without covariates 146.071 146.762 137.75 146.762 100.714 -2 log likelihood with covariates 99.994 103.056 97.636 103.774 63.13 Pseudo R-Squared 0J15 0.298 0.291 0.293 0.373 N 126 126 126 126 126 Events 17 17 16 17 13 * = p<.05 f = p<.IO Base model = lagged dependent variable, percent labor force in industry 1970 and square, service dominance, agricultural population density 1960 and square, and ethnic homogeneity. However, for equations 15 and 16, the 20 percent ethnic dummy is substituted for ethnic homogeneity. All variables in the base model are significant at least at the p = .10 level except where noted. Table 6: Additional tests of macro-social variables: 1970 analysis (Continued) 159 TaWe^ér Continued Variable Model 13 Model 14 Model 15^ Model 16^ Model 17 P 2.310 S.C. 1.314 Ethnic history of lost autonomy * p 0.0787t g o i^ size ...... h.r. 10.072 0.94805 s.e. 0.57604 P 0.0998t Semijperi])hery world _%stem jiosidon h.r. 2.581 "P 0.02239 s.e. 0.01137 P 0.0489* Ajgicultuml exports/totol .1980 h.r. 1.023 "P -1.4e+07 s.e. 6972757 P 0.0463* Foreign o\vned stock/GDP 1973 h.r. 0.000 J 0.00428 s.e. 0.01001 P 0.669 Trade Openness h.r. 1.004 -2 log likelihood without covariates 130.787 146.762 145.39 107.529 146.071 -2 log likelihood with covariates 89.937 101.22 103.555 76.807 102.793 Pseudo R-Squared 0.312 0.310 0.288 0.286 0.296 N 126 126 126 126 126 Events 17 17 17 13 17 * = p<.05 f = p<.10 Base model = lagged dependent variable, percent labor force in industry 1970 and square, service dominance, agricultural population density 1960 and square, and ethnic homogeneity. However, for equations 15 and 16, the 20 percent ethnic dummy is substituted for ethnic homogeneity. All variables in the base model are significant at least at the p = . 10 level except where noted. ■* For lagged dependent variable, p = .126. ^ For service dominance, p=.473. For agricultural population density 1960. p of square = .130. 160 APPENDIX K TABLE? Variable Model 18^° Model 19^' Model 20 Model 21 P 0.07328 s.e. 0.04263 P 0.086f Inf^t mor^ity rate h.r. 1.076 "P ^ .0004 s.e. 0.0002 Infant mortality rate P 0.067t Squared h.r. 1.000 "P 0.54395 s.e. 0.31369 Intensity of political discrimination P 0.083t 1975 h.r. 1.723 "P 0.38428 s.e. 0.14752 Ethnic grievance (all) * group P 0.009* pojpdarion h.r. 1.469 T 0.9449 s.e. 0.4151 P 0.023* Gini income m eq^ity h.r. 2.5720 7 ’^.’6107 s.e. 0.0047 Gini Income Inequality P 0.024* Squared h.r. 0.9890 -2 log likelihood without covariates 150.3 114.384 136.904 125.263 -2 log likelihood with covariates 107.115 74.177 94.796 81225 Pseudo RrSquared 0287 0J52 0J08 0252 N 116 116 116 116 Events 18 15 18 16 * = p<.05 f = p<.10 Table 7: Tests of intervening variables: 1965 analysis (Continued) “ For labor force in industry, p of square =.141. For lagged dependent variable, p=. 162. For 20 percent ethnic dununy, p=.310. 161 Table 7r Contmaerf Variable Model 22 Model 23^- Model 24 Model 25 P 0.2685 s.e. 0.1661 P 0.106 Gini Sectoral Inequality 1970 h.r. 1.3080 P -0.0049 s.e. 0.0027 Gini Sectoral Inequality 1970 P 0.075f Squared h.r. 0.9950 P 31.00 s.e. 45.71 P 0.498 Exports of Primary Products/GDP h.r. 2.92el3 P -133.69 s.e. 210.74 Exports of Primary Products/GDP P 0.526 Squared h.r. 0.0000 P 0.599 s.e. 0.254 Annual average percentage change of P 0.018* total population h.r. 1.820 P 0.73339 s.e. 0.26585 Annual average percentage change in P 0.006* population 15-64/total population h.r. 2.082 -2 log likelihood without covariates 141.733 40.094 150.3 150.3 -2 log likelihood with covariates 99.104 19.639 107.158 105.016 Pseudo R-Squared 0.301 0.510 0.287 0.301 N 116 116 116 116 Events 17 5 18 18 * = p<.05 t = p<.IO Base model = lagged dependent variable, percent labor force in industry 1965 and square, service dominance, agricultural population density 1960 and square, and 20 percent ethnic dummy. All variables in the base model are significant at the p = .10 level except where noted. (Continued) For agricultural density in I960, p of square = .751. For 20 percent ethnic dummy, p = .996. 162 TaWe 7r Continned Variable Model 26 Model 27 Model 28 Model 29^^ 0.50701 s.e. 0.22518 Annual average percentage change in P 0.024* population 15-29 h.r. 1 .6 6 P -0.10788 s.e. 0.06307 P 0.087t Secondary educational enrollment h.r. 0.898 P 0.00171 S.C. 0.000781 Secondary educational enrollment P 0.028* Squared h.r. 1 .0 0 2 P 0.05102 s.e. 0.04524 P 0.259 Democracy-Autocracy h.r. 1.052 P -0.01105 s.e. 0.00827. P 0.182 Democracy-Autocracy Squared h.r. 0.989 P -1.17277 s.e. 1.03748 P 0.258 Repression Index h.r. 0.31 P 0.11812 s.e. 0.12831 P 0.357 Repression Index Squared h.r. 1.125 -2 log likelihood without covariates 150.3 136.692 158.633 91.745 -2 log likelihood with covariates 107.099 89.012 115.981 69.692 Pseudo R-Squared 0287 0.349 0.269 0.240 N 116 116 116 116 Events 18 17 19 11 * = p<.05 f = p<.10 Base model = lagged dependent variable, percent labor force in industry 1965 and square, service dominance, agricultural population density 1960 and square, and 20 percent ethnic dummy. All variables in the base model are significant at the p = . 10 level except where noted. 33 For agricultural density in 1960. p of square = 286. For 20 percent ethnic dummy, p = .721. 163 APPENDIX L TABLE 8 Variable Model 18 Model 19 Model 20 Model 21^-* P 0.04043 s.e. 0.039 P 0.300 Infant mortality rate h.r. 1.041 P -0.0002 s.e. 0.00019 Inhmt mortality rate P 0.264 Squared h.r. 1.000 P 1.71534 s.e. 0.56753 Intensity of political discrimination P 0.002* 1975 h.r. 5.559 P 0.33651 s.e. 0.12123 Ethnic grievance (all) * . roup P 0.005* population h.r. 1.4 P 0.6614 s.e. 0.3973 P 0.096f Gini income inequality h.r. 1.9370 P -0.0075 s.e. 0.0044 Gini Income Inequality P 0.090t Squared h.r. 0.9930 -2 log likelihood without covariates 137.75 100.714 130.787 113.462 -2 log likelihood with covariates 96.097 48.763 85.98 73.008 Pseudo R-Squared 0.302 0.516- 0.343 0.357 N 126 126 126 126 Events 16 13 17 14 * = p<.05 f = p<10 Base Model; See next page. Table 8: Tests of intervening variables: 1970 analysis (Continued) For 20 percent ethnic dummy, p = 294. 164 Table & Continued Variable Model 22 Model 23^* Model 24 Model 25 P 0.3520 s.e. 0.1995 P 0.078t Gini Sectoral Inequality 1970 h.r. 1.4220 P -0.0065 s.e. 0.0036 Gini Sectoral Inequality 1970 P 0.067t Squared h.r. 0.9930 P 84.2961 s.e. 85.1698 P 0.322 Exports of Primary Products/GDP h.r. 4.068e36 P -294.37 s.e. 331.3176 Exports of Primary Products/GDP P 0.374 Squared h.r. 0.0000 P 0.194 s.e. 0.272 Annual average percentage change of P 0.470 total population h.r. 1.214 P 0J6314 s.e. 0.27775 Annual average percentage change in P 0.343 population 15-64/total population h.r. 1.301 -2 log likelihood without covariates 128.665 41.063 137.75 137.75 -2 log likelihood with covariates 91.939 17.066 97.164 96.795 Pseudo R-Squared 0.285 0.584 0.295 0.297 N 126 126 126 126 Events 15 5 16 16 * = p<.05 f = p<.10 Base model = lagged dependent variable, percent labor force in industry 1970 and square, service dominance, agricultural population density 1960 and square, and ethnic homogeneity. However, for model 21, the 20 percent ethnic dummy is substituted for ethnic homogeneity. All variables in the base model are significant at least at the p = . 10 level except where noted. (Continued) In this model, no variable was significant 165 Tablets: Gontinoed Variable Model 26 Model 27 Model 28 Model 29^* P 022656 s.e. 022733 Annual average percentage change in P 0.319 population 15-29 h.r. 1254 P -0.05105 s.e. 0.06875 P 0.458 Secondary educational enrollment h.r. 0.95 P 0.000889 s.e. 0.000753 Secondary educational enrollment P 0.238 Squared h.r. 1.001 P 0.01439 s.e. 0.04574 P 0.753 Democracy-Autocracy h.r. 1.014 P -0.00877 s.e. 0.0091 P 0.336 Democracy-Autocracy Squared h.r. 0.991 P -1.36636 s.e. 0.99068 P 0.168 Repression Index h.r. 0.255 P 0.1642 s.e. 0.12052 P 0.173 Repression Index Squared h.r. 1.178 -2 log likelihood without covariates 137.75 123.391 145.905 102.957 -2 log likelihood with covariates 96.673 75221 102.826 75.445 Pseudo R-Squared 0298 0.390 0.295 0.267 N 126 126 126 126 Events 16 15 17 12 * = p<.05 t = p<.10 Base model = lagged dependent variable, percent labor force in industry 1970 and square, service dominance, agricultural population density 1960 and square, and ethnic homogeneity. However, for model 21, the 20 percent ethnic dummy is substituted for ethnic homogeneity. All variables in the base model are significant at least at the p = .10 level except where noted. 36 For agricultural density in I960, p of square = 210. For ethnic homogeneity, p = 279. 166 APPENDIX M FIGURES 5-16 70 60 I 50 i I 40 0 30 1 g I 20 10 0 0 ■> 3 4 5 6 7 8 9 10 12 Prior Civil Wars — 1965 Analysis 1970 Analysis Figure 5: Independent effect of previous civil wars on civil war onset 167 90 80 70 8 60 U 50 O I 40 u 30 20 1 10 0 10 20 30 40 50 Labor Force in Industry — 1965 Analysis 1970 Analysis Figure 6: Independent effect of labor force in industry 1965 on civil war onset 168 0.2 S 0.16 O I 0.12 > 3 I*. o M 0.08 g 10 20 30 40 50 60 70 Service Dominance 1"^ 1965 Analysis 1970 Analysis Figure 7: Independent effect of service sector dominance on civil war onset 169 14 12 i 10 I I 8 o 6 g 4 0 0 20 40 SO60 100 120 140 160 180 Agricultural Population Density -**- 1970 Analysis -*- 1965 Analysis Figure 8: Independent effect of agricultural population density on civil war onset 170 5 20 •c 20 40 60 80 100 Ethnic Homogeneity 1965 Analysis 1970 Analysis Figure 9: Independent effect of ethnic homogeneity on civil war onset 171 b 5 I ! 4 > *Ü 3 iû 1 & 0 0 20 percent ethnie dummy -*- 1965 Analysis 1970 Analysis Figure 10: Independent effect of 20 percent ethnic dummy on civil war onset 172 60 50 § 40 > "S *0 30 •m "C I 20 1 10 0 0 02 0.4 0.6 0.8 1.2 Ethnie history of lost autonomy * group size —*~ 1965 Analysis 1970 Analysis Figure 11: Independent effect of ethnic history o f lost autonomy * group size on civil war onset 173 10 9 I 8 I 7 > 6 Ü *0 5 •c 4 g 3 Î 7 1 0 0 20 40 60 80 100 Agricultural Exports/Total Exports 1980 1965 Analysis -**-1970 Analysis | Figure 12: Independent effect of agricultural exports/total exports 1980 on civil war onset 174 10 1200 9 1000 8 8 7 I 800 6 > V 5 600 I 4 400 3 200 0 0 3 4 Intensity of political discrimination 1975 1965 Analysis 1970 Analysis Figure 13: Independent effect of intensity of political discrimination 1975 on civil war onset 175 120 _ ICO 80 60 •n 40 20 0 0 3 4 5 6 7 8 9 10 n 12 13 Ethnie grievance (all) * group population 1 -^ 1965 Analysis 1970 Analysis Figure 14: Independent effect of ethnic grievance (all) * group population on civil war onset 176 1.4E+09 2.5E+06 1.2E+09 2.0E+06 12 l.OE+09 I ;> 8.0E+08 1.5E+06 u k.0 ig 6.0E+08 •c l.OE+06 g •= 4.0E+08 01 5.0E+05 2.0E+08 O.OE+00 O.OE+00 10 20 30 40 50 60 70 Gini Income Inequality 1965 Analysis 1970 Analysis Figure 15: Independent effect of Gini coefïïcient of income inequality on civil war onset 177 120 ICO I 80 I> ‘S 60 I g 40 % 1 20 0 10 20 30 40 50 60 70 80 Gini sectoral inequality 1970 — 1965 Analysis -**-1970 Analysis | Figure 16: Independent effect of Gini coefficient of sectoral inequality 1970 on civil war onset 178