Catastrophe or adaptation? Explaining the impacts of resource scarcity and adaptability on political instability

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Request accessible transcript

Transcript (if available)

Content CATASTROPHE OR ADAPTATION? EXPLAINING THE IMPACTS OF RESOURCE SCARCITY AND ADAPTABILITY ON POLITICAL INSTABILITY by Wenyu Li A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (POLITICS AND INTERNATIONAL RELATIONS) May 2011 Copyright 2011 Wenyu Li ii Acknowledgements Every journey has its destination no matter how long it is. The past several years of Ph.D. study has left a number of precious episodes in my memory. The academic training I have received in the University of Southern California has deeply influenced my way of thinking as well as my perspective towards life. The most important thing I have learnt in graduate school is to think independently, which probably is an essential requirement to be a true scholar. Many people have offered me important support during this long journey. First of all, I want to thank my advisor, Dr. Patrick James. Pat is a wonderful mentor and friend. I am always amazed by his super positive attitude towards life and career. He is always willing to hear about the difficulties I have encountered in research and professional life despite his extremely busy schedule. His encouragement and mentoring skills have helped me conquer the frustration and be more confident to resolve various research problems. Besides, I truly appreciate the openness and flexibility Pat gives to students. He consistently supports my plan to receive more training in quantitative research skills and apply them in the dissertation project. Thanks to his openness and encouragement, I have enjoyed the academic freedom to focus on something I am really interested in. I also owe thanks to Dr. Sheldon Kamieniecki, who is now enjoying an environmentally friendly life in Santa Cruz. I entered the program with an interest in his research and he advised my study in the first two years. His door was always open to students and he spent a great amount of time supervising my direct research on iii environmental politics. With his guidance I conducted a series of comparative environmental research, which eventually leads to the cross-national environmental security research in this dissertation. The other faculty members in the University of Southern California have also provided enormous amount of help along the way. Dr. Stanley Rosen and Dr. Shui Yan Tang have both offered me great support in the qualifying exam and dissertation writing. Dr. Rosen is well-known on campus for his familiarity with Chinese culture, and his class on Chinese politics is awesome. Although I end up not choosing China for case analysis based on the quantitative results, he has given me valuable suggestions on fitting China into the research framework. Dr. Tang is an expert on environmental issues. Coming from a public policy background, he has consistently urged me to think about the real word value and policy applications of my research. Through teaching and research, I have got to know many prominent professors in the university. As an extraordinary female scholar, Dr. Alison Renteln is one of the most upright persons I have met. She demonstrates that a woman can succeed both in academic career and family life. I am also grateful to Dr. Jeffery Sellers, who taught me important skills of data cleaning and quantitative research design when I worked on his IMO project. Moreover, I have been very lucky to work with Dr. Jeb Barnes, Dr. Arthur Auerbach, Dr. Anthony Kammas and Prof. Darry Sragow in the last several years. Their passion for teaching and popularity with students set good examples for me. Last but not least, my family and friends are the major sources of power in the whole process. As a companion, my husband, Jay, has shared the happiness and the iv sadness. We do not only offer each other encouragement but also urge each other to make progress. My parents have given me long term support in pursuing the degree oversea and always respect my choice. Moreover, my fellow graduate students in the program have added much more fun to the journey. It is this group of people that make me not feel alone on the road. v Table of Contents Acknowledgements ii List of Tables vii List of Figures ix Abstract x Chapter 1 Introduction 1 Background and Rationale 1 Significance of Research 5 Research Questions 7 Data, Statistical Methods and Case Selection 8 Overview of This Study 9 Chapter 2 A Literature Review: Impacts of Resource Scarcity on Human Society 12 The Debate among Three Schools 12 Empirical Evidence 29 Key Concepts of This Study 36 The Comprehensive Model and Causal Diagram 62 Chapter 3 Data, Statistical Methods and Case Selection 73 Data Sources 73 Statistical Methods 90 Case Selection 106 Chapter 4 The Different Patterns in Two Scarcity Groups: An Application of Auto-regressions Over Two Periods 108 Auto-regressions on Annual Number of Civil Wars Over Two Periods 108 Discussion on the Findings 127 Chapter 5 Adaptability and Environmental Scarcity-Political Instability Nexus 133 Data Screening and Transformation 133 Principal Component Analysis and Regression Analysis 137 Summary of the Results 206 vi Chapter 6 Analysis on Nonconforming Cases and Implications on The Model 208 The Process of Case Selection 208 Analysis on the Three Nonconforming Cases 213 Summary on the Case Analysis 238 Chapter 7 Conclusions and Implications 242 Empirical Findings and the Revised Model 243 Implications for the Theoretical Debate 248 Policy Implications and Future Research 255 Bibliography 261 Appendix A: Regression Analyses Based on Three Scarcity Groups 284 Appendix B: Residual Analysis for Outliers 290 Appendix C: List of Civil Wars and State Failures from 1970 to 1999 291 Appendix D: Country List by Scarcity Group 313 vii List of Tables Table 1: Five Datasets on Civil War 81 Table 2: Descriptive Statistics for Mean Scarcity Groups 116 Table 3: Auto-regressions of Annual Number of Civil Wars Ongoing over Two Periods 117 Table 4: Auto-regressions Using Mean of Natural Capital Per Capita to Divide Groups 119 Table 5: Descriptive Statistics for Median Scarcity Groups 122 Table 6: Auto-regressions of Annual Number of Civil Wars Ongoing over Two Periods 123 Table 7: Auto-regressions Using Median of Natural Capital Per Capita to Divide Groups 124 Table 8: Principal Component Analysis of Economic Development 146 Table 9: Pearson‘s Correlations between Economic Conditions Variables, Environmental Scarcity and Political Instability Variables 149 Table 10: Testing for Interactions and Confounding Effects of Economic Conditions 150 Table 11: The Different Patterns among Low and High Economic Adaptability Groups 151 Table 12: Principal Components of Political Adaptability 161 Table 13: Pearson‘s Correlations between Political Institution Variables, Environmental Scarcity and Political Instability Variables 165 Table 14: Testing for Interactions and Confounding Effects of Political Institutions 166 Table 15: The Different Patterns among Low and High Political Adaptability Groups 167 Table 16: Principal Components of Social Adaptability 175 viii Table 17: Pearson‘s Correlations between Social Fractionalization Variables, Environmental Scarcity and Political Instability Variables 179 Table 18: Testing for Interactions and Confounding Effects of Social Fractionalization 180 Table 19: The Different Patterns of Low and High Social Adaptability Groups 181 Table 20: Principal Components of Demographic Adaptability 192 Table 21: Pearson‘s Correlations between Demographic Characteristics Variables, Environmental Scarcity and Political Instability Variables 196 Table 22: Testing for Interactions and Confounding Effects of Demographic Characteristics 197 Table 23: The Different Patterns among Low and High Demographic Adaptability Groups 198 Table 24: Summary of the Statistical Findings in Chapter 5 207 Table 25: Summary of the Three Nonconforming Cases 212 Table 26: Major Characteristics of the Four Nonconforming Cases 215 Table 27: Testing for Interactions and Confounding Effects of Economic Adaptability (Three Scarcity Groups) 287 Table 28: The different patterns among low, middle and high economic adaptability groups 288 Table 29: List of Civil Wars by Fearon and Laitin 291 Table 30: List of Civil Wars by Correlates of War Project 294 Table 31: List of Civil Wars by Collier and Hoeffler 297 Table 32: List of Civil Wars by Doyle and Sambanis 299 Table 33: UCDP/PRIO Armed Conflict Dataset (Version 4-2008) 303 Table 34: Political Instability Task Force coding of Political Failure 308 Table 35: Country List by Mean and Median Scarcity Group 313 ix List of Figures Figure 1: Venn Diagram of the Relationship among the Three Schools 28 Figure 2: The Hypothesized Causal Diagram 65 Figure 3: Distributions of the Frequency of Annual Number of Civil Wars Ongoing 111 Figure 4: Distribution of Environmental Scarcity 114 Figure 5: Distributions of Renewable and Nonrenewable Resources Per Capita 115 Figure 6: Comparison of Scatterplots of the Two Scarcity Groups Divided by Mean 120 Figure 7: Comparison of Scatterplots of the Two Scarcity Groups Divided by Median 125 Figure 8: Scatter Plots for Two Adaptability Groups of Renewable Resources (Civil War Index is Used) 152 Figure 9: Economic Adaptability as a Confounder between Nonrenewable Resources and Political Instability 157 Figure 10: Political Adaptability as a Confounder between Nonrenewable Resources and Political Instability 171 Figure 11: Demographic Adaptability as a Confounder between Renewable Resources and Political Instability 201 Figure 12: Malawi, Mauritania and Tanzania on the Map of Africa 219 Figure 13: The Revised Causal Diagram 246 Figure 14: Scatter Plots for Three Adaptability Groups 289 x Abstract With the rapid growth of population and human consumption, natural resource scarcity is increasingly considered as an important national security issue. A debate on the causal relationships between resource scarcity and political instability of a state, in particular civil war, has arisen in the last two decades. Although a series of case studies have been conducted by several research projects, quantitative studies on the subject still lag behind and the few existing ones have reported quite inconsistent results. This is because most of them focus on the shortage of certain natural resources and do not take into account the interaction between the natural and human system. This dissertation creatively develops an ―adaptability‖ model bringing four aspects of adaptability respectively as the intermediate variable to explain the linkages between resource scarcity and political instability. Conducting auto-regressions over two periods utilizing five widely-used civil war datasets, this study confirms that the risk of civil war increases when a state‘s resource scarcity reaches some threshold level, such as the mean and median values. It then applies principal component analysis and regression analysis to find that renewable and nonrenewable resource scarcity relate to political instability in quite different ways. The results imply that renewable resource challenge is detrimental to the political stability of states with low economic development level and/or poor political institutions, but not to states with high economic development level and healthy political institutions. However, the tests for nonrenewable resources do not find significant evidence for such relationships. To further evaluate the model, this xi dissertation analyzes Malawi, Mauritania and Tanzania as three worst-predicted cases. It finally identifies cultural contexts and international actors as two important factors that explain their nonconformity with the statistical model. 1 Chapter 1 Introduction Background and Rationale After the end of the Cold War, environmental factors are increasingly addressed by scholars interested in non-traditional security issues. A central debate in the field of ―environmental security‖ today revolves around whether pressures from resource scarcity and environmental degradation can lead to violent conflict. It is suggested that three distinct schools of thought can be identified in the current debate (Homer-Dixon 1999; Kahl 2006). 1 Originally inspired by Malthus‘ (1807) proposition, neo-Malthusians claim that rapid population growth will lead to over-consumption and degradation of resources, and eventually result in scarcities. As resource scarcity aggravates, people will fight for survival (Meadows and Club of Rome 1972 ; Diamond 2005). In contrast, neoclassical economists deny the strict limits to population and prosperity in human society. According to them, technological innovation, market pricing, and cooperation can help us to surpass scarcity (Boserup 1965 ; Simon 1981 ; Lomborg 2001). Compared with the above two schools, less attention in the popular discourse has been received by the alternative political ecologist argument that distributional inequality is the essential cause of scarcity (Durham 1995 ; Moore 1996 ; Suliman 1999 ; Hildyard 1999 ; Peluso and Watts 2001). According to them, scarcity is mainly experienced by poor communities in developing countries. With an expanding agenda, the debate nowadays does not only 1 Some scholars such as Gleditsch (1998) suggest that the major debate in the field is between neo-Malthusians and the cornucopians. Kahl (2006) suggests that political ecologists have provided alternative arguments to the above two schools and Homer-Dixon (1999) also address the arguments of maldevelpment. Thus this study tries to provide a more comprehensive literature review by including works of the resource curse school and the political ecologists. 2 address traditional environmental problems, it also pays increasing attention to newly emerging environmental issues such as climate change and ozone depletion. As the debate goes on, a series of case studies addressing environmental scarcity and civil conflict have been conducted by several academic projects since the 1990s. Some of them have developed hypothetical models to investigate the causal relationships between environmental scarcity and violence. However, the development of quantitative studies on the subject still lags behind. And the few existing ones focus primarily on the shortage of certain natural resources, 2 showing inconsistent statistical results. Since little systematic empirical evidence has been obtained so far to bear on the impact of environmental scarcity on political instability, in particular civil conflict, the generalizability of proposed hypothetical models is being questioned. Furthermore, focusing on specific natural resources, most of the quantitative studies today define environmental scarcity in such a narrow way that they do not take into account human influences on the natural system. In sum, studies in the domain have often ignored many potential important environmental variables as well as the interaction between the human system and the natural environment. To explore a more comprehensive model supported by solid empirical evidence, this project examines the in-depth relationships between environmental scarcity and political instability of a state, with adaptability as an important intermediate variable. The concept of environmental scarcity is used here to capture the scarcity of renewable (i.e. 2 Most of the quantitative studies in the field look at the scarcity of specific resources, such as oil, diamond, gold, fresh water, and so on. Few studies including de Soysa (2002b) and Binningsbø et al (2007) try to evaluate countries‘ resource challenge at a more comprehensive level. 3 arable land, forest, fresh water and other renewable resources) and nonrenewable natural resources (oil, minerals, coal etc.). As the three schools suggest, this scarcity can arise from increased resource demand, resource degradation or depletion, as well as unequal distribution of resources. 3 High environmental scarcity usually imposes high environmental stress on the state, which can lead to various social effects. Originally defined in biology, adaptability or adaptive capacity 4 means ―an ability to become adapted (i.e., to be able to live and to reproduce) to a certain range of environmental contingencies‖ (Gallopí n 2006:300). And adaptation is an instrumental ―feature of structure, function, or behavior of the organism‖ in securing the status of being adapted (Dobzhansky 1968 ; Gallopí n 2006:300). Discussion on adaptability can be found in case analyses addressing environmental scarcity (e.g.Bä chler 1998 ; Homer- Dixon 1999 ; Kahl 2006), but few quantitative studies in the field have tried to measure a state‘s capacity of adapting to environmental challenges and the intermediate role of adaptability in the environmental scarcity-political instability nexus. The available case studies at best indicate that a number of factors can affect a state‘s adaptability, and the role each factor plays varies from case to case. Thus our knowledge about environmental adaptability so far is quite limited and implicit. However, with the help of cross-national quantitative studies, we can figure out the major factors determining a state‘s adaptive capacity from a laundry list of variables. 3 Homer-Dixon (1999) categorizes the different types of scarcities into supply-induced, demand-induced, and structural scarcities according to the different causes. 4 Following Smit and Wandel (2006) and Gallopí n (2006), this study treats adaptability and adaptive capacity as synonymous. 4 Moreover, with its advantage in generalization, quantitative analysis can provide us a more systematic view of interactions among various factors. Therefore, drawing from the broad literature on political instability and civil conflict, this study conducts quantitative analyses considering a number of variables as potential determinants of adaptability. These variables can be categorized into four factors: they are economic development level, political institutions, social fractionalization, and demographic characteristics. Finally, the concept of Political stability is defined as ―the regularity of the flow of political exchanges,‖ and thus ―the more regular the flow of political exchanges, the more stability‖ a state has (Ake 1975:273). Civil conflict, assassinations, coups d‘état, strikes, major government crises, riots, and anti-government demonstrations can impede the regular flow of political exchanges to different degrees, and thus they are all indicators of political instability. While most studies in the field prefer to focus their attention on civil war, this dissertation also looks at those less intense forms of political instability to broaden the range of our dependent variable and get a more complete view of its relations with environmental stress. In general, this study aims to explore the interplay among environmental scarcity, adaptability and political instability. Drawing on major insights from current literature discussing the environmental scarcity-civil conflict nexus, it hopes to uncover more accurate causal mechanisms connecting the three variables and then derive theoretical conclusions. Furthermore, this study departs from widely-used single equation models, which utilize a number of exogenous explanatory variables to predict a single dependent variable. Instead, it brings in relatively novel statistical techniques such as auto- 5 regression and principal component analysis. The application of these methods can also shed light on quantitative studies in the security fields. Significance of Research As a matter of fact, the study hopes to make three contributions to current research on the topic. First, it brings in an in-depth investigation on the role of adaptability in the environmental scarcity-political instability nexus. Previous studies on the topic often address environmental scarcity from a security perspective, and thus their central research question is whether environmental scarcity leads to the occurrence of conflict. Few of them have paid attention to the state and society‘s adaptive capacity to environmental scarcity, and how that works as an intermediate variable in the environmental scarcity-political instability nexus. To fill the gap, the project tries to identify the major factors determining the strength of a state‘s adaptability and figure out the causal diagrams incorporating the role of adaptability based on relevant theories and empirical evidence. Moreover, some methodological insights on how to deal with the relationship between complex and subtle variables in social science research are introduced. 5 Given the mixed results found by current quantitative research on the topic, this study discusses the major causes of the inconsistency and suggests some more sensitive statistical 5 It has been suggested that social science studies often involve complex and subtle variables that are hard to be quantified and measured in accurate ways. For example, Scruggs (2003) notes that specifying fully the complexity implied in a model of environmental outcomes is extremely difficult since the relative social theories are not as exact as to allow us to do so. In this study, the concern applies to the major variables including resource scarcity, political instability and adaptability. 6 methods such as auto-regression to capture the masked association between environmental scarcity and civil conflict. Auto-regression is a fundamental technique in time-series analysis. It basically regresses the value of a variable during the later occasion on it value during the earlier occasion. Compared with other statistical methods, it can simplify complex models and help us to detect the general pattern. Furthermore, Principal Component Analysis (PCA) and regression analysis are brought in to explore the potential confounding relationships and interactions among environmental scarcity, adaptability and political instability. PCA is a popular technique to reduce the dimensionality of a data set of ―a large number of interrelated variables‖, and in the same time retain as much variation in the data as possible (Jolliffe 2002:1). It is used as a central technique to construct the adaptability variables in this study. Finally, the project intends to enrich current research on the topic by bringing in insights on political development, political economy, democratization, ecology and so on. As mentioned above, major research on the topic usually comes from the domain of conflict studies. They mainly use civil conflict as a dependent variable and are more interested in cases where violence does happen. However, this project treats war as an extreme and rare case of political instability, and includes other less intense forms of political instability in the analysis. By broadening the range of the dependent variable and adding adaptability as intermediate variable, this project thus can bring in theories of a variety of domains to improve our understanding of the causal mechanisms and processes. 7 Research Questions Focusing on the interplay among environmental scarcity, adaptability and political instability, this study is intended to explore the following research questions: 1. Is there an association between environmental scarcity and political instability in general? 2. Do different levels of environmental scarcity affect political stability differently? Is it possible that low to fair level scarcity has little impact on political stability but high scarcity can lead to civil conflict and other less intense political instability? 3. What are the major factors determining a state‘s adaptability to environmental scarcity? 4. How does the interplay between environmental scarcity and adaptability affect a state‘s political stability? Can we identify the causal mechanisms connecting them? 5. Can we use the statistical results to explain why resource scarcity is associated with political instability in some countries but not in others? 6. What are the implications of the project for domestic environmental policymaking and international environmental institutions? These questions start from the causal links between environmental scarcity and political instability in general, and then continue to explore more complex causal mechanisms by adding environmental adaptability as an intermediate factor. They also ask about the explanatory power of the statistical results when it comes to specific cases 8 and the policy implications of the findings. Compared with previous studies, this study intends to assess the three-way debate in more depth by answering the above questions. Data, Statistical Methods and Case Selection As described in the previous sections, environmental scarcity, adaptability and political instability are the three major variables of interest in this study. Both quantitative and case analyses are conducted to explore the mechanisms connecting the three variables. In the quantitative investigation, various advanced multivariate statistical techniques are applied. 6 At first, auto-regressions over two periods are used to detect the associations between environmental scarcity and civil conflict since a series of empirical studies have been conducted on them. As mentioned before, auto-regression can simplify complex models to help us detect the general pattern. Thus it is expected to be a more sensitive statistical method to capture the masked association between environmental scarcity and civil conflict. Then PCA is used to reduce the dimensions of various indicators of adaptability. It not only mitigates the problem of multi-collinearity but also calculates component scores which can be used in further regression analyses. After then each of the four aspects of adaptability is tested respectively as an intermediate variable 6 Auto-regression is used at first to detect the masked association between environmental scarcity and civil conflict, and provide explanations for the inconsistent results obtained by previous research. Then adaptability is added as an intermediate variable in the scarcity-political instability nexus. Simple and multiple regressions are used to test the further relationships, with investigation on interactions, confounding effects among different variables. In addition, we also pay great attention to the group differences based on the four aspects of adaptability, including economic development level, political institutions, social fractionalization and demographic characteristics. In the regression analyses, this study uses the component scores for each aspect of adaptability generated by PCA instead of the original variables. More details of the methodology and data selection can be found in Chapter 3. 9 between environmental scarcity and political instability, using regressions with discussion on confounding effects, interactions and group differences. The major data are drawn broadly from widely used data bases on environmental studies, governance, and conflict, including the World Bank, Correlates of War Project, the Penn World Table, etc. On the one hand, those data bases are widely known as trust- worthy and well-maintained. On the other hand, using the same data sources facilitates conversations with other studies in the field. Finally, based on the results of the tested model, three worst-predicted cases will be selected and further case analysis will be performed. It is common for investigators of case studies to ―confront nonconforming cases and account for them by citing factors that are outside their explanatory frameworks‖ (Ragin 2004:137). Thus cases that fit poorly with the model are believed to indicate important variables outside the existing boundaries of the causal model and help us understand better its merits and deficiencies. Ideally, states with high scarcity and low adaptability that maintain political stability would be regarded as candidates for further case analysis due to their poor fit with the model. Overview of This Study This dissertation consists of seven parts and the road map is depicted in the following texts. The first chapter introduces the major research questions of interest and the motivations to explore them, placing this study within the context of the scholarly debate on environmental scarcity. It then highlights the theoretical and empirical 10 contributions this study aims to make and briefly addresses the methodology. Moreover, the end of the introduction part provides an overview of the following chapters. The second chapter reviews the major scholarly work on the relations between environmental scarcity and political instability, and then attempts to develop a hypothesized model bringing in adaptability to more accurately capture the nexus between the two variables. It also establishes a number of theoretical expectations and a set of variables that are associated with the hypothesized model. Before moving on to the empirical analysis, the third chapter provides a justification of the research design, data selection, statistical techniques and case selection method. The following two chapters carefully conduct and explain the statistical analyses of this study. Chapter 4 tries to rethink the mixed results of the relationship between environmental scarcity and internal conflict found by current quantitative research. It applies auto-regressions over two periods as a more sensitive statistical assessment to detect the masked association between the two variables. Based on the general trends detected in auto-regressions, the next chapter goes further to unpack the more complex mechanisms between environmental scarcity and political instability, adding environmental adaptability as an intermediate variable. It is intended to investigate the scarcity-instability nexus respectively for each aspect of adaptability. PCA is used as a major data reduction method for multiple dimensions of adaptability. Moreover, various theoretical expectations and the comprehensive hypothesized model are tested by regressions with discussion on confounding effects, interactions and group difference. 11 Built on the statistical results obtained in the previous two chapters, Chapter 6 evaluates and modifies the proposed model and hypotheses presented in the earlier chapters. Three cases that poorly fit the tested model are selected for an in-depth review to probe for information beyond the quantitative analyses. The purpose is to explore how those countries were able to maintain a relatively peaceful political environment under severe resource scarcity and poor adaptability. Furthermore, the theoretical significance of these nonconforming cases is discussed and their implications on our hypothesized model are also addressed. Finally, the concluding chapter is roughly divided into two sections. The first section provides a brief overview of the results and revises the hypothesized model based on the empirical findings. Then the theoretical implications of the findings and their contributions to the three schools‘ debate are evaluated. In the end, it addresses the policy implications of this study on security and environmental policymaking, and it also highlights some suggestions for future research. 12 Chapter 2 A Literature Review: Impacts of Resource Scarcity on Human Society This chapter brings together literatures of various disciplines that form the theoretical and empirical foundation of this study. It starts with the debate on the relationship between resource scarcity and violent conflict, including the main arguments and empirical evidence from the three schools of thoughts in the field. It also offers concluding remarks on the consent and disagreement in the debate. The second half of the chapter aims to build the theoretical framework of this study. The three core variables—resource scarcity, political instability and environmental adaptability—are defined within the context of this specific study, and it also reviews the available theoretical and empirical findings on the relationships among them. Based on the theoretical foundations, this study proposes an ―adaptability‖ model that establishes new causal mechanisms explaining the relationship between resource scarcity and political instability in a more comprehensive way. Furthermore, a series of propositions regarding the comprehensive model as well as the specific relations within the model are also presented at the end of the chapter to lay out the research track for the following discussion. The Debate among Three Schools The atmosphere of general environmental crisis was first perceived in the 1960s, followed by social movements and an avalanche of environmental writing. Scholars from various disciplines including biologists, ecologists, and economists have joined the 13 debate. 7 In general, their claims can be approximately identified as Neo-Malthusian, neoclassical economist and political ecologist arguments. Depicting distinct pictures of current and future environmental challenges for human beings, the three schools disagree with one another about the casual links between environmental scarcity and violent conflict. The following texts examine the arguments and major scholarly works of each school. Neo-Malthusians It is widely known that a wave of alarmist ―neo-Malthusian‖ literature emerged at the end of the 1960s and the early 1970s. They were inspired by Malthus‘ (1807) argument on arithmetical growth of food production and exponential growth of human population, as well as his belief that the discrepancy would result in food shortage and human misery at some point of human history. Built on Malthusian pessimism in population and environmental catastrophe, Neo-Malthusians claim that rapid population growth, environmental degradation, and increasing resource scarcity can eventually produce violent conflict. In the late 1960s, Ehrlich (1968) boldly predicted in his bestselling book The Population Bomb that overpopulation would certainly lead to a massive disaster in the 7 In the 1960s, public concerns over environmental issues arose. In the United States and other developed countries, an important sign of the movement is the publication of Rachel Carson‘s Silent Spring (1962). The book discusses the problems associated with the use of the insecticide DDT, which can spread through the food chain and poison the environment and human health. A number of books review the history of environmental movement. For example, Kline (2007) and Costain and Lester (1995) focus on the evolution of environmental movement in the United States. Schreurs (2002:32-59) discusses environmental movement in Japan, Germany and the United States. Guha and Martinez (1997) and Guha (2000) address and compare the development of environmentalism in both the western countries and the developing world. 14 next few years. The book sees little hope of avoiding the disaster entirely after examining a variety of crises caused by rapid population growth. Around the same time, another provocative thesis, ―The Tragedy of the Commons‖, was proposed by ecologist Garret Hardin (1968). He depicts the dilemma to sustain a shared limited resource under free market institution because individuals acting in their self-interest will ultimately destroy the resource. 8 In the early 1970s, the Club of Rome, a global think tank dealing with international political issues, published their well-known thesis of ―the limits to growth‖ (Meadows and Club of Rome 1972). 9 Their model tries to explore the relationships between population growth and finite resources by computer simulation, and eventually predicts collapse in the new century. Most recently Diamond (2005), a prominent biologist, examines a series of past civilizations and societies, showing that mismanagement of natural resources is the common factor contributing to collapse in these societies. Furthermore, he concludes that ecological collapse is the ultimate source of other failures. Since the 1990s several extensive research programs have been undertaken in the field of security study to explore the linkage among environmental degradation, resource scarcity and violent conflict, most of which are case studies. The two projects of the 8 Thirty-one years after the publication of Hardin‘s essay, political scientist Ostrom and her colleagues revisit the problems of common-pool resources in the Science Magazine (Ostrom et al. 1999). More of her arguments and critiques of Hardin‘s point can be found in her books (Ostrom 1990 ; Ostrom, Gardner and Walker 1994 ; Ostrom and National Research Council (U.S.). Committee on the Human Dimensions of Global Change. 2002). In general, she discusses institutional arrangements for sustainable common-pool resource management and shows that the collapse of ecosystem is avoidable based on past successful stories. 9 The Club of Rome adopted the assumption of exponential growth of population and five variables were addressed (Meadows and Club of Rome 1972). Later updates of their research can be found in Meadows et al. (1992 ; 2004). In 2008, their predictions are examined by Turner (2008) based on the reality of the past thirty years. 15 Toronto Group—―Environment, Population and Security‖ 10 and ―Environmental Scarcities, State Capacity, and Civil Violence‖ 11 —directed by Homer-Dixon are quite influential in North America (Homer-Dixon 1991, 1994, 1999 ; Homer-Dixon and Blitt 1998). Led by Homer-Dixon, scholars in the groups have carried out a number of case studies, many of which are focused on developing countries such as Haiti, Mexico (Chiapas), the Palestine Authority (Gaza) and Pakistan (Homer-Dixon and Blitt 1998). Resource scarcity, as the core concept in the group‘s work, is a composite variable consisting of three dimensions caused by different factors (Homer-Dixon 1999 ; Homer-Dixon and Blitt 1998). Supply-induced scarcity, for example, emerges when the reduction and degradation of resources are faster than regeneration. Moreover, population growth or increased consumption per capita can produce demand-induced scarcity. Finally, the unequal distribution of resources may result in resource shortages for a population in poverty, and thus cause structural scarcity. Interacting with and reinforcing each other, these scarcities then generate resource capture and /or ecological marginalization (Homer-Dixon 1999:6; Homer-Dixon and Blitt 1998). 12 As Homer-Dixon and his colleague describe, the scarcity aggravate when some elites use their power to grab resources that are anticipated to become scarce. As the 10 The project took place from 1994 to 1996. Further information about the project can be found in their website <http://www.library.utoronto.ca/pcs/eps.htm>. 11 The project was conducted from 1994 to 1998. Its website is <http://www.library.utoronto.ca/pcs/state.htm>. 12 According to Homer-Dixon and Blitt (1998:6), resource capture occurs when ―powerful groups within society, anticipating future shortages due to increased population growth and a decrease in the quantity and quality of the resource, shift resource distribution in their favor, which subjects the remaining population to scarcity‖, and ecological marginalization describes the situation ―when demand-induced and structural scarcities interact to produce supply- induced scarcity: lack of access to resources caused by unequal distribution forces growing populations to migrate from regions where resources are scarce to regions that are ecologically fragile and extremely vulnerable to degradation‖. 16 scarce resources are captured by elites, the weak groups are suffering even greater scarcity. In the meantime, the weakened institutions manipulated by elites cannot respond effectively to grievances of the poor, thus the victims tend to resort to violent means to obtain resources for survival. Another possible outcome produced by resource scarcity is ecological marginalization, which, according to Homer-Dixon (1999), results from human migration. When resource-deprived groups migrate into an area with a fragile eco- system, it is very likely that the environmental pressure in the area will be aggravated and the newly migrated people will have conflicts with the native people. In sum, Homer- Dixon and his colleagues suggest that scarcity of renewable resources such as forest, fisheries, cropland and water can give rise to a series of social effects, including economic decline, social segmentation and human migration. It is through these social effects that environmental scarcity increases the risk of violent conflict. Although the Toronto Group, according to Homer-Dixon‘s works, agree with earlier Malthusians that population and resource pressures in developing countries often generate intensions among social groups, they also contend that ―environmental scarcity does not inevitably or deterministically lead to social disruption and violent conflict‖ (Homer-Dixon and Blitt 1998:7). Instead, they have depicted an obscure and indirect causal pathway from environmental degradation to social violence, where scarcity interacts with political, economic, and other conditions of the society in question (Homer- Dixon 1999:178). States with a strong economy and effective political institutions sometimes are able to avoid falling into such deprivation. Nevertheless, non- 17 representative states and economically poor states are said to be prone to environmentally-induced conflict because they lack the adaptive capacity to overcome the challenges. With a series of case studies, they find that the combination of resource scarcity and limited adaptive capacity has generated or contributed to violent conflicts in a number of regions (Homer-Dixon and Blitt 1998). Furthermore, they conclude that scarcity can hinder institutional and technological adaptation by reducing the supply of ingenuity 13 available in the society (Homer-Dixon 1995b, 1999, 2000). 14 Another well-known research team, the Swiss ―Environmental Conflicts Project‖ (ENCOP) under the direction of Bä chler (1998) starts with slightly different assumptions, with more emphasis on the role of institutions (Swatuk 2004). They suggest a more complex model linking maldevelopment, environmental degradation and violent conflict. On the one hand, they believe that socioeconomic and political factors greatly contribute to civil conflicts triggered by resource degradation. On the other hand, it is argued that environmental scarcity can lead to or exacerbate social and political maldevelopment and therefore challenge the political stability of a state (Bä chler 1998:24). They think that maldevelopment of developing and transitional societies usually results from the ways in which they ―have experienced modernization‖, and the outcome of maldevelopment in these places is typically ―a weak state form imposed on a multi-ethnic society‖, that is 13 Homer-Dixon (1999:111) defines the supply of ingenuity as the amount of ingenuity ―actually delivered and implemented by the economic and social system‖. 14 During the different stages of their projects, Homer-Dixon and his colleagues published a series of articles in the journals of security studies and brought broad attention to the environmental security issue. The major pieces are Homer-Dixon (1991, 1994, 1995a, 1995b). Based on the findings of their projects, they then published several books discussing the topic in a more theoretical and systematic way (Homer-Dixon and Blitt 1998 ; Homer-Dixon 1999, 2000, 2006). 18 ―dominated by one ethnic group dependent on one or a few primary commodities for revenue‖ (Swatuk 2004:9). It is within this weak state form that conflict is most likely to take place. Forty case studies on countries whose ―transformation of society-nature relationships was perceived as serious‖ have been undertaken and eighteen of them have ―crossed the threshold of violence‖, with eight leading to war and ten having ―violent conflicts below the threshold of war‖ (Bä chler 1998:32). The group trace back the environmental roots and associated causal pathways of violent conflicts and wars in countries such as Rwanda to demonstrate the interaction of ethnic, social, political and ecological factors (Bä chler 1999). The project‘s major findings agree with Homer-Dixon‘s conclusion that environmental scarcity does not automatically leads to violent conflict and they also emphasize the importance of political, economic and social context in the causation of a conflict (Bä chler 1998). However, Bä chler and his colleagues go further to find substantial evidence that ―environmental scarcity causes large population movements, which in turn cause conflicts‖ (Bä chler 1998:31). Moreover, Bä chler (1999:98) suggests the possibility of environmental conflicts becoming catalyst for cooperation when ―political compromises are seen as desirable and technical solutions are feasible‖. The neo-Malthusian account that population and resource pressure can bring conflict is challenged by critiques on several grounds. The main criticism focuses on its suffering from a ―demographic and environmental determinism‖ (Kahl 2006:11). Some neo-Malthusian models depict an automatic causal linkage between population and 19 environmental stress, and also assume the causal connections between environmental stress and violent conflict. They thus ―exaggerate the causal importance of demographic and environmental factors, and ignore or downplay crucial intervening variables and processes‖ (Kahl 2006:11). However, revised neo-Malthusian theses proposed by the Toronto Group and ENCOP program deny the simple and straightforward causal linkage between the two. By taking into account a number of intervening variables assessing states‘ adaptive capacity, revised neo-Malthusians seek to provide a more accurate picture of the relationships (Bä chler 1998 ; Homer-Dixon and Blitt 1998 ; Homer-Dixon 1999). But unfortunately, as Kahl (2006:13) suggests, the ―laundry list of important intervening variables‖ identified by them remains underspecified. Additional clarification is needed to explore how these intervening variables interact with environmental pressures to produce violent conflict. 15 Neoclassical Economists The controversy between neoclassical economists and neo-Malthusians has been going on for a time. Current neoclassical economic research challenging neo-Malthusian theses on the relations between environmental scarcity and conflict can be approximately divided into two groups. One is composed of economic optimists, and the other is the newly developed ―resource curse school‖. 16 While the former criticize neo-Malthusian 15 Besides Kahl (2006), Gleditsch (1998) also makes a similar argument in his critique of literature on conflict over resource scarcity. He summarizes the critique into nine points, including the lack of concept clarification and various methodological issues. 16 Here we follow Kahl (2006:13-21) to view the resource curse school as part of the neoclassical economist tradition. Kahl names their major arguments as ―the honey pot hypothesis‖ and ―the resource curse hypothesis‖. 20 pessimism, the latter suggest that resource abundance rather than scarcity is the true cause of domestic violent conflict. Claiming that neo-Malthusians are overly pessimistic about the negative impacts of rapid population growth and environmental degradation, neoclassical economists such as Lomborg (2001) and Simon (1981, 1996) call into question the causal connections between environmental scarcity and civil strife, maintaining that societies rarely fall into the traps of violent conflict thanks to their adaptive capacity. Two main arguments are raised by them to challenge the neo-Malthusian paradigm. First, optimists claim that the analysis of relevant data does not suggest that the world is going to experience a major resource crisis, at least not at a global scale (Lomborg 2001). Second, they are confident that humankind is able to achieve adaptation even if resource challenges do get exacerbated (Urdal 2005). In the first place, it is expected that the demand for scarce resources will decrease with the rising market price. Besides, natural resource scarcity is also believed to be an important catalyst for technological innovations that can reduce future scarcity. For example, Boserup and his colleague (Boserup 1965 ; Boserup and Schultz 1990) point out that high population density actually imposes pressures on societies to accelerate their adaptation to new technologies. Another economist Simon (1981, 1996) believes that population, with its ability to generate human innovation, is actually the solution to resource scarcities. As the innovative process produces more advanced technology, more resources are available to the human society. Instead of preoccupying themselves with the association between resource scarcity and conflict, another group of neoclassical economists, collectively called 21 ―resource curse school,‖ argue since the last decade that abundance of valuable natural resources can become the cause of economic stagnation, corruption, and civil war. According to them, the endowment of natural resources may be a curse rather than a blessing. The school establishes its major analytical framework with Collier and Hoeffler‘s articles (1998) sponsored by the World Bank in the late 1990s. The studies explore the economic causes of civil war based on utility theory, arguing that decision of rebellion is based on the cost-benefit analysis of the rebel groups. That is to say, ―the incentive for rebellion was increasing in the probability of victory, and in the gains conditional upon victory, and decreasing in the expected duration of warfare and the costs of rebel coordination‖ (Collier and Hoeffler 1998:571). Therefore they contend that opportunities for primary commodity predation cause conflict. Collier and Hoeffler go further to unpack the relationship that natural resource endowment is an important determinant of the probability and duration of civil war, and the effect is nonmonotonic. At a lower level, the amount of natural resources is positively associated with the risk of war. This is due to the attraction of the ―taxable base of the economy‖ to rebel groups, since revenue from selling natural resources can provide financial support for the start-up and sustainability of rebellions (Collier and Hoeffler 1998 ; Collier 2000b). But the endowment of natural resources begins to decrease the risk of civil war when it passes some threshold level. Collier and Hoeffler (1998:571) argue that the shifting relationship between the two variables is due to the increased ability of the government to ―defend itself through military expenditure‖ as a result of its ―enhanced financial capacity of the government.‖ 22 In their later works, Collier and Hoeffler (2004a, 2000) extend and revise the theoretical framework to be a more comprehensive one. They propose a theory combining two competing models stating that rebellions are either motivated by ―greed‖ or ―grievance‖. The economic model takes rebellion as ―an industry that generates profits from looting.‖ It thus assumes that rebellions are motivated by the rebel group‘s ―greed‖ for private gain in the process. This to a great extent resembles Collier and Hoeffler‘s earlier work in maintaining that the existence of ―lootable‖ resources provides the opportunities for rebellion via financing arms and labor. As rebel groups care about the profit, the levels of natural resources and the government‘s ability to impede the rebellion together put a constraint on the size of the rebel group and the possible returns to rebellion. According to this economic model, the authors suggest that natural resource abundance increases the likelihood of civil conflict whereas higher opportunity cost of rebellion decreases its probability. In contrast, the ―grievance‖ argument explains the motivation of civil conflict from the perspective of political scientists, claiming that rebellions are results of highly acute grievances of people. According to Collier and Hoeffler (2000), the interplay of a series of factors such as autocratic regimes, highly unequal land and income distribution, and ethnic and religious fractionalization can generate and reinforce the grievance of social groups and then increase the risk of violent conflict. In the meantime, it has also been noticed that grievance-driven rebellions are harder to be put into practice than 23 greed-driven rebellions due to a more serious collective action problems. 17 Moreover, the same factors that generate grievance also aggravate the collective action problem. For example, ethnic and religious fractionalization may not only produce grievance since the marginalized social groups are demanding for social justice, it can also make the coordination of a rebellion more difficult due to the social fragmentation. Therefore, the authors find some of these factors have ambiguous effects on civil war. Collier and Hoeffler (2004a) try to integrate these two models as a common explanation of ―opportunity.‖ They thus bring a concept of ―misperceptions of grievances‖ based on Hirshleifer (1995, 2001). Unlike grievances in the above political scientist model, ―misperceived grievances‖ are not exactly connected to factors like inequality, political rights, and ethnic or religious identity. Although they are misperceived, they can also trigger rebellions, and in turn fighting in violent means can generate genuine grievance and again reinforce the motivation for war. Therefore, according to Collier and Hoeffler (2004a) ―opportunity‖ does not only explain ―greed‖ as the economic motivation of rebellion, it is also consistent with the ―grievance‖ model if the spread and acuteness of ―perceived grievance‖ is concerned. A number of subsequent projects contest the general applicability of this causal story, mainly focused on physical configurations of specific resources. One of the explanations for resource predicament suggests that the effects of resource rents on political stability are indirect, working via the effects of resources on state institutions 17 According to Collier and Hoeffler (2000:13-14), ―grievance-assuagement‖ is a public good not provided by the government. It has a severe free-rider problem in financing because hard core adherents and non-participants experience different levels of grievance. 24 (Fearon and Laitin 2003c ; Fearon 2005 ; Ross 2004). Recent research goes further to advocate a systematic discussion of how the state, regime type and economic institutions enter into this picture (Dunning 2005 ; Collier and Hoeffler 2005). In sum, neoclassical economists question the inevitable causal relations among population growth, environmental degradation, and violent conflict. They have made important contribution to the debate by taking into account the adaptive capacities of markets and societies. However, as Kahl (2006:16) states, the optimists tend to be overly confident about the prospects of adaptation, particularly the adaptability of developing countries. The prospects for economic growth in the context of rapid population expansion rely on ‗the initial level of economic development‘ and ‗the adoption of appropriate economic strategies,‘ both of which are particularly lacking in most of the poor countries (Kahl 2002 ; 2006:17). Moreover, their arguments work much better at explaining the impacts of nonrenewable resources such as oil, minerals, and diamonds than renewable resources (Kahl 2006:19). Political Ecologists As a new subject created by anthropologists in the 1970s (Wolf 1972 ; Cole and Wolf 1974), political ecology represents a third approach in the debate. Its practitioners criticize the neglect of the political dimensions of human and environment interactions in the other two schools by emphasizing the relations of power (Robbins 2004). In general, political ecology is believed to lie more in Marxist tradition and green environmental politics rather than ecology and anthropology (Vayda and Walters 1999 ; Peluso and 25 Watts 2001). The burgeoning field of political ecology has attracted scholars from vary backgrounds and training to query the relationship between economics, politics, and nature. According to political ecologists, previous research on the topic focus on the homeostatic or adaptive processes, treating human communities as homogeneous, autonomous units (Durham 1995 ; Moore 1996). With emphasis on the political dimension of environment and violence, political ecologists assume ―a reciprocal relationship between nature and humans‖ and thus the starting points of their research is ―the relations between users and nature‖ rather than ―a presumed scarcity or precursor ideational factor‖ (Peluso and Watts 2001:27). Most political ecologists share three fundamentally linked assumptions summarized by Bryant and Bailey (1997:28-9): (i) ―costs and benefits associated with environmental change are for the most part distributed among actors unequally‖; (ii) this unequal environmental change ―reinforces or reduced existing social and economic inequalities‖; (iii) it also holds ―political implications in terms of the altered power of actors in relation to other actors.‖ Their research thus tends to reveal cost-benefit analysis, winners and losers, and power relations behind the environmental issue (Robbins 2004). According to political ecologists, scarcity and environmental degradation result from unequal access to common pool resources and the political failure of both domestic and international justice systems (Suliman 1999). For example, Hildyard (1999:14) argues that ―the systematic inequalities that block people‘s access to income, health, 26 education and democratic rights‖ are the major cause ―for the geographical and sociological ‗profile‘ of ecological degradation‖. Consequently, instead of viewing scarcity as a natural limit, political ecologists take it as an artifact experienced by poor communities in developing countries due to the unequal political and economic structure. Following their line of argument, environmental degradation and conflict are not causally linked (Robbins 2004). Like economic optimists, political ecologists believe that population growth and environmental degradation are not important sources of scarcity or violence. To some extent, they also agree with each other that ―resources are constructed rather than given‖ (Robbins 2004:8). Therefore, they believe it is incorrect to view the amount of resources as finite and unchanging, imposing absolute limits on human action. However, neoclassical economists focus on the role of markets, and political ecologists emphasize the unequal distribution of wealth and power. Political ecology also mirrors some neo-Malthusian arguments in that local communities may rise up to challenge the unequal resource distributions in violent confrontation with the state (Kahl 2006). But the two schools disagree with each other on the importance of natural versus social sources of scarcity. Whereas neo-Malthusian accounts often contend that natural sources of scarcity is much more important than social ones, political ecologists concentrate on the latter and do not see demographic explanations as critical predictors of environmental crisis. Although political ecologists make an important contribution to the debate by bringing in the political dimension of environmental change and human-environment 27 interactions, many of them go well beyond to insist that political influences especially political influences from the outside political economic system are always more important than natural conditions, and should always be given priority in research. They thus not only ignore factors other than the political dimension, but also miss the complex and contingent interactions of factors, from which environmental changes often arise. As some scholars suggest, overemphasizing the influence of politics could result in a ―politics without ecology‖ (Vayda and Walters 1999). To summarize, neo-Malthusians and neoclassical economists disagree with each other on the causal links between environmental scarcity and violent conflict. Neo- Malthusians warn us the danger of population growth and environmental degradation, but neoclassical economists tell us that society can adapt to the challenges and maintain its order. They both suffer from simplification and overemphasize one side of the story. Political ecologists provide an alternative which views scarcity as socially constructed. But their weakness also lies in that they always put distributional inequality as the cause of ―illusive‖ scarcity. The relationship among the major arguments of the three schools of thought can be simplified in the Figure 1. 28 Figure 1: Venn Diagram of the Relationship among the Three Schools As the Venn diagram shows, the three schools have established quite different causal mechanisms about the impacts of resource scarcity on political instability, in particular civil conflict. 18 In the meantime, they also share with each other some of the basic assumptions. For example, both neoclassical economists and political ecologists deny the forthcoming of large-scale resource scarcity and emphasize the critical role of human actions on natural resources and political instability. Moreover, political ecology and neo-Malthusian theory tend to agree with each other in that resource pressure is an important trigger for insurgencies in local communities. Finally, the overlapping area between neo-Malthusians and neoclassical economists represents the major casual frameworks proposed by the revised neo-Malthusian scholars. Although they agree with 18 Venn diagrams are useful tools to visually display the logical relations of several sets. They were introduced by John Venn in 1880 (1880). Ruskey and Weston (1997) and Edwards (2004) introduce the history and various facts and figures about Venn diagrams. Political Ecologists Neoclassical Economists Neo-Malthusians 29 traditional neo-Malthusians that scarcity can lead to civil conflict, improved neo- Malthusian models such as those of Homer-Dixon and Bä chler depict much more complex causal links. They take into account the neoclassical economist arguments in claiming that the effects of environmental scarcity on civil wars are not direct and linear, and human ingenuity and adaptability play an important role in mitigating or reversing the trend. They also provide a framework which makes it possible to bring in political ecologist arguments on resource inequality to explain the potential social sources of scarcity. Empirical Evidence Quantitative studies on environmental scarcity were sporadic before the late 1990s. After then more large-N studies have been conducted by scholars in the field. Most of the statistical tests utilize data from four well-known civil war datasets: Collier and Hoeffler (2004b), Fearon and Laitin (2003c), Doyle and Sambanis (2000), and Gleditsch et al. (2002), all of which, in turn, draw upon the pioneering ―Correlates of War‖ intrastate war dataset compiled by Singer and Small (1994). Most of the available statistical analyses on the topic test the relationship between the scarcity of specific resources and political instability, especially civil conflict and civil war. For example, Hauge and Ellingsen (1998) test the effect of deforestation, land degradation, and scarce supply of freshwater on civil war in the period 1980–1992. A modest, but significant effect of the scarcity variables has been found. However, another cross-national time-series study conducted by Urdal (2005) covering the period 1950– 30 2000 does not provide strong support for either the pessimistic or optimistic perspective. According to the author, the overall robustness of the empirical support for both models is low. Instead of addressing the scarcity of each resource, some scholars test a more general argument regarding the effects of resource scarcity on conflict by examining comprehensive indexes measuring environmental scarcity of various resources. De Soysa (2002b) utilizes the natural capital data from the World Bank and applies maximum likelihood analysis to test neo-Malthusian claims on renewable resources and minerals respectively. With data for the entire post-Cold War period, the author concludes that the risk of conflict increases with the abundance of mineral wealth, but is largely unrelated to renewable resources. Evaluating another well-known environmental index—the Ecological Footprint—Binningsbø et al (2007) claim that countries with a heavier footprint, which indicates higher resource consumption, have a substantially greater chance of peace. Biological capacity and the ecological reserve are also found to be predictors of peace, but the results are more fragile. Overall, both analyses using comprehensive environmental indexes provide little support for the neo-Malthusian model of conflict. Moreover, The Task Force‘s work broadens the range of the dependent variable by examining four types of acute instability: revolutionary war, ethnic war, adverse regime change, and genocide or politicide. They found that a country‘s political institutions are crucial determinants of its vulnerability to political instability, and material well-being also have long-term effects on the probability of acute instability. But 31 both the scarcity-grievance approach and the abundance-greed argument receive little evidence from their results. Population, environment, and resource pressures alone are found to be neither necessary nor sufficient causes of violent political instability (Esty, Gladstone, et al. 1998). In the meantime, several methodological problems of these projects have been discussed in the field. De Soysa (2002a) suggests that rates of deforestation and soil degradation, which are reported to be significant predictors in Hauge and Ellingsen‘s research (1998), do not really measure environmental scarcity. Without accounting for the available stock, such environmental variables do not accurately capture the degree of scarcity in each country. Furthermore, Theisen (2008) fails to replicate the study of Hauge and Ellingsen (1998) after ten years. He then provides a reanalysis with new data which shows that only the level of soil degradation is a significant predictor of violent conflict. In general, ―only weak support for an environment-conflict linkage‖ has been found in the reanalysis. For the Task Force‘s analysis, King and Zeng (2001a) identify and correct several methodological problems that ―lead to overly large forecasts, exaggerated assessments of forecasting performance, and biased causal inferences.‖ 19 They also bring in some ―underulitized‖ methods and procedures to improve the analysis. As a matter of fact, more quantitative studies in the field are designed to test the resource curse thesis. In a series of studies, Collier and Hoeffler report (1998, 2002, 2004b, 2005) that there is a positive association between primary commodity exports and 19 More State Failure Task Force Reports can be accessed on their website at <http://globalpolicy.gmu.edu/pitf/pitfdata.htm>, including the overview of Esty et al. (1998), and findings of different phrases such as Esty et al. (1995), Esty et al. (1998), Goldstone et al. (2000), Bates et al. (2003), Goldstone et al. (2010). 32 the danger of civil war in a county. Examining up to 52 civil wars over the period from 1960 through 1999, they find that a state‘s likelihood to have a civil war over the next five years increases with its degree of dependence on natural resources. Their research has attracted much attention in both academia and the policy realm. A number of scholars have tried to replicate Collier and Hoeffler‘s findings (Fearon and Laitin 2003c ; Elbadawi and Sambanis 2002 ; Hegre 2002). However, using different databases of civil wars, their results vary. Ross (2004) has summarized the findings of 14 cross-national econometric studies on ―lootable resources‖ and concluded that there appears to be ―little agreement on the validity of the resource–civil conflict correlation.‖ For example, Hegre (2002) finds that primary commodities become less important when a lower death threshold is used for defining conflict. Fearon and Laitin‘s (2003c) tests reports that neither the ratio of primary commodity exports in GDP nor its square is remotely slightly significant in their model covering the period from 1945 to 1999. The authors thus doubt that primary commodity exports are a good measure of financing potential for rebels. Furthermore, in a more recent study, Fearon (2005) suggests that only substantial production of oil might have an association with the risk of civil war because of ―relatively low state capabilities‖ of oil producers and oil‘s high value relative to its costs of production. Elbadawi and Sambanis (2002) obtain even more ambiguous results which show that the significance of primary commodity exports varies largely across models. They cannot find solid evidence for Collier and Hoeffler‘s thesis since the results are 33 highly dependent on the missing data imputation methods and the procedure of variable transformation. Rather than using civil war onset as the dependent variable, some quantitative studies investigate the influence of primary commodity exports on the duration of civil wars. For example, Doyle and Sambanis (2000) suggest that the share of primary commodity exports in GDP is negatively associated with peace-building success efforts in 124 post-World War II civil wars, which implies that primary commodity exports provide incentives for new wars. However, Collier et al. (2004) examine 55 civil wars between 1960 and 1999 and find no significant correlation between primary commodity exports and civil war duration. The studies listed above have shown inconsistent results regarding environmental scarcity‘s effects on violent civil conflict. As Ross (2004) suggests, there might be two possible sources of variation. First, scholars employ different civil conflict databases for their statistical analyses. Another reason according to Ross (2004) is that ―primary commodities‖ and ―civil war‖ are overly broad variables so that commodities that appear to be strongly linked to conflict and those that do not are both included in primary commodities variable. Moreover, another group of quantitative studies focused on the effects of land maldistribution and income inequality on conflict, which to a great extent test the political ecologist proposition on artificial scarcity due to distributional inequality instead of the absolute amount of resources. According to their theories of relative deprivation (RD) and resource mobilization, inequality creates discontent among urban people and 34 peasants, who are then mobilized by the ―vanguard of urban professional revolutionaries‖ (Muller and Seligson 1987). The cross-national study of Muller and Seligson (1987) finds that income inequality increases the likelihood of mass political violence when other factors such as the repressiveness of the regime, governmental acts of coercion, and level of economic development are taken into account. But land maldistribution turns out to be irrelevant to the risk of political violence. The major explanation according to them is that rural populations who are the main victims of land maldistribution are quite difficult to be mobilized for political protest. Although a series of later studies also find a positive connection between income inequality and armed conflict (Muller et al. 1989 ; Boswell and Dixon 1990 ; Timberlake and Williams 1987), Collier and Hoeffler (2001) fail to find a significant association between the two variables. Besides, Schock‘s (1996) empirical study shows that the relationship between economic inequality and violent conflict is moderated by political opportunity structures. Lichbach (1989) critically reviews the major studies on the linkage between income inequality and conflict. According to his review, contradictory findings have been reported due to the variations in all aspects of research design, including the definitions and measurements of economic inequality and political conflict, the inclusion of cases, time frames, and the specification of control variables like economic development level and regime type. Moreover, he also recommends a combination of different approaches such as statistical modeling, formal modeling and theoretical building as the hope of solving the puzzle. In addition, data on inequality are still very limited and often not 35 comparable (Gissinger and Gleditsch 2000). Some scholars use Gini coefficient of income inequality and land ownership (Collier and Hoeffler 2004b ; Deininger and Squire 1996), others utilize the concentration of income in the top 20% of the population (Gissinger and Gleditsch 2000). Besides the inconsistency in measurement of inequality, the income inequality data are only available for less than 100 countries, with a number of missing values. In sum, empirical research especially large N studies on the resource scarcity-civil conflict nexus have brought in a variety of datasets and control variables to figure out the possible mechanisms between the two variables. Generally speaking, inconsistent results have been found by the studies, most of which are times-series cross-section analyses. But we can reach some broad conclusions based on current findings. First, it is agreed among the major research that different types of resources tend to have different impact on the risk of conflict, especially that the renewable and nonrenewable resources should not be assumed to relate to civil conflict in the same way. Secondly, the results seem to be sensitive to the measurement of resource scarcity and internal conflict. Integrating and comparing the data from various databases is necessary for better understanding of the statistical results. Finally, consent has been reached that important explanatory variables such as the level of economic development, political regime type, land and income inequality, ethnic and religious fractionalization, and population growth cannot be ignored in civil violence studies. Based on the above review of the major theoretical and empirical findings on the topic, the following section starts to build the research framework of this study. Since the 36 available studies do not address the core concepts such as resource scarcity, political instability and environmental adaptability in exactly the same way, it is necessary to clarify the three core variables and the detailed research plan. The chapter then goes further to propose a comprehensive ―adaptability‖ model to explain the causal mechanisms among the three variables. Based on the model, a series of hypotheses are presented in the end for empirical tests in the later chapters. Key Concepts of This Study Major studies of the three schools have depicted the linkage between environmental degradation, natural resource scarcity, and deadly conflict within a country in quite different ways, but the empirical research so far has only provided contradictory evidence for the main arguments of the three schools. This implies that some of the claimed causal linkages might not capture the complete picture. Therefore, a more systematic theoretical framework that bridges the causal mechanisms suggested by different schools and illuminates the inconsistent empirical results is required. In order to fill the gap, this study proposes a more comprehensive causal framework that connects natural resource scarcity to political instability, adding adaptability as the critical intermediate variable. The model draws extensively on the neo-Malthusian arguments regarding the causal relationships between renewable resource scarcity and civil violence. It also brings in the resource curse hypothesis of the neoclassical economists to explain the distinct linkage between nonrenewable resource and rebellion. Besides, insights from the political ecologist assessment on maldistribution 37 and revolution, as well as more recently studies on civil conflict are also incorporated in the framework. Before a detailed explanation of the model is presented, this section will define the three key concepts of this study—scarcity, political instability and adaptability. Scarcity The concepts of ―natural resource scarcity‖ and ―environmental scarcity‖ are used to capture the scarcity of both renewable (i.e. arable land, forest, fresh water and fish stocks) and nonrenewable natural resources (oil, minerals, coal etc.). In this model, renewable and nonrenewable resource scarcities are investigated separately since their different utilities and features are assumed to have different impacts on the risk of political instability. The formation of nonrenewable resources by nature is on a timescale that largely falls behind their consumption rate. 20 Fossil fuels such as oil, natural gas, and coal as well as minerals are examples of nonrenewable resources. In contrast, renewable resources such as clean water and timber can be recreated or recycled during human lifetime. Most renewable resources are the basic supplies for human survival (i.e. drinking water, food), whereas many nonrenewable resources are either ―lootable‖ resources (i.e. diamond, silver, gold) or materials for industrial production (i.e. steel, petroleum). Furthermore, this study intends to talk about the ―objective‖ scarcity or abundance of natural resource rather than resource dependence or ―mispercepted‖ resource scarcity 20 Some suggest that minerals should be viewed as renewable resources since they can be recycled. However, more believe that they are nonrenewable since they cannot be regenerated in human lifetime. This study takes minerals as nonrenewable resources because their economic utilities are more similar to fossil fuels, and often more ―lootable‖ than renewable resources. 38 (de Soysa 2002b). This is not exactly the same as neo-Malthusian perception of scarcity, which is categorized as supply-induced, demand-induced, and structural scarcity based on the distinct ways that they are produced. It is also different from primary commodity exports used by resource curse thesis, which mainly measure a country‘s economic dependence on exporting raw materials. Besides, it is also differentiated from the scarcity due to maldistribution suggested by political ecologists. This study consolidates their theses by identifying a concept of natural resource scarcity that objectively reflects the resource challenge to a country. That is to say, whether a country is scarce in a resource depends on the per capita possession of that particular resource, which does not have to reflect the different resource demands due to factors such as consumption level and exports. For example, the water resource scarcity is measured by fresh water resource per capita of a country, without considering the per capita water consumption level of that particular country. More complex interplays between objective resource abundance/scarcity and human factors are included in the investigation on a country‘s adaptive capacity. Furthermore, this study follows Homer-Dixon‘s note on scarcity that ―all types of environmental depletion or damage‖ are interpreted as ―various forms of scarcity‖ of resources (Homer-Dixon 1999:9). That is to say, water pollution increases the scarcity of fresh water, deforestation leads to forest resource scarcity, and the depletion of minerals increases a country‘s scarcity of mineral resources. Rather than exploring the complex processes that lead to environmental degradation and resource depletion step by step, this 39 study takes them as a whole and only addresses the aggregated results of these interplays as the ultimate indicator of resource scarcity/abundance. Political Instability Most of the studies on the relationship between resource scarcity and internal security are interested in cases of civil war, which is an extremely intense situation of governmental crises. Even for a relatively long history, we can see that wars are rare events that occur infrequently. A great number of time-series cross-section studies on civil war use dichotomous variables such as civil war onset as the dependent variables, which is characterized by thousands of 0‘s (nonevents) and a hundred or even less 1‘s (events of civil war). King and Zeng (2001b) point out a couple of methodological problems with the logit analysis of rare events in international security studies. First, the estimated probabilities of civil war onset are always very small since the sample sizes of those panel studies are in the thousands and the values of the dependent variable (civil war onset) are always in the same direction. Logit analysis is ―suboptimal‖ in analyzing these rare-events data with finite samples. Because it can lead to ―errors in the same direction as biases in the coefficients‖, and thus underestimate the probabilities of civil war onset (King and Zeng 2001b:693). In addition, they suggest that most of the datasets collected by researchers in the field have a large number of observations (i.e. country- year data) with few poorly measured explanatory variables. The large number of observations makes it difficult to collect explanatory variables for all the cases. 40 Given those problems in analysis of rare events, this study decides to expand the dependent variable to be political instability to cover those less intense forms of political and social crises. Alesina and Perotti (1993) summarize that political instability are in general defined in two approaches. The first perspective views political instability as the ―propensity to observe government changes‖, which emphasizes the executive instability of a state (Alesina and Perotti 1993:2). This definition includes both constitutional and unconstitutional executive changes. Since it concerns about the ―propensity‖ to change, it thus can be taken as probabilities distinct from the actual frequency of changes. 21 The other approach, according to Alesina and Perotti (1993), measures political instability based on a variety of indicators of social unrest and political violence. 22 The decision of which definition to use depends on the research issue of interest. While most studies in the field focus on political instability in the latter approach, this study hopes to define the concept in a more comprehensive way to incorporate both dimensions. Therefore political stability is defined as ―the regularity of the flow of political exchanges‖ (Ake 1975), and political instability includes a variety of situations when the regular flow of political exchanges are impeded. A number of events such as assassinations, coups d‘état, strikes, major government crises, riots, and anti-government demonstrations are all indicators of political instability. Civil war is also carefully 21 Studies on inflation such as Cukierman et al. (1992) and Edwards and Tabellini (1994) use this definition of instability. Londegran and Poole (1990) Alesina et al. (1996) further address the interaction between political instability and economic outcome as an endogeneity of the definition. 22 Hibbs (1973) uses principal component analysis to construct an index of political instability following this approach and, a number of studies such as Venieris and Gupta (1986), Gupta (1990), Barro (1991), Ozler and Tabellini (1991), Venieris and Sperling (1994) and Benhabib and Spiegel (1994) also create indices of political instability in their political economic research. 41 investigated because it is an extreme case of the deterioration of a state‘s political order, which indicates high political instability. Moreover, important studies and datasets on civil war are also incorporated in this research since the major theoretical frameworks in the field are based on analyses of those data. 23 Besides the methodological consideration, this concept of political instability also enables us to look at the impact of resource scarcity on less intense social crises. Since few of the studies on resource scarcity-civil war nexus find strong support for their direct connection, it suggests that resource scarcity itself might not be enough to arouse grievances that can ultimate lead to war. Although environmental disputes are not rare in our daily life, most of them seem to be at lower intense level. It is common to see people protest for the building of chemical factories or local residents demonstrate for their mineral resources. However, they seldom escalate into big conflict that can actually be identified as war, which features much higher level of deaths. For instance, more than a year ago a movement in Colombia called on the government to halt the activities of gold mining plans in a protected area (Martí nez 2009). The local farmers and activists wanted to preserve their forests and water resources. Also in 2009, a Brazilian fisherman died in the protest against irregularities in the construction of a gas pipeline in rainforest (Frayssinet 2009). However, neither of them involved mass scale violence and so far it does not see the probability of escalating to war. 23 A measurement of civil war is used as an indicator of the dependent variable in the comprehensive regression analysis, and annual number of civil war ongoing is used in the auto-regressions. More details about how to incorporate civil war datasets and findings in this study are explained in Chapter 3. 42 For some of the well-known civil wars that do involve resource issues, it is still arguable whether resource dispute is the main cause of war. Cases studies on the Rwandan genocide (Bä chler 1999 ; Ohlsson 2000) and the clashes in Kenya and South Africa (Kahl 1998, 2006 ; Homer-Dixon and Blitt 1998) have argued that resource scarcity play a very important part in these violent conflicts through its entanglement with complex factors such as ethnic identities, state failures, and poverty. By broadening the dependent variable, this study is not only able to examine the impact of resource scarcity with more observations of events (political instability instead of civil war), it can also find out whether resource scarcity is more likely to result in lower scale political instability as well as full scale wars. As it can been seen, cross-sectional statistical analyses on less intense forms of political instability other than civil war have been much less. The datasets compiled by State Failure Task Force intend to forecast when state will fail. State failure, which is their major variable of interest, refers to crises of the central government that make it fail to deliver order. Their coding of state failure includes four types of severe political crises: revolutionary wars, ethnic wars, adverse regime changes, and genocides and politicides. While revolutionary wars and ethnic wars are two different types of civil wars, adverse regime changes and genocides/politicides are at lower level in terms of death scale. Adverse regime changes refer to major, adverse shifts in patterns of governance and genocides/politicides are defined as physical termination of enough members of a target group by the authorities. The task force find little evidence for either neo-Malthusian or ―resource curse‖ thesis (Esty, Gladstone, et al. 1998 ; Goldstone et al. 2005), and King 43 and Zeng (2001a) identify several methodological errors in the task force‘s work that lead to exaggerated forecast probabilities of conflict and biased inferences. However, it is still an important project that attempts to collect data on state failures besides civil war. There have been some studies on coup (Belkin and Schofer 2003) and genocides. Few of them focus on the role played by resource scarcity/abundance, though some studies mention the lack of diversification in exports (O'Kane 1987), which to some degree refers to the export of primary commodities in some resource abundant developing countries. Thus by performing systematic analyses on the linkage between resource scarcity and political instability, this study intends to examine the effects of resource scarcity in a more comprehensive and more accurate way. Adaptability In the debate, adaptability is viewed by the optimists as a critical ability of the human society to adjust to resource challenges and it has also been included in the theoretical framework of revised neo-Malthusian and political ecologist frameworks. However, no concise and widely-accepted definitions have been found in current literature. What is worse, most of the quantitative studies do not consider a country‘s adaptability to resource pressures in their models. As Homer-Dixon (1999) and Bä chler (1998) argue in their case studies, resource scarcity itself is not enough of a predictor of political instability, or civil conflict in particular. The capability of a society to adapt to the challenges with ingenuity becomes a very critical factor in the possible causal relationship between resource scarcity and political instability. It would be deviant from 44 reality if we simply look at the direct connections between environmental scarcity and political instability. Hoping to fill the gap, this study introduces and quantifies adaptability to explore more accurate mechanisms better explaining the complex relationship. Based on its original definition in biology, adaptability or adaptive capacity refers to the ―ability to become adapted (i.e., to be able to live and to reproduce) to a certain range of environmental contingencies‖ (Gallopí n 2006). And adaptation is a feature of structure, function, or behavior of the organism that is instrumental in securing the status of being adapted (Dobzhansky 1968 ; Gallopí n 2006). As mentioned before, the concept of adaptability or adaptation is addressed by some case analyses of revised neo- Malthusians, economic optimists and political ecologists (Bä chler 1998 ; Homer-Dixon 1999 ; Kahl 2006). Homer-Dixon and Blitt (1998:ch1) argue that society might be able to adapt to resource scarcity if it can generate enough social and technical ingenuity. According to them, ingenuity is reflected in institutions such as efficient markets and agriculture technologies increasing production. In contrast, adaptation failures like market failure, social friction, and lack of capital do not only reduce the supply of ingenuity, but also increase the requirement of it. Specifically, if a state fails to adapt to environmental scarcity, it then can interact with various contextual factors, resulting in constrained economic and agriculture productivity, social segmentation, migration, disruption of legitimate institutions, etc. Those social effects, independently or collectively, can lead to violent conflict among groups. In order to prevent conflict, 45 societies thus need to understand these links between resource scarcity, negative social effects and violence, and promote adaptation to scarcity (Homer-Dixon 1999). Economic optimists seem to have much more confidence in human society‘s adaptability to resource challenges. They suggest that a resource‘s price tends to rise as it becomes scarcer. People then either look for substitutes or develop new technologies that can increase efficiency, or simply decrease demand for that resource. Resources like petroleum are increasingly viewed as scarce or becoming scarce in the near future (Rees et al. 1989). Its rising price in the international market encourages the searching for alternatives like solar and wind power, and the research of fuel efficient automobiles. In the meantime, consumers tend to drive less and thus reduce the demand as the price is rising. Thus for economic optimists, adaptability is the essential factor that makes the society timely adjust to resource challenge before it becomes a serious crisis. Based on their theories, the key for these adaptations is the market, which provides price signals and incentives for actors. For example, the ―induced innovation‖ theory (Hayami and Ruttan 1985 ; Binswanger 1978) argues that in successful economies, market price accurately reflects the supply and demand of factors such as land, labor and energy. Then market price works as signals stimulating technological innovation to continue the society‘s growth. With its close connection to ecology, political ecology also puts great emphasis on adaptation. Same as the optimists, political ecologists also believe in that human action plays a dominant and subjective role in the nature-human interaction. Besides, they go further to point out that a ―humanized‖ nature is constructed, transformed, and 46 managed by human institutions and actions. The central problem for political ecology is to ―understand the processes by which human beings transform and reshape nature and in the process transform themselves‖ (Biersack 2006:126). Therefore, for political ecologists, human institutions and actions are the sources of artificial resource scarcity and also determinate whether a society can adapt to the environmental challenges. Thus in their theoretical framework, adaptability can even decrease or delete scarcity from its ―humanized‖ sources. Few quantitative studies in the field have tried to measure a state‘s adaptability to environmental challenges and its intermediate role in the environmental scarcity-political instability nexus. The available case studies at best indicate that a number of factors can affect a state‘s adaptability, and the role each factor play varies from case to case. Thus our knowledge about environmental adaptability so far is quite limited and implicit. This study tries to decrease the gap by quantifying and testing adaptability within the context of resource scarcity-political instability nexus. With the advantage in generalization, the quantitative analysis carried out by this study provides us a more systematic view of interactions among various factors. Based on the broad literature on political instability and civil conflict, this project figures out the major determinants of a state‘s adaptive capacity from a laundry list of control variables proposed by relevant research. It is suggested that these control variables moderate the association between resource scarcity and political instability to a different degree. As the concept of resource scarcity depict the availability of resources per capita in an objective sense, a verity of control variables used in security research can 47 be viewed as elements of adaptability. Although there is no consent on the exact factors that can decide a state‘s adaptive capacity of resource pressure, the major variables proposed to be significant predictors of civil conflict can be categorized as economic development level, political institutions, social fractionalization, and demographic characteristics based on the available research on civil conflict and ecology. The following section addresses how each element determines a state‘s capacity of adapting to resource challenge and summarizes the main empirical findings on their impact. Economic Development The level of economic development has been addresses by the theoretical models of each school as an essential factor that determine whether the pressure of resource scarcity can eventually lead to violent conflict in a country. According to revised neo- Malthusians, both financial and human capital availability decide the supply of ingenuity (Homer-Dixon and Blitt 1998). Without necessary financial capital, a country does not have enough funds to support research developing technologies of utilizing alternative resources and increasing resource efficiency. Neither can it build the infrastructures, communication networks that are foundations for technological innovations and governmental responses to resource challenges. Their assessment of efficient markets and market failures also reflects the level of a country‘s economic development to some extent. Generally speaking, more economically developed countries tend to have more mature and well-functioning markets where price signals more reliably reflect the degree of scarcity in the society. 48 It is no surprise that the level of economic development plays an even more critical role in the frameworks of neoclassical economists. The optimists have much stronger confidence in the market prices as effective signals and triggers for technological innovations which enable the human societies to adapt to environmental challenges. Thus for them, a well-functioning market which is also an important indicator of economic development is essential. For the economists of the Resource Curse School, countries with abundant ―lootable‖ resources but low level of economic development tend to rely on the export of primary commodities as their main source of income. Grievances generated by economic inequalities and/or financing opportunities from exports greatly increase the risk of internal conflict among groups. Finally, political ecologists claim maldistribution of land and income inequality as a cause of conflict, both of which are also important indicators of a country‘s level of economic development. Internationally, the maldistribution of resources and income gap between developing and developed countries make conflicts more likely in the developing world. When looking at a specific country, poor groups and areas suffer more from environmental pressures and violence. Therefore, countries with high inequality and groups at the bottom of a highly unequal society are more likely to resort to violence when facing scarcity. Almost all of the empirical studies on internal security test whether the level of economic development significantly predicts the risk of violence. It is often assessed by macroeconomic indicators such as Gross Domestic Product (GDP) per capita, Gross National Income (GNI) per capita, and strong dependency on export of primary 49 commodities. Moreover, unequal land and income distribution are also important measures of economic development which are assumed to influence people‘s tolerance of resource scarcity and a society‘s ability to consolidate when coping with scarcity. The statistical results tend to be significant in most studies, though some contradictory results of economic inequality have been reported. In an early study covering 65 countries from 1800 to 1960, Flanagan and Fogelman (1970:14) find that domestic violence occurs much less in countries at a high level of economic development. After then a number of works have verified the thesis using different datasets (Rapkin and Avery 1986 ; Hauge and Ellingsen 1998 ; Henderson and Singer 2000 ; Collier and Hoeffler 2002 ; Fearon and Laitin 2003c ; Sorli, Gleditsch and Strand 2005). In their comprehensive piece on civil war, Fearon and Laitin (2003c) show that per capita income is a highly significant predictor of civil war onset in different models ―in both a statistical and substantive sense.‖ With an econometric analysis, Collier and Hoeffler (2002) find poor economic performance to be the most significant predictor of conflict in African countries. Consistent results are obtained for conflict in the Middle East by Sorli et al. (2005), drawing a more complex picture. Furthermore, based on Rapkin and Avery‘s (1986) model, the level of domestic economic development is one of the most important factors that mediate the linkage between the effects of world markets and domestic political instability in third world countries. Besides, it is also suggested by Hauge and Ellingsen (1998) that the level of economic development is a more decisive predictor of the incidence of internal violent conflict than environmental factors per se. 50 Some other studies concentrate on the economic conditions of the rebel groups or supporters, exploring the grievance and/or greed arguments. The greed model suggests that civil conflict occurs when a rebel organization is ―financially viable‖ (Collier 2000a ; Collier and Hoeffler 2004b). MacCulloch‘s (2004) surveys of revolutionary support across one-quarter of a million people identify that there is some threshold of people‘s incomes that affect their responses to revolutions. Furthermore, the grievance argument emphasizes the economic discontent in people‘s decision of joining war. Indicators such as household income are tested to be linked to the level of household participation in conflict. Poorer households have higher probability to participate and support an armed group (MacCulloch 2004 ; Justino 2009). The impact of economic inequality on political conflict has also been investigated frequently by scholars. However, inconsistent findings have been reported on the economic inequality-political conflict nexus with approaches such as statistical modeling, formal modeling, and theory building (Lichbach 1989). For instance, Besanç on (2005) suggests that traditionally deprived identity groups are more likely to engage in class or revolutionary wars under conditions of greater economic inequality, but contrary evidence has been found for ethnic conflicts and the results for genocides are ambiguous. Hoping to synthesize the debate, Shock (1996) proposes a conjunctural model combining economic discontent and political opportunity. The evidence then shows that there is a positive relationship between economic inequality and separatist potential on political violence, moderated by political opportunity structures. However, the Gini coefficient 51 estimates of income inequality ―do not come close to either statistical or substantive significance‖ in the analysis of Fearon and Laitin (2003c). Political Institutions Another question that scholars and politicians have discussed for a long time is whether political institutional arrangements can prevent civil conflict in a country. Kahl (2006) synthesizes the research on political institutions of three schools and summarizes it as the concept of ―a state‘s strength.‖ According to Kahl, a state‘s strength refers to ―its ability to actually realize, in the empirical sense, its binding rule-making authority,‖ and it is determined by a state‘s functional capacity and cohesion (Kahl 2006:39). Specifically, Kahl (2006:39) suggests that functional capacity for binding rule- making requires a state‘s coercive power, administrative capacity, and legitimacy of authority, which together enable the state to deter or repress violent revolts of groups induced by environmental scarcity or degradation. Moreover, a state‘s cohesion measures the degree to which its elites are unified or divided based on their interests and strategies. It represents the ability and willingness of elites to unify when collective action is needed to fulfill governmental functions and maintain domestic order when the risk for violence is rising. Thus based on the two criteria, state strength can be mapped along a continuum, where states low in both functional capacity and cohesion can be labeled as ―weak‖, and states high in both determinants are viewed as ―strong‖. The major political institutional characteristics being examined by statistical analyses in the field are regime types (democracy or autocracy), regime change, political rights and civil liberties. 52 The tradition of democracy dates back to ancient Greek city-states 2500 years ago. For a long time political scientists have made efforts to identify the major characteristics of democracy. Since Lijphart (1977, 1999), the presence of proportional representation, multiparty systems, and decentralization have been discussed as the major features of democracy. Civil war researchers believe that democratic states endow citizens with the power to vote, thus they often have less discrimination and repression along cultural or other lines. The political rights and civil liberties associated with democracies lead to lower grievances of contending groups (Lijphart 1977, 1999 ; Fearon and Laitin 2003c). Furthermore, the political changes of in-between regime types, semi- democracies make them even less stable than autocracies (Eckstein and Gurr 1975 ; Rummel 1995). However, the empirical evidence does not support any clear relationship between democracy and civil wars. For instance, Reynal-Querol (2002, 2005) finds that democracy alone is not enough to deter social conflicts. He proposes that the level of inclusiveness of the political system is an important factor in explaining the probability of civil war. The proposition is supported by empirical tests on the probability of civil war in high inclusive systems, such as the proportional representation system and low inclusive systems favoring political exclusion, such as the majoritarian system. Furthermore, quite a few studies find that partially democratic countries are more prone to civil war than coherent democracies and harshly authoritarian states(Henderson and Singer 2000 ; Ellingsen 2000 ; Sambanis 2001 ; Hegre et al. 2001 ; Reynal-Querol 2002). The main reason is that intermediate regimes are still during the process of transition, whereas domestic violence tend to be associated with political change, no 53 matter it is toward greater democracy or greater autocracy (Hegre et al. 2001). Moreover, it is suggested that the democratization process itself may produce conflict (Huntington 1991). The results of logistic regressions by Henderson and Singer (2000) again confirm the finding that semi-democracies are more likely to have civil wars. They point out that semi-democracy has greater impact on civil war onset compared with other political, economic, and cultural factors. More recently, Goldstone et al. (2005) derive a new model using case control methods to identify risk factors for a state‘s probility of experiencing political instability in two years. With global data from 1955 to 2003, they are able to identify that regime type is overwhelmingly the dominant factor behind revolutions, ethnic wars, and adverse regime changes. What needs attention is that their investigation of regime type does not only concern about whether a state is democratic or autocratic. According to them, further information about its patterns of executive recruitment and political participation is needed before we can access a state‘s vulnerability to political instability. Some studies observe the links between democracy and the environment. It is hoped to find out whether democracy affects the environment in positive ways by minimizing scarcity and then lessening the tendency toward civil conflict induced by resource scarcity. It is hypothesized that compared with autocracies, the decentralized decision making and civil liberties associated with democracy have benign political influence on the environment. However, the empirical results tend to be mixed when specific types of resources are tested. Using a dichotomization of the Polity III index, Gleditsch and Sverdrup (2002) find that democracy works positively when environmental 54 degradation such as deforestation and loss of bio-diversity are looked at. But when it comes to climate change, democracy has a negative effect. Since their data are from around 1990, the explanation is that the counteractions to the emerging climate change are still at its early stage. Furthermore, they suggest that the effect of regime type on the environment is reduced when the level of development is controlled. Moreover, Midlarsky (1998) investigates the impact of democracy on six measures of environmental protection or degradation. His multiple regression analysis shows that rather than lowering down the degradation, democracy actually has a significant negative effect on deforestation, carbon dioxide emission, and soil erosion by water. Besides, democracy has a positive impact on protected land area, whereas no significant effect on freshwater availability and soil erosion is found. Although the different findings may be partially attributed to sample sizes and different measures of democracy used, they suggest that the relationship between democracy and environment cannot be viewed unidimensionally. Social Fractionalization Social characteristics such as ethnic and religious fractionalizations are also believed to affect the risk of violent civil conflict. Researchers including Homer-Dixon suggest that social friction, which refers to the fierce competition among narrow interest groups, is one of the most important factors impeding the supply of social ingenuity (Homer-Dixon and Blitt 1998 ; Homer-Dixon 1999). The origin of this argument can be traced back to Olson‘s widely-translated book The Rise and Decline of Nations (1982), where he explores how collective goods are provided in various types of social coalitions. 55 Olson points out that small coalitions hold disproportionate influential political power due to their flexibility, especially in relatively unstable countries. However, these small coalitions usually pursue narrowly defined interest instead of commonwealth, which often hinders the construction of institutions that reflect broader social interest. Combining Olson (1982) and Homer-Dixon (1999), this social friction weakens a society‘s capacity to adapt to environmental challenges by coordinating social activities, talents and resources to make technological innovations. Thus it is hypothesized that societies with high ethnolingusitic and religious heterogeneity have higher risk of violent conflict since their government is believed to be dysfunctional based on the social friction argument. The major evidence reported is that developing countries in general are more ethnically diverse than developed countries, and they suffer much more civil wars. For example, regions such as Africa have high ethnic diversity, and therefore suffer the highest incidents of civil war. Moreover, some security studies such as Carment et al. (2006) point out that ethnic diversity at first compounds the political and economic problems of weak states and leads to civil strife. During the second stage, ethnic conflict can escalate both horizontally and vertically, which means the conflict can become more violent and involve into interstate confrontation. Thus in many cases, the internal and international dimensions of ethnic crisis reinforce each other and can hardly be separated. Contrary to the common expectation, statistical evidence for the social diversity- civil conflict linkage is rare. As one of the few studies that find ethnic division has significant effect on civil war, Reynal-Querol‘s (2002) work uses both religious 56 polarization and animist diversity to explain the incidence of ethnic civil war and concludes that religious difference is a more important predictor of civil war than linguistic difference. More scholars in the field question the expected positive relationship between ethnic diversity and the risk of civil war. It is suggested that hatred based on ethnicity, religion and culture does not provide a good explanation for violent conflict. The results of Fearon and Laitin‘s (2003c) study, for instance, indicate that ethnic and religious fractionalization is substantively and statistically insignificant in predicting the incidence of civil war after controlling for per capita income. They also test several alternative measure of ethnic and religious diversity, including the proportion of the largest group and the number of languages spoken by at least 1%, neither of which is proved to be related to civil war onset. Thus they do not see a direct connection between ethnic diversity and civil war. However, it still does not deny the possibility that ethnic diversity can cause civil war indirectly when it lowers per capita income or state strength. Some studies even go further to develop a countervailing effect of ethnic diversity on civil war, suggesting that pluralist societies actually are safer than homogenous societies. Using data from Africa, Bates (1999) concludes that ethnically diverse societies might have more political protests but less political violence. Ethnic diversity actually lowers rather than increases the risk of violence. Based on Collier and Hoeffler (2004a), the stabilizing effect of ethnic and religious diversity is mainly because it makes rebel cohesion more costly. De Soysa (2002b) reaches a similar conclusion that highly plural 57 societies face less risk. He also suggests that ethnicity increases the probability of conflict when a society is moderately homogenous. Another group of scholars claim that investigation of ethnic diversity should distinguish between dominance and fragmentation. Perhaps ethnic diversity is damaging when it takes the form of dominance, not fragmentation (Arcand, Guillaumont and Guillaumont Jeanneney 2000 ; Collier, Honohan and Moene 2001). According to Collier and his colleagues (Collier and Garg 1999 ; Collier, Honohan and Moene 2001), ethnic diversity is only a problem for public sector organizations but not for the private sector. The patronage power of kin groups has been curbed in firms due to fierce market competition. Therefore even in factionalized societies, the worse public sector performance can be offset by better private sector performance. In contrast, ethnic dominance produces poorer economic policies and performances due to the exploitation of the minority by the majority, and eventually leads to higher risk of large-scale violence (2001). A series of studies done by Collier and Hoeffler (1998, 2004b ; Collier, Honohan and Moene 2001) all support that ethnic dominance significantly increases the risk of civil war, whereas fragmentation reduces the risk. In sum, most studies in the field do not find support for the positive association between ethnic diversity and incidence of civil war. Evidence of a possible pacifying effect of diversity on violent conflict have been reported, very likely to be explained by ethnic fragmentation rather than dominance. However, as Sambanis (2004) points out in his summarizing piece, it might be too quickly for us to write off ethnic fractionalization as a correlate of civil war. The nonsignificance of this variable might be attributed to 58 different coding rules for civil war and the author finds ethnic fractionalization significant in explaining minor insurgency. Besides, conventional research relying on the ethnolinguistic fractionalization index (ELF) is criticized recently by Cederman and Girardin (2007). They introduce a new index N* of ethnonationalist exclusiveness to analyze ethnic configurations and political violence. The index is proved to be highly significant with regression analysis using data on Eurasia and North Africa, which also recommends that ethnic diversity should not be simply ignored in civil war analysis. Demographic Characteristics It is widely acknowledged that demographic factors are closely connected to environmental scarcity and violent conflict. According to the neo-Malthusian scenario (Homer-Dixon and Blitt 1998 ; Homer-Dixon 1999), demographic challenges such as rapid population growth reduce the amount of resources available for each individual, thus it generates demand-induced scarcity. Population growth may also drive supply- induced scarcity because the growth speeds up or worsens environmental degradation and depletion. Population pressure on natural resources thus makes societies more vulnerable to internal conflict, especially for renewable resources because it is more often open access. In contrast, economic optimists argue that resource scarcity such as freshwater, and cropland land scarcity caused by population growth and high population is actually a driving force rather than impediment for innovations and economic development, eventually resulting in long-term peace. For the optimists, the supply of ingenuity and 59 adaptability needs human capital, which is determined by various demographic characteristics including education level. The Resource Curse School tends to examine the demographic characteristics of a country to explore the motivation and opportunities for rebel groups to revolt. It is believed that factors such as high proportion of rural population, high percentage of young male and low literacy level provide demographic foundations for rebellion. Finally, political ecologists are probably more interested in the distribution of population, such as its density in different regions and their percentages in urban and rural areas. According to their argument on distributional inequality of resources, a few rich people take up most of the resources including land, freshwater etc., whereas the poor populations suffer from scarcity. To summarize, all the schools note the importance of demographic characteristics in the resource scarcity-political instability linkage. On one side, demographic variables are considered as the major sources of true resource scarcity or ―illusionary scarcity‖ resulting from distributional inequality. On the other side, demographic characteristics also determine a country‘s capability to adapt to environmental challenges by either providing the human capital for technological innovation or the personnel foundation for rebellion. Based on distinct theoretical frameworks, empirical research on the topic also concentrates on different aspects of a country‘s demographic condition. First, the effect of population, population density and population growth on the risk of civil war is frequently investigated, but so far no agreement has been reached. A stable and highly 60 significant positive effect of population on the risk of conflict has been found in most of the studies (Fearon and Laitin 2003c ; Hegre and Sambanis 2006 ; Theisen 2008). A weak but significant positive relationship between population density and the risk of civil war has been reported (de Soysa 2002a, 2002b ; Raleigh and Urdal 2007 ; Urdal 2008). But as de Soysa (2002b) suggests, the effect of population density cannot be isolated from other factors. Countries simultaneously have the three features—high population density, low level of democracy, and less open to trade—are more likely to suffer violent civil conflict. De Soysa (2002b) further examines the threshold effect of battle deaths. It is concluded that densely populated countries capture smaller civil wars with the threshold of 25 battle-related deaths, probably not for larger-scale civil war such as those with a 1,000 battle-death threshold. In his subnational analysis of Indian states, Urdal (2008) examines population density in the rural area in particular, and finds it increases the risk of conflict. This is consistent with de Soysa‘s (2002b) finding that densely populated rural societies with abundant renewable resource tend to have more civil conflict. Moreover, Raleigh and Urdal‘s (2007) test on population growth complies with Huntington‘s (1996) argument that rapid population growth significantly increase the risk of armed conflict. However, such a relationship between population features and civil war is unobserved in other analyses (Collier and Hoeffler 1998, 2004a ; Hegre and Sambanis 2006). Collier and Hoeffler (2004a) do see the increase of grievances with the heterogeneity produced by larger population. But the connection between this heterogeneity and the risk of civil conflict is not observed. Hegre and Sambanis (2006) 61 confirm that the incidence of civil war is associated with a large population but not population density. Theisen (2008) fails to replicate Hauge and Ellingsen‘s (1998) significant result for population density, his own revised tests show that population size is a significant predictor of the onset of armed civil conflict, but neither population density nor growth is significant. Moreover, according to the opportunistic argument, demographic features affecting the personnel foundation for rebellion are also proved to be important. For example, examining the physical and psychological characteristics of young males, Huntington (1996) suggests that societies with a high percentage of young males are prone to civil violence. According to him, this argument partially explains Islam‘s ―bloody innards‖. However, the statistical results for this thesis are contradictory. Although Urdal (2008) find that youth bulges is associated with increasing risk of civil conflict in his subnational analysis of Indian states, Fearon and Laitin (2003c) find no such significant impact of percentage of young males. They attribute the significant effects of percentage of young males and male secondary schooling rates in some studies to the two variables‘ strong negative correlation with income. Other factors such as high rates of urbanization and large refugee populations are also found to be unrelated to the danger of internal armed conflict (Urdal 2005) . In sum, the above theoretical and empirical studies demonstrate that the importance of adaptive capacity has been widely accepted in the studies of resource- induced civil conflict. But a comprehensive theoretical framework that systematically investigates adaptability, in particular whether and how it mitigates or worsens a 62 country‘s existing resource challenges, and eventually affects the country‘s political instability is required. In most of the quantitative studies, economic, political, social and demographic factors are treated as controlling variables in regressions. Although the analyses have suggested which variables are substantively and significantly stronger predictors of civil violence, they seldom go further to examine how these factors interact with natural resource scarcity to impact political instability. One of the main contributions of this study is exploring the causal mechanisms focused on the intermediate role of various aspects of adaptive capacity to resource scarcity. The Comprehensive Model and Causal Diagram This study takes a state-centric approach to examine the scarcity-adaptability- political instability linkage. The state is used as the unit of analysis. Based on the assumption that renewable and nonrenewable natural resources have different features and utilities for human life, the model differentiates the causal mechanisms for these two types of resources. The causal relationship between renewable resource scarcity and political instability draws extensively from neo-Malthusian arguments, whereas the connection between nonrenewable resource abundance and political instability reflects more neoclassical economist theories. Within these two basic frameworks, the four aspects of adaptability affect the causal relationship in both directions and eventually lower down or increase political instability. According to our hypothetical model, the scarcity-adaptability-political instability causal relations start with a country‘s natural resource scarcity level. For renewable 63 resources, high scarcity puts stress on the state, whereas abundance of nonrenewable resources imposes pressure on the state. Whether a state can handle the pressure from natural resource scarcity/abundance depends on its adaptability, which is determined by its capacity in four aspects: economic development level, political institutions, social fractionalization and demographic characteristics. Countries with higher level of economic development, stronger political institutions, lower social fractionalization and favorable demographic conditions are expected to have higher adaptive capacity, whereas countries with the reverse features are viewed as having lower adaptability. Countries with high adaptability can cope well with the pressure from resource scarcity/abundance, and thus reduce or overcome the risk of political instability. In contrast, low adaptability constrains a country‘s capacity to handle environmental challenges, and eventually resource scarcity/abundance is going to make the political system less stable. In the meantime, poor adaptive capacity can also worsen environmental degradation and resource depletion which can once again increase the pressure on the state, whereas high adaptability reduces the pressure. The model also acknowledges the reverse effects of political instability on environmental situations. 24 But based on previous literature and our major research interest, it is assumed that the reverse effects would be minimum compared with the main effects. Figure 2 shows the tentative causal diagram that depicts the causal relationships among the three major variables. Inside the ellipse, factors such as economic 24 Some studies in international relations suggest that civil war has devastating effects on a state‘s environmental conditions. 64 development, political institutions, social fractionalization, and demographic characteristics are major determinants of a society‘s adaptive capacity. As it shows, adaptability works as an intermediate variable between environmental scarcity/abundance and political instability. Moreover, two types of resources have different associations with political instability, and their connections are drawn separately. 65 Figure 2: The Hypothesized Causal Diagram 66 The upper part of the diagram depicts the causal mechanisms between renewable resources scarcity and political instability. A society with high adaptability usually has better ability to cope with crisis and challenges, due to the listed economic, political, social and demographic features. Thus the scarcity of renewable resources would not shake its political stability, and sometimes a country with high adaptability can even reinforce its political stability after effective reactions. In contrast, environmental scarcity could be detrimental to a society with low adaptability, since the country lacks the necessary economic, political, social and demographic conditions to fight against the environmental challenge. Therefore, in facing of renewable resource scarcity, the political stability of a country with low adaptability tends to be decreased to a different extent. The reverse effects can also exist. The arrows in the diagram demonstrate that the stability/instability of the political system can in turn influence those determinants of adaptability and the factors determining a society‘s adaptability can also affect its environmental scarcity level. However, we expect that the left to right causal effect would be the primary one. The lower part of the causal diagram then describes the relationship between the amount of nonrenewable resources, environmental adaptive capacity and political instability. Same as the model for renewable resources, economic development level, political institutions, social fractionalization and demographic characteristics are determinants of a country‘s adaptive capacity to resource challenges. Countries with high 67 adaptability usually have better ability to cope with challenges, whereas those with low adaptability are often poor at addressing environmental issues. However, we can see that the hypothesized causal mechanisms are different from those of renewable resources based on assumptions of the distinct utility of these two types of resources. Here the abundance rather than scarcity of nonrenewable resources would be environmental challenges to a country‘s political stability. It is thus expected that the abundance of nonrenewable resources would weaken the political stability of a country with low adaptability. The abundance of nonrenewable resources can bring great wealth to a country, which often leads to the country‘s over-reliance on selling of primary commodities. It in a long term is unhealthy for the development of a country. In the meantime, according to the ―greed‖ theory, the wealth from selling of nonrenewable resources is the primary source of funding for most violent factions. But the political stability of a country with high adaptability will not be negatively impacted by its amount of nonrenewable resources. The abundance of ―lootable‖ resources provides a firm platform for the economic taking off for a country with high adaptability, whereas the lacking of nonrenewable resources drives its attention to the international market and achieves the transformation of its economic strategies. Again, we recognize the possible reverse effects, which acknowledge that political instability can in turn influences the various determinants of adaptability and then affects the pressure from a country‘s nonrenewable resources abundance level. With the same assumption, we suppose that the left to right causal effect would be the dominant one. 68 Based on the hypothesized model, this study examines the following hypotheses by both quantitative and case analysis. It starts with a proposition reviewing the linkage between resource scarcity and civil war, which is taken as an extreme scenario of political instability. Since most of the available studies use the incidence or onset of civil war as their dependent variable, it hopes to shed light on the contradictory results reported. 1. Countries with high scarcity of resources are likely to experience more civil wars than those with low resource scarcity. Based on the discussion of resource scarcity in this chapter, we believe that it would be inaccurate and oversimplified to conceive the same causal mechanisms for renewable and nonrenewable resources. For instance, neo-Malthusian suggest the detrimental effects of scarcity on political stability according to their investigation on renewable resources, however, the Resource Curse School find resource abundance rather than scarcity leads to violence. Therefore, the second proposition intends to see if the two types of resources indeed connect with adaptability and political instability in different ways, and the next two propositions go further to test the neo-Malthusian argument on renewable resource scarcity and the resource curse hypothesis on nonrenewable resource abundance. 2. The association between scarcity of renewable resources and political instability is different from the association between nonrenewable resources and political instability. 69 3. Countries with more severe renewable resource scarcity are more likely to be politically unstable than otherwise. 4. Countries with more nonrenewable resource endowments are likely to be more politically unstable than countries with less nonrenewable resources. One of the major contributions of this study is that it brings in adaptability as an important explanatory variable in the causal mechanisms. As discussed earlier, the determinants of adaptability can be categorized into four aspects based on the major control variables in the current literatures. Therefore, the fifth proposition arises: 5. A state‘s adaptive capacity to resource challenges can be evaluated based on its following conditions: 5.1. Economic development. States with higher GDP per capita, higher per capita income and lower income inequality are likely to have higher adaptive capacity to environmental scarcity. 5.2. Political institutions. Countries with more democratic political institutions and civil liberties are politically more accountable to the public, and thus are likely to be better at adapting to environmental pressures. 5.3. Social fractionalization. More fragmented societies have more ethnic and religious fractions, and more languages. Social fractionalization tends to decrease a country‘s environmental adaptive capacity. 5.4. Demographic characteristics. A country‘s adaptability to resource pressures is likely to decrease with its unfavorable demographic characteristics including 70 higher population growth rate, lower percentage of urban population, smaller working population, and lower literacy rate. After the conceptualizing of adaptability, the essential part of this study investigates the intermediate role played by adaptability in the resource scarcity/political instability nexus. Again, the models for renewable and nonrenewable resources are evaluated respectively in the following two propositions. 6. High scarcity of renewable resources and low adaptability together are associated with higher political instability, but countries with high scarcity of renewable resources and high adaptability are not more likely to have higher political instability. This pattern also applies to each type of adaptive capacity. 6.1. For a state with a high economic development level, renewable resource scarcity does not increase its political instability. Otherwise, a state‘s political stability is likely to decrease with the severity of renewable resource scarcity. 6.2. A state with high renewable resource scarcity is likely to be less stable if its political adaptability is low. Otherwise, high renewable resource scarcity does not increase its political stability. 6.3. Higher renewable resource scarcity is more likely to be associated with higher political instability when a state has low social adaptability. Otherwise, a state‘s political stability is unlikely to decrease with the severity of renewable resource scarcity. 71 6.4. A state facing high renewable resource scarcity is likely to be less stable when its demographic adaptability is low. Otherwise, renewable resource scarcity is not likely to reduce the state‘s political stability. 7. The associations between scarcity of nonrenewable resources, adaptability and political instability are different from those of renewable resources. Abundance of nonrenewable resources and low adaptability together are likely to be associated with higher political instability. In particular, 7.1. A state with low economic development level is likely to be less politically stable when it has abundant nonrenewable resources. Otherwise, nonrenewable resource endowments do not make it less stable. 7.2. For a state with low political adaptability, nonrenewable resource abundance is more likely to endanger its political stability. Otherwise, nonrenewable resource abundance does not increase its political instability. 7.3. Nonrenewable resource abundance is more likely to decrease a state‘s political stability if it has high social fractionalization. Otherwise, the state is not likely to be less stable. 7.4. A state with nonrenewable resource abundance is likely to be political unstable if its demographic adaptability is low. Otherwise, political stability of the state does not increase with nonrenewable resource abundance. In sum, this chapter has reviewed the debate among the three schools of thought on the connections between resource scarcity and political instability. It addresses both 72 the theoretical confrontations and available empirical evidence, remarking on the merits and limitations of each school. The second part of this chapter goes further to introduce the hypothesized model of this study. Firstly, it defines the three core variables—scarcity, political instability and adaptability—within the context of this study and examines how they are addressed in the relevant literature. Built on those key concepts, a new ―adaptability‖ model is established to explain the nexus in a more comprehensive way. Moreover, a series of propositions are presented and explained as the roadmap for empirical tests. The next chapter elaborates on the statistical methods, data sources, variable construction and case selection to test the above hypotheses. 73 Chapter 3 Data, Statistical Methods and Case Selection This chapter deals with the methodological issues of this study. First, it introduces and evaluates the data sources and how they are transformed into the major variables of interest. Then the second section explains the statistical techniques utilized to assess the causal relationships proposed in the previous chapter. The quantitative analysis is composed of two parts. The first part applies auto-regressions over two periods to detect the masked association between environmental scarcity and civil wars. Then principal component analysis and regression analysis with focus on interaction and confounding effects are brought in to further explore the relationships among the three variables. Based on the statistical results, this study then investigates several cases in more detail, and the case selection method is explained in the end of this chapter. Data Sources The data used for this study are drawn broadly from databases of environmental studies, governance, economics, and security studies. A number of widely used datasets are scanned. Most of our data are extracted from the World Bank, the replication dataset of Fearon and Laitin (2003c), and several other well-kept databases compiled and maintained by scholars of international security and nonprofit organizations, including the Correlates of War Project (Sarkees and Schafer 2000), Penn World Table (Summers and Heston 1991), Polity IV (Marshall and Jaggers 2007), and Freedom House (2009). The following section addresses in detail the original data sources and the transformations made to construct the variables of interest. 74 Scarcity As mentioned before, this study intends to investigate the objective scarcity/abundance of natural resources. Besides, previous studies suggest that renewable and nonrenewable resources should be analyzed separately regarding their interplay with adaptability and political instability. Additionally, this study does not plan to investigate each type of resource (i.e. water, arable land) respectively. Instead, it hopes to find comprehensive indexes that integrate the scarcity/abundance of multiple resources. Thus the study utilizes the natural capital wealth estimates from the World Bank‘s Wealth Estimates dataset. They measure separately the amount of renewable and nonrenewable resources each country has, and are viewed as one of the most objective indexes on resource scarcity/abundance (de Soysa 2002b). The wealth estimates are compiled by economists in the World Bank. The calculation is based on the idea of classical economics that land, labor and produced capital are the primary factors of production. 25 The estimates measure a country‘s total wealth by adding produced capital, natural capital, and intangible capital together. Capital stocks are calculated by summing up the value of gross investments and subtracting depreciation of produced capital of an initial stock over time (Hamilton and World Bank 2006:22-27). This study utilizes natural capital as the crucial variable measuring the amount of natural resources a country has. According to its definition, natural capital is the sum of 25 According to the World Bank team, produced capital is ―the sum of machinery, equipment, and structures (including infrastructure)‖(Hamilton and World Bank 2006:22). Urban land is considered as produced capital rather than a natural resource in their wealth estimates. 75 nonrenewable resources (including oil, natural gas, coal, and mineral resources), and renewable resources (cropland, pastureland, forested areas, and protected areas). For each type of land or areas listed as renewable resources, the calculation of their capital stocks is based on the values and quantities of products and utilities they can provide (World Bank). For example, forest areas include both areas for timber extraction and non-timber forest products such as hunting, recreation, watershed protection and so on (World Bank). The value of different natural resources is transformed into US dollars for the year 2000. Specially, they are valued by ―taking the present value of resource rents—the economic profit on exploitation—over an assumed lifetime‖ (Hamilton and World Bank 2006:23). The effective lifetime is calculated based on the current exploitation rate of a particular resource. Thus the effective lifetime for renewable resources will not be infinite if not sustainably managed. In general, the estimates tell us ―in standardized values the net worth of the stock of natural resources of any given country in per capita terms‖ (de Soysa 2002b). This study uses the total natural capital stock in the auto-regression analysis, and then investigates renewable and nonrenewable resource capital respectively in regression analysis. The sub-soil assets in natural capital estimates include the major nonrenewable resources, such as oil, natural gas, coal, and various minerals. The stock of renewable resources would be the natural capital estimates minus sub-soil assets. It thus includes timber and non-timber forest assets, cropland, pasture land, and protected areas, which actually sum up values of resources produced on these lands, such as timber, crops, beef, lamb, milk, and wool. The data show that subsoil assets are abundant in the Middle East, 76 North Africa, Europe, Central Asia and Latin America, whereas agricultural land is the important resource in East Asia and the Pacific, South Asia, and Sub Saharan Africa (Hamilton and World Bank 2006:27). Although the natural capital estimates is one of the most comprehensive indices for standardizing values of different resources, the World Bank team also acknowledges that data are lacking for several assets, including subsoil water, diamonds, and fisheries. Besides, they mention that the index does not estimate the explicit value for services provided by ecosystems, such as ―the hydrological functions of forests‖ and ―the pollination services of insects and birds‖, due to data limitations (Hamilton and World Bank 2006:24). But as they suggest, the estimates of the values of cropland and pastureland do capture the values for ecosystem services indirectly. Therefore, it does not affect our estimates of the renewable resource stock. Another problem with the data that indeed affect our analysis to some degree is that the nonrenewable resource wealth for some countries are recorded as zero dollar per capita, which seems impossible for some of them. For example, it is well-known that Iraq has rich oil resources. However, the dataset codes Iraq‘s nonrenewable resource wealth per capita as zero. It could be that the data is missing or unavailable, since the estimate depends on how much a country is exploiting the resource. Around the time the data was calculated, the United Nations actually imposed sanctions on Iraq for invading Kuwait in 1990. Since the sanctions amounted to a near-total financial and trade embargo, there might be virtually no exploitation of oil resources at the time. 77 However, the dataset and its methodology paper do not specify which ones are (a) missing or (b) truly zero based on the calculation. It would be ambiguous to either accept zero values for all the 30 cases or just take them as missing values. As de Soysa (2002b) suggests, the ambiguity associated with those zero values is outweighed by the benefit of gaining more data points. If the 30 cases are discarded, the sample size will be considerably reduced. He thus assigned $5 to all cases with zero value with a nonrenewable component to facilitate the subsequent natural logarithmic transformation. De Soysa (2002b) believes that it does not mean they do not have any lootable wealth even when some countries do not exploit those nonrenewable resources. However, this study chooses not to do that since assigning $5 as a value does not make any meaningful difference in the statistical analysis. Instead, it keeps an eye on those cases to see if anything behaves unusually, and makes some transformations when necessary. 26 Some other environmental indexes such as Ecological Footprint (EF) and Genuine Savings (GS) are also popular with scholars in the field. The EF consists of six main bioproductive areas: cropland, grazing land, forest area, fishing ground, built-up land and ―energy land.‖ 27 Its three main variables are biocapacity per capita, ecological footprint per capita and the difference between the two— ecological reserve per capita (Ewing et al. 2008 ; Kitzes et al. 2008). However, the EF index does not separate renewable resources from nonrenewable ones. Furthermore, its calculation is based on consumption 26 In the statistical analysis, we will add $1 to all the cases before the log transformation is conducted on sub soil assets. This is done so the logged value will be zero. 27 The EF data are calculated and compiled by the Global Footprint Network. Sample data are available at their website <http://www.footprintnetwork.org>. 78 and two of its variables—ecological footprint per capita and ecological reserve per capita—actually capture the consumption level of each country and therefore do not meet our intention to reflect the objective resource scarcity/abundance of each country. The GS index is created by the World Bank, covering 149 countries from 1970 to 2001 (Hamilton et al. 2000; Kunte et al. 1998). As an economic indicator built on the concepts of green national accounts, GS measures ―the true rate of savings in an economy after taking into account investments in human capital, depletion of natural resources and damage caused by pollution.‖ 28 Since it accounts for the investments in human capital and education expenditure, it is more of a synthesized sustainability index rather than measuring the objective amount of natural resources every citizen has in a country. Therefore, neither the EF nor the GS index is used in the final statistical analysis of this study based on our conceptualizing of scarcity in Chapter 2. 29 Political Instability For political instability, this project does not only look at extreme cases of interruption of political stability, such as civil war and armed conflict, it also counts other less intensive forms of political instability, including annual assassinations, coups d‘état, strikes, major government crises, riots, anti-government demonstrations and so on. It brings in three indicators of political instability. The first one is a civil war indicator, 28 This definition is from the World Bank‘s website for Adjusted Net Savings, which is the updated name for ―Genuine Savings‖. 29 Chapter 2 has clarified that this study talks about the ―objective‖ scarcity of natural resource rather than ―mispercepted‖ resource scarcity or scarcity due to maldistribution. That is to say, whether a country is scarce in a resource depends on the per capita possession of that particular resource, without considering the consumption level, exports and inequality of resource distribution within the country. 79 which quantify the extreme form of instability. Most of the relevant datasets provide several measurements of civil war, such as civil war onset, incidence of civil war, number of civil wars ongoing, war duration and number of deaths. The mainstream quantitative analyses of the linkage between resource scarcity/abundance and civil war adopt civil war onset as their dependent variable, which makes sense if the researchers are interested in exploring the factors contributing to the breaking out of war. However, it does not provide enough information about ongoing war status. For example, the Somali Civil War is an ongoing civil war that started in 1991. The coding of the civil war onset for Somali is 1 for the year 1991 but 0 for the later years. Thus it does not reflect the difference between shorter wars and longer wars, which obviously have different damaging effects on a state‘s stability and economy. Furthermore, as a dichotomous variable, civil war onset does not offer enough information about the magnitude of the wars. For instance, some countries might have several wars ongoing at the same time, which is often more detrimental than a single war. The coding of civil war onset cannot incorporate this distinction and therefore excludes some important information. Given the research purpose of this study, we choose number of civil wars ongoing as a principal indicator of political instability. It does not only include information about civil war onset, it also captures information on civil war duration and simultaneous wars. Thus it more accurately measures the substantive disturbance of civil war on a state‘s political system over time. 80 The indicator of number of civil wars in progress is drawn from five standard civil war/conflict datasets: Fearon and Laitin (2003c), Collier and Hoeffler (2004b), Doyle and Sambanis (2000), intrastate war dataset of ―Correlates of War‖ project (Singer and Small 1994), and Gleditsch et al. (2002). Fearon and Laitin‘s (2003c) replication dataset incorporates the civil war indicators of the first four datasets, and thus we only need to add the last measurement of civil war from PRIO‘s database. The five datasets have different definitions of civil war based on the coding rules of accumulated number of deaths and annual number of battle-death. The definitions and coding rules adopted by the five datasets are listed in Table 3.1. 30 Based on their definitions of civil war and coding rules, we can see that except the UCDP/PRIO Armed Conflict dataset, the other four datasets all require at least 1,000 battle-related deaths to qualify as a civil war. Furthermore, the Collier and Hoeffler (2001, 2004b) dataset follows the rule used by the old version COW dataset which requires an annual battle deaths of 1,000. Fearon and Laitin (2003c) set a relative lower standard of 100 annual battle deaths. Compared with the other four datasets, UCDP/PRIO Armed Conflict Dataset adopts a much lower level battle deaths threshold of 25, which captures the lower-scale civil conflicts. 30 Sambanis‘ article (2004) and its supplement tables provide a more comprehensive summary of the definitions and coding rules of civil war in the major scholarly projects. It also includes the main variables that are found to be significantly related to civil war onset. The replication data and notes are available on Sambanis webpage at <http://pantheon.yale.edu/~ns237/index/research.html#Civil>. 81 Table 1: Five Datasets on Civil War Project or Dataset Definition of Civil War and Coding Rules Notes Correlates of War (Singer and Small 1972, 1994 ; Small and Singer 1982 ; Sarkees and Schafer 2000) 31 A civil war is fought between a government and non-government force that (1) must cause at least 1,000 battle deaths; (2) both sides must make organized violent opposition, or the weaker side must inflict upon the stronger opponents at least five percent of the number of fatalities it sustains. List of civil wars from 1816-1997 Doyle and Sambanis (2000) A civil war is defined as ―an armed conflict that meets all the following conditions: (1) causes more than 1,000 deaths overall and in at least a single year; (2) challenges the sovereignty of an internationally recognized state; (3) occurs within the recognized boundary of that state; (4) involves the state as a principal combatant; (5) includes rebels with the ability to mount organized armed opposition to the state; (6) has parties concerned with the prospect of living together in the same political unit after the end of the war.‖ List of civil wars from 1945-1999. Termination of a civil war is defined as signature of a peace treaty or victory by one side 31 More information can be found in the project‘s website <http://www.correlatesofwar.org/>. 82 Table 1, continued Project or Dataset Definition of Civil War and Coding Rules Notes Collier and Hoeffler (2001, 2004b) Adopt the definition of Singer and Small (1982 ; 1994) and data of COW, and updated their dataset for 1992-1999. It also sets the criteria of at least 1,000 annual battle-related deaths. Both sides must suffer at least 5% of fatalities. List of civil wars from 1960-1999, covering 161 countries The UCDP/PRIO Armed Conflict Dataset (Hegre et al. 2001 ; Gleditsch et al. 2002) 32 Civil war (internal armed conflict) is a contested incompatibility that ―occurs between the government of a state and one or more internal opposition group(s) without intervention from other states,‖ resulting in at least 25 battle-related deaths per year (Uppsala Conflict Data Program 2008). List of civil war from 1946-2008 Fearon and Laitin (2003c) A civil war is a ―fighting between agents of (or claimants to) a state and organized, nonstate groups who sought either to take control of a government, to take power in a region, or to use violence to change government policies.‖ It requires (1) more than 1,000 cumulative battle deaths (2) at least 100 battle deaths on both sides (3) annual deaths of at least 100 Country-year dataset from 1945-1999 32 The dataset is described in Gleditsch et al. (2002). Six annual updates have been published in the Journal of Peace Research: in September 2003(Eriksson, Wallensteen and Sollenberg 2003) ; September 2004 (Eriksson and Wallensteen 2004); September 2005 (Harbom and Wallensteen 2005); September 2006 (Harbom, Hogbladh and Wallensteen 2006); September 2007 (Harbom and Wallensteen 2007); and September 2008 (Harbom, Melander and Wallensteen 2008). The dataset‘s website is <http://www.prio.no/CSCW/Datasets/Armed-Conflict/UCDP-PRIO/>. 83 The replication dataset of Fearon and Laitin (2003c) has summarized the civil war onset and number of civil wars in progress of the COW, Doyle and Sambanis (2000) and Collier and Heoffler (2001, 2004b) datasets. Thus we only need to add in the UCDP/PRIO indicators for internal armed conflict. The UCDP/PRIO number of civil wars in progress for the years from 1970 to 1999 is extracted and the mean value for each country is calculated for the later statistical analysis. The Political Instability Task Force has also compiled comprehensive indicators of political instability from various open sources. Reflecting the diverse concerns of the project‘s sponsors, their list includes four types of severe political crises: revolutionary wars, ethnic wars, adverse regime changes, and genocides and politicides. While revolutionary wars are ―episodes of violent conflict between governments and politically organized groups‖, ethnic wars are defined as‖ violent conflict between governments and national, ethnic, religious, or other communal minorities‖ (Marshall, Gurr and Harff 2009). The Cuban revolution of 1956-1959 and the Iranian revolution of 1979 are examples of revolutionary wars, and events like the uprisings that began in 1976 in black townships in South Africa are coded as ethnic warfare. Moreover, adverse regime changes are defined by the Political Instability Task Force as ―major, adverse shifts in patterns of governance‖ (Marshall, Gurr and Harff 2009). For example, the deposition of Saddam Hussein in Iraq in 2003 is coded as an adverse regime change. And the genocidal assault launched against Kosovo‘s civilian population in 1998-1999 is an example of events coded as genocides and politicides. In order to measure a country‘s political 84 instability, this study sums up the four types of state failure. It is assumed that the more political crises a state has the less stable it is. Besides the civil war and political failure projects, a widely used indicator of political instability is taken from the Worldwide Governance Indicators (WGI) project of the World Bank. The project covers 212 countries and territories and measures six dimensions of governance between 1996 and 2006 (Kaufmann, Kraay and Mastruzzi 2009a). They are voice and accountability, political stability and absence of violence, government effectiveness, regulatory quality, rule of law, and control of corruption. The datasets are compiled based on governance data from the relevant public sectors, private sectors, NGOs, and survey of citizens and firms. Its dimension of political stability and absence of violence is used as one of the main indicators of our dependent variable. The values of the indicator range from -2.5 to 2.5. Higher values imply more political stability and less violence, whereas lower values correspond to less political stability and more violence (Kaufmann, Kraay and Zoido-Lobaton 1999). Compared with the first two indicators of political instability mentioned above, the WGI index does not only consider intense types of political crises such as war, genocide and regime change, it also evaluates and differentiates the stability of states in relative peaceful times. Thus it is viewed as a more comprehensive assessment of political instability than the civil war and political failure indicators. Generally speaking, the three indicators of political instability used in this study emphasize different severities of government crises. The composite index of annual number of civil war in progress focuses on the more intense forms of political instability, 85 which involves mass political violence and great challenge to the government. The PITF index summarizes four types of political instability, which do not only include wars, but also take into account less dramatic government changes such as regime change and genocide/politicide. Finally, the political stability and absence of violence scores from the WGI project evaluate the executive instability of states even for those absent of violence. Thus compared with the civil war and PITF indicators, the WGI index tends to be more comprehensive and provides more information for statistical analysis. The three indicators thus have exemplified the two approaches of measuring political instability summarized by Alesina and Perotti (1993), one of which considers the executive instability and the other investigates political violence. 33 Adaptability As discussed in the literature review chapter, economic development, political institutions, social fractionalization and demographic characteristics are proposed as four categories of variables measuring adaptability based on current models developed by research on civil conflict and resource. Data on these indicators are found in various sources, including Fearon and Laitin‘s (2003c) replication dataset, the World Bank, Freedom House, the Penn World Table, etc. The data sources and transformation of each variable for preparation of statistical analysis are explained in this section. 33 In the previous chapter, we mention that according to Alesina and Perotti (1993), political instability is defined in two ways. The first definition emphasizes the executive instability of a state, whereas the second one focuses on social unrest and political violence. The civil war index and the PITF index thus capture the definition of the second approach, and the WGI index complies with the first approach by emphasizing more on the executive side and taking into account the peaceful times. 86 For economic development, this study brings in two groups of macroeconomic indicators. The first group measures the economic development level of a country, including Gross Domestic Product (GDP) per capita, Gross National Income (GNI) per capita, and primary commodities ratio of exports. The second group of indicators measures the economic development inequality of a country. Data for GDP per capita, GNI per capita, primary commodities ratio of experts and GINI index are extracted from the World Development Indicators (WDI) of the World Bank, real GDP per capita is obtained from the Penn World Table. Based on the major literature in the field, the characteristics of a state‘s political institutions include regime type, political freedom, civil liberties, government accountability and effectiveness, rule of law, etc. The well-known Polity IV scores are coded based on the authority characteristics of states and are thus brought in as an important indicator of regime type (democracy or dictatorship). 34 The Polity IV scores range from +10 (strongly democratic) to -10 (strongly autocratic), calculated based on the project‘s composite indexes of autocracy and democracy. 35 Furthermore, the Polity IV project provides a revised Polity IV scores by converting instances of ―standardized authority scores‖ to conventional polity scores which are within the range of -10 to +10. 36 The modification makes it more convenient to use the scores in statistical analysis across 34 Polity IV is the Polity score compiled in the fourth phrase of the project, which is the latest version. 35 It is computed by subtracting the AUTOC (autocracy) score from the DEMOC (democracy) score. 36 The ―standardized authority scores‖ in the original Polity scores are -66, -77, and -88. -66 stands for cases of foreign ―interruption‖ and are treated as ―system missing‖ in the revised scores. -77 means ―interregnum‖ or anarchy, and they are converted to ―0‖, which is a neutral score. -88 are used for cases of ―transition‖ in the original coding, and in the revised scores these cases are prorated across the span of the transition. The methodology paper by Marshal and Jaggers (2007:15-16) explains more about the transformation made in the revised scores. 87 time. Thus this study extracts the revised Polity IV scores from 1970 to 1999 and calculates the mean values for further analysis. In addition, two indicators of freedom from the Freedom House are introduced to compose an important index of political institutions. The organization‘s table of ―Freedom in the World Country Ratings: 1972-2009‖ evaluates both political rights and civil liberties each country gives to citizens. A 1-to-7 scale is assigned to both Political Rights and Civil Liberties, with one representing the most rights and liberties and seven the least. These two indicators are then added together and averaged as a composite index of ―Freedom House Scores‖ for the purpose of this study. Same as their original scale, the composite index still ranges from 1 to 7, with one representing the highest degree of freedom and seven the lowest. Moreover, only data from 1972 to 1999 are extracted and the mean value is calculated for each country. As mentioned in the previous section, the WGI project conducted by the World Bank also provides several good indicators measuring a state‘s quality of political institutions. The five dimensions of governance—voice and accountability, government effectiveness, regulatory quality, control of corruption and rule of law—are chosen as indicators of political institutions in this study. 37 It has been clarified that the WGI calculates both point estimate and the percentile for every indicator. The point estimate is used and its values range from -2.5 to 2.5, with 2.5 representing the best governance quality and -2.5 corresponding to the worst governance (Kaufmann, Kraay and Zoido- 37 Besides the five indicators mentioned here, another one is political stability and absence of violence. It is used as one of the major indicators of the dependent variable. 88 Lobaton 1999). The data are reported by the World Bank every two years from 1996 to 2002, and after then they are compiled every year. Since the time period of interest is from 1970 to 1999, this study only picks the data of 1996 and 1998 and then calculates their average. For social fractionalization, this study brings in the widely-used fractionalization indexes coded by scholars in the field. The best known effort to code ―ethnolinguistic‖ groups was conducted by a team of Soviet ethnographers in the early 1960s, and later it was published as Atlas Narodov Mira (Soviet Union. Glavnoe upravlenie geodezii i kartografii. et al. 1964). Language is the major standard to define group according to the Soviet team, but sometimes the notion of race as well as national origin are also investigated. Their research has been adopted by social scientists for decades to produce estimates for ethnic fractionalization. Fearon and Laitin‘s (2003c) replication dataset includes the estimate of ethnic fraction based on the Soviet Atlas, with estimates for missing values in 1964. Since language is an important measurement of ethnic fractionalization, the study also takes the number of major languages spoken in a country as one of the main indicators of social fractionalization. Besides, Fearon and Laitin‘s (2003c) replication dataset also provides their own indicator of ethnic fractionalization originated from Fearon‘s (2002) APSA paper. Based on ―a list of 820 ethnic groups in 160 countries that made up at least 1% of each country‘s population in the early 1990s,‖ the new indicator measures cultural 89 fractionalization by the structural distance between languages (Fearon 2002). It is believed to be more appropriate in conflict studies compared with the Atlas‘ indicator. 38 Moreover, religion fractionalization is also viewed as an important indicator of a country‘s social fractionalization. One of the widely-used indexes comes from R. Quinn Mecham‘s work (Fearon and Laitin 2003a). He constructed a list of religions by country, and their percentage of adherents according to the CIA Factbook and many other sources. All of the above four indicators of social fractionalization are available in Fearon and Laitin‘s replication dataset (2003c). Finally, the fourth dimension of adaptability—demographic characteristics— include relatively broad types of indicators of a country‘s population. Variables such as population, population density, population growth, urban and rural population, youth population, male population, and education level are widely-used in the previous statistical analyses of civil conflict. The first three variables measure a country‘s demographic stress and its annual change in general. Population stress does not only directly impact a country‘s resource scarcity (the amount of resource per capita decreases as population increases); it is also an important determinant of the country‘s available ingenuity according to the economic optimists. 39 38 Besides the two indicators of ethnic fractionalization, Ted R. Gurr and his collaborators developed an array of variables coding group characteristics, situations, and experiences in their list of ―minorities at risk‖ (MAR) (Gurr 1996). Alesina et al. (2003) attempted to construct measures of ethnic, linguistic, and religious fractionalization based on a sample of about 190 countries. Roeder (2010) made a series of fractionalization measures for 1961 and 1985 based almost entirely on Soviet ethnographic sources. 39 The optimists argue that people are the main impetus for technology advancement and larger populations in general can elicit more innovation. 90 The latter four variables describe the country‘s population composition in terms of urbanization, age, gender and education. According to the theories and empirical analysis presented by neo-classic economists, these factors are believed to impact a country‘s supply of labor force, ingenuity, as well as supply of the personnel for insurgency. All of the above demographic variables are available in the World Bank‘s WDI project. Statistical Methods As the literature review chapter shows, time series cross-section studies in the field actually have found inconsistent results about the scarcity-civil conflict nexus, even though they use interconnected datasets and similar statistical techniques. At first, the inconsistent results suggest that the application of a correlation coefficient might be inadequate to explore the relationship, because its power may not be sensitive enough to detect the relationship. That is to say, correlation coefficient may not effectively detect significant association when it actually exists. The main cause for the loss of detection power is the complexity of the analytical model. A number of confounding variables are expected to mediate the relationship between environmental scarcity and civil war. As Scruggs (2003) notes, social theories about environmental outcomes are not so exact as to allow us to specify fully the complexity implied in a model. Thus it is almost impossible for us to analyze all the confounding variables at this stage. Secondly, some of the data are only available for a specific year and they measure some characteristics of a country that are stable for a relatively long term. For example, 91 the data for renewable and nonrenewable resource per capita and GINI index were compiled in the 1990s. The original dataset does not keep annual record of them due to the unavailability of the sub-indexes and their relative slow shifting. Considering the characteristics of the data and the research topic, this study decides to use mean values to do a cross-sectional study. Therefore, to better capture the association between environmental scarcity and civil war, this study uses auto-regressions on the number of civil wars ongoing over two periods in the first part of the quantitative analysis. Auto-regression is a fundamental technique in time-series analysis. Compared with other statistical methods, auto- regression can simplify complex models and help us to detect the general pattern. After then, multiple regressions will be brought in to further explore the possible interaction and confounding relationships among environmental scarcity, adaptability and political instability. The following text explains the mathematical model and application of the two statistical methods in detail. Auto-regression Over two Periods The original idea to introduce auto-regression over two periods into this study comes from the research findings on prior war and recurring wars. Prior war is tested to be a highly significant predictor of civil war onset in the major studies (Fearon and Laitin 2003c). Furthermore, a series of studies have found that some countries tend to have recurring wars (Walter 2004 ; Collier, Hoeffler and Soderbom 2008). Thus, it appears plausible to assume that countries with more number of civil wars ongoing in the 92 previous year tend to have more number of civil wars in the current year. Since most of the datasets on civil war keep record of time-series country-year data, 40 it is convenient for us to investigate the ebb and flow of civil war incidence in different countries. This study thus uses the first order auto-regression AR(1) model for each country and then compares their patterns. Since the purpose is looking at a whole list of countries rather than one particular country, higher order models of auto-regression are extremely difficult to analyze. 41 The first-order autoregressive process, or AR(1) process, is represented in the autoregressive form as 42 (3.1) Where [ ] [ ] And [ ] Basically, the first order auto-regressive model regresses the value of the later occasion over the value of the earlier occasion. Specifically, the time period covered by this study is from 1970 to 1999. Since an AR (1) model is adopted, the 30 years are divided into two periods: 1970-1984 and 1985-1999. Each period is fifteen years long. 40 All the civil war datasets used in this study can be transformed into time-series cross-section format, and they all have variables such as civil war onset, number of civil war in progress, and civil war ongoing for each country-year. 41 As Greene (2003) and Wooldridge (2009) suggest, higher order auto-regressive models are very difficult to analyze even when the time-series of one particular subject is looking at. Thus it would be impractical to perform higher order auto-regressions for a cross-national dataset. 42 The formula is from (Greene 2003:257-258). 93 Thus time t is the period of 1985-1999, and time t-1 is the period 1970-1984. Then the annual number of civil wars ongoing in the period of 1970-1984 is used as the predictor of the period of 1985-1999. According to our hypotheses, the auto-regressive patterns for countries with low resource scarcity are different from those with high resource scarcity. In order to detect that difference, all the countries on the list are divided into two groups by the threshold value (median/mean) of environmental scarcity index. 43 Countries with environmental scarcity below the threshold value are in the ―low scarcity‖ group and the rest are in the ―high scarcity‖ group. After conducting auto-regressions of annual number of civil conflict on the two groups separately, it then compares their R-squares and the shape of the two groups‘ best fitting lines. Models with bigger R-squares tend to better fit the linear pattern, and scatter plots can also provide visual evidence of their pattern. The hypothesis would be supported if both the R-square and scatter plot imply that the countries in the high scarcity group shows a more positive linear pattern than the low scarcity group in the auto-regression analysis. 44 Principal Component Analysis PCA is used as a central technique to compose the adaptability variables in this study. The basic idea of PCA is to reduce the dimensionality of a dataset of a large 43 In the auto-regression analysis, the mean is used as the threshold value to divide the two groups and then the median is used. 44 Actually further information beyond the hypothesis can be obtained from the auto-regression analysis. Chapter 4 explores this in more detail and discusses the theoretical implications after the statistical results. Visual representations are also presented. 94 number of interrelated variables, and in the same time retain as much variation in the data as possible (Jolliffe 2002:1). Data reduction is achieved by transforming the original variables to the principal components, which are uncorrelated linear combinations of the variables. The first few of these ordered principal components often retain most of the variation in all of the original variables, whereas the last few principal components account for very little variation, which shows near-constant linear relationships among the original variables (Jolliffe 2002:3). Therefore the last few principal components can be discarded without great loss of information. Suppose the complete dataset has m variables, and the variances of the m random variables and the structure of the covariances or correlations between the m variables are of interest. PCA can help us to find out whether the total variability of the m variables can be accounted for largely by a smaller set of k linear combinations of them. When PCA is applicable, it means that the k components can retain almost as much variation as it is in the original m variables. The original m variables can then be replaced by the k components, which results in a working dataset of k components (Larose 2006:2-3). This project applies PCA as a data reduction method to extract and simplify the useful information from a series of variables measuring the same aspect of environmental adaptability, which is the intermediate variable in the model. Drawing on the broad literature of political instability and civil conflict, it classifies the widely-tested control variables into four categories—economic development level, political institutions, social fractionalization, and demographic characteristics. 95 As addressed in the literature review chapter, a number of studies in the field have tested the association between these variables and political instability, particularly civil conflicts. In general, most of the economic and political variables are found to be significant predictors of civil conflicts, whereas mixed results are reported for some demographic and social variables. This study thus includes the mostly-tested control variable of these four categories in the quantitative analysis and test which of them are significant confounders in the resource scarcity-political instability nexus, and which of them have significant interaction with environmental scarcity. PCA can either base on correlation or covariance matrices. Jolliffe (2002) suggests that the results of analysis based on correlation matrices for different sets of random variables are more directly comparable than those based on covariance matrices. The major reason is that principal components obtained from covariance matrices analysis are very sensitive to the units of measurement used for each variable. If there are large differences between the variances of variables of interest, the first few principal components tend to be dominated by variables with the largest variances. Thus principal components on a covariance matrix are not a good choice when the variables consist of measurements of different types. Furthermore, Jolliffe (2002) argues that it is easier to informally compare the results of different analyses with correlation matrices. The main reason is that the sizes of variance of principal components have the same implications for different correlation matrices of the same dimension, but not for different covariance matrices (Jolliffe 2002:22-24). 96 Our dataset involves quite different types of measurement of adaptability. Even within each aspect of adaptability, it involves measurements from different databases using very different scales. For example, both Polity IV scores and Freedom House scores measure the political institutions of states. However, they have quite distinct scales. Polity IV scores ranges from -10 (strongly autocratic) to +10 (strongly democratic), and Freedom House scores are measured on a 1 to 7 scale, with one representing the highest degree of freedom and seven the lowest. Therefore, PCA is conducted in this study based on correlation rather than covariance matrices. The identifying of components underlying the correlated variables also greatly facilitates further analysis downstream, such as regression analysis, classification, and so on (Larose 2006:8). It is known that the problem of multi-collinearity is one of the major difficulties with the usual least squares estimators. It is often indicated by large correlations between subsets of the variables and its existence can make variances of some of the estimated regression coefficients become very large, leading to unstable and misleading results (Jolliffe 2002:167). Moreover, singularity occurs when the variables are redundant. For example, one of the variables is singular with the other variables if its subscale scores are also included in the correlation matrix. Both multi-collinearity and singularity indicate that the variables of interest contain redundant information so that not all of them are needed in the same analysis (Tabachnick and Fidell 2001:83). 45 In most statistical analyses, the existence of multi-collinearity and singularity can cause both logical and statistical problems. Redundant variables tend to inflate the size of 45 Multi-collinearity or singularity can be attributed to either bivariate or multivariate correlations. 97 error terms and thus weaken the analysis. Thus they are recommended to be omitted unless the aim is to investigate variable structure or repeated measures of the same variable (Tabachnick and Fidell 2001:84). 46 Furthermore, singularity prohibits, and multi-collinearity renders unstable, matrix inversion (Tabachnick and Fidell 2001:84). Matrix inversion is required in many calculations. However, singular matrices produce determinants equal to zero that cannot be used as a divisor in calculations. 47 The existence of multi-collinearity creates near-zero determinants, and ―the sizes of numbers in the inverted matrix fluctuate widely with only minor changes in the sizes of the correlations in R‖ when divided by a near-zero determinant (Tabachnick and Fidell 2001:84). Therefore, the statistical solution based on the unstable inverted matrix is also fluctuating. For most forms of factory analysis, the determinant of R and eigenvalues associated with factors need to be investigated. 48 It is suggested that multi-collinearity or singularity might be present if the determinant is less than 0.00001. Then the Squared Multiple Correlation (SMCs) for each variable is recommended to be checked (Bollen 46 Generally speaking, statistical techniques such as PCA, factor analysis, and structural equation modeling are focused on the structure of variables. Repeated measures are often obtained from longitudinal experimental designs. They are response outcomes measured on the same experimental unit over a period of time (Tabachnick and Fidell 2001). ANOVA is usually used to analyze repeated measures. 47 The determinant of a matrix is a scale factor for measure associated to any square matrix. Calculation of determinants becomes complicated as the dimension of a matrix gets larger. 48 Eigenvalue is a concept of linear algebra that connects with eigenvector. Eigenvalues are a special set of scalars in a finite-dimensional vector space, which is also called characteristic root, proper value or latent root (Marcus and Minc 1988:144). The vector is called an eigenvector of that matrix. The scalar by which each eigenvector is multiplied is the eigenvalue corresponding to the specific eigenvector. 98 1989). The SMC is used to indicate the proportion of variance explained by the model. 49 In SMC analysis, the variable of interest serves as the dependent variable with the rest variables as independent variables. Singularity is present if SMC is one and multi- collinearity is present if SMC is close to one. Although singularity or extreme multi-collinearity is a problem for most types of factory analysis, neither of them is a problem in PCA because matrix inversion is not required (Tabachnick and Fidell 2001:589; Jolliffe 2002:167). 50 Moreover, PCA is interested in investigating the structure of variables to pick up variables that are highly correlated with one another and identify the same components they measure. Thus multi- collinearity and singularity is not a major concern in PCA. This study applies PCA to get one component for each aspect of adaptability, and then utilizes the component scores to do further regression analysis. On the one hand, PCA can give us a single component summarizing the information of several variables in the same category, which enables us to further examine confounding effects and interactions, and divide groups based on the component scores. On the other hand, the application of PCA also mitigates the problem of multi-collinearity, which is a very common problem in social science research. Researchers have been using PCA to construct indexes that summarize various variables capturing the same phenomena. A good example of such an effort is political 49 The multiple correlations indicate the magnitude of the relationship between a dependent variable and the set of independent variables. If y is the dependent variable and and are independent variables, the multiple correlation is calculated as √ 50 There are different forms of factory analysis, such as principal factors, image factor extraction, maximum likelihood factor extraction, alpha factoring, un-weighted least squares factoring, generalized least squares factoring, and so on. (Tabachnick and Fidell 2001:612-613). 99 scientist Douglas Hibbs (1973), who conducts cross-national studies of collective violence utilizing the method of PCA to construct an index of socio-political instability. The author defines mass political violence as ―anti-system‖ events having ―political significance‖ and involving ―collective or mass activity.‖ According to these criteria, six domestic conflict events are taken into account. They are riots, anti-government demonstrations, political strikes, assassinations, armed attacks and deaths. With the help of PCA, Hibbs (1973) reduces the six events to two dimensions of mass political violence—collective protest (incorporates the first three events) and internal war (includes the last three events)—that constitute the dependent variables of his study. Following that tradition, Alesina and Perotti (1993) also construct their own index of socio-political instability (SPI) with PCA. The indicators taken into account by their index are diverse in terms of content and scales, including number of politically motivated assassinations, the fraction of people killed in domestic mass violence, number of successful and unsuccessful coups, and a dummy variable for regime type. Those variables with different scales are averaged of annual values over the sample period and then PCA is utilized to reduce the dimensions and get the SPI scores, which are used in the later regression analysis. Another political scientist, Dipak Gupta (1990, 2008) uses factor analysis in his econometric studies of political instability and violence. For example, in his book The Economics of Political Violence: The Effect of Political Instability on Economic Growth (1990), Gupta condenses ten indicators of political instability into one variable— quotient—and then performs correlation analysis. In his later research on terrorist groups, 100 Gupta (2008) again applies factory analysis to explore how different terrorist activities are associated with each other. The results show that different types of dissent groups tend to pick particular types of activities with careful consideration of their own ideology, expertise, opportunity and the general modus operandi. Based on the factor loadings, the author divides the terrorist groups into five categories: ideological terrorists, professional terrorists, anomic terrorists, hooligan terrorists, and vigilante terrorists. 51 PCA can also be used in his study to achieve the research goal. Some scholars have expressed the concern that the reduction of a number of indictors into one of a few components might be difficult to grasp and the statistical results based on these components are difficult to interpret (Kim 1992). 52 This is not a major concern for our PCA of the four aspects of adaptability because the picking of indicators and how many components to extract are not solely based on the quantitative relationship. More importantly, it is mainly based on the theoretical frameworks and previous empirical research. The main indicators of each type of adaptability are chosen if the major empirical works in the field have found it to be significantly correlated with resource scarcity or political instability. 51 The author mainly focuses on the first three categories of terrorist group activities. According to Gupta (2008:80-81), terrorist attacks such as suicide bombing are defined as ideological because the act requires dedicated members to give their lives. They are either inspired by ideological fever like Hamas, religious extremism like al-Qaeda, or leaders‘ personal charisma like the LTTE. Professional terrorists, in contrast, carry out bombings and car bombings that involve many specialized skills. The IRA and the ETA are examples of this category. Anomic terrorists such as FARC and the Abu Sayyaf Group usually take hostages to make monetary gains. They often thrived in states with weaken central government and operate in anomie or lawlessness. 52 Kim‘s (1992) review of Gupta (1990) expresses this concern. He points out that the way Gupta collapses ten indicators into one quantified dependent variable is ―difficult to grasp,‖ and ―the results of correlations with other variables are difficult to interpret.‖ He also criticizes that the procedure is arbitrary and misleading. Although this study does not agree with most of Kim‘s criticism, it does try to be more careful when it comes to the interpretation of correlations based on the components. 101 Once the indicators are compiled, PCA is conducted to see if they are of a one- component structure, which shows how closely they cluster together. The quantitative criteria of judging how many components should be extracted will be discussed and the output of SPSS will show the extracted components after specific criteria are set up. 53 If the PCA shows that for example, the indicators for economic development form more than one component, then the factor loading of each indicator will be further investigated. Indicators with low factor loadings on the first component are very likely to be dropped if they are not the core indicators of economic development level based on their definitions and previous literature. But for indicators proved to be highly significant and important in previous empirical research, we will be very cautious to drop them off even if their factor loadings are relatively low. 54 What needs to be noted is that this study only applies PCA to reduce the dimensions of the intermediate variable—adaptability—but not the indicators of the dependent variable. On the one hand, it is a consideration of the large number of indicators that can be found for adaptability and their natural categories based on definitions. It makes the following regression analysis much more convenient. On the other hand, the three indicators of political instability are analyzed separately instead of creating a composite score with PCA. The main reason is that they each measures different level of political instability and therefore it can retain more information if the 53 The specific quantitative criteria for extracting reasonable numbers of components are discussed in Chapter 5 where PCA is conducted. In the meantime, the results of statistical analyses by SPSS are also explained carefully. 54 How we decide to drop off or include which indicators is also addressed in more detail in Chapter 5. In general, this study is trying to balance the theoretical and statistical criteria. 102 three indicators of political instability are put in the model separately. It is then possible to check whether different levels of political instability are associated with resource scarcity and adaptive capacity in different ways. For example, some scholars suggest that resource scarcity often leads to protests and demonstrations but seldom results in mass violence. 55 In Chapter 2, a hypothesis is made to address whether resource scarcity is more likely to induce lower-scale political instability than mass political violence. 56 Keeping the three separate indexes thus help us to explore the relationships on different levels of political instability. Interaction and Confounding Effects of Regressions The investigation on interaction and confounding effects is trying to assess whether additional variables may affect the association between the major variables of interest. In general, interaction exists when the relationship of interest is different at different values of the extraneous variable(s), and confounding is present if the inclusion or exclusion of an extraneous variable meaningfully influences the interpretations of the relationship of interest (Kleinbaum, Kupper and Muller 1998:186-187). The main difference between the two is that the assessment of interaction often employs a statistical test on the relationship of interest at different levels of the extraneous variable in addition to subjective evaluation. It is often ―evaluated by using statistical tests about product terms involving basic independent variables in the model‖ 55 A couple of examples are given in Chapter 2. 56 Hypothesis 7 addresses this concern. 103 (Kleinbaum, Kupper and Muller 1998:186-187). In contrast, the assessment of confounding compares a crude estimate (without the extraneous variable(s)) with an adjustment estimate (with the extraneous variable(s) of an association). Confounding is present and the extraneous variable(s) must be included in the model if the crude and adjusted estimates are found to be meaningfully different. Thus a comparison of estimates rather than a statistical test is required according to the definition of confounding (Kleinbaum, Kupper and Muller 1998:186-187). In practice, the same dataset might be considered for both confounding and interaction. In general, it has been agreed that interaction should be assessed before confounding. The evaluation of confounding is recommended only when meaningful interaction does not exist (Kleinbaum, Kupper and Muller 1998:188). Therefore, this study starts with the examination of interaction. Interaction among independent variables can generally be described in terms of a regression model that involves product terms. Suppose a study is intended to explore the relationship between an independent variable and a dependent variable Y, controlling for the possible interaction effects of another independent variable , then the model can be summarized as follows (Kleinbaum, Kupper and Muller 1998:191): (3.2) In the specific model, the variables are as follows Y: political instability : the intercept : slopes for 104 : renewable/nonrenewable resource scarcity : adaptability group (four aspects, respectively) : the product term of resource scarcity and adaptability (interaction) E: the error term According to Kleinbaum et al. (1998), a two-step test need to be performed to assess interaction for this model First, a hypothesis of is evaluated. If the interaction is non-significant, the model should then be re-fit without the interaction (without ). If the interaction is found to be significant, that data are then stratified by (adaptability group), and as the second step, further analyses are done separately within each subgroup (separately for low and high scarcity group). In the presence of significant interaction between resource scarcity and adaptability group, estimates of main effects derived from analysis of the entire list of countries do not tell the whole story. Thus it is better off splitting the countries and reporting the relationship separately for each adaptability group. Assuming that no strong interaction is presence in the model, it then goes on to assess the confounding effects. As mentioned above, is the independent variable and the dependent variable. There is also a third variable , which is the possible confounder. It intends to compare a crude estimate of the -Y relationship with an adjusted estimate that accounts for . The crude models which ignores the effect of the control variable can be expressed as 57 (3.3) 57 The formulas and explanation are also based on Kleinbaum et al. (1998). 105 And the adjusted model (3.4) In the specific model, the variables are Y: political instability : the intercept : slopes for : renewable/nonrenewable resource scarcity : adaptability group (four aspects, respectively) E: the error term A crude estimate of the -Y relationship is the estimated coefficient of ( ̂ ) based on model (3.3) and the adjusted estimated coefficient of ( ̂ ) is obtained from model (3.4), which accounts for -Y. In general, if the coefficient of the variable meaningfully changes when the variable is included, it can be concluded that confounding is present. That is, ̂ ̂ (3.5) As represented by the sign in expression (3.5), whether the two estimates each describes a different interpretation of the -Y association in question is up to a subjective decision. A statistical test is neither required nor appropriate (Addison and Murshed 2002 ; Kleinbaum, Kupper and Muller 1998:194). It is recommended that the researchers should focus on the extraneous variables believed to be reasonably predictive to the dependent variable. 106 Case Selection Once the quantitative analysis is completed, three cases that poorly fit the hypothetical model verified by PCA and multiple regressions will be selected to perform an in-depth case study. It is common for investigators of case studies to confront nonconforming cases to identify missing factors and revise their explanatory framework (Ragin 2004). Since quantitative analysis has an advantage in generalization, looking at the nonconforming cases is known to be more efficient than further investigation on normal cases. In the case analysis, in-depth investigation on the three countries‘ environmental scarcity, state adaptive capacity based on their political, economic, social and demographic conditions, and political stability will be conducted. By investigating nonconforming cases, it can evaluate the merits and deficiencies of the model. Poorly- fitted cases may also help to improve the hypothesized model by adding factors that have been ignored in the statistical analysis. Ideally, a state under high scarcity, low adaptability but maintains its political stability might be indicated as a candidate for further analysis. The selection of cases will also consider the availability of data and the author‘s knowledge of the particular country. In sum, this study utilizes broad data sources representing the quantitative efforts made by scholars in the field. It also breaks through the routine statistical methods that lead to inconsistent results regarding political instability. By introducing auto-regressions and PCA, this study intends to detect the general trend with more sensitive statistical techniques. The concentration on interaction and confounding effects in multiple 107 regressions addresses the role of adaptability in a novel way. Moreover, analysis on nonconforming cases complements the statistical results to better explore the causal mechanisms among the major variables. The next chapter begins the quantitative analysis with auto-regressions. 108 Chapter 4 The Different Patterns in Two Scarcity Groups: An Application of Auto-regressions Over Two Periods This chapter starts with reviewing current literature on the relationships between environmental scarcity and civil conflicts. It focuses on why inconsistent results have been reported in previous time-series cross-section studies. Then it introduces auto- regression to successfully detect the different patterns of annual number of civil wars/conflicts ongoing in countries with high scarcity and those with low scarcity, using five widely-known datasets of civil conflicts. Based on the group differences, it concludes that resource scarcity is significantly associated with the incidence of civil conflicts. Auto-regressions on Annual Number of Civil Wars Over Two Periods The available quantitative studies have found inconsistent results regarding the association between resource scarcity and political instability. Most of them compile time-series cross-section data and use each country-year as the unit of analysis. Their independent variables often include indicators of environmental scarcity and other control variables measuring political, economic, ethnic, religious and demographic characteristics of a country, and civil war onset is widely used as the dependent variable. 58 Furthermore, logit analysis is the most commonly used statistical method in 58 Studies such as Sambanis (2004), Ross (2004), Theisen (2008) review the major variables scholars have used to explore the relationships between environmental scarcity and civil war. 109 the field. Some studies also conduct probit and maximum likelihood analysis to test their hypotheses about the connections between resource scarcity and civil conflicts. 59 To better capture the association between environmental scarcity and civil wars, this study brings in auto-regressions on annual number of civil wars ongoing over two periods in the first part of the quantitative analysis. The measurements of civil conflicts are drawn from five popular civil war/conflict datasets: Fearon and Laitin (2003c), Collier and Hoeffler (2004b), Doyle and Sambanis (2000), intrastate war dataset of ―Correlates of War‖ project (1994), and Gleditsch et al. (2002). Fearon and Laitin‘s (2003c) replication dataset includes the first four indicators, and the last measurement of civil war from PRIO‘s database. 60 Based on the availability of data, the analysis covers the period from 1970 to 1999. The thirty years are divided into two periods: 1970-1984 and 1985-1999, each covering fifteen years. Considering the fact that war is a rare event for the world as a whole, this study looks at relatively longer period rather than examine country-year data. Since it is mainly interested in civil conflicts in each of the fifteen-year term, this study chooses to use annual number of civil wars on going as the indicator of a country‘s state 59 Logit analysis is used to predict a binary dependent variable from a set of independent variables. It makes no assumptions on the distributions of the independent variables (Tabachnick and Fidell 2001:517). Probit analysis is highly related to logit analysis. The main difference is that the logit analysis assumes an ―underlying qualitative‖ dependent variable, whereas probit analysis assumes that the dependent variable is normally distributed (Tabachnick and Fidell 2001:535-536) 60 Fearon and Laitin‘s (2003) replication dataset is available at <http://www.stanford.edu/group/ethnic/publicdata/publicdata.html>, and the dataset proposed by Gleditsch et al. (2002) is available at PIRO‘s website <http://www.prio.no/CSCW/Datasets/Armed-Conflict/UCDP-PRIO/Armed- Conflicts-Version-X-2009/>. 110 of war in each period. This indicator is also better than civil war onset at describing the situations when some countries have several civil wars going on at the same time. 61 Auto-regression is introduced to test the patterns of annual number of civil war ongoing in the two periods. Compared with other statistical methods, such as ordinary least squares (OLS), auto-regression can simplify complex social science models and help to detect the general patterns. It is a good tool to find the masked relationship between variables utilizing datasets from several different databases, since it shows the different patterns which are not as sensitive to minor measurement difference as correlation coefficient is. As discussed before, the estimate of Natural Capital per capita calculated by the World Bank‘s Policy and Economics Team is used as the indicator of a country‘s resource scarcity. It includes both nonrenewable resources (subsoil assets) and renewable resources (timber, non-timber forests, cropland, pasture land and protected areas), and is considered as one of the most comprehensive indicators of resources. Distributions of the Major Variables The distributions of the major variables used in the auto-regressions are displayed in Figure 3, 4 and 5. Based on the five indicators of annual number of civil wars ongoing, Figure 3 shows that the distribution of annual number of civil wars in 1970-1984 is very similar to the distribution of annual number of civil wars in 1985-1999. In each period, most of the countries in the world do not have any civil wars. For example, developed 61 The reasons and considerations for using this indicator of civil war are explained in detail in Chapter 3. 111 countries such as Australia, Belgium, Denmark, Italy, Japan, Norway and the United States experienced no large-scale civil war or conflict in the thirty-year period. Most developing countries such as Brazil, Benin, Cuba, Fiji and Ecuador were also able to maintain a relatively stable political order. Figure 3: Distributions of the Frequency of Annual Number of Civil Wars Ongoing 112 Figure 3, continued 113 Figure 3, continued This explains why scholars have found that war including civil war is a rare event in international relations in general. The highly skewed distributions also imply that some important information can be lost if a way to differentiate the patterns is not figured out. It is believed that logit analysis is not the best way to analyze the dichotomous variable of civil war onset since the time-series data often feature thousands of 0‘s (nonevents) and a hundred or even less 1‘s (events of civil war) (King and Zeng 2001b). 62 Figure 4 shows the distribution of the indicator of environmental scarcity— natural capital per capita. Its distribution is also highly skewed to the right. As the histogram shows, a large number of countries have zero or close to zero dollar of natural capital per capita and very few countries sit in the right part. It can be seen that most 62 This is discussed in Chapter 2 in the section of the concept of political instability. 114 countries suffer a serious to moderate level of environmental scarcity, whereas only a few countries enjoy the abundance of resources. Figure 4: Distribution of Environmental Scarcity Natural Capital Per Capita The discrepancy still holds when the two types of resources are examined separately in Figure 5. A large part of the countries are scarce in renewable or nonrenewable resources. The discrepancy is even more serious for nonrenewable resources: several countries have abundant nonrenewable resources, whereas most countries in the world have none or quite limited nonrenewable resources. For example, several countries including Kuwait and United Arab Emirates have abundant oil resource and thus the value of their nonrenewable resources per capita is higher than $100,000. However, a number of countries such as Afghanistan and Bhutan have little to no nonrenewable resources. 115 Figure 5: Distributions of Renewable and Nonrenewable Resources Per Capita Renewable Resources Per Capita Nonrenewable Resources Per Capita Statistical Results of Auto-regressions All the countries in Fearon and Laitin‘s (2003) list are divided into two groups by the mean value of natural capital per capita. Countries with Natural Capital per capita below the threshold value are in the ―high scarcity‖ group and the rest are in the ―low scarcity‖ group. Firstly, auto-regressions of annual number of civil wars ongoing of the two groups are conducted separately, and then it compares the R-squares and the shape of the best fitting lines of the two groups to see if they have different patterns. Before auto-regressions are conducted, a variable named ―mean group‖ is created. 63 Countries with natural capital per capita less than or equal to the mean value of $7980 is coded as ―1‖ (low natural capital per capita), and countries with natural capital per capita bigger than $7980 is coded as ―2‖ (high natural capital per capita). Table 2 63 Based on the definition, the mean value of the natural capital per capita is $7980. 116 presents the descriptive statistics of the different indicators of number of civil wars in the two groups divided by the mean scarcity value. Table 2: Descriptive Statistics for Mean Scarcity Groups High Scarcity Group Low Scarcity Group N Mean ($) Median ($) SD N Mean ($) Median ($) SD FL Annual Number of Civil Wars Ongoing 1970-1984 94 .188 .000 .390 26 .033 .000 .121 FL Annual Number of Civil Wars Ongoing 1985-1999 94 .292 .000 .511 26 .059 .000 .160 SD Annual Number of Civil Wars Ongoing 1970-1984 94 .194 .000 .389 26 .010 .000 .052 SD Annual Number of Civil Wars Ongoing 1985-1999 94 .210 .000 .349 26 .044 .000 .110 Col Annual Number of Civil Wars Ongoing 1970-1984 94 .117 .000 .250 26 .015 .000 .078 Col Annual Number of Civil Wars Ongoing 1985-1999 94 .134 .000 .279 26 .041 .000 .131 COW Annual Number of Civil Wars Ongoing 1970-1984 94 .118 .000 .301 26 .015 .000 .078 COW Annual Number of Civil Wars Ongoing 1985-1999 94 .165 .000 .374 26 .010 .000 .049 PRIO Annual Number of Civil Wars Ongoing 1970-1984 121 .243 .000 .590 30 .058 .000 .163 PRIO Annual Number of Civil Wars Ongoing 1985-1999 121 .292 .000 .633 30 .096 .000 .233 The results of auto-regressions are summarized in Table 3. Three models have been tested separately using the five indicators of annual number of civil wars ongoing. F-tests of the three models are all highly significant and the R-squares of the three models using the same indicator are quite close. Model 1 uses the annual number of civil wars ongoing in 1970-1984 as independent variable to see if it can predict the annual number 117 of civil wars in 1985-1999. The results are highly significant for all the five indicators of civil war, which implies that countries with more civil wars in the past are likely to have more civil wars in the later period. Table 3: Auto-regressions of Annual Number of Civil Wars Ongoing over Two Periods Model 1 Model 2 Model 3 Intercept .076 ** .158* .143 FLwars7084 1.091*** 1.074*** 1.433** Mean Group (Natural Capital per capita) -.067 -.054 Interaction (Mean Group* Flwars7084) -.350 Explain R 2 .668 .687 .688 P value for F test .000 .000 .000 Intercept .092*** .156* .147* SDwars7084 .599*** .583*** 1.343 Mean Group (Natural Capital per capita) -.059 -.051 Interaction (Mean Group* Sdwars7084) -.756 Explain R 2 .409 .442 .445 P value for F test .000 .000 .000 Intercept .062** .103 .088 Colwars7084 .753*** .570*** 1.283* Mean Group (Natural Capital per capita) -.035 -.023 Interaction (Mean Group* Colwars7084) -.695 Explain R 2 .354 .270 .279 P value for F test .000 .000 .000 Intercept .060* .136 .119 Cowwars7084 .902*** .836*** 1.729** Mean Group (Natural Capital per capita) -.070 -.055 Interaction (Mean Group* Cowwars7084) -.877 Explain R 2 .469 .479 .487 P value for F test .000 .000 .000 Intercept .110** .195 .206 Priowars7084 .694*** .687*** .519 Mean Group (Natural Capital per capita) -.070 -.080 Interaction (Mean Group* Priowars7084) .165 Explain R 2 .413 .415 .415 P value for F test .000 .000 .000 *p<.05 **p < .01 ***p<.001 118 Model 2 adds the grouping variable divided by the mean value of natural capital per capita. As Table 3 shows, the annual number of civil wars ongoing in 1970-1984 is still highly significant, but the grouping variable is not significant. Finally when the interaction term is added in Model 3, the annual number of civil wars ongoing in 1970- 1984 is not a significant predictor any more in the tests using two of the indicators. For the rest of the three indicators, the annual number of civil wars ongoing in 1970-1984 is still a significant predictor of the annual number of civil wars ongoing in 1985-1999. However, the significance level has changed from 0.001 to 0.01 and 0.5. The statistical results of Model 3 suggest that the strong association between previous and more recent civil conflicts is weakened when the interaction term is added. It implies that there might be some connections between natural capital per capita and the pattern of annual number of civil wars in progress. To better uncover the pattern, this chapter goes further to conduct auto-regressions of annual number of civil wars ongoing over two periods in each of the group separately and then compare their results of F-tests and R 2 . Moreover, the scatter plots of the auto-regressions are also compared in Figure 6 to confirm the difference with statistical visualization. The results of these auto-regressions are listed in Table 4, including the correlation coefficient, F-test and R 2 value of each auto-regressive model. Based on the table, the F-rests for auto-regressions are highly significantly for the high scarcity group (F ranges from 35 to 199, p<.001). However, the F-tests for the low scarcity group are only significant for two of the civil war indicators at .01 level (FLwars and PRIOwars), but not for the other three indicators. 119 The comparison of R 2 values also demonstrates a clear difference between the two groups. For the high scarcity group, except Colwars (R 2 = .278), previous annual number of civil wars ongoing can explain more than 40% of the variance of more current annual number of civil wars ongoing for all the other four indicators of civil wars (R 2 = .684, .429, .470, and .406). Furthermore, for each of the five indicators, the R 2 for high scarcity group is much bigger than the R 2 for the low scarcity group. Table 4: Auto-regressions Using Mean of Natural Capital Per Capita to Divide Groups Auto-regressions Mean Groups High Scarcity Group Low Scarcity Group Beta F-test R 2 Beta F-test R 2 FLwars8599 = FLwars7084 .827*** 199.306*** .684 .554** 10.618** .307 SDwars8599 = SDwars7084 .655*** 69.108*** .429 -.081 .158 .007 Colwars8599 = Colwar7084 .527*** 35.458*** .278 -.064 .099 .004 Cowwars8599 = Cowwars7084 .686*** 81.609*** .470 -.040 .038 .002 PRIOwars8599 = PRIOwars7084 .637*** 81.354*** .406 .593** 15.200** .352 *p<.05 **p < .01 ***p<.001 To better illustrate the different patterns, Figure 6 provides the scatter plots of the above auto-regressions with their best fitting lines. The horizontal axis is the annual number of civil wars ongoing in 1970-1984 and the vertical axis is the annual number of civil wars ongoing in 1985-1999. Both linear and quadratic regression lines are showed in the plots. The visual tests also turn out to be very significant. We can see that there is a very obvious linear trend in the high scarcity groups, where the quadratic lines and linear regression lines are quite close with high R 2 values (ranges from .278 to .684). 120 However, the scatter plots for low scarcity countries are much more sporadic. The regression lines are either not as well-fitting as quadratic lines or almost flat. Although the R 2 changes vary when different datasets are used, they all show a much more linear pattern in countries with high environmental scarcity than countries with low scarcity. Thus the scatter plots again confirm the statistical results summarized in Table 4. Figure 6: Comparison of Scatterplots of the Two Scarcity Groups Divided by Mean High Scarcity Group Low Scarcity Group FL Civil Wars SD Civil Wars 121 Figure 6, continued Col Civil Wars COW Civil Wars PRIO Civil Wars 122 Since the distribution of scarcity is highly skewed, the median is probably a more appropriate measure of centrality (Tabachnick and Fidell 2001:81). To confirm the findings above, it then uses the median of natural capital per capita as the threshold value to divide all the countries into two groups and conduct auto-regression analysis again. A new grouping variable named ―median group‖ thus is created and coded according to the median of natural capital per capita. As the distribution is highly skewed, the median differs greatly from the mean (median is $3460, mean is $7986). Again, the results of auto-regressions for all the countries and two scarcity groups are showed in Table 6 and 7. The results of Table 6 are to a large extent consistent with the findings in Table 3. Table 5: Descriptive Statistics for Median Scarcity Groups High Scarcity Group Low Scarcity Group N Mean Median SD N Mean Median SD FL Annual Number of Civil Wars Ongoing 1970-1984 60 .205 .000 .390 60 .103 .000 .311 FL Annual Number of Civil Wars 1985-1999 60 .312 .000 .474 60 .171 .000 .454 SD Annual Number of Civil Wars Ongoing 1970-1984 60 .208 .000 .417 60 .101 .000 .268 SD Annual Number of Civil Wars 1985-1999 60 .232 .000 .358 60 .116 .000 .267 Col Annual Number of Civil Wars Ongoing 1970-1984 60 .146 .000 .271 60 .043 .000 .161 Col Annual Number of Civil Wars 1985-1999 60 .152 .000 .277 60 .076 .000 .230 COW Annual Number of Civil Wars Ongoing 1970-1984 60 .156 .000 .348 60 .036 .000 .145 COW Annual Number of Civil Wars 1985-1999 60 .202 .000 .413 60 .060 .000 .223 PRIO Annual Number of Civil Wars Ongoing 1970-1984 76 .279 .000 .648 75 .132 .000 .387 PRIO Annual Number of Civil Wars 1985-1999 76 .133 2.000 .744 75 .114 .000 .289 123 Table 6: Auto-regressions of Annual Number of Civil Wars Ongoing over Two Periods Model 1 Model 2 Model 3 Intercept .076 ** .121 .214* FLwars7084 1.091*** 1.082*** .499* Median Group (Natural Capital per capita) -.031 -.091 Interaction (Median Group* Flwars7084) .420** Explain R 2 .668 .684 .708 P value for F test .000 .000 .000 Intercept .092*** .165* .168* SDwars7084 .599*** .586*** .564** Median Group (Natural Capital per capita) -.054 -.057 Interaction (Median Group* Sdwars7084) .017 Explain R 2 .409 .443 .443 P value for F test .000 .000 .000 Intercept .062** .086 .100 Colwars7084 .753*** .573*** .407 Median Group (Natural Capital per capita) -.018 -.027 Interaction (Median Group* Colwars7084) .132 Explain R 2 .354 .268 .271 P value for F test .000 .000 .000 Intercept .060* .113 .121 COWwars7084 .902*** .836*** .726* Median Group (Natural Capital per capita) -.041 -.046 Interaction (Median Group* Cowwars7084) .096 Explain R 2 .469 .475 .476 P value for F test .000 .000 .000 Intercept .110** .379** .273* PRIOwars7084 .694*** .671*** 1.218*** Mean Group (Natural Capital per capita) -.177* -.103 Interaction (Median Group* Priowars7084) -.434** Explain R 2 .413 .435 .466 P value for F test .000 .000 .000 *p<.05 **p< .01 ***p<.001 124 Table 7: Auto-regressions Using Median of Natural Capital Per Capita to Divide Groups Auto-regressions Median Groups High Scarcity Group Low Scarcity Group Beta F-test R 2 Beta F-test R 2 FLwars8599 = FLwars7084 .758*** 77.548*** .572 .917*** 206.899*** .841 SDwars8599 = SDwars7084 .678*** 49.236*** .459 .601*** 32.743*** .361 Colwars8599 = Colwar7084 .526*** 22.238*** .277 .470*** 16.408*** .221 COWwars8599 = COWwars7084 .692*** 53.283*** .479 .598*** 32.261*** .357 PRIOwars8599 = PRIOwars7084 .784*** 64.475*** .466 .350*** 20.586*** .220 *p<.05 **p < .01 ***p<.001 The results reported in Table 7 need some explanation. The tests for the civil war indicator from Fearon and Latin‘s (2003c) dataset do not show significantly different patterns in the two groups. As it indicates, the F-tests for both groups are highly significantly (F = 77.548 and 206.899, p<.001), and the R 2 for both groups are also very high (R 2 = .572 and .841). However, for all the other four indicators, the F-test value of the high scarcity group is bigger than the one of the low scarcity group, and the R 2 for the high scarcity group is also higher than the low scarcity group. Again the scatter plots are showed in Figure 7 for comparison. The difference is not as large as the tests based on mean value, but the linear patterns in the high scarcity group are still stronger than the ones in the low scarcity group except for FLwars. The major reason that FLwars shows different patterns from the other indicators may be attributed to its coding of civil war. Same as the other four data sets, it uses a cumulative deaths criterion of 1,000. Besides, it also requires at least 100 deaths per year to code an 125 ongoing war, which places a much higher constraint on coding compared with the other datasets. 64 Figure 7: Comparison of Scatterplots of the Two Scarcity Groups Divided by Median High Scarcity Group Low Scarcity Group FL Civil Wars SD Civil Wars 64 Priowars requires 25 battle deaths per year per incompatibility, and the other datasets either has no criterion for number of death per year. 126 Figure 7, continued Col Civil Wars COW Civil Wars PRIO Civil Wars 127 Discussion on the Findings In sum, the results of auto-regressions have showed that the association between the annual number of civil wars ongoing in 1970-1984 and the annual number of civil wars ongoing in 1985-1999 is quite different in the two scarcity groups. For countries with high environmental scarcity, the annual numbers of civil wars ongoing are very stable. There is a significant positive linear relationship between previous and more current estimates of annual number of civil war in progress. 65 That is to say, in the high resource scarcity group, countries had more civil wars in the past are having more civil wars today, and are also likely to have more in the future. The fact that the estimate of annual number of civil wars ongoing in the previous period is a significant predictor of the later period suggests that they both might relate to or are caused by the same factor. Since resource scarcity is used as the criterion of dividing two groups, the comparison of linear patterns in the two groups implies that resource scarcity is very likely the factor that leads to civil wars in countries in the high scarcity group. A number of countries, such as Philippines, Cambodia, Chad, El Salvador, Ethiopia, Lebanon, Mozambique, Nicaragua, South Africa, and Sudan are typical examples that illustrate this linear relationship. Their estimates of natural capital per capita are lower than both the mean and median values, thus are categorized as the high scarcity group. For example, Philippines had more than one civil war ongoing every year in the period of 1970-1984 for four of the indicators, and the estimate of annual number 65 As Table 4 and Table 6 show, all the linear models for the high scarcity groups are significant. 128 of civil wars ongoing for the left indicator is .87. 66 For the second period from 1985- 1999, Philippines is again estimated to have one or more than one civil war in progress annually according to four of the indicators, and .80 for the other indicator. 67 Thus a consistent pattern of annual number of civil wars ongoing is observed for the two periods, which shows that Philippines‘ risk of civil war is high. Case studies on civil wars of Philippines have identified the causal mechanisms between environmental scarcity and violent conflict. 68 Kahl (2006:66) argues that the spread of communist insurgency in the Philippines was rooted in ―the tremendous demographic and environmental stress‖ on the society since 1960s. The pressures firstly created ―escalating poverty and inequality‖ and then serious economic crisis in the 1980s, both of which imposed significant strains on the Philippine state. As the weakened state could not response effectively to the challenges, societal groups resorted to ―self-help strategies‖ to ensure their own survival, which resulted in an explosion of communist insurgency and eventually a nationwide civil war (Kahl 2006:66). South Africa is another example that demonstrates a consistent pattern of annual number of civil wars ongoing in the two time periods. As a country with high resource scarcity, South Africa is estimated to have civil war ongoing in the two periods for three indicators and no civil wars for the other two indicators. Based on the coding of Fearon 66 The estimates of annual number of civil wars ongoing for Philippines in the period 1970-1984 are as follows: FLwars (1.87), SDwars (1.73), Colwars (.87), COWwars (1.47), and PRIOwars (2). 67 For the period from1985-1999, the annual number of civil wars ongoing Philippines had is similar to the previous period. They are FLwars (1.67), SDwars (1.33), Colwars (.80), COWwars (1), and PRIOwars (1.73). 68 For example, Homer-Dixon (1994) and Kahl (2006) use Philippines as a case to support their proposition of the causal mechanisms. 129 and Latin (2003c), SDwars and PRIOwars, South Africa had more than zero or even more than one civil wars ongoing annually in both periods. However, according to Collier and Heoffler‘s and COW datasets, South Africa had no civil war in the thirty years. 69 The main reason for the difference is the different definitions and coding rules the different datasets use for civil war. COW and Collier and Hoeffler‘s datasets require 1000 annual battle-deaths, whereas the other three datasets use lower threshold levels. 70 But for each indicator, the annual number of civil wars in progress for South Africa is consistent across the thirty years. Previous case studies in the field have also analyzed the linkages between resource scarcity and civil violence in South Africa. Percival and Homer-Dixon (1998b, 1998a) examine the state-society relations in South Africa and demonstrate that environmental scarcity is the major contributor to social instability in the context of apartheid. In South Africa, resource pressures led to lower agricultural productivity, which increased migration to urban areas and stress on the urban environment. During this process, the state failed to effectively cope with these strains due to its undermined ability to serve the need of the society, which eventually resulted in rising grievances within society. Since South Africa was under the transition from minority rule, the grievance was more likely to be expressed in violent means (Percival and Homer-Dixon 69 For the period of 1970-1984, the estimates of annual number of civil wars ongoing in South Africa are as follows: FLwars (.13), SDwars (.60), Colwars (0), COWwars (0), and PRIOwars (1.20). The estimates for 1985-1999 are FLwars (.67), SDwars (.67), Colwars (0), COWwars (0), and PRIOwars (.53). 70 The other datasets either use a lower threshold of annual battle-deaths (i.e. PRIO dataset requires 25 annual battle- deaths, Fearon and Laitin dataset requires 100 annual battle-deaths) or cumulative deaths (i.e. Doyle and Sambanis dataset sets the criteria of 1,000 overall deaths). More details about each dataset‘s definition and coding rules on civil war are explained in Chapter 2. 130 1998a:109-110). The case of South Africa clearly demonstrates that countries in the process of democratization and social transitions have greater possibility to be destabilized by environmental challenges. However, countries with low environmental scarcity do not show a significant pattern for the annual number of civil wars they have. The linear patterns between previous and more current annual number of civil wars in progress are less significant compared with the high scarcity group, and some results based on the groups divided by mean are even not significant at all. 71 It indicates that previous annual number of civil wars in progress is not a consistent predictor of the annual number of civil war ongoing in the later period. In other words, for countries with low scarcity, the annual number of civil conflicts they had in the past fail to predict the annual number of civil wars they each has later. It is thus suggested that resource scarcity is not a significant predictor of the annual number of civil wars ongoing in countries with low resource scarcity. It is probably other political, economic and social factors not considered in the auto- regression that cause civil wars in these countries. Most of the countries in the lower scarcity groups did not experience any civil war in the period from 1970 to 1999. 72 Several African countries such as Algeria and Congo are rich in nonrenewable resources and therefore are in the low scarcity group. They indeed had civil wars from 1970-1999, and they demonstrate inconsistency regarding the annual number of civil wars ongoing in the first and second fifteen-year periods. For 71 Table 4 and Table 6 show the results. 72 Countries like Australia, Canada, Costarica, Finland, etc. 131 instance, all five indicators show that Algeria did not suffer any civil wars from 1970- 1984, but in the period of 1985-1999, Algeria had more than zero wars ongoing annually based on the same indicators except the one coded by COW project. 73 After its independence from France in 1962, Algeria was in relative peace in the 1970s and 1980s. As the government cracked down on the Islamist rebel groups, the fight escalated into Algerian civil war around 1992. It ended in the early 2000s with a government victory, but low-level fighting is still going on in some areas. As Mortimer (1996:19) suggest, Algeria‘s political crisis and civil war results from its ―flawed transition from single-party to democratic politics‖. 74 The process of democratic transition requires dialogues among diverse social forces to address pluralism and liberties. But the Algerian state failed to deliver such a dialogue between secularists and Islamists, which resulted in an armed conflict ―between the military establishment and the extremist wing of the Islamic movement‖ (Mortimer 1996:19). Therefore, it can be concluded that low level of resource scarcity does not necessarily increase the risk of civil war, but when scarcity reaches some threshold level, it then endangers a country‘s political stability and produces civil wars. The statistical evidence of group difference in terms of linear patterns and significance of F-tests and R 2 ‘s suggest that mean value is a critical threshold for the relationship between resource 73 For the period of 1970-1984, the annual number of civil wars ongoing in Algeria are all zero, and the estimates for 1985-1999 are FLwars (.53), SDwars (.40), Colwars (.60), COWwars (0), and PRIOwars (.60) 74 More analysis of the Algerian Civil War is provided in Roberts (1995), Mortimer (1996) and Addi (1998). Roberts (1995) maintains that in order to reach ―a valid, internally reached solution‖ for the war, the Algerian army and France needs to reconsider their positions. Mortimer (1996) explains the war by analyzing the democratic transition. Addi (1998) points out that the army‘s place above civil law rather than the Islamist rebels is the key player in Algeria's crisis. 132 scarcity and civil war. The findings still hold to a large degree when the median value is used to divide the scarcity group, though it is not as strong as the results based on mean. In conclusion, the results of the auto-regression analysis imply that environmental scarcity is an important factor for understanding civil war. While countries with high environmental scarcity suffer from more civil wars for a long time, countries with low environmental scarcity have fewer and more randomly distributed civil wars. Our findings suggest that it is promising to divide countries into two groups based on their degree of environmental scarcity, and then compare the patterns between those with low and high scarcity. Most scholars with similar interest in environmental scarcity do not take the group differences into consideration, and thus they fail to detect the hidden association between environmental scarcity and civil wars. Based on the interesting findings of auto-regressions, this study goes further to confirm the more complex relations between environmental scarcity and political instability, adding environmental adaptability as an intermediate variable. The comprehensive model, focused on interaction and confounding effects, is tested by regression analysis in the next chapter. 133 Chapter 5 Adaptability and Environmental Scarcity-Political Instability Nexus This chapter goes further to test the more complex relations between environmental scarcity and political instability, adding environmental adaptability as an intermediate variable. Based on the proposed theoretical model, four aspects of adaptive capacity are considered important for environmental challenges. They are economic conditions, political institutions, social fractionalization and demographic characteristics. Each aspect of adaptability is tested respectively as an intermediate variable in the environmental scarcity-political instability nexus in this section. The chapter utilizes Principle Component Analysis (PCA) as the data reduction method for multiple indicators of each type of adaptability, and then tests the theoretical expectations by regression analysis followed by a discussion on confounding effects, interactions and group difference. Data Screening and Transformation Before the core statistical analysis is conducted, the study performs data screening to examine the basic characteristics of the data. It starts with an investigation on the pattern of the missing data. Since missing values can skew the distributions of variables, it is crucial to understand if they are randomly missing or following some pattern. In general, randomly missing values pose less serious problems, whereas non-randomly missing values can be misleading even though only a few of them exist. To say it in a brief way, the pattern of missing values can affect the generalizability of the statistical 134 results (Tabachnick and Fidell 2001:58). Most of the statistical softwares provide some instruments to test if the values are randomly missing and detect the possible patterns. SPSS 17.0 provides a test of Missing Value Analysis (MVA) to investigate the patterns of missing data and then help users to make decisions about handling them (SPSS Inc.). Instead of treating incomplete data with simple techniques such as listwise deletion, pairwise deletion, or mean substitution, this study utilizes the Expectation Maximization (EM) algorithm to impute missing data. The EM method assumes that missing values are Missing at Random (MAR), which means that the missing data is missing conditional on some other independent variables observed in the dataset, not on the dependent variable of interest (Schafer 1997:10; Scheffer 2002:153). Another more stringent condition is Missing Completely at Random (MCAR), which requires that the mechanism of the missing data does not depend on any variable observed in the dataset (Scheffer 2002:153). It is valid to delete the cases with missing values only if this very rare condition is satisfied (Rubin 1976). Moreover, replacing missing values by the observed mean for that particular variable is another widely-used and easy to implement method to handle missing data in social sciences. However, as Little and Rubin (1987) suggest, mean substitution is a poor choice under most circumstances since it tends to decrease the variance of the variable with missing values. The main reason for the reduced variance is due to its underrepresentation of extreme values. Thus mean substitution should generally be avoided as a method of handling missing values given the danger of generating biased estimates of variances and covariances (McKnight et al. 2007:177). 135 Compared with case deletion and mean substitution, the EM method offers a more reasonable approach to impute missing data, as long as the MVA tests and the preliminary analysis can provide adequate evidence for the randomness of missing cases (Tabachnick and Fidell 2001:66) . The MVA tests are performed on the datasets using SPSS, and evidence has been found that the missing values are MAR not MCAR. Thus EM algorithm is believed to be appropriate for handling the missing data. Specifically, the EM method is an iterative procedure for missing data (Tabachnick and Fidell 2001). It calculates the maximum likelihood estimates through two steps: the E-step and the M-step. First, the E-step estimates the missing data given the observed data and current estimates of the model parameters. According to the estimates of the missing data from the E-step instead of the actual missing data, the M- step then computes the parameters that maximize the expected likelihood function. It is well-known that the EM method as imputation for missing values is less biased than listwise deletion and mean substitution and more accurate than pairwise deletion (Tabachnick and Fidell 2001). As another important step for multivariate analysis, the major variables are then screened to see if they satisfy the basic assumptions underlying most multivariate procedures and statistical tests such as normality, linearity and homoscedasticity. Although normality of the variable is not always required for analysis, Tabachnick and Fidell (2001:73) suggest that the solution is usually better if all the variables are normally distributed, otherwise the solution is often degraded. 136 Therefore, before multivariate analysis is conducted, researchers usually assess the normality of variables by statistical or graphical methods. Skewness and kurtosis are two widely used indicators for assessing the normality of distributions. Skewness measures the symmetry and kurtosis has to do with the peakedness of a distribution (Tabachnick and Fidell 2001:73). They both are zero for a perfectly normal distribution. 75 Furthermore, the histogram of each variable is also checked to see if it is close to a bell- shaped distribution or not. Bell-shaped distributions are in general closer to normal, whereas non bell-shaped distribution might require further transformation. 76 It has been found that the distributions of some variables in the model differ moderately or greatly from normal. One common way to handle such skewness is to apply a transformation to the variable, such as the square root or ln transformation. After the transformation of skewed variables is performed, the distributions often better meet the assumptions, and the interpretability will also be improved. In this study, ln transformation is conducted before the multivariate analysis in order to make those independent variables of heavily skewed distributions closer to bell-shaped normal distribution. For these abnormally distributed variables, ln-transformed values are used instead of their original values in the statistical analysis followed. As addressed in the methodology chapter, this study is focused on the time period from 1970-1999 due to the availability of data. Furthermore, for the time-series data, we 75 Perfectly normal distributions are very unlikely in reality, especially in social science research. The distributions are close to normal if the skewness and kurtosis are small. 76 Usually researchers perform ln, log, square or square-root transformation on the skewed variables to make the distributions closer to normal distribution, at least bell-shaped distribution. As normality is one of the basic assumptions of many multivariate analysis methods, highly skewed distributions of variable will violate the assumption and produce inaccurate statistical results. 137 use the mean values of those indicators across the thirty-year period. It is believed that simplifying the model using mean values can help us better detect the general trend when dealing with complex social science mechanisms. 77 Principal Component Analysis and Regression Analysis The previous chapter has detected the masked connection between environmental scarcity and civil conflict using auto-regressions over two periods. This section goes further to test the comprehensive model adding environmental adaptability as an intermediate variable, and enlarges the concept of civil conflicts to political instability to incorporate less intense forms of political crises, such as coups d‘état, strikes, major government crises, riots and so on. In the following text, Principal Component Analysis (PCA) is used as a data reduction method for various indicators of the four aspects of adaptability respectively. After then regression analysis is applied to examine the confounding effects and interactive relationship. In the meantime, all the countries in the list are divided into groups based on their values of each type of adaptability, and then further exploration of group difference regarding the scarcity-political instability nexus is performed. These statistical analyses together offer a comprehensive test of the proposed model and hypotheses. 77 Chapter 2 explains the concerns and provides justifications for conducting cross-sectional analysis using mean values in detail. 138 Before the PCA for each aspect of adaptability is carried out, factorability of the correlation matrix R is required to be checked to see if PCA is appropriate for the data structure of interest (Tabachnick and Fidell 2001:589). This involves several steps. First, the sizes of the correlations need to be investigated before further analysis is conducted. The use of PCA is in question if none of the bivariate correlations between variables exceeds .30. Low correlations probably imply that there is little to analyze regarding component relationship. 78 However, high bivariate correlations are still not sufficient proof that the correlation matrix contains components. Further examination of partial correlations matrices is required before PCA can be reasonably applied (Tabachnick and Fidell 2001:589). SPSS provides two tests for examining the partial correlation matrices. Bartlett‘s test of sphericity is a ―sensitivity test of the hypothesis that the correlations in a correlation matrix are zero‖ (Tabachnick and Fidell 2001:589). However, it is not recommended as the sole standard because of its dependence on N. Another test is Kaiser‘s measure of sampling adequacy (1970 ; 1981), which is ―a ratio of the sum of squared correlations to the sum of squared correlations plus sum of squared partial correlations‖ (Tabachnick and Fidell 2001:589). The value approaches 1 as partial correlations are small and 0 when the partial correlations are big. In general, the values of .6 and above are required for good PCA. This study first applies PCA to construct a one-component structure for each aspect of adaptability, and then utilizes the component scores to do further regression 78 The sizes of correlation coefficients between variables show the magnitude of their relationship. 139 analyses. On the one hand, PCA can give us a single component summarizing the information of several variables and enable us to examine confounding effects and interactions, and divide groups based on the component scores. On the other hand, the application of PCA also mitigates the problem of multi-collinearities, 79 which is a very common problem in large-N social science research. Specifically, the statistical analysis in this section starts with PCA for each aspect of adaptability. Take the economic conditions as an example. At first, a variety of economic variables used by scholars in political instability and environmental studies are listed, such as GDP per capita, GNI per capita, GINI index and so on. It then picks up the mostly tested ones and conducts PCA to see how many components can be extracted. The PCA is performed several times and modified until the results suggest that only one component should be retained. During this process, some variables are dropped and some are added. According to Larose (2006:9-11), there are four main criteria for deciding how many components should be retained: 1) Check the communalities to retain the first several components that explain 70-80% of the total variation. 80 2) Only those components with eigenvalues greater than 1 are kept in the model. 81 79 If the variables of interest are very highly correlated (.90 and above) with each other in the correlation analysis, then there is a problem of multi-collinearity. That means those variable contain redundant information and therefore not all of them are needed in the same analysis (Tabachnick and Fidell 2001:82-83). 80 For each variable investigated in PCA, its communality is the sum of the squared loadings for all the retained components (Dunteman 1989:58). It shows the amount of variance in a particular variable that is accounted for by the retained components. The outputs of the PCA always list those components with higher communalities at first. 81 The definition of Eigenvalue is explained in Chapter 3. 140 3) Check the scree plot, look for the ―elbow‖. 82 4) Consider whether the component has a sensible and useful interpretation. Since the aim is to get a reasonable one-component structure of the variables, this study tries different compositions of the economic variables given the first three criteria. But the fourth criterion is also of great importance. Hence, this study tends to retain those variables that have been confirmed as significant by a number of previous studies. For instance, as far as economic conditions are considered, GDP per capita has been addressed by many scholars as a highly significantly predictor of civil conflict and environmental crises. Therefore, it would be extremely risky to drop crucial variables like this. Once the appropriate component structure is decided, PCA can calculate a component score for each case based on the variable retained. Specifically, PCA assigns weights to each variable and the component scores are actually scales developed by combing the variables in terms of these weights (Rummel 1967). The component scores then are used in place of the original variables in further regression analysis since they reduce the dimensionalities while retaining the meaningful variation in the original data. In this part of statistical analysis, the major measurement of resource scarcity is still extracted from the Natural Capital dataset of the World Bank. But it now goes further to examine the two categories of resources—renewable and nonrenewable—respectively. The scarcity of each type of resources is tested in the statistical models separately to see 82 In the scree plot, the components are arranged in descending order along the abscissa with eigenvalue as the ordinate (Tabachnick and Fidell 2001:621). Thus, the eigenvalue is highest for the first component and decreasing for the next ones. The elbow in the scree plot shows the change of slope, which means the eigenvalue for the components on the left of the point suddenly decreases to small values. 141 if their associations with political instability and adaptability differ due to the resource of interest. As it is addressed in the previous chapters, this study looks not only at extreme cases of interruption of political stability, such as civil war and armed conflict, it also counts other less intense forms of political instability, including annual assassinations, coups d‘état, strikes, major government crises, riots, anti-government demonstrations and so on. Besides the annual number of civil war ongoing from the five datasets appeared in auto-regression analysis, the Worldwide Governance Indicators (WGI) of political instability are also tested as dependent variables in this section. Moreover, the Political Instability Task Force has also collected data for four types of severe political crises: revolutionary wars, ethnic wars, adverse regime changes, and genocides and politicides. The numbers of these four types of political crises are summed up as one of the main indicators of political instability. Based on the hypothesized model, the interplay between environmental scarcity and adaptability affects a state‘s political stability. And the patterns for renewable and nonrenewable resource scarcities are different from each other. A society with high adaptability usually has better capability to cope with crises and challenges, due to the proposed economic, political, social and demographic features. Thus it is believed that the scarcity of renewable resources would not shake its political stability. Instead, sometimes a country with adequate adaptability can even reinforce its political stability after effective reactions. In contrast, environmental scarcity could be detrimental to a 142 society with low adaptability, since the country lacks the necessary economic, political, social and demographic conditions to fight against the environmental challenges. Therefore, in front of renewable resource scarcity, the political stability of a country with low adaptability will be decreased to a different extent. The reverse effects may also exist, which means political instability can in turn influence those determinants of adaptability and then the factors determining a society‘s adaptability can also affect its environmental scarcity level. However, this study is more interested in the former effects based on the tradition of international security studies. Thus it follows the literature in the field in assuming that the former effects rather than the reverse effects would be the primary one. Same as the model for renewable resources, the economic, political, social and demographic conditions determinate a country‘s adaptive capacity to nonrenewable resource scarcities. Countries with high adaptability usually have high capacity to cope with challenges, whereas those with low adaptability are often poor at coping with environmental issues. However, the hypothesized causal mechanisms of nonrenewable resources are different from those of renewable resources based on assumptions of the distinct utilities of these two types of resources. For nonrenewable resources, it is more likely the abundance rather than the scarcity would impose environmental challenges on a country‘s political stability according to the major studies in the field. It is thus expected that the abundance of nonrenewable resources would weaken the political stability of a country with low adaptability. The abundance of nonrenewable resources can bring great 143 wealth to a country, which often leads to the country‘s over-reliance on selling primary commodities. This in the long term is unhealthy for the country‘s development. In the meantime, according to the ―greed‖ theory, the wealth from selling nonrenewable resources is the primary source of funding the violent factions. 83 However, the political stability of a country with high adaptability is not necessarily negatively impacted by its amount of nonrenewable resources. The abundance of oil and minerals, for instance, provides firm platform for the economic taking off of a country with high adaptability, whereas the lack of these nonrenewable resources might drive its leadership‘s attention to the international market, looking for opportunities to transform its economic strategies. 84 Again, this study recognize but does not focus on the possible reverse effects, which acknowledge that political instability can in turn influence the various determinants of adaptability and then affect the country‘s nonrenewable resources abundance/scarcity level. The following analysis explores the hypothesized model in detail. Economic Conditions This section firstly uses economic conditions as intermediate variables to examine the relationship between resource scarcity and political instability. Several economic variables such as Gross Domestic Product (GDP) per capita have been tested to be 83 Chapter 2 explains this theory advocated by some neoclassical economists. 84 For example, Japan developed with an international strategy, importing raw materials and exporting industrial products (Balassa, Noland and Institute for International Economics (U.S.) 1988 ; Calder 1988). The strategy is greatly determined by its limited land and nonrenewable resources. 144 significantly associated with political instability. The resulting scree plots and Eigen values are examined, and the interpretability is also discussed. After several experiments, it picks up three indicators of economic conditions—GDP per capita, Gross National Income (GNI) per capita and GINI index—from the World Bank‘s World Development Indicators, and the estimate of Real GDP per capita from the Penn World Table 6.1. Initially, GDP growth, GDP per capita growth and foreign direct investment (% of GDP) (FDI) are also included in the model. But the determinant turns out to be .000 when all those variables are included, and PCA finds a three-component structure. The analysis shows that GDP per capita, real GDP per capita and GNI per capita cluster together, whereas GDP growth and GDP per capita growth cluster together. Besides, FDI and GINI index are a little bit far from the two clusters. 85 In general, variables within a cluster are approximately measuring the same factor, whereas variables far away from each other on the scree plot do not share the same components. The first cluster is taken as the baseline since it incorporates the biggest number of variables. After then the other variables are included in the following PCAs one by one. It appears that GINI index is the closest to the main cluster compared with the other variables according to their factor loadings and scree plots. 86 Besides, the other three variables are always far from the first cluster and a one-component structure is not 85 The new growth theories have suggested that inequality has impact on economic growth of a country. More discussion and empirical tests of the theories can be found in Aghion et al. (1999), Barro (2000), Moene and Wallerstein (2001), and Abhijit and Esther (2003). 86 The four variables create a straightforward one-component structure, with the GINI index showing a relative lower factor loading. The following table and text will explain this in more details. 145 detected. 87 Given its theoretical implications and empirical evidence found by previous research, the GINI index is thus kept in the component as a critical measurement of income inequality. This concept generally reflects a country‘s distribution of resources among individuals and it is of crucial importance to political ecologists‘ arguments. 88 The results of PCA of economic adaptability are listed in Table 8. The factor loadings reported are the correlation coefficients between the variables and components. And the squared factor loading is the percentage of variance in the variable that is explained by a component. Furthermore, for a given variable, its communality is the sum of the squared loadings for all the retained components (Dunteman 1989:58). The communality shows the amount of variance in a particular variable that is accounted for by all the retained components together. If no components are dropped, then the communality should equal 1.0 which means 100% of the variance of the given variable is accounted. Since a stable one-component structure is desired for each aspect of adaptability in this study, the factor loadings showed are just for a single component. Researchers generally use factor loading level such as .4 for the central component and .25 for other components as cutoff levels (Raubenheimer 2004). Component scores are also offered by SPSS and the results listed in Table 8 are weights assigned to each variable for calculating the scores. The regression method is selected 87 Two components are found when GDP growth and GDP per capita growth are added in the model; similar results are found for FDI (% GDP). It then can be concluded that they are not closed clustered with the first three indicators of economic development. 88 Political ecologists suggest that it is the distribution of natural resources rather than the absolute amount of them that leads to political instability. 146 when the scores are calculated since it is the widely used one (Tabachnick and Fidell 2001). Table 8: Principal Component Analysis of Economic Development Component and Variables Factor Loading Score Coefficient Sampling Adequacy Test of Sphericity Compt1: Economic Development .742 .000 GDP per capita (ln) .979 .324 GNI per capita (ln) .982 .325 Real GDP per capita (ln) .971 .322 GINI index -.390 -.129 As Table 8 shows, ln GDP per capita, ln GNI per capita and ln real GDP per capita have factor loadings higher than .9. The factor loading of GINI index is much lower than the first three variables, but it is still close to .4 (-.390). The main reason GINI index has lower factor loading than the first three economic variables can be traced back to their definitions. According to the World Bank, GDP is ―the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products,‖ and GNI is ―the sum of value added by all resident producers plus any product taxes (less subsidies) not included in the valuation of output plus net receipts of primary income (compensation of employees and property income) from abroad (World Bank).‖ 89 To put it in an easy way, their major difference is that GNI includes ―net income from abroad,‖ whereas GDP excludes it. Furthermore, real GDP is a 89 Both definitions are from database of World Development Indicators. 147 given year‘s nominal GDP stated in the based-year price level. Basically it is normal GDP adjusted for price changes and inflation. Therefore, GDP per capita, real GDP per capita and GNI per capita are closely related macroeconomic measures based on their definitions and calculating methods. A number of security-oriented studies have found that economic development, in particular GDP per capita is a highly significantly predictor of civil conflict outbreak (Fearon and Laitin 2003c). Furthermore, GNI is also a widely used index measuring economic development. For instance, household income is an important factor predicting the probability of household participating and supporting an armed insurgence in combat areas (Justino 2009). GINI index is the most widely used measure of income inequality. It varies between 0 (complete equality) and 1 (complete inequality). It is quite different from the above three variables, thus is not so closely related to them. However, this study decides to retain it as a main indicator of economic adaptability because the nexus between economic inequality and civil wars have also been exhaustively examined by political scientists and economists. As some scholars suggest, countries with more economic inequality are more likely to suffer civil violence, especially genocides (Besanç on 2005). The major reason is that the production of grievances due to economic inequality makes the traditionally deprived identity groups more likely to engage in conflict (Besanç on 2005 ; Schock 1996). With the regression method, PCA generates component scores based on these four variables. The score has a normal distribution with mean zero and standard deviation one. 148 It sums up the information of variation of the four economic variables. To test the hypothesis that the associations between environmental scarcity and political instability tend to be different for countries with high adaptability and those with low adaptability, this study divides all the countries into two groups based on their values of economic component scores. Specifically, the 66th percentile of economic component score is the dividing point. Countries with component scores below and equal to the 66th percentile are assigned to group1 (low economic adaptability group), and those above the 66 th percentile are assigned to group 2 (high economic adaptability group). Thus a grouping variable differentiating countries with low from those with high economic adaptability is created. The reason to choose the 66th percentile as the threshold value is based on experiments. Originally, the countries are divided into three groups based on the 33rd and the 66th percentiles of the economic adaptability component scores. Although some interesting results have been found, this study further discovers that the statistical results are more significant and consistent when the low and middle economic adaptability groups are combined together. 90 After conducting a series of regressions using the three indicators of political instability—wars, PITF political instability, and World Bank political instability estimate, it finds that the interaction between renewable resource scarcity and economic adaptability group is a significant predictor of political instability. 90 The results of analysis based on three adaptability groups are provided in Appendix A. 149 Table 9: Pearson’s Correlations between Economic Conditions Variables, Environmental Scarcity and Political Instability Variables GDPpc (ln) GNIpc (ln) Real GDPpc (ln) GINI Econgroup Renew (ln) Nonren (ln) Wars PITF wars WB Polin GDPpc (ln) 1.0 .986** .940** -.236** .811** .377** .414** -.316** -.353** .615** GNIpc (ln) 1.0 .938** -.267** .829** .383** .414** -.326** -.356** .636** RealGDPpc(ln) 1.0 -.303** .790** .386** .460** -.322** -.362** .612** GINI 1.0 -.402** -.129 -.037 .055 .071 -.261** Econgroup 1.0 .291** .249** -.309** -.318** .566** Renew (ln) 1.0 .361** -.225** -.220** .324** Nonrenew (ln) 1.0 -.095 -.105 .197* Wars 1.0 .832** -.538** PITF wars 1.0 -.564** WB Polin 1.0 *p<.05 ** p < .01 150 Table 10: Testing for Interactions and Confounding Effects of Economic Conditions Civil Wars PITF Political Instability WB Political Instability Model1 Model2 Model3 Model1 Model2 Model3 Model1 Model2 Model3 Intercept .707*** .748*** 2.520*** 32.149*** 33.116*** 112.831*** -2.689*** -2.940*** -7.891*** Renewable Resources per capita (ln) -.069** -.045 -.276** -3.128** -2.049* -12.422** .317*** .170* .814** Adaptability Group (66% ) -.169*** -1.223** -7.343*** -54.764** 1.038*** 3.984*** Interaction (Renew*Adapt Group) .136** 6.112** -.380* Explain R 2 .054 .115 .153 .049 .125 .167 .105 .348 .377 P value for F test .004 .000 .000 .005 .000 .000 .000 .000 .000 Intercept .204*** .433*** .424*** 9.681*** 19.465*** 20.682*** -.465*** -1.786*** -1.846*** Subsoil Assets per capita (ln) -.008 -.002 .000 -.424 -.323 -.588 .055* .017 .030 Adaptability Group (66% ) -.193*** -.185* -7.987*** -8.923* 1.120*** 1.156*** Interaction (Subsoil*Adapt Group) -.002 .191 -.009 Explain R 2 .009 .096 .096 .011 .109 .120 .039 .323 .324 P value for F test .232 .000 .001 .185 .000 .000 .012 .000 .000 *p<.05 ** p < .01***p<.001 151 Table 11: The Different Patterns among Low and High Economic Adaptability Groups Civil Wars PITF Political Instability WB Political Instability Adaptability Group (66%) Group1 Group2 Group1 Group2 Group1 Group2 Intercept 1.297*** .073 58.067** 3.302 -3.907*** .076 Renewable Resources per capita (ln) -.140** -.004 -6.310** -.198 .435*** .055 Explain R 2 .067 .003 .077 .003 .107 .012 Intercept .238*** .053 11.759*** 2.836* -.689*** .467** Subsoil Assets per capita (ln) -.001 -.003 .397 -.206 .020 .011 Explain R 2 .000 .007 .008 .021 .006 .003 *p<.05 ** p < .01***p<.001 152 Figure 8: Scatter Plots for Two Adaptability Groups of Renewable Resources (Civil War Index is Used) Economic Development Political Institutions Social Fractionalization Demographic Characteristics Low High 153 Figure 8, continued Economic Development Political Institutions Social Fractionalization Demographic Characteristics Low High 154 As Table 10 shows, renewable resource per capita is a significant predictor of political instability based on the simple regressions in Model 1. However, renewable resource per capita is no long significant or less significant (from p<.001 to p<.05) when economic adaptability group is added in Model 2. Furthermore, Model 3 goes further to add the interaction between renewable resource per capita and the dummy group to the regressions. All the variables are highly significant and the R 2 are also improved for all the three indexes of political instability. This means that it does not make much sense to talk about the separate effects of renewable resource scarcity and economic conditions on political instability, since their impacts on political instability do not operate independently. The statistical results suggest that the data should be stratified by the value of economic adaptability, and the relationship should be then explored separately within each group. Table 11 summarizes the results of simple regressions between renewable resources and political instability when the two groups are considered respectively. Generally speaking, the association between renewable resource scarcity and political instability is significant for the low economic adaptability group (Group 1), but not for the high economic adaptability group. Besides, the R 2 of the low economic adaptability group is also much higher than those of the other groups. The results are consistent for all the three indexes of political instability. The comparison of scatter plots in Figure 8 confirms the conclusion that there is more linear trend in the low than in the high economic adaptability group. 155 The statistical results suggest that high economic adaptability can effectively mitigate the deteriorative impacts of environmental pressure, whereas low to fair level of economic adaptability in general cannot save the country from the crisis. Japan is a very strong example to illustrate the effectiveness of economic adaptability. Its renewable resource per capita is below the 25 percentile, 91 which indicates that the country faces great pressure from renewable resource scarcity. However, Japan is quite strong economically. Its GDP per capita and GNI per capita rank top on the list, and it is among the top 5% countries in the world as far as income equality is concerned. The strong economic conditions of Japan make it very capable at adapting to resource challenging. According to the data, Japan does not have any types of civil conflict during the period 1970-1999, and rank as one of the top fifteen politically stable countries based on the World Development Indicators. However, countries with low economic adaptability are not as lucky when facing renewable resource scarcity. For instance, Afghanistan also suffers from great renewable resource scarcity. Meanwhile, the country is economically weak and thus is poor at adapting to environmental challenges. During the period from 1970 to 1999, it had civil wars among different ethnic factions and insurgencies followed by Soviet invasion. Generally speaking, the country is in great danger of violent conflict and regime crisis. The same regression tests are conducted on the model of nonrenewable resources. In model 1, nonrenewable resource per capita is a significant predictor for the political instability indexes based World Bank data, but not for the civil war composite scores and 91 The 25th percentile is $1492. 156 PITF index. This to some degree implies that the scarcity of nonrenewable resource is more likely to be associated with less intense forms of political instability instead of severe crises and full-scale wars. Once the grouping variable of economic adaptability is added in model 2, nonrenewable resource per capita is no longer significantly associated with any of the political instability indexes. The further analysis in Model 3 suggests that the interaction between nonrenewable resource per capita and the group variable is not a significant predictor of political instability. The simple regressions within low and high economic scarcity groups in Table 11 also imply that no significant difference can be found between the two groups, which confirms the earlier finding that no significant interactions exist and thus it would not be a productive way to divide them into groups for further analysis. Since no strong interaction has been found, it is time to look at the possible confounding effects. For the World Bank indicator which is significantly correlated with nonrenewable resource scarcity, the association is no longer significant after controlling for the variable of economic adaptability group. 92 Adjustment for economic adaptability group does not only result in insignificance of the relationship, it also leads to a substantially lower slope estimate, which has changed more than 20% from the unadjusted model. According to the bivariate correlation tests summarized in Table 9, economic adaptability group is both significantly correlated with nonrenewable resource 92 The indicator of annual civil war number and PITF index are not significantly correlated with nonrenewable resource scarcity in the simple regression. This might suggest that nonrenewable resource scarcity is associated with lower lever of political instability rather than civil conflict and full scale wars. 157 per capita and political instability. 93 Thus it is a confounder of the association between nonrenewable resource per capita and the World Bank indicator of political instability. Although nonrenewable resource scarcity is significantly related to the annual number of civil war or the PITF index in neither model 1 nor model 2, its slope changes substantially (from -.008 to -.002, and from -.424 to -.323). Thus it is consistent with the confounding effect identified using the other two political instability measurements. Hence the relationship among the three variables can be simplified in Figure 9. Figure 9: Economic Adaptability as a Confounder between Nonrenewable Resources and Political Instability It indicates that nonrenewable resource scarcity appears to be associated with political instability, especially less intense forms of political instability, when a country‘s 93 They are all significant at .01 level. Less intense Forms of Political Instability Economic Adaptablity Nonrenewable Resources Scarcity 158 economic adaptability is not considered in the data analysis. However, as a matter of fact, the significant connections between the two are mainly caused by economic development level. Countries scarce in nonrenewable resources often have lower economic adaptability whereas worse economic conditions are associated with political instability. In sum, the statistical results in this section indicate that the associations among resource scarcity, economic adaptability and political instability are different when renewable and nonrenewable resources are considered respectively. The association between renewable resource scarcity and political instability depends on the level of economic adaptability. Countries with strong economic power can effectively adapt to renewable resource scarcity and thus no serious challenge to its political stability is placed on their political systems. However, countries with weak economic conditions cannot adapt to severe renewable resource challenges, their political stability is then shaken or worsen by severe renewable resource scarcity. The relationships do not work the same way when nonrenewable resources are investigated. For one of the political instability indicators, significant confounding effects but no strong interactions are found among nonrenewable resource scarcity, economic development and political instability. It thus suggests that the relationship between political instability and nonrenewable resource scarcity is weaker than the one with renewable resource scarcity. The possible association between the two is more likely to be attributed to the close connection between economic development level and political instability since countries with scarce nonrenewable resources often have low economic development level. 159 Three hypotheses regarding economic adaptability are put forward in Chapter 2. They are as follows: 5.1. Economic development. States with higher GDP per capita, higher per capita income and lower income inequality have higher adaptive capacity to environmental scarcity. 6.1. For a state with a higher economic development level, renewable resource scarcity does not increase its political instability. Otherwise, a state‘s political stability is likely to decrease with the severity of renewable resource scarcity. 7.1. A state with low economic development level is more likely to be less politically stable when it has abundant nonrenewable resources. Otherwise, nonrenewable resource endowments do not make it less stable. Drawing on the statistical analysis in this section, it is reasonable to conclude that Hypothesis 5.1 on the main determinants of economic adaptability is supported by the results of PCA. Furthermore, the significant interaction found in the renewable resource model proves Hypothesis 6.1 to be right. However, no evidence of Hypothesis 7.1 is found in the analysis. The confounding effects offer some weak evidence that nonrenewable resource scarcity rather than abundance is fairly associated with political instability due to the effect of poor economic development. Political Institutions Based on the previous theoretical and empirical research, political institutions are believed to have an important effect on the relationship between environmental scarcity 160 and political instability. Following the same procedures for investigating economic development, this section examines the intermediate role played by political institutions. A series of widely used indicators of democracy, political freedom, civil liberties, and governance quality are brought in as potential measurements of a state‘s quality of political institutions. As it is explained in the methodology chapter, scores compiled by the Polity IV project, composite scores of political freedom and civil rights based on the Freedom House data set, and five indicators calculated by the World Bank‘s World Governance Indicators project are found to cluster together as one component that measures the political adaptability of a country. The five indicators from the World Bank are Government Effectiveness, Regulatory Quality, Rule of Law, Control of Corruption, and Voice and Accountability. 94 Both the sampling adequacy and sphericity tests imply that PCA is appropriate for analyzing the data set. Moreover, the scree plot and Eigen values both show that the six variables consist a one-component structure, which measures the political adaptability of a country. The factor loadings and component scores for the variables are summarized in Table 12. As Table 12 shows, the factor loadings of the seven variables are all quite high, with six bigger than .9 and one bigger than .8. Thus according to the results, the correlation coefficients between the variables and the component are high. That is to say, 94 WGI have six indicators in total. In this study, the one measures political stability and absence of violence is used as an indicator of the dependent variable. 161 all the six variables are typical indicators of the component—the healthiness of political institutions. Table 12: Principal Components of Political Adaptability Component and Variables Factor Loading Score Coefficient Sampling Adequacy Test of Sphericity Compt 2: Political Institutions .887 .000 Polity IV .823 .141 Freedom House Score -.916 -.157 Government Effectiveness .945 .162 Regulatory Quality .912 .156 Rule of Law .942 .161 Control of Corruption .919 .157 Voice and Accountability .936 .160 Based on both the correlation matrix in Table 13 and the factor loadings, the five WGI indicators are highly correlated with each other (all the correlation coefficients are bigger than .8 except one is .787). Besides, their factor loadings are also bigger than .90. Moreover, Polity IV score and Freedom House score are highly correlated with each other (-.935), but their bivariate correlations with the WGI indictors are slightly lower (between .60 and .80). The factor loading of Polity IV score is also lower than the others (.823 whereas the other are bigger than .910). Except the Freedom House score, other variables all have positive factor loadings and score coefficients with the component. The main reason is that the increase of all the other variables indicates high political 162 adaptability, but higher Freedom House score implies lower freedom, which corresponds to lower political adaptability. 95 Another point that needs to be mentioned is that among the five WGI indicators, Voice and Accountability is the most highly correlated with Polity IV score and Freedom House score. 96 This statistical evidence can also be explained with their definitions. Polity IV score measures the authoritarian characteristics of countries, which shows their regime types are closer to democracy or dictatorship. The Freedom House score is a composite index of Political Rights and Civil Liberties indicators coded by the organization. As the grant and protection of political rights and civil liberties is a critical element of a democratic regime, Polity IV and Freedom House score are highly correlated (-.935, p<.01). The Voice and Accountability indicator from the WGI project measures ―the extent to which a country‘s citizens are able to participate in selecting their government, as well as freedom of expression, association, and the press‖ (Kaufmann, Kraay and Mastruzzi 2009b). This definition clearly shows that the Voice and Accountability indicator mainly measures citizens‘ rights to participate in governance, as well as the amount of political freedom and civil rights they have. Thus what Voice and Accountability indicator measures is also quite similar to the content of Polity IV and Freedom House scores. 95 As it is explained in Chapter 3, the Freedom House score ranges from one to seven, with one representing the highest degree of freedom and seven the lowest. 96 The bivariate correlation coefficients of Voice and Accountability with Polity IV score and Freedom House score are both no less than .80. 163 The other four WGI indicators, such as Government Effectiveness, evaluates the quality of a state‘s public services, including its capacity and independence from political pressures, and the quality of its policy formulation (Kaufmann, Kraay and Mastruzzi 2009b). Another indicator named Regulatory Quality intends to assess the government‘s ability to ―promote private sector development‖ by making sound policies and regulations. Besides, the indicator of Rule of Law measures ―the extent to which agents have confidence in and abide by the rules of society,‖ including the investigation on important rights, judiciary and crime control agencies (Kaufmann, Kraay and Mastruzzi 2009b). Finally, Control of Corruption evaluates ―the extent to which public power is exercised for private gain‖ by accounting the number of corruptions and elite ―capture‖ of the state (Kaufmann, Kraay and Mastruzzi 2009b). Comparing the definitions of these four indicators with Voice and Accountability, it can be seen that Voice and Accountability more directly assesses the essential features of democracy. Hence it is closer to what the Polity IV score and Freedom House score are measuring compared with the other four WGI indicators. Same as the previous analysis on economic development level, the regression method is then used to calculate the component scores which summarize the information from the six variables. The countries are again divided into groups according to their component scores of political institutions. Through this way, it is then possible to test the proposed hypotheses that expect different associations between resource scarcity and political instability for countries with high political adaptability and those with low political adaptability. 164 As it is mentioned before, two different ways of dividing groups have been tested: one is to divide the countries into three groups based on the 33rd and 66th percentile of their political adaptability component score; the other method is obtaining two groups using the 66th percentile of the political component score as the threshold value. The statistical results followed suggest that the two-group division exceeds the three-group division in terms of significance and consistency. Therefore, countries with component scores below and equal to the 66th percentile are assigned to group1 whereas those above the 66th percentile are assigned to group 2 for the final analysis. 97 In this way, countries in Group 1 are assumed to have lower political adaptability than those in Group 2 since higher component score indicates better political institutions and thus higher political adaptability. 97 The results of the regression analyses based on three adaptability groups are provided in Appendix A. 165 Table 13: Pearson’s Correlations between Political Institution Variables, Environmental Scarcity and Political Instability Variables Polity IV FH GE RQ RL CC VA Polgroup Renew (ln) Nonren (ln) Wars PITF wars WB Polin Polity IV 1.0 -.935** .648** .644** .652** .619** .800** .755** .384** .140 -.114 -.192* .511** FH 1.0 -.773** -.775** -.774** -.737** -.880** -.779** -.424** -.213** .234** .296** -.634** GE 1.0 .873** .947** .948** .830** .760** .305** .284** -.313** -.344** .784** RQ 1.0 .851** .816** .856** .703** .319** .257** -.301** -.339** .725** RL 1.0 .946** .832** .769** .301** .263** -.306** -.362** .804** CC 1.0 .787** .739** .281** .235** -.308** -.362** .764** VA 1.0 .792** .426** .203** -.269** -.339** .727** Polgroup 1.0 .292** .194* -.159* -.208** .561** Renew(ln) 1.0 .348** -.234** -.214** .322** NonR (ln) 1.0 -.091 -.090 .186* Wars 1.0 .832** -.551** PITF wars 1.0 -.558** WB Polin 1.0 *p<.05 ** p < .01 166 Table 14: Testing for Interactions and Confounding Effects of Political Institutions Civil Wars PITF Political Instability WB Political Instability Model1 Model2 Model3 Model1 Model2 Model3 Model1 Model2 Model3 Intercept .725*** .741*** 2.045** 31.368*** 32.504*** 91.633** -2.714*** -2.968*** -7.153*** Renewable Resources per capita (ln) -.072** -.063* -.233* -3.039** -2.381* -10.104* .318*** .171* .718** Adaptability Group (66% ) -.063 -.839* -3.860* -39.842* 1.037*** 3.528** Interaction (Renewable*Adapt Group) .100 4.552 -.322* Explain R 2 .055 .063 .084 .046 .069 .089 .104 .342 .363 P value for F test .003 .006 .003 .006 .003 .002 .000 .000 .000 Intercept .202*** .316*** .286* 9.399*** 15.515*** 16.151** -.480*** -1.836*** -2.009*** Subsoil Assets per capita (ln) -.008 -.006 .001 -.363 -.408 -.054 .052* .023 .059 Adaptability Group (66% ) -.093 -.068 -4.769* -5.296 1.108*** 1.251*** Interaction (Subsoil*Adapt Group) -.005 .106 -.029 Explain R 2 .008 .029 .030 .008 .051 .052 .035 .321 .323 P value for F test .251 .098 .193 .257 .016 .040 .018 .000 .000 *p<.05 ** p < .01***p<.001 167 Table 15: The Different Patterns among Low and High Political Adaptability Groups Civil Wars PITF Political Instability WB Political Instability Adaptability Group (66%) Group1 Group2 Group1 Group2 Group1 Group2 Intercept 1.206*** .367 53.158*** 11.949 -3.625*** -.097 Renewable Resources per capita (ln) -.133** -.033 -5.831** -1.000 .395** .073 Explain R 2 .076 .024 .074 .019 .091 .018 Intercept .218*** .150 10.855*** 5.558* -.758*** .494* Subsoil Assets per capita (ln) -.004 -.009 -.436 -.330 .030 .001 Explain R 2 .002 .010 .013 .010 .017 .000 *p<.05 ** p < .01***p<.001 168 A series of regressions using the three indicators of political instability—wars, PITF political instability, and the World Bank political instability estimate—are conducted and their results are summarized in Table 14 and Table 15. Generally speaking, the results here are quite similar to what we have obtained for economic adaptability. First, significant interaction effects have been found for renewable resources. According to the results in Table 14, renewable resource per capita is a highly significant (p<.01 or p<.001) predictor of political instability in the bivariate regression analyses. When the categorical variable of adaptability group is added in the regression of Model 2, renewable resource per capita (ln) becomes less significant (p<.05). Then Model 3 brings in the interaction between renewable resource per capita and the political adaptability group variable. It turns out that the interaction is a significant predictor for the World Bank indicator of political instability (p<.05). For Model 3 of the annual number of civil wars indicator and the PITF index, the interaction is not strictly significant, but is significant at an edge level (p<=.062 and p<=.066). The comparison of R 2 of the three models demonstrates that a more complete model tend to better explain the variance of the dependent variable. Model 2 has bigger R 2 than Model 1, and Model 3 has the biggest R 2 among the three. 98 The evidence thus implies that the interaction between renewable resource scarcity and economic adaptability group is quite important in the prediction of political instability, and it would be misleading if only their separate effects are considered. 98 This pattern is consistent for the regression analyses using three different indicators of political instability. 169 The significant interaction effect confirms that it is a correct way to stratify the data by the score of political adaptability, and then analyze the relationships between renewable resource scarcity and political instability within each group. The results for group analysis are summarized in Table 15. For countries in the low political adaptability group (Group 1), renewable resource scarcity is significantly associated with political instability. That is to say, the risk of political instability increases with the severity of renewable resource scarcity. But such a pattern is not found for the high adaptability group (Group 2) since there is no significant association between the two variables. Furthermore, although none of the R 2 for the simple regressions within the two groups is big (<.1), the R 2 for the low adaptability group is much bigger than that of the high adaptability group. This indicates a more linear relationship between renewable resource scarcity and political instability exists in countries with low political adaptability. The pattern is further confirmed by the scatter plots in Figure 8. Based on the statistical results, it can be concluded that political adaptability is an important intermediate variable between renewable resource scarcity and political instability. Countries with low to fair level of political adaptability are not able to effectively respond to the challenges imposed by renewable resource scarcity and thus their political stabilities are in danger. On the contrary, higher political adaptability can help a country to effective handle the quest for renewable resource without putting the state in serious political crisis. The following regression tests for nonrenewable resources find quite different results from those of renewable resources. Table 14 shows that nonrenewable resource 170 per capita is significantly associated with the political instability indexes compiled by the World Bank in the simple regressions of Model 1, but it is insignificant for the Civil War indicator and the PITF index. When political adaptability group is added in Model 2, nonrenewable resource per capita is no longer a significant predictor of any of the political instability indicators. Furthermore, analysis in Model 3 demonstrates that the interaction between nonrenewable resource scarcity and political adaptability group is not significantly related to political instability. Thus it does not appear to be a productive way to stratify the data based on nonrenewable resource scarcity group for further analysis. The simply regressions within each group in Table 15 again confirms that there is no significant difference between the low and high political adaptability groups in terms of their linear patterns and R 2 . The next step is to investigate the possible confounding effects. In Model 1, nonrenewable resource scarcity is significantly related to the World Bank indexes of political instability, but not any more when political adaptability group is added in Model 2. 99 In the meantime, the slope estimate of nonrenewable resource per capita in the adjusted model (Model 2) has changed more than 20% from the unadjusted model (Model 1). 100 Since political adaptability group is significantly correlated with nonrenewable resource scarcity and three political instability indicators according to 99 The annual civil war number indicator and PITF indicator are not significantly correlated with nonrenewable resource scarcity in the simple regressions. This might suggest that nonrenewable resource scarcity is associated with lower levers of political instability rather than full scale wars and severe political crises. 100 From .052 to .053. 171 Table 13 (p<.05 or p<.01), it is a confounder of the association between nonrenewable resource per capita and the World Bank‘s political instability indicator. Although nonrenewable resource per capita is not significantly related to the civil wars indicator or PITF index, its slope estimates change 20% in the adjusted Model 2 for civil war indicator and 12% for PITF index. Thus it in general is consistent with the confounding effect discussed above. Figure 10 simplifies the relationship among the three variables, with political adaptability as a confounder. Figure 10: Political Adaptability as a Confounder between Nonrenewable Resources and Political Instability In sum, nonrenewable resource scarcity appears to be correlated with less intense forms of political instability but not severe political instability in bivariate analysis. More in-depth investigation finds that the significant effect is due to confounders such as political institutions. The statistical relationship implies that countries facing Less Intense Forms of Political Instability Political Adaptablity Nonrenewable Resources Scarcity 172 nonrenewable resource scarcity also tend to have lower political adaptability, and worse political adaptability increase the risk of less intense forms of political instability. Therefore, the connections between nonrenewable resource scarcity and political instability tend to be weak. Based on the above statistical results, the role played by political adaptability in the resource scarcity/political instability nexus is quite similar to the findings of economic adaptability in the previous section. First of all, different patterns among resource scarcity, economic adaptability and political instability are identified when renewable and nonrenewable resources are considered respectively. The impact of renewable resource scarcity on political instability is influenced by a country‘s political adaptability level. Renewable resource scarcity does not increase a country‘s risk of political instability if its political institutions are strong and healthy. Otherwise, the country is not able to adapt to severe renewable resource challenges relying on its poor political institutions and thus will face serious political crisis. The linkages are much weaker when it comes to nonrenewable resource scarcity. Only analysis based on one of the three political instability indicators has found significant confounding effect of political adaptability in the nonrenewable resource scarcity/political instability nexus but the interaction is not significant for any of the indicators. Thus it implies that even the existing association between the nonrenewable resource scarcity and less intense forms of political instability is most likely due to the confounding effects of other factors, such as political institutions. As we can see, 173 countries with high nonrenewable resource scarcity often have weaker political institutions, which are highly correlated with political instability. The relevant hypotheses on political adaptability presented in Chapter 2 are listed as follows. Each of them is evaluated according to the statistical conclusions reached in this section. 5.2. Political institutions. Countries with more democratic political institutions and civil liberties are political more accountable to the public, and thus are likely to be better at adapting to environmental pressures. 6.2. A state with high renewable resource scarcity is likely to be less stable if its political adaptability is low. Otherwise, high renewable resource scarcity does not increase its political stability. 7.2. For a state with low political adaptability, nonrenewable resource abundance is more likely to endanger its political stability. Otherwise, resource abundance does not increase its political instability. The PCA in this section has demonstrated a positive association between those political institution indicators and a country‘s political adaptability. Variables such as regime type, civil liberties, and governance qualities are all pointing to the measurement of political adaptability. Therefore, Hypothesis 5.2 on the main determinant of political adaptability is supported. Besides, the positive association between renewable resource scarcity and political instability in countries with low political adaptability expected in Hypothesis 6.2 is supported by the significant interaction found in the regression analysis. But the 174 association among nonrenewable resource abundance, political institutions, and political instability proposed in Hypothesis 7.2 is not found. Nonrenewable resource scarcity rather than abundance is found to be weakly associated with political instability due to poor political institutions. Social Fractionalization Besides economic development and political institutions, variables measuring social fractionalization are also widely tested by scholars as potential predictors of political instability, especially civil conflict. The review in the earlier chapter shows that scholars are mainly interested in the effects of ethnic and religious fractionalization on the risk of civil conflict. As it is reported, most empirical studies in the field do not find statistical support for a positive association between ethnic and religious diversities and incidence of civil war. Other researchers have found a possible pacifying effect of diversity on violent conflict, which is suggested to be explained by ethnic fragmentation rather than dominance. 101 According to them, ethnic dominance rather than ethnic fragmentation is contributing to civil violence. Based on the existing literature, four widely-used indexes of social fractionalization from Fearon and Laitin‘s (2003c) data set are included in the analyses. The two ethnic fraction estimates are based either on Soviet Atlas 102 or Fearon‘s (2002) APSA paper. The religious fractionalization indicator is estimated based on the online 101 More details about the findings on social fractionalization can be found in Chapter 2. 102 The CIA website provides an introduction of the Soviet Atlas at < https://www.cia.gov/library/center-for-the-study- of-intelligence/kent-csi/vol10no2/html/v10i2a04p_0001.htm>. 175 CIA Fact book (Central Intelligence Agency). 103 Besides ethnic and religious diversity, the number of languages widely spoken in a country is also believed to be an important indicator of social fractionalization (Fearon and Laitin 2003c). The first three variables range from 0 to 1, with 0 indicates lowest level of fractionalization, and 1 indicates the highest, and the number of languages is a discrete variable. All of the four variables are included in the PCA on social fractionalization. It turns out that they form a nice one-component structure based on the Eigen values and scree plot. The factorability tests and statistical results of PCA on social fractionalization are listed in Table 16. Table 16: Principal Components of Social Adaptability Component and Variables Factor Loading Score Coefficient Sampling Adequacy Test of Sphericity Compt 3: Social Fractionalization .745 .000 Soviet Atlas Ethnic Fractionalization .901 .354 Fearon Ethnic Fractionalization .879 .345 Number of Languages .825 .324 Religious Fractionalization .531 .209 The sampling adequacy test is bigger than .7 and the sphericity test is significant, both of which imply that PCA is appropriate for analyzing the variable and data structure. Furthermore, as showed in Table 16, the factor loadings for the two ethnic fractionalization indexes and number of languages are quite high (≥.8). The factor loading for religious fractionalization, however, is much lower (.531). As it is mention 103 The online book is available at < https://www.cia.gov/library/publications/the-world-factbook/>. 176 before, factor loading level such as .4 is often used for the central component (Raubenheimer 2004). Therefore religious fractionalization is still considered to be an important measurement of social fractionalization that should be kept in the model. The difference in factor loadings suggest that ethnic fractionalization and number of languages are more closely cluster together, whereas religious fractionalization does not measure quite the same thing as they do. This is also confirmed by the bivariate correlation analyses in Table 17, which shows that the bivariate correlations between religious fractionalization and the other three variables are lower than .5. Thus it is suggested that ethnicity is often closely linked with specific languages. More ethnically diverse countries tend to have many more languages spoken. It is a plausible argument since language is an important symbol for ethnic identity. On the one hand, ethnic groups often develop their own languages for communication through the long process of evolution. On the other hand, language is also an important way for members of ethnic groups to differentiate themselves from outsiders and build their own identity. The solidarity of the ethnic group is also reinforced by the creation and exercise of their own languages. However, religious diversity does not exactly overlap with ethnic and language diversity. This also appears to be empirically plausible because the spread of religions and languages does not follow the same way. It seems that some small and local religions tend to attach with specific ethnic groups within a specific region. But the spread of the three largest religions has transcended ethnic and national borders. 104 104 The three largest religions with the biggest population of followers are Christianity, Islam and Buddhism. 177 Moreover, large countries such as the United States and China might only have one official or several ethnological languages, but they can have many more religious sectors. Moreover, the extension and acceleration of globalization tend to result in less diversity in languages since the integration of languages can help to lower down the transaction costs. However, its impact on religious diversity seems to be less obvious. Although the mainstream religions such as Christianity have extended influence with the spread of western culture and life style, new religions and religious sectors are still being created every day mostly within small communities. This is not the same as what happens to languages, since the costs of creating and adopting new languages are much higher than those of creating and adopting a new religion. Again, the regression method is used to calculate the component scores of social fractionalization. For all the four variables, higher values indicate higher social fractionalization, which is expected to imply lower level of social adaptability based on their definitions and our earlier hypothesis. Thus higher component scores of social fractionalization represent lower social adaptability to environmental challenges in the hypothesized model. Following the procedures conducted for economic development level and political institutions, the countries are divided into two groups based on their social fractionalization component scores. The purpose of this step is to test the major hypotheses exploring the different associations between resource scarcity and political instability for countries with low social and those with high social adaptability. 178 As it is mentioned before, dividing the countries into two groups instead of three groups can lead to more significant and obvious statistical results. Therefore, the 66 th percentile of social adaptability component score is used as the threshold value. 105 That is to say, countries with component scores below and equal to the 66 percentile are assigned to Group1, whereas those above the 66th percentile are assigned to Group 2. What needs to be noted is that Group 1 is the high adaptability group and Group 2 is the low adaptability group since higher social fractionalization is assumed to result in lower social adaptability to resource scarcity based on our earlier hypotheses. 106 In the meantime, the grouping variable of social adaptability is created and then a series of regression analyses are performed to further explore the possible interaction and confounding effects as well as group difference. 105 Originally, the countries are divided into three groups based on the 33rd and the 66th percentiles of their social adaptability component scores. The results of the regression analyses based on three adaptability groups are provided in Appendix A. 106 Although some studies have found inconsistent evidence for the linkage between social fractionalization and political instability, scholars in general test hypotheses that expect positive associations between the two variables. Furthermore, as mentioned in Chapter 2, some scholars suggest that ethnic fragmentation and domination should be considered separately since they may correlate with political instability differently. However, not much evidence is found and not many good measurements can be found. Thus it is still more common to investigate ethnic and religious fractionalization in general. 179 Table 17: Pearson’s Correlations between Social Fractionalization Variables, Environmental Scarcity and Political Instability Variables Ethfrac Ef Numlan Relfrac Socgroup Renew (ln) Nonren (ln) Wars PITF wars WB Polin Ethfrac 1.0 .771** .661** .324** .800** -.165* -.128 .251** .269** -.287** Ef 1.0 .596** .326** .713** -.179* -.100 .188* .211** -.374** Numlan 1.0 .295** .677** -.189* -.164* .283** .306** -.286** Relfrac 1.0 .410** -.037 -.024 -.009 .005 .010 Socgroup 1.0 -.160* -.173* .130 .221** -.238** Renew(ln) 1.0 .351** -.209** -.229** .317** NonR (ln) 1.0 -.075 -.149 .194* Wars 1.0 .791** -.524** PITF wars 1.0 -.530** WB Polin 1.0 *p<.05 ** p < .01 180 Table 18: Testing for Interactions and Confounding Effects of Social Fractionalization Civil Wars PITF Political Instability WB Political Instability Model1 Model2 Model3 Model1 Model2 Model3 Model1 Model2 Model3 Intercept .672*** .548** .164 31.371*** 20.733* .031 -2.643*** -1.890** -1.428 Renewable Resources per capita (ln) -.065** -.060* -.010 -3.035** -2.605** .102 .312*** .282*** .222 Adaptability Group (66% ) .063 .387 5.074* 22.521 -.386* -.775 Interaction (Renewable*Adapt Group) -.043 -2.293 .051 Explain R 2 .044 .053 .056 .045 .087 .092 .101 .136 .137 P value for F test .008 .013 .029 .007 .001 .002 .000 .000 .000 Intercept .196*** .085 .142 9.48*** 1.951 3.941 -.455*** .158 -.198 Subsoil Assets per capita (ln) -.007 -.005 -.018 -.379* -.422 -.888 .054* .044* .127* Adaptability Group (66% ) .077 .037 5.396* 4.007 -.423** -.175 Interaction (Subsoil*Adapt Group) .010 .342 -.061 Explain R 2 .006 .020 .023 .009 .062 .063 .038 .081 .091 P value for F test .347 .207 .309 .0235 .007 .016 .014 .001 .002 *p<0.05 ** p < 0.01***p<0.001 181 Table 19: The Different Patterns of Low and High Social Adaptability Groups Civil Wars PITF Political Instability WB Political Instability Adaptability Group (66%) Group1 Group2 Group1 Group2 Group1 Group2 Intercept .551** .937 22.553** 45.074 -2.203*** -2.978* Renewable Resources per capita (ln) -.052* -.095 -2.191* -4.485 .273*** .324 Explain R 2 .040 .197 .055 .037 .097 .062 Intercept .179*** .216** 7.948*** 11.955*** -.373* -.548** Subsoil Assets per capita (ln) -.008 .002 -.546 -.204 .066* .005 Explain R 2 .010 .000 .033 .002 .055 .000 *p<.05 ** p < .01***p<.001 182 The results of these regressions using the three indicators of political instability— wars, PITF political instability, and World Bank political instability estimate—are showed in Table 18 and Table 19. It seems that the role played by social fractionalization in the resource scarcity/political instability nexus is not quite similar to those played by economic and political adaptability, especially for renewable resources. First of all, the interaction between renewable resource scarcity and social adaptability group is not found to be a significant predictor for political instability. This is quite different from the models for economic and political adaptabilities, which proved to have significant interaction effects. 107 As it can be seen in Table 18, renewable resource per capita is a highly significant (p<.01 or p<.001) predictor of political instability in the unadjusted Model 1. There is no clear pattern of change when the group variable of social adaptability group is added in Model 2. Their significance level varies when different indicators of political instability are used. 108 When the interaction between renewable resource per capita and social adaptability group is included in Model 3, none of the predictors is significant any more. Although the R 2 of the three models are increasing gradually, there is no persuasive evidence that a more complete model can better predict the dependent variable. It thus suggests that there is no significant difference between the low and high social adaptability groups regarding their linear patterns. 107 As it is showed in the previous two sections, the interactions are significant for renewable resource models. 108 For example, when civil war index is used, the significance of renewable resource per capita (ln) changes from p<.01 to p<.05, but for the other two indicators, the significances level of renewable resource per capita (ln) in Model 2 are the same as those in Model 1. 183 As Table 19 shows, further regression analyses within each group are also conducted to confirm the insignificant results reported in Table 18. Although the interaction is found to be insignificant in Model 3, the difference between the two groups seems to be obvious according to Table 19. For all the three indicators of political instability, renewable resource per capita (ln) is a significant predictor of political instability when Group 1 is considered. But none of the simple regressions of Group 2 is significant. Since Group 1 is the high adaptability group and Group 2 is the low adaptability group, the results suggest that there is significant linear pattern between renewable resource scarcity and political instability for countries with high social adaptability (low social fractionalization), but not for countries with low social adaptability (high social fractionalization). However, further investigation on the R 2 of the two groups does not demonstrate a consistent changing direction when different indicators of political instability are used as the dependent variable. 109 In general, there is some evidence suggesting that renewable resource scarcity is significantly associated with political instability in countries with low social fractionalization, and no significant association between the two variables can be found for countries with high social fractionalization. However, the evidence of group difference appears to be weak since the interaction is not tested to be a significant predictor in Model 3 of Table 18 and the R 2 of the simple regressions of the two groups do not change in a consistent way. The inconsistent pattern is further implied by the 109 When civil war index is used as the dependent variable, the R 2 of Group 1 is smaller than the R 2 of Group 2, but for PITF and World Bank indicators of political instability, the R 2 of Group 1 is bigger than the R 2 of Group 2. 184 scatter plots in Figure 8, which do not show a clear linear pattern in one group and non- linearity in the other group. Therefore, only weak evidence of group difference is found based on the analyses so far. The study then goes further to investigate the possible confounding effect of social fractionalization in the models. In Model 1 of Table 18, renewable resource per capita (ln) is significantly related to all the three indicators of political instability, and its significance level has changed in Model 2 which is adjusted for the variable of social adaptability group. 110 If the civil war index is used as the dependent variable, the slope estimate of nonrenewable resource per capita in Model 2 has changed only 8% from its slope estimate in Model 1, which cannot be viewed as significant difference based on the widely-used criterion. 111 Furthermore, the bivariate correlation analyses show that the group variable is significantly associated with renewable resource per capita (ln) but not with the civil war index. Thus social adaptability group is not a significant confounder in the relationship between renewable resource per capita (ln) and civil war number index. The PITF and World Bank indicators of political instability are significantly correlated with social adaptability group. Comparing with the results for the civil war number index, it implies that social adaptability group might be a better predicator of less intense forms of political instability instead of full scale civil wars. When these two indicators are used as the dependent variable, the slope estimate of nonrenewable 110 As we addressed earlier, the significance level has changed in different directions when different indicators of political instability is used in the model. When civil war index is used, the significance of renewable resource per capita (ln) change from p<.01 to p<.05, but for the other two indicators, the significances level of renewable resource per capita (ln) in Model 2 are the same as those in Model 1. 111 Scholars think that at least 12% change of coefficient is enough for the proof of confounding effect. 185 resource per capita (ln) in the adjusted model (Model 2) has changed 11% and 19% respectively from the unadjusted model (Model 1). This does suggest that social adaptability group has some confounding effect in the association between renewable resource scarcity and political instability, but the evidence tends to be weak. 112 In general, no obvious interaction or confounding effects have been found for the relationship among renewable resource scarcity, political instability and social adaptability group. Renewable resource scarcity appears to be correlated with political instability in bivariate analyses, and the significant association still holds when social adaptability is added in. It then can be concluded that renewable resource scarcity is a better predictor of political instability compared with social fractionalization and social fractionalization does not play a significant intermediate role in the relationship between renewable resource scarcity and political instability. What these statistical results suggest for real cases is that social fractionalization does not effectively determine a country‘s adaptability to renewable resource challenge. When facing renewable resource scarcity, countries with less social fractionalization are not necessarily more capable of handling this challenge and sustaining their domestic political stability. Then the possible interaction and confounding effects for the nonrenewable resource models are checked. Model 1 in Table 18 shows that nonrenewable resource per capita is significantly related to PITF and World Bank indicators of political instability 112 Stronger evidence of the existence of confounding effect often requires more than 20% change of the slope estimates. 186 but not the annual civil war number index. Model 2 is adjusted for the grouping variable of social adaptability, and nonrenewable resource scarcity is not a significant predicator for the PITF index of political instability any more. 113 Finally, the interaction between renewable resource scarcity and social adaptability group variable is tested to be insignificant at all for Model 3 no matter which political instability indicator is used. Hence, no interaction effect can be concluded in the model. Table 19 shows the bivariate association between nonrenewable resource scarcity and political instability in Group 1 and Group 2 separately. None of the simple regressions is significant except that nonrenewable resource per capita (ln) is a significant predictor of the World Bank indicator of political instability when Group 1 is investigated (p≤.5). Thus there is no clear evidence that the association between nonrenewable resource scarcity and political instability is significantly different in the two groups. This again confirms the insignificance of interaction variable in Model 3. Besides, the scatter plots in Figure 8 provide visual support that no clear patterns can be figured out in any of the two groups. The analysis of confounding effect of social adaptability group requires further investigation of the bivariate correlations in Table 17. It tells us that social adaptability group is significantly correlated with nonrenewable resource per capita (ln), and PITF and World Bank indicators of political instability, but not the annual civil war number index. Then Model 2 in Table 18 using the PITF and World Bank indicators are checked. 113 Nonrenewable resource per capita (ln) is still insignificant for the annual civil war number index in Model 2, but it is significant for World Bank indicator of political instability. 187 It shows that the slope estimates of nonrenewable resource per capita (ln) have changed 11% and 2% respectively from its values in Model 1. Although annual civil war number index does not significantly correlate with social adaptability group or nonrenewable resource per capita (ln), the change of slope estimate in Model 2 from Model 1 is bigger than 20%, which is consistent with the trend when using the other two indicators. 114 Similar to those reached for political and economic adaptabilities, the best conclusion can be made is that the possible association between nonrenewable resource scarcity and political instability is not direct. It can be partially explained by their significant linkage with social fractionalization. That is to say, nonrenewable resource scarcity and high social fractionalization tend to coexist in the same countries, and the two factors together can endanger the countries‘ political stability. 115 However, the evidence for this association as well as confounding effect is quite weak. In sum, the role played by social fractionalization in the resource scarcity/political instability nexus is not exactly the same as those of economic development level and health of political institutions. Again it has been found that renewable and nonrenewable resources tend to have different association with social fractionalization and political instability. First of all, unlike economic development level and political institution, social fractionalization does not play a significant intermediate role between renewable resource scarcity and political instability. That is to say, a country‘s degree of social 114 The change of slope estimate is 29%. 115 The signs of the correlations of social adaptability group in Model 2 also provide important information for the positive relationship between social fractionalization and political instability. 188 fractionalization does not effectively influence its capability to cope with renewable resource challenge, and therefore low level of social fractionalization is not a decisive factor that can reduce the country‘s risk of political instability. The model identified for nonrenewable resource, is similar to but weaker than those of economic and political adaptability. The statistical analyses tell us that social fractionalization is a possible confounder in the relationship between nonrenewable resource scarcity and political instability. This is better supported for the PITF index, which implies some moderate association between nonrenewable resource scarcity and political instability. When the association is unpacked, it then can be seen that it is partially due to the confounding effect of other factors, in particular social fractionalization in this model. Finally, it comes back to the proposed hypotheses on social adaptability to resource challenge. 5.3. Social fractionalization. More fragmented societies have more ethnic and religious fractions, and more languages. Social fractionalization tends to decrease a country‘s environmental adaptive capacity. 6.3. Higher renewable resource scarcity is likely to be associated with higher political instability when a state has low social adaptability. Otherwise, a state‘s political stability is unlikely to decrease with the severity of renewable resource scarcity. 189 7.3. Nonrenewable resource abundance is more likely to decrease a state‘s political stability if it has high social fractionalization. Otherwise, the state is not likely to be less stable. Hypothesis 5.3 concerns the variables that measure social fractionalization level and their implications on a country‘s social adaptability level to resource challenges. In general, the PCA in this section has confirmed ethnicity, religion, and language as critical measurements of social fractionalization and their close connections with one another. Moreover, the factor loadings provide strong evidence that they measure social fractionalization in the same direction, which are negatively associated with a country‘s social adaptability. However, neither the analysis of interaction nor confounding effects have found evidence for the positive association between renewable resource scarcity and political instability in countries with low social adaptability. Thus Hypothesis 6.3 is not supported and no significant effect of social adaptability has been reported. Some fragile evidence of confounding effect of social adaptability is found for the nonrenewable resource model. However, it is nonrenewable resource scarcity rather than nonrenewable resource abundance that is weakly associated with political instability due to high social fractionalization. Therefore, Hypothesis 7.3 is not supported. Demographic Characteristics Demographic features of a country are believed to be closely connected with its environmental conditions. Some studies examine the impact of demographic and resource 190 stress together on a state‘s political stability. 116 The earlier literature review in this study has demonstrated the attention given to demographic conditions in the discussion of resource-induced civil conflict. But it has also been concluded that inconsistent empirical results are reported for most of the demographic variables. Current security studies mainly include demographic variables such as population, population density, population growth, urban and rural population, youth population, male population, and education level in their statistical analyses of civil conflict. The first three variables measure a country‘s demographic stress and transitional trend in general and the latter four variables describe the country‘s population composition in terms of urbanization, age, gender and education. Originally, this study brings in all the above variables for the PCA to find their structure. These data are extracted from the World Bank‘s World Development Indicators, which is believed to be one of the most comprehensive and widely cited cross- national data set. Eventually, the total population of each country is excluded in the PCA considering the fact that the independent variables—renewable resource per capita and nonrenewable resources per capita—have already incorporated the information of total population in a country. The number of population is then one of the main determinants of a country‘s resource scarcity level. Therefore, it can result in double count of information if total population is investigated again in the statistical model. The PCA finds a two-component structure when all the other six variables are included in the model. The factor loadings and component plot show that except 116 For example, Homer-Dixon and Blitt (1998), Homer-Dixon (1999), Kahl (2006). 191 population density, population growth and female percentage of population, the other three variables—percentage of urban population, percentage of population ages from 15 to 64, and literacy rate—cluster together, with factor loadings higher than .75. Therefore, a series of tests have been performed to see if a nice one-component structure can be found when one or two variables chosen from population density, population growth and female percentage are removed from the model. It turns out that female population percentage measures quite different information from the three clustered variables. Thus it must be removed from the PCA model; otherwise, the group of variables always form a two-component structure. Moreover, either population growth or density can form a one-component structure with the three clustered variables, if they are not kept in the model at the same time. The comparison of tests of factorability and factor loadings suggest that it is better to retain population growth in the model instead of population density. The factorability tests are significant when either of them is kept in the model, but the factor loading of population growth is much higher than that of population density. 117 Therefore, the final four variables that are kept in the PCA model to summarize a country‘s major demographic characteristics are population growth, percentage of urban population, percentage of population ages from 15 to 64, and literacy rate. The statistical criteria and results of the PCA are listed in Table 20. First of all, the two factorability tests show that the sampling adequacy is bigger than .7, and the sphericity test is also 117 The factor loadings are -.735 compared with .351. 192 significant. Thus it is appropriate to use PCA to analyze the data and explore the variable structure. Table 20: Principal Components of Demographic Adaptability Component and Variables Factor Loading Score Sampling Adequacy Test of Sphericity Compt 4: Demographic Characteristics .724 .000 Population Growth -.735 -.260 Urban Population (% of total) .778 .275 Population ages 15-64 (% of total) .929 .328 Ln literacy .908 .320 Moreover, the factor loadings of the four variables showed in Table 20 are all higher than .70, among which the loadings of percentage of populations ages 16-64 and literacy rate are higher than .90. It should be mentioned that different from the other three variables, the factor loading of population growth is a negative number. It means that population growth measures demographic adaptability in a negative way. That is to say, higher population growth reduces a country‘s demographic adaptability to possible environmental challenges. This is understandable based on the Neo-Malthusian belief that rapid population growth tends to increase a country‘s demand for resources, and thus imposes higher stress on its environmental and political system and less capable at handling resource scarcity. The other three variables are positively correlated with demographic ability. For example, the percentage of urban population measures a country‘s extent of urbanization. As an important indicator of modernization, it also suggests a country‘s concentration of 193 labor force for industrial production and technology development, and thus might be the personnel foundation of human ingenuity and adaptability. The percentage of population ages from 15 to 64 is a critical indicator of a country‘s potential labor force, since the age range covers the life span that is the major component of work force. The bigger the population pool is, the more labor force the country has. If the percentage is low, it means that less people in the country are involved or capable of involvement in production, and each of the working population has to support more people who are mainly consuming instead of producing, such as children and elderly people. An aging society, for example, is also believed to be less capable in terms of technological innovation. Besides, education level is also a major measurement of population quality, which is the foundation of a country‘s industrial and technological development. A number of literatures on modernization have addressed this. 118 As far as environmental scarcity and political instability are concerned, literacy rate indicate the educational preparation for labor force and possibly danger for insurgency. Firstly, illiterate population is not good source of labor for industrial production, especially in more advanced high technology sector. The adaptability to increasing environment and resource scarcity requires innovative technologies and institutions that demands great human ingenuity. For example, the development of fuel efficient cars and the exploration of alternative energy all require advanced knowledge and training. The working force in these sectors and the 118 For example, Inglehart (1997) uses education level as an important indicator of modernization and postmodernization culture. 194 industrial sectors utilizing these new technologies also increase their expectations of professional workers nowadays. Secondly, a country with large illiterate population is believed to in greater danger of insurgency. The possible reasons are that illiterate population are more likely to be in poverty and marginalized in the industrialization and globalization process. Unemployment and poverty are more likely to result in grievances and thus these people are more likely to join insurgence and resort to violent means to change their disadvantaged social economic statutes. 119 Once the four variables have been finalized for PCA analysis, the regression method is then used to calculate the component scores of demographic adaptability. Higher demographic score is associated with lower population growth, higher urbanization, higher rate of population as potential work force, and higher literacy rate. Thus higher demographic component score indicates higher demographic adaptability to environmental and resource challenges based on our assumptions. To continue the further regression analyses, all the countries are divided into two groups based on the 66 th percentile of the demographic adaptability component score. Originally, three groups are divided according to the 33 rd and the 66 th percentile of demographic adaptability component score. However, comparison of the statistical results suggests that dividing the countries into two groups instead of three groups shows more significant and obvious difference between groups. The two group strategy is thus adopted and the 66 th percentile score is used as the threshold value. 120 Therefore, 119 The grievance argument is elaborated in Chapter 2 by neoclassical economists. 120 The results of the regression analyses based on three adaptability groups are provided in Appendix A 195 countries with component scores below and equal to the 66 th percentile are assigned to Group 1, which are of low to fair level demographic adaptability. Countries with component score above the 66 th percentile are assigned to Group 2, and they are assumed to have high demographic adaptability to resource scarcity. 121 Meanwhile, a grouping variable of demographic adaptability is created, with ―1‖ for countries with low demographic adaptability and ―2‖ for countries with high demographic adaptability. With the group variable, it is then possible to conduct a series of regression analyses to further explore whether the association between resource scarcity and political instability differs between the low and high adaptability groups. The assessment of group difference is achieved by examining the possible interaction and confounding effects respectively for renewable and nonrenewable resources. Table 22 and Table 23 list the results of regressions for demographic adaptability using the three indicators of political instability—Civil War Index, PITF political instability, and the World Bank political instability estimate. 121 Some studies have found inconsistent evidence for the linkage between those demographic variables and political instability. However, most scholars in general test hypotheses that expect associations between them similar to our hypothesis. 196 Table 21: Pearson’s Correlations between Demographic Characteristics Variables, Environmental Scarcity and Political Instability Variables Popgr Urbpop Pop1564 Liter (ln) Demogroup Renew (ln) Nonren (ln) Wars PITF WB Polin Popgr 1.0 -.279** -.648** -.495** -.682** -.478** -.031 .109 .138 -.360** Urbpop 1.0 .671** .613** .655** .205** .358** -.258** -.323** .430** Pop1564 1.0 .702** .841** .269** .257** -.238** -.318** .582** Liter (ln) 1.0 .596** .269** .257** -.256** -.359** .463** Demogroup 1.0 .327** .189* -.255** -.307** .512** Renew(ln) 1.0 .274** -.175* -.162* .270** NonR (ln) 1.0 -.089 -.149 .192* Wars 1.0 .791** -.540** PITF wars 1.0 -.535** WB Polin 1.0 *p<.05 ** p < .01 Note: The correlation coefficients for some of the pairs of variables are different from the previous table due to EM imputation for missing values conducted for each table. But the significance levels in general are the same. 197 Table 22: Testing for Interactions and Confounding Effects of Demographic Characteristics Civil Wars PITF Political Instability WB Political Instability Model1 Model2 Model3 Model1 Model2 Model3 Model1 Model2 Model3 Intercept .576** .596** 1.187 23.849** 24.315** 44.195 -2.229*** -2.363*** -3.019 Renewable Resources per capita (ln) -.053* -.031 -.108 -2.075 -.872 -3.475 .259*** .111 .196 Adaptability Group (66% ) -.141** -.501 -7.641*** -19.753 .958*** 1.357 Interaction (Renewable*Adapt Group) .047 1.567 -.052 Explain R 2 .031 .075 .080 .022 .099 .102 .073 .274 .275 P value for F test .026 .002 .005 .059 .000 .001 .001 .000 .000 Intercept .202*** .394*** .402*** 9.483*** 19.281*** 20.557*** -.461*** -1.681*** -1.646*** Subsoil Assets per capita (ln) -.008 -.004 -.005 -.380 -.350 -.622 .053* .027 .020 Adaptability Group (66% ) -.157** -.163 -7.763*** -8.785* .996*** .968*** Interaction (Subsoil*Adapt Group) .001 .208 .006 Explain R 2 .008 .067 .067 .009 .103 .104 .037 .272 .272 P value for F test .261 .004 .012 .236 .000 .001 .015 .000 .000 *p<0.05 ** p < 0.01***p<0.001 198 Table 23: The Different Patterns among Low and High Demographic Adaptability Groups Civil Wars PITF Political Instability WB Political Instability Adaptability Group (66%) Group1 Group2 Group1 Group2 Group1 Group2 Intercept .686 .185 24.442 4.690 -1.662 -.304 Renewable Resources per capita (ln) -.062 -.015 -1.908 -.342 .145 .093 Explain R 2 .015 .016 .008 .008 .012 .031 Intercept .239*** .075 11.773*** 2.988* -.678*** .290 Subsoil Assets per capita (ln) -.004 -.003 -.414 -.206 .026 .031 Explain R 2 .002 .003 .010 .017 .010 .021 *p<0.05 ** p < 0.01***p<0.001 199 The statistical results from the two tables imply that the role of demographic adaptability in the resource scarcity/political instability nexus is more similar to social rather than economic and political adaptability. As far as renewable resource is concerned, the interaction between renewable resource scarcity and demographic adaptability group is not significantly related to any of the three indicators of political instability. Based on the unadjusted Model 1 of Table 22, the bivariate association between renewable resource per capita and political instability is significant for civil war index and the World Bank indicator (p<.05 and p<.001). However, renewable resource per capita (ln) is no longer significant when the grouping variable of demographic adaptability is added in. Model 3 goes further to add the interaction between the two variables in the tests. Although the R 2 of Model 3 is bigger than those of Model 1 and Model 2, none of the predictors is significant in the regressions of Model 3. Thus the insignificance of interaction suggests that the patterns between renewable resource scarcity and political instability for low and high demographic groups are not significantly different from each other. The simple regressions within each group further confirm the findings. Table 23 summarizes the results using the three different indicators of political instability. It shows that renewable resource per capita (ln) is not a significant predictor of political instability in neither the low demographic nor the high demographic adaptability group. Thus 200 investigating the bivariate association within each group cannot help us to find further information. 122 As usual, the study moves on to test the possible confounding effect of demographic adaptability when interaction is found to be insignificant in the regression model. It has been noticed that renewable resource per capita (ln) is a significant predictor in the simple regressions of Model 1 for two of the political instability indicators, but not in Model 2 adjusted for the grouping variable of demographic adaptability. Since demographic adaptability group is significantly correlated with both renewable resource per capita (ln) (.327, p<.01) and the three indicators of political instability (p<.01), 123 the change of significance level indicates that demographic adaptability group might be a confounder in the bivarate association. By checking the slope estimates of renewable resource per capita (ln) in Model 1 and Model 2, we find that the slope estimate change is bigger than 20%. 124 Thus the confounding effect of demographic adaptability group has been confirmed for two of the political instability indicators. The slope estimate when the PITF indicator is used also changes 42% from Model 1 to Model 2, which is consistent with the patterns when the other two indicators are used. Therefore, the significant association between renewable resource scarcity and political instability is not direct. Both of the two variables are significantly related to demographic adaptability group, which is one of the 122 This is different from social adaptability, where the bivariate associations within the two groups are different from each other when renewable resource is concerned. 123 The bivariate correlation tests are listed in Table 21. 124 The slope estimate change for the civil war index is 42% and 57% for the World Bank indicator. 201 major contributors of their association. That is to say, countries with renewable resource scarcity often have low demographic adaptability and low demographic adaptability is associated with high political instability. Figure 11 again simplifies the statistical relationships among the three variables in a chart. Figure 11: Demographic Adaptability as a Confounder between Renewable Resources and Political Instability Then the study continues to investigate the possible interaction and confounding effects regarding nonrenewable resource scarcity. Table 22 shows that the bivariate relationship between nonrenewable resource per capita (ln) and political instability is significant for the World Bank indicator of political instability (p <.05), but not for the civil war and PITF indexes. The implication is that nonrenewable resource scarcity might be a significant predicator of less intense forms of political instability rather than more intense instability such as full scale wars. When grouping variable of demographic Political Instability Demographic Adaptablity Renewable Resources Scarcity 202 adaptability is added in Model 2, nonrenewable resource per capita (ln) is no longer significantly associated with the World Bank indicator of political instability and the grouping variable is highly significant (p<.001 or p<.001). Model 3 adds in the interaction between nonrenewable resource per capita (ln) and demographic adaptability group, and it is found to be insignificant for all of the three political instability indicators. Although the comparison of R 2 suggest that a more complete model might better explain the variance of the dependent variable, 125 the interaction term is not found to be significant based on the three models. To confirm the conclusion reached for interaction, Table 23 shows that simple regressions of political instability on nonrenewable resource scarcity in Group 1 and Group 2 separately. It turns out none of the bivariate regressions is significant and it again confirms that no significant interaction exists in the model. Figure 8 shows the scatter plots for the two groups based on demographic adaptability, and no clear linear patterns can be visualized in neither of the two groups. To check the possible confounding effects, the study then looks at the bivariate correlation tests in Table 21. It is suggested that demographic adaptability group is significantly correlated with nonrenewable resource per capita (ln) and the three indicators of political instability. As mentioned before, the statistical results for Model 1 and Model 2 demonstrate that nonrenewable resource per capita (ln) is only a significant predictor for the World Bank indicator. Further analysis finds its change of slope estimate 125 The R 2 of Model 3 is equal to or bigger than the R 2 of Model 2 and Model 1. 203 from Model 1 to Model 2 is bigger than 20%. 126 Thus demographic adaptability group is a confounder between nonrenewable resource scarcity and the World Bank indicator of political instability. Although neither annual civil war number index nor PITF index is significantly correlated with nonrenewable resource group, the change of slope estimate of nonrenewable resource per capita (ln) from Model 1 to Model 2 of the civil war index is also bigger than 20%. 127 Thus it is consistent with the trend when the World Bank political instability indictor is used as the dependent variable. What we can conclude so far is that some evidence has been found that demographic adaptability group is a confounder in the association between nonrenewable resource per capita and political instability, in particular less intense forms of political instability. The apparent statistical connection between nonrenewable resource scarcity and political instability, no matter significant or insignificant, is actually due to the influence of other factors, such as demographic characteristics of the country. Since countries with high nonrenewable resource scarcity often have low demographic adaptability, they are more likely to have lower political stability. To summarize, the statistical results found for demographic adaptability in the resource scarcity/political instability nexus are not exactly the same as those of social, economic or political adaptability. First of all, same as the findings of the previous three sections, the renewable and nonrenewable resource scarcities have a different relationship 126 The slope estimate of nonrenewable resource per capita (ln) in Model 1 is .053, and .027 for the adjusted Model 2. 127 From -.008 to -.004. 204 with political instability and demographic adaptability. When renewable resource is concerned, demographic adaptability is a significant confounder in the relationship between renewable resource scarcity and political instability, including both less intense forms of political instability and full scale wars. This is different from the significant interaction effects found when political or economic adaptability is investigated in the prediction of renewable resource scarcity. No significant group difference can be found when the association is analyzed separately within each demographic adaptability group, but the confounding effect shows that it is illusive to accept the bivariate correlation between nonrenewable resource scarcity and political instability. As a matter of fact, the relationship between the two is indirect, and demographic adaptability is one of the major contributors that bridge the two variables and results in significant association. A weaker confounding effect has been reported for the models of nonrenewable resource scarcity. In a strict sense, demographic adaptability group is a significant confounder only for the association between nonrenewable resource scarcity and the World Bank political instability indicator since the other two political instability indexes are not significantly related to nonrenewable resource scarcity. It can only conclude that demographic adaptability group is a more significant predictor of political instability than nonrenewable resource scarcity. Then it is time to compare the statistical findings with our hypotheses regarding demographic adaptability to resource challenges. 205 5.4. Demographic characteristics. A country‘s adaptability to resource pressures is likely to decrease with its unfavorable demographic characteristics including higher population growth rate, lower percentage of urban population, smaller working population, and lower literacy rates. 6.4. A state facing high renewable resource scarcity is likely to be less stable when its demographic adaptability is low. Otherwise, the state‘s political stability is not likely to be associated with renewable resource scarcity. 7.4. A state with nonrenewable resource abundance is likely to be political unstable if its demographic adaptability is low. Otherwise, political stability of the state does not increase with nonrenewable resource abundance. The PCA test addresses Hypothesis 5.4 on the variables measuring demographic adaptability to resource scarcity. It has been confirmed that population growth, percentage of urban population, percentage of potential labor force, and literacy rate are core indicator for a country‘s demographic foundation for effective adaptation to environmental challenges. Besides, their factor loadings show that their relationship with demographic adaptability is consistent with the direction expected based on previous literature. Hypothesis 6.4 is weakly supported since the grouping variable of demographic adaptability is a significant confounder for some indicators. However, since the interaction is not significant in the tested model, it is more accurate to say that countries with low demographic adaptability are more likely to have renewable resource scarcity 206 and higher political instability. Even weaker evidence for confounding effect has been found for nonrenewable resource scarcity. Since it is nonrenewable resource scarcity rather nonrenewable resource abundance that is possible to be associated with political instability and low demographic adaptability, Hypothesis 7.4 is not supported. Summary of the Results The above statistical analysis tests our hypothesized model regarding the four aspects of adaptability. As it shows, the ―greedy‖ hypothesis regarding the connection between nonrenewable resource abundance and political instability are not supported. Considering the significant confounding effects of some types of adaptability, weak evidence indicates that nonrenewable resource scarcity rather abundance might be a potential predictor of political instability. However, the evidence is too weak for us to reach any statistical conclusion. This actually explains the inconsistent findings in previous security study on nonrenewable resources. Since no consistent and highly significant results can be found, the statistical results are more dependent on the specific resources and data source the researchers use. The hypothesized model appears to fit much better when renewable resource scarcity is investigated. The interaction between adaptability and renewable resource scarcity is proved to be significant when political and economic adaptability is investigated separately. Furthermore, when each one of economic, political and social adaptability is used as the grouping variable, the association between renewable resource 207 scarcity and political instability is significantly different between the high and low adaptability groups. The findings can be summarized in Table 24. Table 24: Summary of the Statistical Findings in Chapter 5 Renewable Resource Nonrenewable Resource Interact Confound Group Difference Interact Confound Group Difference Economic Adaptability S - S NS PS NS Political Adaptability S - S NS PS NS Social Adaptability NS NS S NS NS NS Demographic Adaptability NS PS NS NS PS NS S: significant NS: nonsignficant PS: significant when some indicators of political instability are used It can be seen that our model fits renewable resources much better than nonrenewable resources. Except for demographic adaptability, renewable resource scarcity is significantly associated with political instability in countries with high adaptability, but not for those with low adaptability. Based on these statistical results, the next chapter selects three non-conforming cases to further examine the established model and enrich our understanding of the casual mechanisms. 208 Chapter 6 Analysis on Nonconforming Cases and Implications on The Model Based on the statistical findings in the previous two chapters, this chapter goes further to evaluate and modify the proposed model by investigating the causal mechanisms qualitatively. In order to account for variables that are possibly ignored by the quantitative analysis, it selects three African cases that poorly fit the tested model for in-depth case study. The analysis suggests that the non-confrontational traditional culture and pressures from the international community are two important factors that contributed to the political stability of the three states. This chapter then goes further to discuss the theoretical significance of these findings and their implications on the hypothesized model, emphasizing the necessity of addressing ―thick‖ and qualitative variables, such as culture and history, in a more complex model. The Process of Case Selection As part of the research design, three worst-predicted cases are selected based on the statistical findings and then further case analysis is performed on them. Confronting nonconforming cases is common in case studies to account for factors that are outside the core explanatory framework (Ragin 2004). The important variables identified outside the hypothesized model can then help us better understand the merit and deficiency of the model. Moreover, this study decides to choose three rather than more anomalies for in- depth analysis given the scope of the research. As addressed in the introduction chapter, the main purpose and contribution of this study is to bring new variables and 209 nontraditional statistical techniques to testify the hypothesized comprehensive model. Therefore, the core of this study is the quantitative analysis, and the case analysis works more as a complementary section to improve the plausibility of the statistical model. Therefore, it would be too much for this project to elaborate on a large number of anomalous cases. As far as this study is concerned, two out of the four types of adaptabilities are found to best fit the hypothesized model—economic and political adaptabilities. Technically, there are two ways to identify the outliers in the tested model regarding these two types of adaptabilities. The first way is to start with the theoretical framework. As mentioned before, the hypothesized model assumes and the statistical results have confirmed that countries with low political and economic adaptability are at higher risk of political instability when they have severe scarcity of renewable resources. Thus a nonconforming case would be a country having high renewable resource scarcity and low political and economic adaptability simultaneously, but is able to maintain its political stability. That is to say, the country did not experience any full-scale civil wars or severe political regime crisis during the period from 1970-1999. Another method intends to find outliers according to the statistical criterion. Basically, it conducts residual analysis of the regression test, and then identifies the outliers by checking the observed residuals and residual plots to assess the appropriateness of a model (Kleinbaum, Kupper and Muller 1998:216-227). If some of the observed residuals are largely different from zero, and the residual plots show that those cases are not randomly scattered around zero, then those cases should be considered 210 as potential outliers. More statistical rules have been developed by researchers to identify outliers and perform diagnostics to access the impact of the identified cases on the soundness of the model. Appendix 3 summarizes the major procedures and rules of thumb in outlier diagnostics. This study originally identifies outliers using both the theoretical and statistical ways. It turns out that the outliers identified according to the two methods at most have one case in common, which suggests that those outliers might not poorly fit the theoretical model in a statistical sense. 128 As discussed in the methodology chapter, highly accurate and valid data are extremely difficult to obtain in social science research. Thus it is a more practical goal if this study performs statistical analysis to find out the general trend within the available data. The results would be quite strong if analyses using different indicators of the dependent and independent variables all point to the same pattern. Following this intention of capturing the big picture, the theoretical method is a better way to identify nonconforming cases of the tested model compared with residual analysis. Moreover, the statistical method might not be able to pick up the most poorly-fitted cases considering the quality of the data. 129 That is to say, small inaccuracy in raw data can result in huge impacts on the statistical analysis. Hence, this 128 The country that may be found as nonconforming case by both theoretical and statistical methods is Benin. However, Benin had one adverse regime change in 1972 according to the PITF coding. Therefore, Benin does not quite qualify as an outlier according to the theoretical method. 129 It actually happens to this study. Most of the outlier cases identified by residual analysis comply with our theoretical model. 211 study chooses the theoretical method instead of the statistical method to find nonconforming cases. 130 At first, all the countries in the group of low economic adaptability are picked up. Then within this group, countries with high renewable resource scarcity (renewable resource per capita is lower than the median value) are identified. As the third step, the political instability of these countries is checked. If a country in this list has no civil wars, no PITF political failures, and its World Bank political instability indicator is close to or bigger than zero, then it implies that this country is able to maintain its political stability throughout the thirty-year period. Such a country does not comply with our tested model which identifies a positive relationship between renewable resource scarcity and political instability in countries with low economic adaptability. The same procedures are performed on political adaptability to pick up outlying cases that do not conform to the statistical model established. Following the procedures, Malawi, Mauritania and Tanzania are picked up as three of the most poorly-fitted countries in the low economic adaptability group. Table 25 gives a summary of the three countries based on the major variables used in the economic and political adaptability models. As Table 25 shows, the component scores of economic adaptability of Malawi, Mauritania, and Tanzania are all less than zero, thus they are in the low economic adaptability group. The three countries‘ political adaptability component scores are also 130 Even residual analysis eventually has to go back to the theoretical model to check if the cases are indeed outliers. 212 below zero, and are less than the median value. 131 Therefore, all of the three countries also belong to the low political adaptability group. Table 25: Summary of the Three Nonconforming Cases Country Malawi Mauritania Tanzania Renewable Resource Per Capita $785 $1671 $1212 Nonrenewable Resource Per Capita 0 $1311 $3 Estimated No. of Civil Wars Per Year 0 0 0 PITF Political Instability 0 0 0 WB Political Instability -0.19 0.39 -0.20 Economic PCA Score -1.55 -0.92 -1.37 Political PCA Score -0.39 -0.57 -0.62 Note: Except the PCA Scores and Renewable Resource Per Capita, the other variables are the thirty-year average values. As far as renewable resource per capita is concerned, the three countries are all below the median value, with two below the 25th percentile and one below the 10th percentile. 132 According to the tested model in the previous chapter for economic and political adaptability, these countries are expected to be at great risk of political instability. That is to say, the three indicators of political instability should demonstrate some severe political crisis in these countries. However, as it is summarized in Table 25, none of the three countries had a full- scale civil war according to the five datasets of civil conflicts. Neither do they experience any types of political failures coded by the PITF team, including revolutionary wars, ethnic wars, adverse regime changes, and genocides and politicides. In addition, their 131 The median score is -.16. 132 The median of renewable resource per capita is $2305, the 25 th percentile is $1492, and the 10 th percentile is $785. 213 scores of the World Bank political instability indicator are close to or bigger than zero. Thus it is suggested that these three countries‘ political systems are relatively stable during the time period of 1970 to 1999, and there was no serious challenge to their stability in general. Analysis on the Three Nonconforming Cases To explore why the tested model does not work well for these three countries, Table 26 summarizes the major geographic, demographic, political and socio-economic characteristics of them. 133 This section then examines the three countries in more depth and conducts comparison to find out if they have some characteristics in common that are left out in the statistical model. By recognizing and analyzing these variables, we can further revise the model and better understand the underlying causal mechanisms. Before the specific contexts of the three countries are addressed, it is helpful to set up the historical background of the time period that is covered in this study. The period of 1970-1999 covers a big part of the Cold War and it also witnesses the sudden collapse of the communist bloc led by the Soviet Union. The first two decades feature the rivalry between the two superpowers and their competition in military building and constructing alliance. Although the antagonism between the two blocs continued, the bipolarity structure also made the world a relatively peaceful environment after the World War II. Most of the countries in the world were associated with either the communist or the 133 The data in Table 6.3 are from the online World Factbook of CIA, accessed in November 2009. 214 capitalist bloc and were able to keep their domestic stability backed up by the two superpowers. 134 However, the swift collapse of the Soviet Union in the early 1990s caught the world by surprise. As a chain reaction, the world in 1990s experienced a dramatic transformation which brought forth a number of new independent countries from the Soviet Union. In the meantime, the era features a wave of institutional transformation that a number of countries shifted from their previous communist or dictatorial rules to democracy. 135 134 For example, Gaddis (2006, 1997) talks about Cold War history. 135 Works such as Huntington (1991) explain the democratization wave at the end and post-Cold War. Przeworski (1991) and Haggard and Kaufman (1995) analyze the political economy of democratic transition. Studies such as Jaggers and Gurr (1995) conduct empirical analysis on the democratic transformation. Shin (1994) synthesizes the significant theoretical and empirical findings on the topic. 215 Table 26: Major Characteristics of the Four Nonconforming Cases Country Malawi Mauritania Tanzania Independence 1964 from UK 1960 from France Tanganyika: 1961 (from UK- administered UN trusteeship); Zanzibar: ber 1963 (from UK); the two merged as Tanzania in 1964 Location Southern Africa Northern Africa Eastern Africa Population 14,268,711 3,129,486 41,048,532 Area (sq km) 118,484 1,030,700 947,300 Border Countries Mozambique, Tanzania, Zambia Algeria, Mali, Senegal, Western Sahara Burundi, Democratic Republic of the Congo, Kenya, Malawi, Mozambique, Rwanda, Uganda, Zambia Terrain narrow elongated plateau with rolling plains, rounded hills, some mountains mostly barren, flat plains of the Sahara; some central hills plains along coast; central plateau; highlands in north, south Natural Resources limestone, arable land, hydropower, unexploited deposits of uranium, coal, and bauxite iron ore, gypsum, copper, phosphate, diamonds, gold, oil, fish hydropower, tin, phosphates, iron ore, coal, diamonds, gemstones, gold, natural gas, nickel 216 Table 26, continued Country Malawi Mauritania Tanzania Land Use arable land: 20.68% permanent crops: 1.18% other: 78.14% arable land: 0.2% permanent crops: 0.01% other: 99.79% arable land: 4.23% permanent crops: 1.16% other: 94.61% Labor force – by occupation agriculture: 90% industry and services: 10% agriculture: 50% industry: 10% services: 40% agriculture: 80% industry and services: 20% GDP per capita $800 $2,100 $1,300 Regime Type multiparty democracy military junta Republic Urbanization 19% 41% 25% Number of Major Ethnic groups 11 3 4 Religions Christian 79.9%, Muslim 12.8%, other 3%, none 4.3% Muslim 100% mainland - Christian 30%, Muslim 35%, indigenous beliefs 35%; Zanzibar - more than 99% Muslim Number of Major Languages > 7 6 > 5 Military expenditure (% of GDP) 1.3% 5.5% 0.2% 217 Table 26, continued Country Malawi Mauritania Tanzania International Disputes disputes with Tanzania over the boundary in Lake Nyasa (Lake Malawi) and the meandering Songwe River remain dormant Mauritanian claims to Western Sahara remain dormant Tanzania still hosts more than a half-million refugees, mainly from Burundi and the Democratic Republic of the Congo; disputes with Malawi over the boundary in Lake Nyasa (Lake Malawi) and the meandering Songwe River remain dormant Source: CIA- The World Factbook, the World Bank, Fearon and Laitin (2003c). 218 Therefore, it is reasonable to expect that a number of countries would have experienced some type of political transition due to the impact of the ending of the Cold War. This section pays particular attention to the 1990s, examining whether any transformation occurred in these countries, and how they were able to peacefully achieve or prevent the transition without damaging their political stability. The major characteristics of each country depicted in Table 26 are examined in comparison with the tested models regarding economic and political adaptabilities. First, these characteristics are briefly addressed as they are compared across the three countries to see if some features are shared by the three countries. In the meantime, with the tested model in mind, it aims to explore any features of these countries that might not have been considered in the tested model. It can be seen from Table 26 that the three countries are all in Africa. Moreover, they are all relatively new countries that became independent from the United Kingdom or France in the 1960s. Mauritania and Tanzania are both middle-sized countries in terms of population and area, whereas Malawi has a much smaller population and area. Mauritania is the biggest among the three in terms of area, but the smallest when population is considered. Probably the main reason is that most of its land is flat plains of the Sahara, which does not provide suitable living conditions for human beings. Mauritania only has 0.2% of arable land and 4.23% of Tanzanian‘s land is arable. Malawi has the highest level of population density, and it also has the largest percentage of arable land among the three (21%). 219 Figure 12: Malawi, Mauritania and Tanzania on the Map of Africa Figure 12 illustrates the location and relative size of the three countries in the map of Africa. It shows that Mauritania is in northwestern Africa, and Malawi and Tanzania are bordering with each other in the eastern part of Africa. According to the data, the three countries all face a scarcity of renewable resources and neither do they have rich nonrenewable resources. In general, Mauritania has relative more renewable and nonrenewable resources per capita than the other two. Malawi and Tanzanian are Mauritani a Tanzania Malawi 220 estimated to be under great resource pressure that may not provide adequate basic requirements for people‘s survival. In the meantime the scarcity of resources also impedes the countries‘ economic development. The eventual results, as it is expected, are great crises of their political systems. As far as economic power is concerned, all the three countries are struggling in poverty. Their GDP per capita are far below the median value of the world, ranked close to 200 or even more behind. Malawi and Tanzania have a low rate of urbanization (19% and 25%). Compared with them, Mauritania is much more advanced in terms of urbanization (41%), which is close to some rapidly developing countries like China (43%). More than half of the labor force in each country is working on agriculture, which indicates the weakness of industry sectors, especially for Malawi (90%) and Tanzania (80%). None of the three countries is homogenous with one overwhelmingly dominant ethnic group. Instead, they are all diverse countries with several major ethnicities. The languages used by people are also very diverse, besides the fact that they all have more than five major languages; they also have many local languages. Moreover, as far as religion is concerned, Mauritania is very homogenous, with 100% Muslim. Malawi is dominated by Christianity (79.9%), whereas in Tanzania, the percentages of Christian (30%) and Muslim (35%) are very close. Thus it can be concluded that the three countries in general are diverse in terms of ethnicity and languages, however, their religious composition is much less fragmented. 221 The three countries put different emphasis on military building. Malawi and Tanzania have a low expenditure on military. Tanzania only spends about 0.2% of its GDP on military. However, the spending on military for Mauritania is much higher (about 5.5%) and it ranks fourteenth in the world. It can be seen that among these three countries, their military spending increase with the size of territory. Due to the impact of the ending of the Cold War, the political institutions of the three countries changed from 1970-1999. Malawi was a one-party rule dictatorship in the first two decades, and shifted to a multi-party democracy in the early 1990s. Mauritanian and Tanzanian also experienced some political transition in the 1990s. 136 What should be noted is that these countries were able to maintain their political stability even in the international context of rapid political transition. Mauritania continues to be ruled by military junta today; Malawi and Tanzanian experienced some type of democratic transition in the 1990s. Therefore, Malawi and Tanzanian were not only able to achieve their political transformation in a peacefully manner, their new regimes were also able to stabilize the situation and effectively control the whole country. To solve this puzzle, the following section examines the history of the three countries focused on the period from 1970-1999, aiming to find out the main explanations for their political stability under dictatorship and democratic transition. It is suggested that the international context contributed to the stability of these newly independent countries in the 1970s and 1980s. In the 1990s, the international community also worked as a catalyst for change and at the same time as a pacifier. Moreover, the 136 The political transition will be discussed in-depth in the following sections. 222 specific cultural contexts of these countries also contribute to the long term political stability. Malawi The predecessor of Malawi is the Nyasaland, a British protectorate established in 1891. Colonial Nyasaland witnessed a radical change in its critical political organization—the Nyasaland African Congress in the 1950s. It increasingly took a confrontational approach to ―achieve its avowed aims‖ (Pike 1969 ; Short 1974 ; Rotberg 1965). This transition, according to many scholars, was brought by a new breed of young politicians with western education, such as Henry Chipembere, Kanyama Chiume, and later, Dunduzu Chisiza (Kalinga 1996). It is suggested that Dr. Kamuzu Banda was encouraged by this group of nationalist politicians to return home to lead the struggle for decolonization (Venter 1995:155). In 1959, decolonization movement in Nyasaland became country-wide disturbances and eventually led to its independence from the United Kingdom as Malawi in 1964 (Central Intelligence Agency). Soon after Malawi became independent, Dr. Banda established his one-party rule by stringent control of political, social and cultural behavior in the country. Since then Malawi experienced three decades of one-party rule under Dr. Banda (Central Intelligence Agency). He built a remarkably high degree of legitimacy and was widely respected as well as feared by the people (Forster 1994). His rule and regime was maintained with the help of a small group of his close advisers (Venter 1995:155). 223 However, Forster (1994) points out that the long-term stability in Malawi was not simply based on political repression by the Banda regime. 137 He argues that the importance of culture should not be underestimated, especially how it was manipulated to be the basis of the Banda regime‘s legitimacy. Although the President appeared to be anti-ideological and pragmatic, there were ―nativistic elements‖ and a constant reaffirmation of a particular ―African culture‖ in his public oratory (Forster 1994:483). Banda did not only establish his leadership of the independence by unifying the African nationalists and sympathizers (McCracken 1998), he also incorporated a particular version of ―African tradition‖ into the country‘s modern political culture. 138 Besides the authoritarian rule and cultural nationalism, the earlier economic development of Malawi since its independence also played a part in the stabilization of the country during the late 20 th century. At the time of its independence, Malawi government was insolvent and it lacked the natural, human or financial resources for survival. However, since its independence, the country has experienced a structural change of economy which resulted in rapid and wide-spread economic growth based on national income per capita data. 139 It is suggested that external financial support, especially the British aid, contributed largely to the economic development of Malawi 137 Forster (1994) agrees with Finer (1976) that power without authority is inadequate for the long-term stability of military regimes. 138 According to Forster (1994), Banda expressed the importance of traditionally ascribed values on family relationships, favorable view of ordinary villagers in contrast to the educated. Banda also adapted traditional dances to establish it as a crucial feature of Malawian cultural nationalism. 139 Kydd and Christiansen (1982) investigates the nature of the structural change in Malawian economy since its independence. According to them, the rapid growth of large-scale agriculture is a salient feature of the change. 224 (Kydd and Christiansen 1982). The country‘s capacity to develop would have been severely undermined without foreign aid. With the end of Cold War in the 1990s, the apparently impregnable Banda regime suddenly collapsed between 1992 and 1994. The transition started from the publication of the Roman Catholic bishops‘ pastoral letter in March 1992, and then the revulsion against dictatorship spread the country and became a popular movement. In May 1994, Malawi held its first genuine multi-party elections and transitioned into a multi-party democracy (McCracken 1998). 140 It is suggested that the Western creditor states of Malawi acted as crucial impetus of the transformation. With the end of the Cold War, African dictators were much less crucial as anticommunist bulwarks to the Western democracies, in particular Britain and the United States. In contrast, Malawi‘s human rights record made it an embarrassment to American and European statesmen (Newell 1995). Besides open criticism, the United States and the European Union reversed their previous policy on Malawi after 1989 and imposed pressure on the Banda regime by freezing financial aid to the country (Lapido, Noel and Mouftaou 1997). Besides the external pressures, Malawi also faced serious internal economic challenges. Since the early 1980s, Malawi began to experience an economic downturn, which accelerated the realignment of its political elites‘ interests (McCracken 1998). Furthermore, the ending of one-party rule and embracing of multi-party election in 140 Many studies address the politics and economics of Malawi since 1991. For example, Lapido et al. (1997) reviews the achievements and problems of implementation of the reform, and summarizes the lessons learned. Donge (1995) discusses the spirit of consensus politics in the process of Malawi‘s democratization. Kaspin (1995) examines the ethnic groups in Malawi and their impacts on the democratic transition of the country. 225 Zambia also worked as a catalyst for political transition in Malawi. Therefore, the international and domestic factors together impose challenges to the Banda regime and eventually resulted in the regime‘s response to the Catholic clerics‘ demand for greater governmental accountability to the Malawian people (Mitchell 2002). The bishops opened the valve and encouraged the criticism of the Malawi government from inside. 141 The period discussed in the model is from 1970-1999. It is understandable that the first two decades tend to be stable under the dictatorial rule of Dr. Banda and ritualized paternalism (Mitchell 2002). In the last several years Malawi suddenly experienced the collapse of dictatorship and started the transition to multi-party democracy. Although there were some riots and disturbances followed the Pastoral Letter (Newell 1995), Malawi was still able to keep a relatively stable domestic environment even during the transitional period of 1990s. Previous research finds that the political culture of Malawi is of crucial importance to the stability of the country in the transitional period. Banda‘s earlier policies for national development and political stability help Malawians to build and realize their common identity (Kaspin 1995). Although regional fragmentation existed in the democratic process, there is always a persistent search in Malawian political culture for ―consensus politics,‖ which intends to build maximum coalition in the political process rather than democratization ―dominated merely by regional fragmentation‖ (Donge 1995:229). The rationale of national unity was adhered to by the major parties 141 The bishops of the Catholic Church issued a Pastoral Letter in March I992, which is the first public criticism of the government in the country (Newell 1995). 226 and grass-roots, and therefore the intimidation and irregularities were marginal in the democratizing process of the country and did not influence the results (Donge 1995). 142 Mauritania The Islamic Republic of Mauritania became independent from France in 1960 after decades of colonial rule. The country was ruled by the heavy hand of President Taya since the coup in 1984 (N'Diaye 2006). After about two decades of flawed presidential elections, President Taya was deposed by a bloodless coup in 2005 and the country was expected to begin a democratic transition under the supervision of a military council. However, another military junta deposed the first freely elected president of Mauritania and General Abdelaziz took power as a civilian president in 2009. 143 Since the statistical analysis covers the period from 1970-1999, this study mainly focuses on the Mauritanian history right after its independence and the rule under President Taya. Mauritania is a vast country in West Africa with large desert territory. The population density in Mauritania is low and people are struggling at the poverty level (Pazzanita 1992). What is worth to be mentioned is that Mauritania is the only African country whose government continued to tolerate slavery in the 1990s (Human Rights Watch/Africa 1994 ; Lydon 2005). Mauritania also distinguishes itself in Africa as an 142 Donge (1995) gives a very detailed description of the transition in the 1990s, including the passing of the constitution and the election. 143 In April 2007, Sidi Ould Cheikh Abdallahi was elected as the president, but the military junta led by General Abdelaziz deposed him in August 2008. Mr. Abdelaziz got himself elected as Mauritania‘s civilian president in July 2009 (All is rather easily forgiven: A coup-maker becomes a civilian president 2009). 227 Islamic republic with a more than three million population that share the same Islamic religion. Mauritania is made up of three major ethnic groups (N'Diaye 2006). 144 The dominant ethnic-cultural group in the country is the lighter-complexioned nomadic Arab- Beydane or Moors. The Beydane are the descendants of Arab tribes migrating from the Arabian Peninsula, and they mainly live in the north, center, and east of the country. Today they still keep a highly hierarchical and tribalized society and identify with the Arab world and Islamic rules. Close to one-third of the Mauritanian population is composed of the Beydane, and they control the political and economic power (N'Diaye 2006). The second group is the Black Mauritanians who are composed of four non- Arabized black ethnic groups (the Halpulaar, the Soninké , the Wolof, and the Bambara). Slightly less than one-third of the Mauritanian population is black and they live along the country‘s southern borders with Malian and Senegalese (N'Diaye 2006). In recent years, as the Beydane try to impose Arabic as the national language and to expel black farmers from the arable land along the Senegal River, the black Mauritanians have attempted to establish their cultural identity in the country and struggle for political and economic equity (Human Rights Watch/Africa 1994). The third ethnic group is composed of descendants of enslaved black Africans, who are called the Haratines and Abeed (N'Diaye 2006). They are the largest group in 144 Alfred Gerteiny (1967:46-56;88-101) gives a detailed description of tribal and ethnic life in Mauritania. He investigates both the Moorish population and the Black Mauritanians. 228 Mauritania in terms of population. Some of them are freed whereas some are still enslaved. Since they share the language and Islamic culture with their Beydane masters, they identify culturally and psychologically as allies of Beydance. Although there is some evidence that Haratines are increasingly emerging as a political and social force fighting against persistent slavery, their demand for greater political and economic power is still absent (N'Diaye 2006). Ethnic tensions have existed in Mauritania since its independence. The most important one is between the Beydane and the Black Mauritanians. It reached a high level in the mid-1980s. It was marked by the denunciation of discrimination by the Black Mauritanian intellectuals and the failed coup against the president by some Black Mauritanian officers in 1986 (Jourde 2001). In the meantime, these ethnic tensions are interwoven with political and economic struggles across ethnic boundaries. However, as it is suggested by Pazzanita (1999), after about four decades of its independence, Mauritania still holds a ―distinctive way of life and political culture‖ that rejects violence, and embraces ―tribalism and detachment from centralized authority.‖ That is to say, the distinctive and isolated political culture in Mauritania might be the major contributor to the survival of the authoritarian governance in the country. Compared with many African countries, the social structure of Mauritania is relatively more homogeneous rather than factious. The opportunities for substantial development and industrialization in Mauritania were few after its independence from France in the 1960s. Due to its scarce natural resources, and deeply embraced tribalism, the nation had to rely on foreign assistance and 229 recognition of non-Arab African countries to solidify the legitimacy of the state (Pazzanita 1999). 145 Although politically dominated by Arabs, Mauritania did not have cordial relations with Arab states since most of them supported Morocco‘s territory claim on Mauritania (Thompson and Adloff 1980:223-6). Therefore, in the early days under its first President, Daddah, Mauritania relied heavily on France for political, economic and military support. It also made efforts to build close diplomatic relations with sub-Saharan African states, some of which were also colonies of France (Handloff, Curran and Library of Congress. Federal Research Division. 1990:24-6). However, Mauritanian diplomacy moved towards the wider Arab world after Morocco‘s abandonment of the territorial claim and the establishment of formal diplomatic relations between the two countries in 1970 (Thompson and Adloff 1980:223- 6). The Daddash regime was seeking more independence from France and eventually became a member of the Arab League in 1973. 146 Politically, the Daddah regime is viewed as a ―benign dictatorship.‖ It continued its efforts to co-opt various antecedent political parties and absorb different ethnic and regional interests since mid-1950s (Pazzanita 1999). In 1961, President Daddah claimed the end of multi-party system and several smaller parties were melded into the Parti du people mauritanien (PPM). According to Pazzanita (1996), it was a move that owed more to Daddah‘s ―not-inconsiderable talents as a conciliator and conscious focus of national 145 Studies of the Western Sahara, such as Thompson and Adloff (1980) and Hodges (1983) discuss Mauritanian political economy and politics with regard to the war in the desert. Besides, Seddon (1996) provides a very good summary of Mauritania‘s political economy from the 1960s to 1990s. 146 French subventions in Mauritania‘s budget ended in 1963 and the presence of French troops in Mauritania ended in 1966 (Thompson and Adloff 1980:51-3). 230 cohesion‖ rather than ―outright repression.‖ Since then the PPM tightened its control on the political and economic institutions of the state, and Mauritania, in effect was a one- party state (Seddon 1996). With its adeptness at maintaining such a quasi-dictatorship, the Daddah regime had firm control with little incident until 1975. 147 Mauritania‘s economy showed relative prosperity at the time due to its deposits of iron ore and fisheries (Seddon 1996). However, President Daddah‘s decision to get into the Western Sahara conflict with little input from the public or governmental officials soon led to the end of his regime. 148 A bloodless coup in 1978 deposed President Daddah and replaced him with the governance of an army group. The military government signed a peace agreement abandoning the claims to Western Sahara, and extracted itself from the Western Saharan conflict (Legum 1982). However, ethnic tensions and tribal rivalries in Mauritania became aggravated with the advent of military rule (Pazzanita 1999). Tribal affiliation and hierarchical social order were given great importance again by the military juntas. As Pazzanita (1999, 1996) suggests, the emphasis on tribe and ethnic identity to some extent had a positive effect. For example, the ―extensive networks of mutual assistance‖ slightly alleviated the desperate economic straits from later 1970s to early 1980s. Moreover, Pazzanita (1996, 147 Although there were sporadic strikes and demonstrations related to the ethnic tensions between Arab and black Mauritanians, the regime in general was stable in the decade, and there was no detrimental ethnic conflict (Pazzanita 1996). 148 Pazzanita (1996) suggests that a number of troubling problems actually existed underneath the stability in Mauritania at the time. For example, the government‘s failure to establish a universal system of education, the party‘s isolation from the general public, the lack of ubiquitous ethnic schisms in the nation, etc. But the foreign policy decision to annex part of Western Sahara quickened the weakening of the regime. 231 1999) mentions that tribes and ethnic groups also provided a relatively safe environment for internal interest group activities. There was much more freedom to express opinions opposing the policies of the military rulers with little fear of retaliation within one‘s own tribal and ethnic group. After the coup in 1978, Mauritania experienced some democratic stirrings. A draft constitution was promulgated by Colonel Heydallah, with the purpose of being elected as the president presiding over a cabinet and bureaucracy that were dominated by civilians (Pazzanita 1996). However, the plans were forced to be abandoned when another coup attempt occurred in 1981. Eventually, Colonel Heydallah was deposed by a peaceful coup led by Colonel Taya in 1984, who is a ―relatively apolitical figure of purely nationalistic orientation‖ according to Pazzanita (1999). At first the Taya presidency showed its willingness to democratize with loosening repressions and a crackdown on corruption (Legum 1982). However, the hope for democratization was dashed as the ethnic tensions between Moors and Black Mauritanians worsened greatly since 1987. 149 An international conflict erupted in the region close to the Senegal Rival involving Senegal and Mauritania in 1989. 150 The ethnic tensions triggered a counter-response from the government that actually resulted in serious human rights abuse (Human Rights Watch/Africa 1994). It struck down riots with iron hand, and reconfirmed a Moor-centric mindset. Furthermore, President Taya allied 149 An important group is the Forces de Liberation Africaine de Mauritanie (FLAM). It claimed to represent the interests of disadvantaged non-Arab B lack Africans in Mauritania. FLAM's led a planned coup against Ould Taya in 1987 and made sporadic armed attacks mainly in the Senegal River (Pazzanita 1999). 150 Summaries of the Senegal-Mauritania dispute can be found in Ron (1991) and Magistro (Magistro 1993). 232 with Bidhan extremist movements and supported the Saddam Hussein regime in the Second Gulf War (Jourde 2001). 151 Therefore, Ba‘thi and other pan-Arab elements were allowed to express their anti-Black African feeling more freely, which worsened the tensions between the two ethno-cultural groups. These internal tensions forced President Taya to begin political liberalization as the country was moving towards even more dangerous directions (Seddon 1996). Besides the fear of growing internal discontent and tensions, the international pressures from foreign lending agencies and governments are also believed to play an essential role in Mauritania‘s move towards a multi-party system. As it is noted, President Taya‘s first proposal for pluralism was compiled after a visit by French Foreign Minister Roland Dumas in 1991. During his visit, Dumas gave diplomatic pressure to Ould Taya urging him to demoratize (Ould-Mey 1998:49). After the beginning of its democratic transformation, Mauritania continued to seek and receive monetary assistance from international lending agencies and donor states. For example, in 1994 the IMF approved a loan of $23.2 million to support the government‘s second annual ESAF program, and in the same year, a meeting of donor states in Paris agreed to provide $235 million in support of Mauritanian economic reform (Seddon 1996). Foreign financial assistance therefore became a major force that impelled Mauritania to economic development and democratic transformation. 151 For more information of the origins and consequences of the 1990-1991 Gulf Crisis and the role of Mauritania in the crisis, refer to studies such as Khalidi (1991), Sayigh (1991), Mattar (1994) and Marianne (2002). 233 During its democratic transformation in the 1990s, the Taya regime did not suffer any serious political crisis, nor did it suffer degeneration. Although the stability of Mauritanian society provided assistance for transformation, the democratization in Mauritania was ―hampered by poverty and low levels of literacy and political involvement‖ (Pazzanita 1996). The oppositional polity in Mauritania remained weak, and the essential elements of a genuine democracy are still absent or at an embryonic stage of development. Tanzania Tanzania was merged as one nation by mainland Tanganyika and the islands of Zanzibar and Permbain in 1964 (Central Intelligence Agency). The two constituent parts were both colonies of Britain and became independent in the early 1960s. Julius Nyerere, the president of Tanganyika became the president of the United Republic of Tanzania and the country began its thirty years of one-party rule. Following the wave of democratization, Tanzania started its democratic transition to a multi-party system in the early 1990s. External forces played an important part in Tanzanian history of the last century. First, as Wilson (1989a) suggests, the creation of the nation of Tanzania was the work of foreigners. The establishment of Tanzania cannot be detached from the international context of the Cold War and the international strategy of the Lyndon Johnson Administration of the United States. 234 In order to protect its investment in southern Africa and resist the penetration from some North African countries such as Algeria and Egypt, the United States was planning to build a belt of control across central and east Africa in the 1960s (Wilson 1989a). A socialist Zanzibar would destroy the shield against communism. Thus the United States took actions in the region right after the revolution. On one hand, the United States tried to push Britain into military intervention in Zanzibar, one the other hand, it made efforts to create a ―neutral East Africa Federation‖ (Wilson 1989a). 152 Eventually, the United States settled for a union of Zanzibar with pro-western Tanganyika in April 1964 (Wilson 1989b). A joint constitution was approved in 1977 and the ruling party Chama Cha Mapinduzi (CCM) were composed by the ruling parties of the two parts (Skinner 2005:16). 153 Julius Nyerere, who was the president of Tanganyika, remained president of the United Republic of Tanzania. During the late 1960s and the 1970s, President Nyerere established a series of ―African socialist‖ policies. In 1967, he issued the socialist Arusha Declaration, announced forced relocations to collective farms, and nationalized the country‘s industrial and business sectors. The poor implementation of these policies left the country one of the poorest in the world (Empire's Children 2007). In the meantime, the foreign policies of the Nyerere Administration tended to put principles before advantages, which strained its relations with the West (Skinner 152 Wilson (1989a) surveyed the documents at the president Johnson library in Texas and found that the Johnson Administration tried to get the British into military intervention in Zanzibar, and it also collided with the top leaders of Kenya and Tanganyika, including secret conversations, bribes and suggested murders. 153 The ruling parties of the two part were Tanganyika African National Union (TANU) of Tanganyika and Afro- Shirazi Party of Zanzibar. 235 2005:19). In the early days, Tanzania received funding from Britain and Germany to improve its agriculture and develop industry. However, Nyerere‘s socialist policies and growing links with communist countries worsened its relations with the West. Tanzania lost its principle aid donor when it broke off relations with Britain in 1965, condemning the latter‘s role in Southern Rhodesia and arm selling to South Africa (Skinner 2005). Relations with neighboring Uganda and Kenya also deteriorated. Nyerere took an uncompromising stand against the brutal regime of Idi Amin in Uganda in the late 1970s. Tanzania successfully defeated the invasion of Uganda, however, the war still ruined the country (Skinner 2005:19; Ferreira 1996). Due to its poor economic policies and ideology-centered foreign policies against the West, Tanzania experienced severe economic decline and financial crisis in the late 1970s and early 1980s. Facing serious poverty, Tanzania started a process of gradual reforms in the mid-1980s (Muganda 2004). 154 It formally adopted an economic recovery plan in 1986. According to Muganda (2004), Tanzania has made significant achievements in its economic reform. It has achieved macroeconomic stability and there is an overall improvement in real incomes. Various studies on Tanzania have implied that there is a sustained decline in poverty in the country. 155 154 For discussion on the structural adjustment in Tanzania, refer to Hyden and Karlstrom (1993), Agrawal et al. (1993), Ferrerira (1996), and Muganda (2004). 155 For example, the survey on poverty of Tanzania by Ferreira (1996) finds that household poverty declined by about 22% between 1983 and 1991. The household poverty is further estimated to decline by approximately 28% between 1994 and 2002. Moreover, Sarris and Tinios (1995) compare real household expenditures from two national surveys. They suggest that real per capita expenditure in both rural and urban areas were higher in 1991 than in 1976, and the level of poverty declined in 1991. 236 With the progress of economic recovery, political reform also started to take place in Tanzania. After thirty years of one-party rule, Tanzania began its democratic transition to a multi-party system in the early 1990s. It separated the ruling party CCM from the government, passed legislations for multi-party transition, broke the link between business and the ruling party, and allowed more freedom of the press and association (Tripp 2000). The first democratic presidential and parliamentary elections of Tanzania were held in 1995 (Research and Education for Democracy in Tanzania (REDET) Project 1997 ; Mushi and Mukandala 1997), during which the ruling party won the elections. Irregularities in the voting were claimed by some international observers. By the end of the 1990s the leaders of Tanzania showed little inclination to move toward consolidating democracy (Bratton and Van de Walle 1997:235). 156 Tanzania was under one-party rule for about thirty years. Although the country suffered from severe poverty, it was able to keep a relatively stable political environment. One of the reasons for the endurance of the authoritarian regime is the conformist political culture in the country (Research and Education for Democracy in Tanzania (REDET) Project 1997). The colonial history in Tanzania destroyed the seed of pluralist democracy and cultivated an obedient political culture. While a strong civil society is required for a working democracy, Tanzanian civil society was weak right after its 156 According to Przeworski (1991), democracy is consolidated ―when most conflicts are processed through democratic institutions, when nobody can control the outcomes ex post and the results are not predetermined ex ante, they matter within some predictable limits, and they evoke the compliance of the relevant political forces.‖ Linz and Stepan (1996) identifies five arenas as characteristics of a consolidated democracy, including civil society, political society, economic society, rule of law and bureaucracy. 237 dependence (Mushi and Mukandala 1997). The priorities of the nation were to cope with poverty, ignorance, and disease. Therefore, a centralized authoritarian regime claiming the mission to sweep away these problems was established and maintained within the cultural context of Tanzania. During its one-party rule, the administration took strong actions against the emergence of civil society in both rural and urban areas (Mushi and Mukandala 1997) . The international community also played an important part in the economic and political reform of Tanzania. The country‘s economic adjustment program was supported by the International Monetary Fund and the World Bank, as well as foreign assistance (Agrawal et al. 1993). It is suggested that most of the foreign assistance was used for investment instead of consumption, and the dependency on foreign assistance did not result in deterioration in domestic savings performance in Tanzania (Agrawal et al. 1993). Furthermore, donors and the SAP imposed pressure on the Tanzanian government for increasing popular participation in politics (Mushi and Mukandala 1997). It is found that the inflows of foreign direct investment (FDIs) on former state-own enterprises (SOE) and green field investment are of great importance in the country‘s economic development and change (Ngowi 2009). 157 157 The impacts of FDI inflows are discuss in BoT et al. (2002, 2004). Bigsten et al. (1999) addresses foreign aid and the economic reform in Tanzania. 238 Summary on the Case Analysis The brief review of the history of the three countries has depicted the political succession, economic development and social order of them in the late 20 th century. As it shows, the three countries were able to maintain its political stability both under the authoritarian rule and the democratic transformation. These countries kept its unification and centralized leadership despite some marginal demonstrations and riots. Their regime change took a relative peacefully and bloodless way instead of severe violence. Based on the case analysis in this chapter, several features shared by the three countries but not considered in the hypothesized statistical model might help to explain why they can maintain political stability when facing serious resource scarcity as well as poverty and poor political institutions. All the three countries are in Africa. Although they have different ethnic and religious compositions, according to Forster (1994) they all share an ―African culture.‖ In the early half of the 20 th century, these African countries still embraced very traditional ways of life. They were the world‘s poorest countries and their political systems were still centered on tribalism, and some even retained slavery. In addition, colonial rule also have profound impact on the evolvement of their political institutions. They were colonies of Britain and France until the mid-29 th century. With the ending of the World War II and the withering of former empires, these countries became independent in the early 1960s. After their political independence, they all endorsed a patriarchal leader and were controlled under one-party rule. Banda in Malawi, Taya in Mauritania, and Nyerere in Tanzania are all patriarchal leaders who 239 maintained their authoritarian rule in the country for decades. In the meantime, these countries began to build their national identities by reviving or perpetuating selected aspects of their ―African tradition‖ in order to reinforce their unity and mitigate the feeling of inferiority (Forster 1994). 158 However, political leaders resorted to the ―African tradition‖ in quite different ways. Nyerere embraced socialism in Tanzania, and Banda based his leadership on nationalism but was hostile to socialism. Compared with the other two countries, Mauritania has a higher level of ethnic tensions due to the split between Arab Beydane and Black African population. Complying with its foreign policy since Daddash, Mauritania started to establish a national identity that bridged Arab and Sub-Sahara Africa. Although the friction between the two ethnic groups continued to exist, the bridging strategy was also inherited by the later president Taya. This specific ―African culture‖ that incorporates the tribalism tradition and colonial experience, followed by the identity-building of modern nation-states, thus summarizes the major characteristics of the three countries‘ political culture. On the one hand, the tribalism and colonial history are critical reasons for their weak foundation of democracy and maintenance of long-term authoritarian rule. On the other hand, in order to facilitate their economic survival and adaptation to the modern world, the authoritarian leaders of these countries tried to enhance their national identities by strengthening aspects of their tradition. Therefore, the traditional tribal culture and colonial experience 158 This is similar to Linton‘s argument that ―nativistic movements‖ can occur in culture-contact situations. They can be seen as ―a conscious, organized attempt on the part of a society‘s members to revive or perpetuate selected aspects of its culture.‖ 240 work together with more contemporary nationalism to retain the patriarchal political system and the unification of the country. Secondly, external factors also played a crucial role in supporting the countries‘ economy and peaceful political transformation. As former colonies, the three countries relied heavily on the financial assistance from donor states, especially Britain, France and the United States, at the time of their independence. As Tanzania adopted socialist policies and strained its relations with the West shortly after its independence, the country suffered severe economic decline partially due to the loss of principle foreign aid. Furthermore, external pressures which link financial aid with human rights also worked as a crucial catalyst for these countries‘ beginning of democratic transformation. The foreign assistance did not only support the economic recovery in the countries, it also worked as an impetus for the international community to press these authoritarian states‘ political reform. As poor states dependent on foreign aid and loans, the three African states can hardly ignore the condemnation and pressures. They thus were forced to respond to the demand for more democratic participation both within and outside the countries. Therefore, on the one hand, the traditions in the region feature a conformist culture, which embraces isolation and tribalism with relatively weak basis for democratization. It did not only delay the democratization, it also made it less confrontational, as the national politics kept searching consensus politics. On the other hand, the external actors worked as pacifiers, not just financially. The international 241 community works as inspectors that make sure the central leadership not taking large- scale violent means to cope with the dissents within the countries. In sum, the specific cultural context and the international actors are two important factors that contributed to the political stability of the three states. Facing with high resource scarcity challenges as well as poor economic performance and political institutions, they were able to maintain a relatively peaceful political environment thanks to the non-confrontational traditional culture and pressures from the international community. It suggests that the hypothesized model might be revised when some more ―thick‖ and qualitative variables, such as culture and history, can be addressed. 242 Chapter 7 Conclusions and Implications The Copenhagen Climate Conference caught tremendous media and public attention at the end of 2009. Representatives of 192 countries gathered together to reach agreement on fighting against climate change. It is now believed that grand change of the earth‘s climate system is leading to a series of environmental disasters and resource scarcity. One of the major concerns in the new century is that climate change and overpopulation may result in severe water scarcity and ultimately causes local armed conflicts, rather than international armed confrontations (Sullivan 2009). Experts have noted that the problem is likely to emerge in sub-Saharan Africa, Southeast Asia and some East Asian countries in coming decades (Sullivan 2009). This dissertation is aimed at explaining the in-depth political effects of such environmental challenges and accessing their impacts on national security. To improve our understanding of the resource scarcity-political instability connections, this study has developed a new model adding four aspects of adaptability as the intermediate variable respectively. It has recognized the different utilities of renewable and nonrenewable resources that result in their distinct linkage with political instability. It is argued that countries with high adaptive capacity can effectively cope with resource challenges and thus maintain their political stability, whereas those with low adaptability fail to do so. This chapter summarizes the findings of statistical and case analyses in the previous chapters, and then revisits the theoretical debate discussed in Chapter 2, addressing how this study sheds light on the debate among three schools. In 243 the end, it discusses the policy implications of this study and highlights some suggestions for future research. Empirical Findings and the Revised Model The previous chapters have conducted both statistical and case analyses to examine the relationships between resource scarcity and political instability, with environmental adaptability as the intermediate variable. First, although previous research on the topic has reported inconsistent results, this study suggests that there are indeed significant connections between resource scarcity and civil conflict. Utilizing five widely-used datasets in the field, the auto-regressions on annual number of civil wars ongoing over two periods have showed a more linear pattern in countries with high resource scarcity than those with low resource scarcity. That is to say, previous annual number of civil wars ongoing better predicts the annual number of civil wars ongoing in the later period when the research subjects are countries under high resource pressure. The significant group difference thus indicates that resource scarcity endangers a country‘s political stability by increasing the risk of civil war when it reaches some threshold level, such as the mean and median values. While the annual number of civil wars suffered by countries with high environmental scarcity can be largely explained by the environmental stress on them, countries with low environmental scarcity have fewer and more randomly distributed civil wars. The interesting findings with auto-regression analysis indicate that resource scarcity is an important variable contributing to political instability. To further unpack the 244 complex relationships, this study then applies PCA and regression analyses to examine the hypothesized model in more detail. It creatively brings in adaptability as an intermediate variable to explain the environmental scarcity-political instability nexus. Based on the literature, four aspects of adaptive capacity—economic development level, political institutions, social fractionalization and demographic characteristics—are examined respectively in the model. In addition, renewable and nonrenewable resource scarcities are examined separately since these two types of resources are of different utilities to the human society. 159 The investigation of interactions, confounding effects and group difference eventually supports the hypothesis that renewable and nonrenewable resource scarcity do associate with political instability in quite different ways. It has been found that economic development level and political institutions are important factors that influence the links between renewable resource scarcity and political instability. Countries with high economic development level and/or healthy political institutions are not necessarily politically unstable even if their renewable resource scarcity is high. But renewable resource challenge would be detrimental to the political stability of countries with low economic development level and/or poor political institutions. Following the same procedures, the analyses of nonrenewable resources do not find such a pattern. Although each of the four aspects of adaptability is sometimes found 159 Chapter 3 explains the different utilities of renewable and nonrenewable resources. Previous qualitative and quantitative studies in the field have also established different causal mechanisms for these two types of resources regarding their linkage with political instability. For example, neo-Malthusian arguments are mainly applied to renewable resources (Homer-Dixon and Blitt 1998 ; Homer-Dixon 1999), and the resource curse school always looks at nonrenewable resources that are ―lootable‖ (Ross 1999 ; Ross 2004). More information about this can be found in the literature review of Chapter 2. 245 to be a confounder in the nonrenewable resource scarcity-political instability association, 160 the weak and fragile statistical evidence is inadequate for reaching any conclusion regarding the relationships. Based on the statistical results, we revised the hypothesized causal diagram and display the significant relationships in Figure 13. Compared with the hypothesized causal diagram in Chapter 2 (Figure 2), nonrenewable resource abundance has been removed from the model of Figure 13 since the statistical results are not found to be significant. Moreover, social fractionalization and demographic characteristics are also taken off from the model for renewable resource. Hence, the revised causal diagram indicates that the hypothesized causal mechanisms better fit the renewable resource scarcity-political instability nexus. Moreover, economic development level and political institutions are crucial determinants of a country‘s adaptability to environmental challenges. What should be mentioned is that the revised model could be further adjusted if additional data analysis brings in new information. 160 The results are summarized at the end of Chapter 5 and Table 24, which show that some aspects of adaptability are confounders in the nonrenewable resource scarcity-political instability association when particular indicators of political instability are used. 246 Figure 13: The Revised Causal Diagram 247 As a complement of the quantitative analysis, this study also conducts case analysis on three worst-predicted cases to account for factors that are outside the core explanatory framework. It turns out that three African countries—Malawi, Mauritania and Tanzania—are identified as nonconforming cases. They all have high renewable resource scarcity and low political and economic adaptability simultaneously, but to a large extent are able to maintain their political stability. By reviewing the history of the three countries in the late 20 th century, this study has found that cultural contexts and international actors are two important factors that have contributed to their political stability. The non-confrontational traditional culture and pressures from the international community, especially Britain, France, the United States and international financial organizations, together helped to maintain these countries‘ political stability both in the earlier period of authoritarian rule and the later democratic transformation. Therefore, a more comprehensive causal mechanism might also consider these two variables when specific countries are investigated in more depth. The findings also illustrate the necessity of synthesizing both qualitative and quantitative research, since culture and history are usually considered as the ―thick‖ contexts which are difficult to address with statistical tools. 161 161 Political scientists have addressed the importance of these ―thick‖ variables and suggested ways to synthesize qualitative and quantitative methods. Important methodological articles and discussion include in Ragin (1987), King et al. (1994), Brady and Collier (2004), and George and Bennett (2005). 248 Implications for the Theoretical Debate The empirical and theoretical findings of this study have important implications on the debate among the three schools from Chapter 2. Neo-Malthusians, Neoclassical economists and Political ecologists hold quite different views about the connections between environmental scarcity and political instability. Introducing the crucial role of adaptability, this study has presented a new model that synthesizes the essential arguments of the three schools. This section revisits the theoretical perspectives, examining to what extent the findings support each theory, and then discusses their analytical and empirical limitations. Neo-Malthusians Inspired by the Malthusian belief in the connections between population and environmental catastrophe, Neo-Malthusians argue that rapid population growth, environmental degradation, and increasing resource scarcity impose high stress on the state‘s political stability and eventually result in violent conflict. The findings of this study suggest that neo-Malthusians are right in contending that resource scarcity causes conflict when each country‘s total amount of resources is considered. 162 However, their theories are limited in accounting for the crucial intervening variables and processes that explain the causal mechanisms in a more plausible way. 163 They also fail to identify the 162 This is supported by the findings in the auto-regression analysis of Chapter 4. 163 The intervening variables are the four aspects of adaptability. 249 distinction between renewable and nonrenewable resources in the analytical and theoretical framework. As discussed in Chapter 2, most neo-Malthusian arguments have been criticized for their demographic and environmental determinism, which takes for granted the causal linkage between population and environmental stress and violent conflict. The exaggeration of the causal importance of environmental factors thus makes them skim some crucial intervening variables and simplify the causal mechanisms. Most of the neo- Malthusians, such as Ehrlich (1968), Hardin (1968), the Club of Rome (1972), and Diamond (2005), focus more on the tragic effects of environmental scarcity, rather than how states can cope with these challenges based on their domestic conditions. This study initially proposed that economic development level, political institutions, social fractionalization and demographic characteristics are the critical intermediate variables, and eventually the first two factors have been confirmed to have significant impacts on the resource scarcity-political instability nexus. What should be noted is that the revised neo-Malthusians represented by Homer- Dixon (1999) and Bä chler (1998) have indeed addressed the conception of adaptation in their case studies. For example, Homer-Dixon and Blitt (1998:Ch1) suggest that a society‘s capability of adaptation is determined by its social technical ingenuity, which is reflected in institutions such as market efficiency and agricultural technologies. Bä chler (1998) also emphasizes the role of institutions, acknowledging that resource scarcity can impose challenge on political stability by causing or exacerbating social and political maldevelopment in the country. However, as Kahl (2006) points out, their account of 250 adaptation has done little more than providing a laundry list of intervening variables such as market failure and social friction. It remains underspecified how these intervening variables interact with environmental pressures to produce violent conflict. Drawing on the major variables proposed by the previous research, this study has been able to sort the list of variables into four categories and examine the interplay among them by PCA. Hence it drives further the discussion on the intervening variables by identifying a series of political and economic indicators as the determinants of adaptation. In addition, it has also figured out the threshold values of economic and political adaptabilities above which countries would be able to achieve successful adaptation. Another limitation of the neo-Malthusian studies is that most of them fail to distinguish the analytical framework of renewable resources from that of nonrenewable resources. As suggested by this and other empirical studies, the neo-Malthusian hypothesis that resource scarcity-generated grievances lead to violence better depicts the picture of renewable resources, but it does not work well for nonrenewable resources. As a matter of fact, this study has found little evidence for the impacts of nonrenewable resource scarcity on political instability. According to the discussion in Chapter 2 and Chapter 5, the difference can be partially explained by the distinct utilities of the two types of resources. 164 Although scholars like Homer-Dixon (1999) and Kahl (2006) note in their case studies that only renewable resources are considered, their works appear to 164 Most renewable resources are the basic supplies for human survival (i.e. drinking water, food), whereas most nonrenewable resources are either ―lootable‖ resources (i.e. diamond, silver, gold) or materials for industrial production (i.e. steel, petroleum). 251 be limited because of the absence of a discussion on nonrenewable resources and an account for the difference. Their research thus fails to take the opportunity to test the theoretical frameworks in a more comprehensive way. Neoclassical Economists Challenging neo-Malthusian theses on the relations between environmental scarcity and conflict, neoclassical economists can be approximately divided into two groups with different focus. The one composed of economic optimists confront with neo- Malthusians directly by contending that the human society is always able to adapt to resource challenges with the help of market mechanisms and technological development. The other one, called ―resource curse school‖ has raised concerns about abundance of valuable natural resources, which, according to them, can lead to economic stagnation, corruption and civil war. The findings of this study imply that the economic optimists are wise to take into account the adaptive capacities of markets and societies in the analytical framework. Neoclassical economists such as Lomborg (2001), Simon (1981, 1996) and Boserup (1965) have suggested that market mechanisms can reduce the demand for scarce resources and promote the adoption of new technologies which help humankind to adapt to the resource challenges. However, they appear to be overconfident about the market by assuming that the price signals are always timely and the market responses are never laggard or impeded. Our statistical analysis has suggested that the economic development level and the quality of political institutions are critical factors determining a country‘s 252 capability to handle resource challenges. While they emphasize the importance of economic factors, economist optimists do not pay adequate attention to the impacts of political institutions on environmental adaptability. According to our findings, variables measuring political institutions such as regime type, civil rights and political freedom, government effectiveness, regulatory quality, rule of law, control of corruption, and voice and accountability are also crucial determinants of a country‘s adaptive capacity to resource scarcity. At first, the quality of political institutions can facilitate or impede the building of market mechanism in a country. Secondly, political institutional arrangements may also influence the flow of information in the market system, and finally, the public and private sectors‘ responses to market signals are also influenced by political institutions. Therefore, economic optimists depict an oversimplified story by ignoring the importance of political institutions in their resource scarcity-adaptation thesis. Compared with the optimists, resource curse proponents incorporate a broader set of variables in their models. They have conducted a number of econometric analyses, accounting for the impacts of political, economic, social and demographic factors on the resource abundance-civil war nexus. According to them, the abundance rather than the scarcity of ―lootable‖ resources is considered as an important cause for civil war. Same as the neo-Malthusian model, the analytical and theoretical frameworks proposed by the resource curse school also suffer from narrowness by focusing solely on valuable nonrenewable resources such as oil, minerals, and diamonds. Their analysis thus does not directly confront the neo-Malthusian these by leaving aside renewable resources. 253 Moreover, even for nonrenewable resources, econometric studies of the resource curse school sometimes have identified different patterns for each resource. 165 In order to test the resource curse hypotheses, this study designed statistical analyses to examine the linkage between nonrenewable resource abundance and political instability, with each of the four aspects of adaptability as the intermediate variable. It turns out that inadequate evidence has been found for significant associations among the three variables. The results to some extent also explain why inconsistent findings have been reported by the school. 166 Although nonrenewable resource abundance might be associated with civil war in some way when each resource is considered separately, no clear patterns can be found when they are considered together. In some cases, weak evidence has suggested that nonrenewable resource scarcity, rather than abundance is more likely to relate to political instability. 167 Hence, it has been implied that the resource curse thesis may only apply to certain types of ―lootable‖ resources. A more complex model with more intervening variables might be required to better explain the impact of nonrenewable resource scarcity/abundance on political instability. Political Ecologists 165 Some the econometric studies of the resource curse school evaluate resources such as oil, diamond, and metals respectively. They find their model work well on some resources but not the others. Ross (2004) addresses this inconsistency when different resources are investigated. 166 Chapter 2 summarizes the inconsistent findings and discusses the major causes. 167 When testing the models for economic development level and political institutions, this study finds that nonrenewable resource scarcity, rather than abundance is more likely to relate to political instability. The analysis of social fractionalization and demographic characteristics does not produce significant results, but the results still suggest that the relations between resource abundance and political instability are positive if any linkage does exist. 254 In contrast to neo-Malthusians and neoclassical economists, political ecologists provide an alternative explanation which views scarcity as socially constructed. Rather than debating on the absolute amount of resources as an important cause of civil conflict, political ecologists suggest that the unequal distribution of resources is the main source of instability. They thus focus more on the political dimension of environmental change and human-environment interactions Since income inequality is believed to be a major indicator of resource distribution in societies, 168 this study takes into account the maldistribution of income both across and within country. Internationally, GDP per capita and GNI per capita account for the income gap between developed and developing countries and GNI index measures the income inequality within a country. According to the PCA and regression analysis, both international and domestic income equalities are crucial determinants of a country‘s adaptive capacity to environmental challenges, and wealthier countries with more equal income distribution are better at adapting to renewable resource challenges. The results to a large extent support the political ecologist argument on the political dimensions of the story. Resource scarcity does not only tell a story of the natural system but also the human system when it results in violence. The maldistribution of resources and income, which is viewed as socially constructed resource scarcity by political ecologists, does contribute to the crisis. 168 As discussed in Chapter 2, a number of studies, such as Muller and Seligson (1987 ; 1988), Timberlake and Williams (1987), Boswell and Dixon (1990), Schock (1996), Gissinger (2000) and Besanç on (2005) use income inequality as an indicator of resource distribution and examine its association with political conflict. 255 However, advocating that structures of the world system and political economy are always more important than natural conditions, political ecologists tend to overemphasize the influences of the political economic system. This study suggests that neo-Malthusians get much of the picture right by arguing that resource scarcity causes conflict in general. But the political ecologist perspective totally rejects the neo- Malthusian argument by ignoring the ―objective‖ measurement of resource scarcity. 169 Thus it does not acknowledge all the supply-induced scarcity and part of the demand- induced scarcity proposed by neo-Malthusians (Homer-Dixon 1999 ; Homer-Dixon and Blitt 1998). Moreover, as Kahl (2006:243) points out, political ecologists fail to clearly specify a causal mechanism for civil strife. According to Kahl (2006:243), they provide a laundry list of political and cultural variables, but have done little theoretical work and hypothesis testing. Policy Implications and Future Research Since the publication of Thomas Malthus‘ (1807) essay on population growth, the debate on the social effects of natural resource scarcity has been going on for over two centuries. With the evolvement of human society and rapid growth of population, the demand for natural resources has developed to a much larger scale. In the meantime, the advancement of technology driven by ingenuity also provides the humankind many more tools to utilize natural resources in more efficient ways. Therefore, the debate continues 169 As defined by this study, objective measurement of resource scarcity means measurement based on the absolute amount of resources per capita in each country, rather than the possession and consumption of the resources based on income distribution. For neo-Malthusians such as Homer-Dixon (1999), this should include the supply-induced scarcity, and part of the demand-induced scarcity, which merely results from the increase of population. 256 nowadays since we have both increased concerns about resource scarcity with the population explosion and confidence on the human ingenuity as the technology is developing at a high pace. In the policy arena and public discourse, increasing concerns have been raised as the resource constraints grow. Recently the National Intelligence Council released a new report, Global Trends 2025: A Transformed World (2008), in which a group of the World‘s best strategic thinkers are trying to identify the key dynamics that may shape the international system in the next twenty-five years. According to the strategic analysis, ―access to relatively secure and clean energy sources and management of chronic food and water shortage‖ is of crucial importance to a number of countries in the next fifteen to twenty years, as the world‘s population will increase over a billion by 2025 (National Intelligence Council 2008:41). Besides, the report suggests that the deepening of urbanization and the expansion of the middle class in the developing countries will also enlarge the resource demand. Moreover, the impending climate change is believed to be an important factor that may further complicate and exacerbate the already stressed resource issues. As temperatures and the sea level rise, the area of arable land and fresh water is likely to decrease, at the same time, the frequency and severity of extreme weather events and natural disasters such as floods, droughts, and tsunami will increase in many regions. In response to this long-lasting debate and its increasing salience today, the theoretical and empirical findings of this study imply some policy lessons. First of all, resource scarcity, especially renewable resource scarcity, indeed challenges the political 257 stability of developing countries. Since most developing countries have lower economic development levels and less mature political institutions, they are less capable to adapt to environmental stress produced by renewable resource scarcity. To maintain its political stability and avoid civil violence, the country should either lower its renewable resource scarcity or increase its economic and political adaptability. It appears difficult for developing countries to ease their renewable resource scarcity by reducing the demand, since their per capita resource consumption is already very low compared with developed countries. 170 In the meantime, they also need resources to achieve the economic modernization. Furthermore, it is less likely for most of them to adopt resource efficient technologies, giving their financial and technological conditions. Hence, the only way they might successfully decrease resource scarcity is keeping an eye on the population. Thus policies ought to be taken to prevent the explosion of population as the amount of resources available is limited. Besides, the findings of this study suggest that policy makers can also keep the political stability and prevent civil conflict of a country by increasing its economic and political adaptability. Specifically, the economic power of a country, which can be measured by a series of indicators such as GDP per capita and GNI per capita, is an important determinant of the country‘s capability to cope with renewable resource challenges. This to some extend support the arguments presented by developing countries in the negotiation of Kyoto Protocol that their top priority is to achieve economic 170 As Appendix D shows, most of the countries in the high scarcity group are developing countries. 258 modernization and then can take more responsibility for production of . 171 In addition, this study suggests that income equality is also an important economic factor that determines whether a country can successfully handle renewable resource challenges. According to political ecologists, a relative equal income distribution can decrease the socially-constructed scarcity. Moreover, the polarization of income distribution is an important indicator of an unhealthy economic structure, but it is usually ignored by countries during their industrialization process. Such a society often features high tensions between groups with distinct socio-economic status and the inequality also impedes the creativity and optimization of human ingenuity. 172 The establishment of healthy political institutions is another measure that can be taken to maintain political stability. As we see, a democratic regime with more political freedom and civil rights is the foundation of a healthy political system. And more indicators measuring different aspects of the government such as its effectiveness, regulations, rule of law, corruption control, and accountability should also be evaluated. Political adaptability to renewable resource challenges thus requires investigation to the synthesized governing power of the regime. Hence the establishment of democracy in some developing countries is not sufficient condition for them to avoid resource-induced 171 Oberthü r and Ott (1999:24-29) summarize the positions of developing countries in the negotiation of the Kyoto Protocol, including China and India. Gupta (2000) analyses the nature and key elements of developing countries‘ negotiating position. Toman et al. (2003:259-332) have collected some studies discussing developing countries, especially India‘s international climate policy. 172 A number of studies including Clarke (1992), Persson and Tabellini (1994), Alesina and Rodrik (1994), Perotti (1996) and Banerjee and Duflo (2003) find inequality is negatively correlated with economic growth. Zweimü ller (2000) explains the mechanisms in more detail. He argues that inequality has an impact on economic growth through its effect on ―the level and the dynamics of an innovator‘s demand,‖ which suggests that inequality impedes the exercise of human ingenuity. 259 political crisis. Actually some studies even suggest that transitional democracies are more prone to civil conflict than both mature democracies and hardcore autocracies. 173 Thus there is still a long way ahead for countries just begin its democratic transformation to get rid of the trap of civil conflict. In addition, the case analysis of Malawi, Mauritania and Tanzania implies that external factors also play an important role in sustaining a country‘s political stability. Using financial aid and political pressure as tools, donor states and international organizations such as IMF, the World Bank, and the United Nations can support the economic development of poor developing countries. In the meantime, they can also press for political transformation in these places and work as monitors for the transitional process. The statistical analysis shows that the hypothesized model does not work well for nonrenewable resources and thus the core thesis of the resource curse school is not supported. Future research on this topic might conduct more in-depth studies on nonrenewable resources by introducing more diverse measurements of nonrenewable resource scarcity/abundance. For example, each nonrenewable resource might be investigated in the model separately to see if it can work for some of them. Some studies have used the ratio of primary commodity exports on total GDP as an indicator of nonrenewable resource abundance. 174 The reasonableness of this indicator is a subject of controversy. This study does not utilize it because it does not comply with our intention 173 For example, Hegre et al. (2001) conclude that intermediate regimes are most prone to civil war by examining the period of 1816–1992. 174 For example, Collier and Heoffler (2004b) use that . 260 of measuring the ―objective‖ resource pressure. This indicator can be included in the future tests on nonrenewable resources. With more diverse data sources and measurements, future research should aim to identify an analytical model that explains the causal mechanisms between nonrenewable resource scarcity and political instability which is different from the one identified for renewable resource scarcity in this study. The statistical model in this study has tried to bring in a number of variables measuring scarcity from four aspects. However, the case analysis suggests that some factors may be still missing in the core explanatory framework, especially the so-called ―thick‖ variables such as historical and cultural elements. Since our analysis implies that a more comprehensive and accurate causal model on the linkages between resource scarcity and political instability should also consider these ―thick‖ factors, the synthesizing of qualitative and statistical analysis is critical for the advancement of the systematic studies in the field. That is to say, researchers in the field should make efforts to quantify those ―thick‖ variables in functional ways. Each variable might be dismantled into several dimensions and coded respectively. Then a composite score of multiple dimensions can be developed to address these concepts in a more accurate way. 261 Bibliography Abhijit, V. Banerjeeand Duflo Esther. 2003. "Inequality and Growth: What Can the Data Say?" Journal of Economic Growth no. 8 (3):267. Addi, Lahouari. 1998. "Algeria's Army, Algeria's Agony." Foreign Affairs no. 77 (4):44- 53. Addison, Tonyand S. Mansoob Murshed. 2002. "Credibility and Reputation in Peacemaking." Journal of Peace Research no. 39 (4):487-501. doi: 10.1177/0022343302039004007. Aghion, Philippe, Eve Caroliand Cecilia Garcí a-Peñ alosa. 1999. "Inequality and Economic Growth: The Perspective of the New Growth Theories." Journal of Economic Literature no. 37 (4):1615-1660. Agrawal, Nisha, Zafar Ahmed, Michael Meredand Roger Nord. 1993. Structural Adjustment, Economic Performance, and Aid dependency in Tanzania. IMF Working Paper. Ake, Claude. 1975. "A Definition of Political Stability." Comparative Politics no. 7 (2):271-283. Alesina, AFand R Perotti. 1993. Income distribution, political instability, and investment. NBER. Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlatand Romain Wacziarg. 2003. "Fractionalization." Journal of Economic Growth no. 8 (2):155- 194. Alesina, Alberto, Sule Özler, Nouriel Roubiniand Phillip Swagel. 1996. "Political instability and economic growth." Journal of Economic Growth no. 1 (2):189- 211. Alesina, Albertoand Dani Rodrik. 1994. "Distributive Politics and Economic Growth." The Quarterly Journal of Economics no. 109 (2):465-490. All is rather easily forgiven: A coup-maker becomes a civilian president. 2009. The Economist. Arcand, Jean-Louis, Patrick Guillaumontand Sylviane Guillaumont Jeanneney. 2000. "How to make a tragedy: on the alleged effect of ethnicity on growth." Journal of International Development no. 12 (7):925-938. 262 Bä chler, Gü nther. 1998. "Why environmental transformation causes violence: A synthesis." Environmental Change and Security Project Report no. 4:24-44. ______. 1999. Violence through Environmental Discrimination: Causes, Rwanda Arena, and Conflict Model: Kluwer Academic Publishers. Balassa, Bela A., Marcus Nolandand Institute for International Economics (U.S.). 1988. Japan in the World Economy. Washington, DC: Institute for International Economics. Banerjee, Abhijit V.and Esther Duflo. 2003. "Inequality and Growth: What Can the Data Say?" Journal of Economic Growth no. 8 (3):267-299. Barro, RJ. 2000. "Inequality and Growth in a Panel of Countries." Journal of Economic Growth no. 5 (1):5-32. Barro, Robert J. 1991. "Economic growth in a cross section of countries." The Quarterly Journal of Economics:407-443. Bates, Robert H. 1999. "Ethnicity, capital formation, and conflict." Social capital initiative working paper no. 12. Bates, Robert H., David L. Epstein, Jack A. Goldstone, Ted Robert Gurr, Barbara Harff, Colin H. Kahl, Kristen Knight, Marc A. Levy, Michael Lustik, Monty G. Marshall, Thomas M. Parris, Jay Ulfelderand Mark R. Woodward. 2003. Political Instability Task Force Report: Phase IV Findings. Belkin, Aaronand Evan Schofer. 2003. "Toward a Structural Understanding of Coup Risk." Journal of Conflict Resolution no. 47 (5):594-620. doi: 10.1177/0022002703258197. Benhabib, Jessand Mark Spiegel. 1994. "The role of human capital and political instability in economic development." Journal of Monetary economics no. 34:143-173. Besanç on, Marie L. 2005. "Relative Resources: Inequality in Ethnic Wars, Revolutions, and Genocides." Journal of Peace Research no. 42 (4):393-415. Biersack, A. 2006. "Reimagining Political Ecology: Culture/Power/History/Nature." Reimagining Political Ecology:3–42. Bigsten, Arne, Deogratias Mutalemwa, Yvonne Tsikataand Samuel Wangwe. 1999. Aid and Reform in Tanzania. Goteborg University and Economic and Social Research Foundation. 263 Binningsbø , Helga Malmin, Indra de Soysaand Nils Petter Gleditsch. 2007. "Green giant or straw man? Environmental pressure and civil conflict, 1961-99." Population and Environment no. 28 (6):337. Binswanger, Hans. 1978. "Induced innovation: technology, institutions, and development." In Induced Technical Change: Evolution of Thought, edited by HP Binswanger and VW Ruttan. Johns Hopkins University Press Baltimore. Bollen, Kenneth A. 1989. Structural Equations with Latent Variables, Wiley series in probability and mathematical statistics. Applied probability and statistics. New York: Wiley. Boserup, Esterand T. Paul Schultz. 1990. Economic and demographic relationships in development, Johns Hopkins studies in development. Baltimore: Johns Hopkins University Press. Boserup, Esther. 1965. The Conditions of Agricultural Growth: the Economics of Agrarian Change under Population Pressure. London: Earthscan Publications. Boswell, Terryand William J. Dixon. 1990. "Dependency and Rebellion: A Cross- National Analysis." American Sociological Review no. 55 (4):540-559. BoT, NBS, TICand ZIPA. 2002, 2004. Zanzibar Investment Report: Report on Foreign Private Investment in Zanzibar. Brady, Henry E.and David Collier. 2004. Rethinking social inquiry : diverse tools, shared standards. Lanham, Md.: Rowman & Littlefield. Bratton, Michaeland Nicolas Van de Walle. 1997. Democratic experiments in Africa : regime transitions in comparative perspective, Cambridge studies in comparative politics. Cambridge, U.K. ; New York, NY, USA: Cambridge University Press. Bryant, Raymond L.and Siné ad Bailey. 1997. Third World political ecology. London ; New York: Routledge. Calder, Kent E. 1988. "Review: Japanese Foreign Economic Policy Formation: Explaining the Reactive State." World Politics no. 40 (4):517-541. Carment, David, Patrick Jamesand Zeynep Taydas. 2006. Who Intervenes?: Ethnic Conflict and Interstate Crisis. Columbus: Ohio State University Press. Carson, Rachel, Lois Darlingand Louis Darling. 1962. Silent Spring. Boston, Cambridge, Mass.: Houghton Mifflin ;Riverside Press. 264 Cederman, Lars-Erikand L. U. C. Girardin. 2007. "Beyond Fractionalization: Mapping Ethnicity onto Nationalist Insurgencies." American Political Science Review no. 101 (01):173-185. doi: doi:10.1017/S0003055407070086. Central Intelligence Agency. "The World Factbook." In (accessed November, 2009). Clarke, George R. G. 1992. More Evidence on Income Distribution and Growth, Policy research working papers WPS 1064. Washington, D.C. (1818 H St. NW Washington DC 20433): Country Economics Dept., World Bank. Cole, John W.and Eric R. Wolf. 1974. The hidden frontier; ecology and ethnicity in an Alpine valley, Studies in social discontinuity. New York: Academic Press. Collier, Paul. 2000a. "Economic Causes of Civil Conflict and their Implications for Policy ". ______. 2000b. "Rebellion as a Quasi-Criminal Activity." Journal of Conflict Resolution no. 44 (6):839-853. doi: 10.1177/0022002700044006008. Collier, Pauland Ashish Garg. 1999. "On kin groups and wages in the Ghanaian labour market." Oxford Bulletin of Economics and Statistics no. 61 (2):133-151. Collier, Pauland Anke Hoeffler. 1998. "On economic causes of civil war." Oxford Economic Papers no. 50 (4):563-573. ______. 2000. Greed and Grievance in Civil War World Bank Policy Research Working Paper http://ssrn.com/abstract=630727. ______. 2001. Data Issues in the Study of Conflict. In The Conference on Data Collection on Armed Conflict. Uppsala. ______. 2002. "On the Incidence of Civil War in Africa." Journal of Conflict Resolution no. 46 (1):13-28. doi: 10.1177/0022002702046001002. ______. 2004a. "Greed and grievance in civil war." Oxf. Econ. Pap. no. 56 (4):563-595. doi: 10.1093/oep/gpf064. ______. 2004b. "Greed and Grievance in Civil War." Oxford Economic Papers no. 56. ______. 2005. "Resource Rents, Governance, and Conflict." The Journal of Conflict Resolution no. 49 (4):625-633. Collier, Paul, Anke Hoefflerand Mans Sö derbom 2004. "On the Duration of Civil War." Journal of Peace Research no. 41 (3):253-273. doi: 10.1177/0022343304043769. 265 Collier, Paul, Anke Hoefflerand Mans Soderbom. 2008. "Post-Conflict Risks." Journal of Peace Research no. 45 (4):461-478. doi: 10.1177/0022343308091356. Collier, Paul, Patrick Honohanand Karl Ove Moene. 2001. "Implications of Ethnic Diversity." Economic Policy no. 16 (32):129-166. Costain, W. Douglasand James P. Lester. 1995. "The Evolution of Environmentalism." In Environmental Politics and Policy : Theories and Evidence, edited by James P. Lester, xii, 386 p. Durham: Duke University Press. Cukierman, Alex, Sebastian Edwardsand Guido Tabellini. 1992. "Seigniorage and Political Instability." The American Economic Review no. 82 (3):537-555. de Soysa, Indra. 2002a. "Ecoviolence: shrinking pie, or honey pot?" Global Environmental Politics no. 2 (4):1-34. ______. 2002b. "Paradise Is a Bazaar? Greed, Creed, and Governance in Civil War, 1989-99." Journal of Peace Research no. 39 (4):395-416. Deininger, Kand L Squire. 1996. A new data set measuring income inequality. World Bank. Diamond, Jared M. 2005. Collapse : How Societies Choose to Fail or Succeed. New York: Viking. Dobzhansky, Theodosius. 1968. "Adaptness and Fitness." In Population biology and evolution; proceedings of the international symposium, June 7-9, 1967, Syracuse, New York, edited by Richard C. Lewontin and Syracuse University., 109-121. Syracuse, N.Y.: Syracuse University Press. Donge, Jan Kees van. 1995. "Kamuzu's Legacy: The Democratization of Malawi: Or Searching for the Rules of the Game in African Politics." African Affairs no. 94 (375):227-257. Doyle, Michael W.and Nicholas Sambanis. 2000. "International Peacebuilding: A Theoretical and Quantitative Analysis." The American Political Science Review no. 94 (4):779-801. Dunning, Thad. 2005. "Resource Dependence, Economic Performance, and Political Stability." Journal of Conflict Resolution no. 49 (4):451-482. doi: 10.1177/0022002705277521. Dunteman, George H. 1989. Principal Components Analysis, Sage university papers series. Quantitative applications in the social sciences no. 07-069. Newbury Park: Sage Publications. 266 Durham, William H. 1995. Political Ecology and Environmental Destruction in Latin America. In The Social Causes of Environmental Destruction in Latin America, edited by Michael Painter and William H. Durham. Ann Arbor: University of Michigan Press. Eckstein, Harryand Ted Robert Gurr. 1975. Patterns of Authority : a Structural Basis for Political Inquiry, Comparative Studies in Behavioral Science. New York: Wiley. Edwards, A. W. F. 2004. Cogwheels of the Mind : the Story of Venn Diagrams. Baltimore, Md.: Johns Hopkins University Press. Edwards, Sebastianand Guido Tabellini. 1994. "Political Instability, Political Weakness, and Inflation: An Empirical Analysis." In Advances in Econometrics: Sixth World Congress, edited by Christopher A. Sims. Cambridge, NY, Melbourne: Cambridge University Press. Ehrlich, Paul R. 1968. The Population Bomb, A Sierra Club-Ballantine book. New York,: Ballantine Books. Elbadawi, Ibrahimand Nicholas Sambanis. 2002. "How Much War Will We See? Explaining the Prevalence of Civil War." The Journal of Conflict Resolution no. 46 (3):307-334. Ellingsen, Tanja. 2000. "Colorful Community or Ethnic Witches' Brew? Multiethnicity and Domestic Conflict during and after the Cold War." The Journal of Conflict Resolution no. 44 (2):228-249. Empire's Children. Country Histories: Independence for Zanzibar 2007 [cited 2009-10- 23. Available from http://channel4.empireschildren.co.uk/category/chapters/index.php?chapter=472& cat=3. Eriksson, Mikaeland Peter Wallensteen. 2004. "Armed Conflict, 1989-2003." Journal of Peace Research no. 41 (5):625-636. doi: 10.1177/0022343304047568. Eriksson, Mikael, Peter Wallensteenand Margareta Sollenberg. 2003. "Armed Conflict, 1989-2002." Journal of Peace Research no. 40 (5):593-607. doi: 10.1177/00223433030405006. Esty, Daniel C., Jack A. Gladstone, Ted Robert Gurr, Barbara Harff, Marc Levy, Geoffrey D. Dabelko, Pamela T. Surkoand Alan N. Unger. 1998. The Statefailure Taskforce Report: Phase Il Findings. Esty, Daniel C., Jack A. Goldstone, Ted Robert Gurr, Pamela T. Surkoand Alan N. Unger. 1995. Working Papers: State Failure Task Force Report. 267 Esty, Daniel C., Jack Goldstone, Ted Robert Gurr, Barbara Harff, Pamela T. Surko, Alan N. Ungerand Robert Chen. 1998. "The State Failure Project: Early Warning Research for US Foreign Policy Planning." In Preventive Measures: Building Risk Assessment and Crisis Early Warning Systems, edited by John L. Davies and Ted Robert Gurr, 27–38. Boulder, CO and Totowa, NJ: Rowman and Littlefield. Ewing, B., A. Reed, S.M. Rizk, A. Galli, M. Wackernageland J. Kitzes. 2008. "Calculation Methodology for the National Footprint Accounts, 2008 Edition." In: Oakland: Global Footprint Network. Fearon, James D. 2005. "Primary Commodity Exports and Civil War." Journal of Conflict Resolution no. 49 (4):483-507. doi: 10.1177/0022002705277544. Fearon, James D. . 2002. Ethnic Structure and Cultural Diversity around the World. In the Annual Meetings of the American Political Science Association. Boston. Fearon, James D.and David D. Laitin. 2003a. Additional Tables for "Ethnicity, Insurgency, and Civil War". ______. 2003b. "Additional Tables for Ethnicity, Insurgency, and Civil War." Unpublished manuscript, Department of Political Science, Stanford University, Palo Alto, CA. ______. 2003c. "Ethnicity, Insurgency, and Civil War." The American Political Science Review no. 97 (1):75-90. Ferreira, M. Louisa. 1996. Poverty and Inequality During Structural Adjustment in Rural Tanzania. World Bank Policy Research Department Working Paper 1641. Finer, Samuel. E. 1976. The Man on Horseback : The Role of The Military in Politics. 2nd enlarged ed, Peregrine books. Harmondsworth ; Baltimore [etc*: Penguin. Flanigan, William H.and Edwin Fogelman. 1970. "Patterns of Political Violence in Comparative Historical Perspective." Comparative Politics no. 3 (1):1-20. Forster, Peter G. 1994. "Culture, Nationalism, and the Invention of Tradition in Malawi." The Journal of Modern African Studies no. 32 (3):477-497. Frayssinet, Fabiana. 2009. Brazil: Murder, Death Threats Amid Environmental Protests. http://www.ipsnews.net/news.asp?idnews=46981. Freedom House. Methodology: Freedom in the World 2009 [cited Nov 3, 2009. Available from http://www.freedomhouse.org/template.cfm?page=351&ana_page=341&year=20 08. 268 Gaddis, John Lewis. 1997. We Now Know: Rethinking Cold War History. Oxford, New York: Clarendon Press ; Oxford University Press. ______. 2006. The Cold War: A New History. London: Penguin Press. Gallopí n, Gilberto C. 2006. "Linkages between vulnerability, resilience, and adaptive capacity." Global Environmental Change no. 16 (3):293-303. George, Alexander L.and Andrew Bennett. 2005. Case Studies and Theory Development in the Social Sciences, BCSIA studies in international security. Cambridge, Mass.: MIT Press. Gerteiny, Alfred G. 1967. Mauritania, Praeger library of African affairs. London: Pall Mall Press. Gissinger, Rand NP Gleditsch. 2000. "Globalization and conflict: welfare, distribution, and political unrest." Journal of World Systems Research no. 5 (2):275-300. Gleditsch, Nils Petter. 1998. "Armed Conflict and The Environment: A Critique of the Literature." Journal of Peace Research no. 35 (3):381-400. doi: 10.1177/0022343398035003007. Gleditsch, Nils Petterand Bjorn Otto Sverdrup. 2002. "Democracy and the Environment." In Human Security and The Environment: International Comparisons, edited by Edward A. Page and Michael Redclift, 45. Edward Elgar Publishing. Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenbergand Havard Strand. 2002. "Armed Conflict 1946-2001: A New Dataset." Journal of Peace Research no. 39 (5):615-637. Goldstone, JA, RH Bates, TR Gurr, M Lustik, MG Marshall, J Ulfelderand M Woodward. 2005. A global forecasting model of political instability. Goldstone, Jack A., Robert H. Bates, David L. Epstein, Ted Robert Gurr, Michael B. Lustik, Monty G. Marshall, Jay Ulfelderand Mark Woodward. 2010. "A Global Model for Forecasting Political Instability." American Journal of Political Science no. 54 (1):190-208. Goldstone, Jack A., Ted Robert Gurr, Barbara Harff, Marc A. Levy, Monty G. Marshall, Robert H. Bates, David L. Epstein, Colin H. Kahl, Pamela T. Surko, John C. Ulfelderand Alan N. Unger. 2000. State Failure Task Force Report: Phase III Findings. Greene, William H. 2003. Econometric analysis. 5th ed. Upper Saddle River, N.J.: Prentice Hall. 269 Guha, Ramachandra. 2000. Environmentalism : A Global History, Longman world history series. New York: Longman. Guha, Ramachandraand Juan Martí nez Alier. 1997. Varieties of Environmentalism : Essays North and South. London: Earthscan Publications. Gupta, Dipak K. 1990. The economics of political violence : the effect of political instability on economic growth. New York: Praeger. ______. 2008. Understanding terrorism and political violence : the life cycle of birth, growth, transformation, and demise, Cass series on political violence. New York, NY: Routledge. Gupta, Joyeeta. 2000. "North-South Aspects of the Climate Change issue: towards a Negotiating Theory and Strategy for Developing Countries." International Journal of Sustainable Development no. 3 (2):115-135. Gurr, Ted Robert. 1996. "Minorities, nationalists, and ethnopolitical conflict." In Managing global chaos: Sources of and responses to international conflict, edited by Crocker Chester A., Fen Osler Hampson and Pamela R. Aall, 53-78. United States Institute of Peace. Haggard, Stephanand Robert R. Kaufman. 1995. The Political economy of democratic transitions. Princeton, N.J.: Princeton University Press. Hamilton, Kirkand World Bank. 2006. Where is the Wealth of nations? : Measuring Capital for the 21st Century. Washington, D.C.: The World Bank. Handloff, Robert Earl, Brian Dean Curranand Library of Congress. Federal Research Division. 1990. Mauritania, a country study. 2nd ed, Area handbook series. Washington, D.C.: The Division : For sale by the Supt. of Docs., U.S. G.P.O. Harbom, Lotta, Stina Hogbladhand Peter Wallensteen. 2006. "Armed Conflict and Peace Agreements." Journal of Peace Research no. 43 (5):617-631. doi: 10.1177/0022343306067613. Harbom, Lotta, Erik Melanderand Peter Wallensteen. 2008. "Dyadic Dimensions of Armed Conflict, 1946--2007." Journal of Peace Research no. 45 (5):697-710. doi: 10.1177/0022343308094331. Harbom, Lottaand Peter Wallensteen. 2005. "Armed Conflict and Its International Dimensions, 1946-2004." Journal of Peace Research no. 42 (5):623-635. doi: 10.1177/0022343305056238. 270 ______. 2007. "Armed Conflict, 1989-2006." Journal of Peace Research no. 44 (5):623- 634. doi: 10.1177/0022343307080859. Hardin, Garrett. 1968. "The Tragedy of the Commons." Science no. 162 (3859):1243- 1248. Hauge, Wencheand Tanja Ellingsen. 1998. "Beyond Environmental Scarcity: Causal Pathways to Conflict." Journal of Peace Research no. 35 (3):299-317. doi: 10.1177/0022343398035003003. Hayami, Yand VW Ruttan. 1985. Agricultural development: An international perspective. Baltimore: Johns Hopkins University Press Hegre, H. 2002. "Toward a democratic civil peace? Democracy, political change, and civil war, 1816?992." American Political Science Review no. 95 (01):33-48. Hegre, Harvard, Tanja Ellingsen, Scott Gatesand Nils Petter Gleditsch. 2001. "Toward a Democratic Civil Peace? Democracy, Political Change, and Civil War, 1816- 1992." The American Political Science Review no. 95 (1):33-48. Hegre, Havardand Nicholas Sambanis. 2006. "Sensitivity Analysis of Empirical Results on Civil War Onset." Journal of Conflict Resolution no. 50 (4):508-535. doi: 10.1177/0022002706289303. Henderson, Errol A.and J. David Singer. 2000. "Civil War in the Post-Colonial World, 1946-92." Journal of Peace Research no. 37 (3):275-299. doi: 10.1177/0022343300037003001. Hibbs, Douglas A. 1973. Mass political violence: a cross-national causal analysis, Comparative studies in behavioral science. New York,: Wiley. Hildyard, Nicholas. 1999. "Blood, babies, and the Social Roots of Conflict." In Ecology, Politics, and Violent Conflict, edited by Mohamed Suliman. London: Zed Books. Hirshleifer, Jack. 1995. "Theorizing about Confict." In Hand book of Defense Economics, edited by K. Hartley and T. Sandler, 165-89. Amsterdam: Elsevier Science. ______. 2001. The dark side of the force : economic foundations of conflict theory. Cambridge, UK ; New York: Cambridge University Press. Hodges, Tony. 1983. Western Sahara : The Roots of A Desert War. Westport, Conn.: L. Hill. Homer-Dixon, TFand J Blitt. 1998. Ecoviolence: Links among environment, population and security. Lanham, MD: Rowman & Littlefield Pub Inc. 271 Homer-Dixon, Thomas F. 1991. "On the Threshold: Environmental Changes as Causes of Acute Conflict." International Security no. 16 (2):76-116. ______. 1994. "Environmental Scarcities and Violent Conflict: Evidence from Cases." International Security no. 19 (1):5-40. ______. 1995a. "Environmental Scarcities, State Capacity, and Civil Violence." Bulletin of the American Academy of Arts and Sciences no. 48 (7):26-33. ______. 1995b. "The Ingenuity Gap: Can Poor Countries Adapt to Resource Scarcity?" Population and Development Review no. 21 (3):587-612. ______. 1999. Environment, Scarcity, and Violence. Princeton, N.J.: Princeton University Press. ______. 2000. The ingenuity gap. 1st ed. New York: Knopf. ______. 2006. The upside of down : catastrophe, creativity, and the renewal of civilization. Washington: Island Press. Human Rights Watch/Africa. 1994. Mauritania's Campaign of Terror: State Sponsored Repression of Black Africans: New York: Human Rights Watch. Huntington, Samuel P. 1991. The Third Wave: Democratization in the Late Twentieth Century, The Julian J. Rothbaum distinguished lecture series v. 4. Norman: University of Oklahoma Press. ______. 1996. The clash of civilizations and the remaking of world order. New York: Simon & Schuster. Hyden, Goranand Bo Karlstrom. 1993. "Structural adjustment as a policy process: The case of Tanzania." World Development no. 21 (9):1395-1404. Inglehart, Ronald. 1997. Modernization and postmodernization : cultural, economic, and political change in 43 societies. Princeton, N.J.: Princeton University Press. Jaggers, Keithand Ted Robert Gurr. 1995. "Tracking Democracy's Third Wave with the Polity III Data." Journal of Peace Research no. 32 (4):469-482. Jolliffe, Ian T. 2002. Principal Component Analysis: Secaucus, NJ, USA: Springer- Verlag New York, Incorporated. Jourde, Cé dric. 2001. "Ethnicity, Democratization, and Political Dramas: Insights into Ethnic Politics in Mauritania." African Issues no. 29 (1/2):26-30. 272 Justino, Patricia. 2009. "Poverty and Violent Conflict: A Micro-Level Perspective on the Causes and Duration of Warfare." Journal of Peace Research no. 46 (3):315-333. doi: 10.1177/0022343309102655. Kahl, Colin. 1998. "Population Growth, Environmental Degradation, and State- Sponsored Violence: The Case of Kenya, 1991-93." International Security no. 23 (2):80-119. ______. 2002. "Demographic change, natural resources and violence: The current debate." Journal of International Affairs no. 56 (1):257. ______. 2006. States, Scarcity, and Civil Strife in the Developing World. Princeton: Princeton University Press. Kaiser, Henry. 1970. "A second generation little jiffy." Psychometrika no. 35 (4):401- 415. Kaiser, Henry F. 1981. "A Revised Measure of Sampling Adequacy for Factor-Analytic Data Matrices." Educational and Psychological Measurement no. 41 (2):379-381. doi: 10.1177/001316448104100216. Kalinga, Owen J. M. 1996. "Resistance, Politics of Protest, and Mass Nationalism in Colonial Malawi, 1950-1960. A Reconsideration (Ré sistance, contestation politique et nationalisme de masse au Malawi (1950-1960). Une ré é valuation)." Cahiers d'É tudes Africaines no. 36 (143):443-454. Kaspin, Deborah. 1995. "The Politics of Ethnicity in Malawi's Democratic Transition." The Journal of Modern African Studies no. 33 (4):595-620. Kaufmann, Daniel , Aart Kraayand Massimo Mastruzzi. 2009a. Governance Matters VIII: Aggregate and Individual Governance Indicators, 1996-2008. World Bank Policy Research Working Paper No. 4978. Kaufmann, Daniel, Aart Kraayand Massimo Mastruzzi. 2009. Governance Matters 2009: Learning From Over a Decade of the Worldwide Governance Indicators. The Brookings Institution 2009b [cited August 26 2009]. Available from http://www.brookings.edu/opinions/2009/0629_governance_indicators_kaufmann .aspx. Kaufmann, Daniel, Aart Kraayand Pablo Zoido-Lobaton. 1999. Aggregating Governance Indicators World Bank Policy Research Department Working Paper No. 2195. Khalidi, Walid. 1991. "The Gulf Crisis: Origins and Consequences." Journal of Palestine Studies no. 20 (2):5-28. 273 Kim, Quee-Young. 1992. "Review: The Economics of Political Violence: The Effect of Political Instability on Economic " Contemporary Sociology no. 21 (1):45. King, Gand L Zeng. 2001a. "Improving Forecasts of State Failure." World Politics:623- 658. King, Gary, Robert O. Keohaneand Sidney Verba. 1994. Designing Social Inquiry : Scientific Inference in Qualitative Research. Princeton, N.J.: Princeton University Press. King, Garyand Langche Zeng. 2001b. "Explaining Rare Events in International Relations." International Organization no. 55 (03):693-715. doi: doi:10.1162/00208180152507597. Kitzes, J, A Galli, SM Rizk, A Reedand M Wackernagel. 2008. Guidebook to the national footprint accounts, 2008 Edition. Oakland (CA), Global Footprint Network. Kleinbaum, DG, LL Kupperand KE Muller. 1998. Applied regression analysis and other multivariable methods. 3th ed: PWS Publishing Co. Boston, MA, USA. Kline, Benjamin. 2007. First along the river : a brief history of the U.S. environmental movement. 3rd ed. Lanham, Md.: Rowman & Littlefield. Kydd, Jonathanand Robert Christiansen. 1982. "Structural change in Malawi since independence: Consequences of a development strategy based on large-scale agriculture." World Development no. 10 (5):355-375. Lapido, Adamolekun, Kulemeka Noeland Laleye Mouftaou. 1997. "Political transition, economic liberalization and civil service reform in Malawi." Public Administration & Development (1986-1998) no. 17 (2):209. Larose, Daniel T. . 2006. Data mining methods and models: Hoboken, NJ: Wiley-IEEE Press. Legum, Colin. 1982. Africa contemporary record; annual survey and documents. New York etc.: Africana Pub. Co. etc. Lichbach, Mark Irving. 1989. "An Evaluation of 'Does Economic Inequality Breed Political Conflict?' Studies." World Politics no. 41 (4):431-470. Lijphart, Arend. 1977. Democracy in plural societies : a comparative exploration. New Haven: Yale University Press. ______. 1999. Patterns of democracy: Government forms and performance in thirty-six countries. New Haven: Yale University Press. 274 Linz, Juan J.and Alfred C. Stepan. 1996. Problems of democratic transition and consolidation : southern Europe, South America, and post-communist Europe. Baltimore: Johns Hopkins University Press. Little, Roderick J. A.and Donald B. Rubin. 1987. Statistical Analysis with Missing Data, Wiley series in probability and mathematical statistics. Applied probability and statistics. New York: Wiley. Lomborg, Bjø rn. 2001. The Skeptical Environmentalist : Measuring the Real State of the World. Cambridge ; New York: Cambridge University Press. Londregan, John B.and Keith T. Poole. 1990. "Poverty, The Coup Trap, and the Seizure of Executive Power." World Politics no. 42 (2):151-183. Lydon, Ghislaine. 2005. "Slavery, Exchange and Islamic Law: A Glimpse from the Archives of Mali and Mauritania." African Economic History (33):117-148. MacCulloch, Robert. 2004. "The Impact of Income on the Taste for Revolt." American Journal of Political Science no. 48 (4):830-848. Magistro, John V. 1993. "Crossing over: Ethinicity and Transboundary Conflict in the Senegal River Valley (Ethnicité et conflit frontalier dans la vallé e du fleuve Sé né gal)." Cahiers d'É tudes Africaines no. 33 (130):201-232. Malthus, Thomas. R. 1807. An essay on the principle of population. 4th ed. London,: Printed for J. Johnson by T. Bensley. Marcus, Marvinand Henryk Minc. 1988. Introduction to Linear Algebra. New York: Dover Publications. Marianne, Marty. 2002. "Mauritania: Political Parties, Neo-patrimonialsm and Democracy." Democratization no. 9 (3):92-108. Marshall, Marshall G., Ted Robert Gurrand Barbara Harff Harff. 2009. PITF Problem Set Codebook. Marshall, Monty G.and Keith Jaggers. 2007. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2006. Dataset Users‘ Manual. Center for Global Policy George Mason University. 19. Martí nez, Helda. 2009. Colombia: Gold vs Preservation in the Central Mountains. Global Issues (July 13). Mattar, Philip. 1994. "The PLO and the Gulf Crisis." Middle East Journal no. 48 (1):31- 46. 275 McCracken, John. 1998. "Democracy and Nationalism in Historical Perspective: The Case of Malawi." African Affairs no. 97 (387):231-249. McKnight, PE, KM McKnight, AJ Figueredoand S Sidani. 2007. Missing data: A gentle introduction: The Guilford Press. Meadows, Donella H.and Club of Rome. 1972. The Limits to Growth: A Report for the Club of Rome's Project on the Predicament of Mankind. New York,: Universe Books. Meadows, Donella H., Dennis L. Meadowsand Jø rgen Randers. 1992. Beyond the limits : confronting global collapse, envisioning a sustainable future. Post Mills, Vt.: Chelsea Green Pub. Co. Meadows, Donella H., Jø rgen Randersand Dennis L. Meadows. 2004. The limits to growth : the 30-year update. White River Junction, Vt: Chelsea Green Publishing Company. Midlarsky, Manus I. 1998. "Democracy and the Environment: An Empirical Assessment." Journal of Peace Research no. 35 (3):341-361. Mitchell, Maura. 2002. ""Living Our Faith:" The Lenten Pastoral Letter of the Bishops of Malawi and the Shift to Multiparty Democracy, 1992-1993." Journal for the Scientific Study of Religion no. 41 (1):5-18. Moene, Karl Oveand Michael Wallerstein. 2001. "Inequality, Social Insurance, and Redistribution." The American Political Science Review no. 95 (4):859-874. Moore, Donald S. 1996. "Marxism, Culture, and political Ecology: Environmental Struggles in Zimbabwe‘s Eastern Highlands." In Liberation Ecologies : Environment, Development, Social movements, edited by Richard Peet and Michael Watts, xii, 273 p. London ; New York: Routledge. Mortimer, Robert. 1996. "Islamists, Soldiers, and Democrats: The Second Algerian War." Middle East Journal no. 50 (1):18-39. Muganda, Anna. 2004. Tanzania‘s Economic Reforms-and Lessons Learned. Paper read at "Scaling Up Poverty Reduction: A Global Learning Process and Conference", May 25–27, 2004, at Shanghai. Muller, Edward N. 1988. "Inequality, Repression, and Violence: Issues of Theory and Research Design." American Sociological Review no. 53 (5):799-806. Muller, Edward N.and Mitchell A. Seligson. 1987. "Inequality and Insurgency." The American Political Science Review no. 81 (2):425-451. 276 Muller, Edward N., Mitchell A. Seligson, Hung-der Fuand Manus I. Midlarsky. 1989. "Land Inequality and Political Violence." The American Political Science Review no. 83 (2):577-596. Mushi, Samuel S.and Rwekaza S. Mukandala. 1997. Multiparty Democracy in Transition: Tanzania's 1995 General Elections. N'Diaye, Boubacar. 2006. "Mauritania, August 2005: Justice and Democracy, or Just Another Coup?" African Affairs no. 105 (420):421-441. National Intelligence Council, WASHINGTON DC. 2008. "Global Trends 2025: A Transformed World." Newell, Jonathan. 1995. "'A Moment of Truth'? The Church and Political Change in Malawi, 1992." The Journal of Modern African Studies no. 33 (2):243-262. Ngowi, Honest Prosper. 2009. "Economic Development and Change in Tanzania Since Independence: The Political Leadership Factor." African Journal of Political Science and International Relations no. 3 (4):259-267. O'Kane, Rosemary H. T. 1987. The Likelihood of Coups. Aldershot ; Brookfield, USA: Avebury. Oberthü r, Sebastianand Hermann Ott. 1999. The Kyoto Protocol : International Climate Policy for the 21st Century. New York: Springer. Ohlsson, Leif. 2000. Livelihood Conflicts: Linking Poverty and Environment as Causes of Conflict. Stockholm: Environmental Policy Unit, Swedish International Development Cooperation Agency. Olson, Mancur. 1982. The rise and decline of nations : economic growth, stagflation, and social rigidities. New Haven: Yale University Press. Ostrom, Elinor. 1990. Governing the commons : the evolution of institutions for collective action, The Political economy of institutions and decisions. Cambridge ; New York: Cambridge University Press. Ostrom, Elinor, Joanna Burger, Christopher B. Field, Richard B. Norgaardand David Policansky. 1999. "Revisiting the Commons: Local Lessons, Global Challenges." Science no. 284 (5412):278-282. Ostrom, Elinor, Roy Gardnerand James Walker. 1994. Rules, games, and common-pool resources. Ann Arbor: University of Michigan Press. 277 Ostrom, Elinorand National Research Council (U.S.). Committee on the Human Dimensions of Global Change. 2002. The drama of the commons. Washington, DC: National Academy Press. Ould-Mey, Mohameden. 1998. "Structural adjustment Programs and Democratization in Africa: the Case of Mauritan." In MultipartyD emocracya nd Political Change: Constraints to Democratization in Africa, edited by John Mukum Mbaku and Julius O. Ihonvbere, p.49. Aldershot, England and Brookfield, VT: Ashgate. Ozler, Sand G Tabellini. 1991. External debt and political instability. NBER. Pazzanita, Anthony G. 1992. "Mauritania's Foreign Policy: The Search for Protection." The Journal of Modern African Studies no. 30 (2):281-304. ______. 1996. "The Origins and Evolution of Mauritania's Second Republic." The Journal of Modern African Studies no. 34 (4):575-596. ______. 1999. "Political Transition in Mauritania: Problems and Prospects." Middle East Journal no. 53 (1):44-58. Peluso, Nancy Leeand Michael Watts. 2001. Violent Environments. Ithaca: Cornell University Press. Percival, Valerieand Thomas F. Homer-Dixon. 1998a. "The Case of South Africa." In Ecoviolence : links among environment, population and security, edited by Thomas F. Homer-Dixon and Jessica Blitt, 109-146. Lanham, MD: Rowman & Littlefield. ______. 1998b. "Environmental Scarcity and Violent Conflict: The Case of South Africa." Journal of Peace Research no. 35 (3):279-298. Perotti, Roberto. 1996. "Growth, income distribution, and democracy: What the data say." Journal of Economic Growth no. 1 (2):149-187. Persson, Torstenand Guido Tabellini. 1994. "Is Inequality Harmful for Growth?" The American Economic Review no. 84 (3):600-621. Pike, John G. 1969. Malawi: a political and economic history, Pall Mall library of African affairs. London,: Pall Mall P. Przeworski, Adam. 1991. Democracy and the market : political and economic reforms in Eastern Europe and Latin America, Studies in rationality and social change. Cambridge ; New York: Cambridge University Press. 278 Ragin, Charles. 2004. "Turning the Tables: How Case-Oriented Research Challenges Variable-Oriented Research." In Rethinking Social Inquiry : Diverse Tools, shared standards, edited by Henry E. Brady and David Collier, xx, 362 p. Lanham, Md.: Rowman & Littlefield. Ragin, Charles C. 1987. The Comparative Method : Moving beyond Qualitative and Quantitative Strategies. Berkeley: University of California Press. Raleigh, Clionadhand Henrik Urdal. 2007. "Climate change, environmental degradation and armed conflict." Political Geography no. 26 (6):674-694. Rapkin, David P.and William P. Avery. 1986. "World Markets and Political Instability within Less Developed Countries." Cooperation and Conflict no. 21 (2):99-117. doi: 10.1177/001083678602100203. Raubenheimer, J. 2004. "An item selection procedure to maximise scale reliability and validity." SA Journal of Industrial Psychology no. 30 (4). Rees, William E., Sharon Bailey, University of British Columbia. Centre for Human Settlements.and University of British Columbia. School of Community and Regional Planning. 1989. Planning for sustainable development : a resource book. Vancouver, B.C., Canada: UBC Centre for Human Settlements. Research and Education for Democracy in Tanzania (REDET) Project. 1997. Political Culture and Popular Participation in Tanzania. Dar es Salaam: Mkuki na Nyota Publ. . Reynal-Querol, Marta. 2002. "Ethnicity, Political Systems, and Civil Wars." The Journal of Conflict Resolution no. 46 (1):29-54. ______. 2005. "Does democracy preempt civil wars?" European Journal of Political Economy no. 21 (2):445-465 Robbins, Paul. 2004. Political ecology: A critical introduction: Wiley-Blackwell. Roberts, Hugh. 1995. "Algeria's Ruinous Impasse and the Honourable Way out." International Affairs (Royal Institute of International Affairs 1944-) no. 71 (2):247-267. Roeder, Philip G. 2010. Ethnolinguistic fractionalization (ELF) indices, 1961 and 1985. In http//: weber. ucsd. edu\~ proeder\ elf. htm>, accessed January. Ron, Parker. 1991. "The Senegal-Mauritania Conflict of 1989: A Fragile Equilibrium." The Journal of Modern African Studies no. 29 (1):155-171. 279 Ross, Michael L. 2004. "What Do We Know about Natural Resources and Civil War?" Journal of Peace Research no. 41 (3):337-356. Ross, ML. 1999. "The political economy of the resource curse." World Politics:297-322. Rotberg, Robert I. 1965. The rise of nationalism in Central Africa; the making of Malawi and Zambia, 1873-1964. Cambridge,: Harvard University Press. Rubin, Donald B. 1976. "Inference and Missing Data." Biometrika no. 63 (3):581-592. Rummel, R. J. 1967. "Understanding Factor Analysis." The Journal of Conflict Resolution no. 11 (4):444-480. ______. 1995. "Democracy, Power, Genocide, and Mass Murder." The Journal of Conflict Resolution no. 39 (1):3-26. Ruskey, Frankand Mark Weston. 1997. "A survey of Venn diagrams." Electronic Journal of Combinatorics no. 4. Sambanis, Nicholas. 2001. "Do Ethnic and Nonethnic Civil Wars Have the Same Causes?: A Theoretical and Empirical Inquiry (Part 1)." Journal of Conflict Resolution no. 45 (3):259-282. doi: 10.1177/0022002701045003001. ______. 2004. "What Is Civil War? Conceptual and Empirical Complexities of an Operational Definition." The Journal of Conflict Resolution no. 48 (6):814-858. Sarkees, Meredith Reidand Phil Schafer. 2000. "The Correlates of War Data On War: an Update To 1997." Conflict Management and Peace Science no. 18 (1):123-144. doi: 10.1177/073889420001800105. Sarris, Alexander H.and Platon Tinios. 1995. "Consumption and poverty in Tanzania in 1976 and 1991: A comparison using survey data." World Development no. 23 (8):1401-1419. Sayigh, Yezid. 1991. "The Gulf Crisis: Why the Arab Regional Order Failed." International Affairs (Royal Institute of International Affairs 1944-) no. 67 (3):487-507. Schafer, Judi L. 1997. Analysis of Incomplete Multivariate Data: Chapman & Hall/CRC. Scheffer, Judi L. 2002. "Dealing With Missing Data." Research Letters in the Information and Mathematical Sciences no. 3 (1):153-160. Schock, Kurt. 1996. "A Conjunctural Model of Political Conflict: The Impact of Political Opportunities on the Relationship between Economic Inequality and Violent Political Conflict." The Journal of Conflict Resolution no. 40 (1):98-133. 280 Schreurs, Miranda A. 2002. Environmental politics in Japan, Germany, and the United States. Cambridge, UK ; New York: Cambridge University Press. Scruggs, Lyle. 2003. Sustaining Abundance: Environmental Performance in Industrial Democracies, Cambridge studies in comparative politics. Cambridge, UK ; New York: Cambridge University Press. Seddon, David. 1996. "The Political Economy of Mauritania: An Introduction." Review of African Political Economy no. 23 (68):197-214. Shin, Doh Chull. 1994. "On the Third Wave of Democratization: A Synthesis and Evaluation of Recent Theory and Research." World Politics no. 47 (1):135-170. Short, Philip. 1974. Banda. London, Boston,: Routledge & Kegan Paul. Simon, Julian Lincoln. 1981. The Ultimate Resource. Princeton, N.J.: Princeton University Press. ______. 1996. The ultimate resource 2. [Rev. ed. Princton, N.J.: Princeton University Press. Singer, J. Davidand Melvin Small. 1972. The Wages of War, 1816-1965: A Statistical Handbook. New York,: Wiley. ______. 1994. "Correlates of War Project: International and Civil War Data, 1816-1992." Ann Arbor, MI: ICPSR no. 9905. Skinner, Annabel. 2005. Tanzania & Zanzibar. 2nd ed: Cadogan Guides. Small, Melvinand J. David Singer. 1982. Resort to Arms : International and Civil Wars, 1816-1980. [2nd ed. Beverly Hills, Calif.: Sage Publications. Smit, Barryand Johanna Wandel. 2006. "Adaptation, adaptive capacity and vulnerability." Global Environmental Change no. 16 (3):282-292. Sorli, Mirjam E., Nils Petter Gleditschand Havard Strand. 2005. "Why Is There So Much Conflict in the Middle East?" Journal of Conflict Resolution no. 49 (1):141-165. doi: 10.1177/0022002704270824. Soviet Union. Glavnoe upravlenie geodezii i kartografii., Solomon Il ich Bruk, V. S. Apenchenkoand Institut *etnografii imeni N.N. Miklukho-Makla*i*a. 1964. Atlas narodov mira. Moskva: Glavnoe upravlenie geodezii i kartografii Gosudarstvennogo geologicheskogo komiteta SSSR : Institut *etnografii i m. N.N. Miklukho-Makla*i*a Akademii Nauk SSSR,. 281 SPSS Inc. SPSS Missing Values 17.0. http://support.spss.com/ProductsExt/SPSS/Documentation/SPSSforWindows/inde x.html. Suliman, Mohamed. 1999. Ecology, Politics and Violent Conflict. London ; New York: Zed Books. Sullivan, Colin. 2009. "Future Scarcity Seen Sparking Local Conflicts, Not Full-Scale Wars " The New York Times, December 23, 2009. Summers, Robertand Alan Heston. 1991. "The Penn World Table (Mark 5): an expanded set of international comparisons, 1950-1988." The Quarterly Journal of Economics no. 106 (2):327-368. Swatuk, Larry A. 2004. Environmental Security in Practice: Transboundary Natural Resources Management in Southern Africa. Paper read at presentation in Section 31 of the Pan-European Conference on International Relations, at The Hague. Tabachnick, Barbara G.and Linda S. Fidell. 2001. Using Multivariate Statistics. 4th ed: MA: Allyn & Bacon. Theisen, Ole Magnus 2008. "Blood and Soil? Resource Scarcity and Internal Armed Conflict Revisited." Journal of Peace Research no. 45 (6):801-818. doi: 10.1177/0022343308096157. Thompson, Virginiaand Richard Adloff. 1980. The Western Saharans: background to conflict. London and Totowa, NJ.: Rowman & Littlefield Publishers. Timberlake, Michaeland Kirk R. Williams. 1987. "Structural Position in the World System, Inequality, and Political Instability." Journal of Political and Military Sociology no. 15 (1):1-15. Toman, Michael A., Ujjayant Chakravortyand Shreekant Gupta. 2003. India and global climate change : perspectives on economics and policy from a developing country. Washington, DC: Resources for the Future. Tripp, Aili Mari. 2000. "Political Reform in Tanzania: The Struggle for Associational Autonomy." Comparative Politics no. 32 (2):191-214. Turner, Graham M. 2008. "A comparison of The Limits to Growth with 30 years of reality." Global Environmental Change no. 18 (3):397-411. Uppsala Conflict Data Program. 2008. "UCDP/PRIO Armed Conflict Dataset Codebook." Uppsala Conflict Data Program (UCDP) and Center for the Study of Civil Wars, International Peace Research Institute, Oslo (PRIO). 282 Urdal, Henrik. 2005. "People vs. Malthus: Population Pressure, Environmental Degradation, and Armed Conflict Revisited." Journal of Peace Research no. 42 (4):417-434. doi: 10.1177/0022343305054089. ______. 2008. "Population, Resources, and Political Violence: A Subnational Study of India, 1956-2002." Journal of Conflict Resolution no. 52 (4):590-617. doi: 10.1177/0022002708316741. Vayda, Andrew P.and Bradley B. Walters. 1999. "Against political ecology." Human Ecology no. 27 (1):167-179. Venieris, Yiannis P.and Dipak K. Gupta. 1986. "Income Distribution and Sociopolitical Instability as Determinants of Savings: A Cross-Sectional Model." The Journal of Political Economy no. 94 (4):873-883. Venieris, Yiannis P.and Samuel M Sperling. 1994. "Saving and Sociopolitical Instability in Developed and Less-Developed Nations." Journal of economic development no. 19 (2). Venn, John. 1880. "On the Diagrammatic and Mechanical Representation of Propositions and Reasonings." Philosophical Magazine Series 5 no. 10 (59):1 - 18. Venter, Denis. 1995. "Malawi: the Transition to Multi-party Politics." In Democracy and political change in Sub-Saharan Africa, edited by John A. Wiseman, xii, 238 p. London ; New York: Routledge. Walter, Barbara F. 2004. "Does Conflict Beget Conflict? Explaining Recurring Civil War." Journal of Peace Research no. 41 (3):371-388. Wilson, Amrit. 1989a. "Revolution and the 'Foreign Hand' in Tanzania." Economic and Political Weekly no. 24 (19):1031-1033. ______. 1989b. US foreign policy and revolution : the creation of Tanzania. London, Winchester, MA, USA: Pluto Press Wolf, Eric. 1972. "Ownership and political ecology." Anthropological Quarterly:201- 205. Wooldridge, Jeffrey M. 2009. Introductory econometrics : a modern approach. 4th ed. Mason, OH: South Western, Cengage Learning. 283 World Bank, Policy and Economics Team – Environment Department. A Guide to Valuing Natural Resources Wealth. Zweimü ller, Josef. 2000. "Schumpeterian Entrepreneurs Meet Engel's Law: The Impact of Inequality on Innovation-Driven Growth." Journal of Economic Growth no. 5 (2):185-206. 284 Appendix A: Regression Analyses Based on Three Scarcity Groups The following results are obtained in model tests based on three scarcity groups instead of two. The threshold values are the 33th and the 66th percentiles of the economic adaptability component scores. Economic Conditions Regressions using all three indicators of political instability suggest that economic adaptability group is a significant confounder in the relationship between renewable resource scarcity and political instability. As Table 27 shows, renewable resource is a significant predictor of political instability based on the simple regression. But it is no long significant when the dummy variable of economic group is added to the model. And the coefficient has changed more than 20% from the unadjusted model. Since economic adaptability group is both significantly associated with renewable resource and political instability according to the correlation test, it is a confounder of the association between renewable resource and political instability. Thus the relationship among the three variables can be represented in Figure 14. Furthermore, the regression test of nonrenewable resources shows that economic adaptability is a significant confounder for two of the political instability indicators (PITF and World Bank indicators), but not for the other one. To further examine the confounding relationship, it conducts simple regression between renewable resources and political instability for each group. Table 28 summarizes the results for each political instability indicator. It shows that the association 285 between renewable resources scarcity and political instability is significant for the middle group (Group 2), but insignificant for the low and high economic adaptability group. The comparison of R-squares also demonstrates the different patterns among groups. Eyeball test of the scatter plots in Figure 14 can also tell us the more linear trend in the middle economic adaptability group. But for nonrenewable resources, such group different is not found based on Table 28. Political, Demographic and Social Adaptability With similar regression tests, the study finds that political adaptability is a confounder of the association between renewable resource and political instability. Moreover, the tests of nonrenewable resources and political instability show that political adaptability is also a significant confounder for two of the indicators. When each group is investigated, the association between renewable resources scarcity and political instability is significant for the low political scarcity group, but insignificant for the middle and high groups. The comparison of R-squares and eyeball test based on the scatter plots in Figure 14 also confirm the different patterns. But a clear pattern for nonrenewable resources and political instability for any group are not found. For social fractionalization, about 10%-15% change of the coefficient of renewable resources has been found using three different indicators of political instability. But social fractionalization is not significantly correlated with nonrenewable resources, which implies no confounding relation is possible. As for the association between renewable resources and political instability within each group, the statistical 286 results indicate that the association is significant for the high group (Group 1), but insignificant for the low and high social adaptability group. The comparison of R-squares and eyeball test of the scatter plots in Figure 3 can also tell us the different patterns. But no clear pattern can be detected for any of the three groups about the relation between nonrenewable resources and political instability. For demographic characteristics, the coefficient has changed greatly from the unadjusted model. Hence, demographic adaptability is a confounder of the association between renewable resource and political instability. Similar to the results for economic adaptability, the regression test of nonrenewable resources suggests that demographic adaptability is a significant confounder for two of the political instability indicators (PITF and World Bank indicators), but not for the other one (civil wars). However, when it investigates the association between renewable resources and political instability within each group, no clear pattern can be detected for any of the three groups. Neither can it find a clear pattern for nonrenewable resources. 287 Table 27: Testing for Interactions and Confounding Effects of Economic Adaptability (Three Scarcity Groups) Civil Wars PITF Political Instability WB Political Instability Model1 Model2 Model3 Model1 Model2 Model3 Model1 Model2 Model3 Intercept .707*** 0.676*** 1.596* 31.347*** 29.961*** 75.925** -2.689*** -2.482*** -3.617 Renewable Resources per capita (ln) -.069** -.041 -.165 -3.087** -1.816 -7.986* .317*** .127 .279 Adaptability Group -.094** -.449 -4.228*** -21.974* .631*** 1.069 Interaction (Renew*Adapt Group) .047 2.368 -.058 Explain R 2 .054 .107 .118 .056 .121 .136 .105 .360 .361 P value for F test .004 .000 .000 .000 .000 .000 .000 .000 .000 Intercept .204*** .389*** .326*** 9.963*** 17.682*** 18.126*** -.465*** -1.598*** -1.462*** Subsoil Assets per capita (ln) -.008 .001 .019 -.598* -.202 -.329 .055* -.003 -.042 Adaptability Group -.113 -.079 -4.704*** -4.940** .690*** .618*** Interaction (Subsoil*Adapt Group) -.009 .060 .018 Explain R 2 .009 .092 .098 .026 .106 .107 .039 .345 .348 P value for F test .232 .000 .001 .041 .000 .000 .012 .012 .000 *p<.05 ** p < .01***p<.001 288 Table 28: The different patterns among low, middle and high economic adaptability groups Civil Wars PITF Political Instability WB Political Instability Adaptability Group Group1 Group2 Group3 Group1 Group2 Group3 Group1 Group2 Group3 Intercept .896 1.743*** .073 46.125 70.381** 3.302 -2.117 -4.147*** .076 Renewable Resources per capita (ln) .-.087 -.196** -.004 -4.707 -7.856* -.198 .174 .480** .055 Explain R 2 .015 .161 .003 .032 .113 .003 .011 .174 .012 Intercept .230** .237** .053 13.372*** 8.576* 2.836* -.832*** -.339 .467** Subsoil Assets per capita (ln) .011 -.006 -.003 -.650 .053 -.206 -.003 -.009 .011 Explain R 2 .007 .003 .007 .017 .000 .021 .000 .002 .003 *p<0.05 ** p < 0.01***p<0.001 289 Figure 14: Scatter Plots for Three Adaptability Groups Economic Conditions Political Failure Social Fractionalization Demographic Characteristics Low Middle High 290 Appendix B: Residual Analysis for Outliers Suppose there are n observations, the ith residual reflects the difference between the observed and the predicted values. It is calculated as 175 ̂ : the observed value ̂ : the predicted value There are several types of residuals based on different ways of calculate them. Standardized Residuals Studentized Residuals √ √ Jackknife residuals √ √ √ 175 The formulas are ased on Kleinbaum, Kupper and Muller (1998). 291 Appendix C: List of Civil Wars and State Failures from 1970 to 1999 Table 29: List of Civil Wars by Fearon and Laitin Country War years Case name AFGHANISTAN 1978-92 Mujahideen AFGHANISTAN 1992- v.Taliban ALGERIA 1992- FIS ANGOLA 1975- UNITA ANGOLA 1992- FLEC(Cabinda) ARGENTINA 1973-77 ERP/Montoneros AZERBAIJAN 1992-94 Nagorno-Karabagh BANGLADESH 1976-97 ChittagongHills/ShantiBahini BOSNIA 1992-95 Rep.Srpska/Croats BURUNDI 1972-72 Hutuuprising BURUNDI 1988-88 Org.massacresonbothsides BURUNDI 1993- Hutugroupsv.govt CAMBODIA 1970-75 FUNK CAMBODIA 1978-92 KhmerRouge,FUNCINPEC,etc CENTRALAFRICANREP. 1996-97 Factionalfighting CHAD 1965- FROLINAT,various... CHAD 1994-98 RebelsinSouth CHINA 1991- Xinjiang COLOMBIA 1963- FARC,ELN,etc CONGO 1998-99 Factionalfighting CROATIA 1992-95 Krajina CYPRUS 1974-74 Cypriots,Turkey DEM.REP.CONGO 1977-78 FLNC DEM.REP.CONGO 1996-97 AFDL(Kabila) DEM.REP.CONGO 1998- RCD,etcv.govt DJIBOUTI 1993-94 FRUD ELSALVADOR 1979-92 FMLN ETHIOPIA 1974-92 Eritrea,Tigray,etc. ETHIOPIA 1997- ALF,ARDUF(Afars) GEORGIA 1992-94 Abkhazia GUATEMALA 1968-96 URNG,various GUINEABISSAU 1998-99 Mil.faction HAITI 1991-95 Mil.coup INDIA 1952- N.Eastrebels 292 Table 29, continued INDIA 1982-93 Sikhs INDIA 1989- Kashmir INDONESIA 1965- OPM(WestPapua) INDONESIA 1975-99 E.Timor INDONESIA 1991- GAM(Aceh) IRAN 1978-79 Khomeini IRAN 1979-93 KDPI(Kurds) IRAQ 1961-74 KDP,PUK(Kurds) JORDAN 1970-70 Fedeyeen/Syriav.govt LAOS 1960-73 PathetLao LEBANON 1975-90 variousmilitias LIBERIA 1989-96 NPFL(Taylor),INPFL(Johnson) MALI 1989-94 Tuaregs MOLDOVA 1992-92 DniestrRep. MOROCCO 1975-88 Polisario MOZAMBIQUE 1976-95 RENAMO NEPAL 1997- CPN-M/UPF(Maoists) NICARAGUA 1978-79 FSLN NICARAGUA 1981-88 Contras NIGERIA 1967-70 Biafra PAKISTAN 1971-71 Bangladesh PAKISTAN 1973-77 Baluchistan PAKISTAN 1993-99 MQM:Sindhisv.Mohajirs PAPUAN.G. 1988-98 BRA(Bougainville) PERU 1981-95 SenderoLuminoso PHILIPPINES 1968- MNLF,MILF PHILIPPINES 1972-94 NPA PORTUGAL 1961-75 Angola PORTUGAL 1962-74 Guinea-Bissau PORTUGAL 1964-74 Mozambique RUSSIA 1994-96 Chechnya RUSSIA 1999- ChechnyaII RWANDA 1990- RPF,genocide SENEGAL 1989- MFDC(Casamance) SIERRALEONE 1991- RUF,AFRC,etc. SOMALIA 1981-91 SSDF,SNM(Isaaqs) SOMALIA 1991- post-Barrewar SOUTHAFRICA 1983-94 ANC,PAC,Azapo 293 Table 29, continued SRILANKA 1971-71 JVP SRILANKA 1983- LTTE,etc. SRILANKA 1987-89 JVPII SUDAN 1963-72 AnyaNya SUDAN 1983- SPLA,etc. TAJIKISTAN 1992-97 UTO TURKEY 1977-80 Militia-izedpartypolitics TURKEY 1984-99 PKK UGANDA 1981-87 NRA,etc. UGANDA 1993- LRA,WestNile,etc. UK 1969-99 IRA VIETNAM,S. 1960-75 NLF YEMEN 1994-94 SouthYemen YEMENPEOP.REP. 1986-87 FactionofSocialistParty YUGOSLAVIA 1991-91 Croatia/Krajina ZIMBABWE 1972-79 ZANU,ZAPU ZIMBABWE 1983-87 Ndebeleguer‘s Source: Fearon and Laitin (2003b) 294 Table 30: List of Civil Wars by Correlates of War Project War Name YrBeg1 YrEnd1 YrBeg2 YrEnd2 Laos vs. Pathet Lao of 1963 1963 1973 Sudan vs. Anya Nya 1963 1972 Guatemala vs. Indians 1966 1972 Chad vs. Frolinat of 1966 1966 1971 Nigeria vs. Biafrans 1967 1970 Burma vs. Ethnic Rebels 1968 1980 Thailand vs. Communists 1970 1973 Cambodia vs. Khmer Rouge of 1970 1970 1975 Jordan vs. Palestinians 1970 1970 Guatemala vs. Leftists of 1970 1970 1971 Pakistan vs. Bengalis 1971 1971 Sri Lanka vs. Janatha Vimukthi-JVP 1971 1971 Philippines vs. Moros 1972 1980 Burundi vs. Hutu of 1972 1972 1972 Philippines vs. NPA 1972 1992 Zimbabwe vs. Patriotic Front 1972 1979 Pakistan vs. Baluchi Rebels 1973 1977 Chile vs. Pinochet Led Rebels 1973 1973 Ethiopia vs. Eritrean Rebels 1974 1991 Iraq vs. Kurds of 1974 1974 1975 Lebanon vs. Leftists of 1975 1975 1990 Angola vs. UNITA of 1975 1975 1991 Ethiopia vs. Somali Rebels 1976 1977 1978 1983 Guatemala vs. Leftists of 1978 1978 1984 Ethiopia vs. Tigrean Liberation Front 1978 1991 Afghanistan vs. Mujahedin 1978 1992 Iran vs. Anti-Shah Coalition 1978 1979 Nicaragua vs. Sandinistas 1978 1979 Cambodia vs. Khmer Rouge of 1978 1978 1991 El Salvador vs. Salvadorean Democratic Front 1979 1992 Mozambique vs. Renamo 1979 1992 Chad vs. Frolinat of 1980 1980 1988 Nigeria vs. Muslim Fundamentalists of 1980 1980 1981 Uganda vs. National Resistance Army 1980 1988 Iran vs. Mujaheddin 1981 1982 295 Table 30, continued Peru vs. Shining Path 1982 1995 Nicaragua vs. Contras 1982 1990 Somalia vs. Clan Factions 1982 1997 Burma vs. Kachin Rebels 1983 1995 Sri Lanka vs. Tamils 1983 -888 Sudan vs. SPLA-Garang Faction 1983 -888 Nigeria vs. Muslim Fundamentalists of 1984 1984 1984 Colombia vs. M-19 & Drug Lords 1984 -888 Iraq vs. Kurds & Shiites 1985 1993 India vs. Sikhs & Kashmiros 1985 -888 Yemen People's Republic vs. Leftist Factions 1986 1986 Sri Lanka vs. JVP 1987 1989 Burundi vs. Hutu of 1988 1988 1988 Liberia vs. Anti-Doe Rebels 1989 1990 Rumania vs. Anti-Ceaucescu Rebels 1989 1989 Rwanda vs. Tutsi 1990 1993 Sierra Leone vs. RUF 1991 1996 Yugoslavia/Serbia vs. Croatians 1991 1992 Turkey vs. Kurds 1991 -888 Burundi vs. Tutsi Supremacists 1991 1991 Georgia vs. Gamsakurdia & Abkaz 1991 1994 Azerbaijan vs. Nagorno-Karabakh 1991 1994 Bosnia/Herzogovina vs. Serbs 1992 1995 Algeria vs. Islamic Rebels 1992 -888 Tadzhikistan vs. Popular Democratic Army 1992 1997 Liberia vs. NPFL & ULIMO 1992 1995 Angola vs. UNITA of 1992 1992 1994 Zaire vs. Rebels 1993 1993 Burundi vs. Hutu of 1993 1993 -888 Cambodia vs. Khmer Rouge of 1993 1993 1997 Russia vs. Chechens 1994 1996 Rwanda vs. Patriotic Front 1994 1994 Yemen vs. South Yemen 1994 1994 Pakistan vs. Mohajir 1994 1995 Uganda vs. Lords Resistance Army 1996 -888 296 Table 30, continued Liberia vs. National Patriotic Forces 1996 1996 Iraq vs. KDP Kurds 1996 1996 Zaire vs. Kabila-ADFL 1996 1997 Congo vs. Denis Sassou Nguemo 1997 1997 297 Table 31: List of Civil Wars by Collier and Hoeffler Country Start of the war End of the war Afghanistan Apr-78 Feb-92 Afghanistan May-92 Ongoing Algeria May-91 Ongoing Angola Feb-61 Nov-75 Angola Nov-75 May-91 Angola Sep-92 Ongoing Azerbaijan Apr-91 Oct-94 Bosnia Mar-92 Nov-95 Burma/Myanmar 68 Oct-80 Burma/Myanmar Feb-83 Jul-95 Burundi Apr-72 Dec-73 Burundi Aug-88 Aug-88 Burundi Nov-91 ongoing Cambodia Mar-70 Oct-91 Chad Mar-80 Aug-88 Columbia Apr-84 ongoing Congo 97 Oct-97 Cyprus Jul-74 Aug-74 El Salvador Oct-79 Jan-92 Ethiopia Jul-74 May-91 Georgia Jun-91 Dec-93 Guatemala Jul-66 Jul-72 Guatemala Mar-78 Mar-84 Guinea-Bissau Dec-62 Dec-74 India 84 94 Indonesia Jun-75 Sep-82 Iran Mar-74 Mar-75 Iran Sep-78 Dec-79 Iran Jun-81 May-82 Iraq Jul-74 Mar-75 Iraq Jan-85 Dec-92 Jordan Sep-70 Sep-70 Laos Jul-60 Feb-73 Lebanon May-75 Sep-92 Liberia Dec-89 Nov-91 Liberia Oct-92 Nov-96 Morocco Oct-75 Nov-89 298 Table 31, continued Mozambique Oct-64 Nov-75 Mozambique Jul-76 Oct-92 Nicaragua Oct-78 Jul-79 Nicaragua Mar-82 Apr-90 Nigeria Jan-66 Jan-70 Nigeria Dec-80 Aug-84 Pakistan Mar-71 Dec-71 Pakistan Jan-73 Jul-77 Peru Mar-82 Dec-96 Philippines Sep-72 Dec-96 Romania Dec-89 Dec-89 Russia Dec-94 Aug-96 Russia Sep-99 Ongoing Rwanda Oct-90 Jul-94 Sierra Leone Mar-91 Nov-96 Sierra Leone May-97 Jul-99 Somalia Apr-82 May-88 Somalia May-88 Dec-92 Sri Lanka Apr-71 May-71 Sri Lanka Jul-83 ongoing Sudan Oct-63 Feb-72 Sudan Jul-83 ongoing Tajikistan Apr-92 Dec-94 Turkey Jul-91 ongoing Uganda Oct-80 Apr-88 Vietnam Jan-60 Apr-75 Yemen May-90 Oct-94 Yemen, People's Rep. Jan-86 Jan-86 Yugoslavia Apr-90 Jan-92 Yugoslavia Oct-98 Apr-99 Dem. Rep. of Congo Sep-91 Dec-96 Dem. Rep. of Congo Sep-97 Sep-99 Zimbabwe Dec-72 Dec-79 Collier and Hoeffler (2001, 2004b) 299 Table 32: List of Civil Wars by Doyle and Sambanis Country name Year Start Year End Conflict Afghanistan 1978 1992 Mujahideen, PDPA Afghanistan 1992 1996 Taliban v. Burhanuddin Rabbani Afghanistan 1996 2001 United Front v. Taliban Algeria 1992 Ongoing FIS, AIS, GIA, GSPC Angola 1975 1991 UNITA Angola 1992 1994 UNITA Angola 1997 2002 UNITA Angola 1994 1999 Cabinda; FLEC Argentina 1975 1977 Montoneros, ERP, Dirty War Azerbaijan 1991 1994 Nagorno-Karabakh Bangladesh 1974 1997 Chittagong Hills/Shanti Bahini Bosnia 1992 1995 Rep. Srpska/Croats Burundi 1972 1972 Hutu uprising Burundi 1988 1988 Org. massacres on both sides Burundi 1991 Ongoing Hutu groups v. govt Cambodia 1970 1975 FUNK; Khmer Cambodia 1975 1991 Khmer Rouge, FUNCINPEC, etc Central African Republic 1996 1997 Factional fighting Chad 1965 1979 FROLINAT, various ... Chad 1980 1994 FARF; FROLINAT Chad 1994 1997 FARF; FROLINAT Colombia 1978 Ongoing FARC, ELN, drug cartels, etc Congo – Brazzaville 1993 1997 Lissouba v. Sassou-Nguesso Congo – Brazzaville 1998 1999 Cobras v. Ninjas Congo-Zaire 1977 1978 FLNC; Shabba 1 & 2 Congo-Zaire 1996 1997 AFDL (Kabila) Congo-Zaire 1998 2001 RCD, etc v. govt Croatia 1992 1995 Krajina, Medak, Western Slavonia Cyprus 1974 1974 TCs; GCs; Turkish invasion Djibouti 1991 1994 FRUD El Salvador 1979 1992 FMLN Egypt 1994 1997 Gamaat Islamiya; Islamic Jihad Ethiopia 1974 1991 Eritrean war of independence Ethiopia 1978 1991 Ideological; Tigrean Ethiopia 1976 1988 Ogaden; Somalis 300 Table 32, continued Georgia 1991 1992 South Ossetia Georgia 1992 1994 Abkhazia (& Gamsakhurdia) Guatemala 1966 1972 Communists; Guatemala 1978 1994 Communists; Indigenous Guinea-Bissau 1998 1999 Vieira v. Mane mutiny Haiti 1991 1995 Cedras v. Aristide India 1989 Ongoing Kashmir India 1984 1993 Sikhs India 1989 Ongoing Naxalites (CPI-M; PWG; MCC) India 1990 Ongoing Assam; Northeast States Indonesia 1976 1978 OPM (West Papua) Indonesia 1975 1999 East Timor Indonesia 1990 1991 Aceh Indonesia 1999 2002 Aceh Iran 1978 1979 Khomeini Iran 1979 1984 KDPI (Kurds) Iraq 1961 1970 KDP, PUK (Kurds) Iraq 1974 1975 KDP, PUK (Kurds) Iraq 1985 1996 Kurds; Anfal Iraq 1991 1993 Shiite uprising Israel 1987 1997 Intifada; Palestinian conflict Jordan 1970 1971 Fedeyeen/Syria v. govt Kenya 1991 1993 Rift valley ethnic violence Laos 1960 1973 Pathet Lao Lebanon 1975 1991 Aoun; militias; PLO; Israel Liberia 1989 1990 Doe v. rebels Liberia 1992 1997 NPLF; ULIMO; NPF; LPC; LDF Liberia 1999 Ongoing anti-Taylor resistance Mali 1990 1995 Tuaregs; Maurs Moldova 1991 1992 Transdniestria Morocco/Western Sahara 1975 1991 Polisario Mozambique 1976 1992 RENAMO; FRELIMO Myanmar/Burma 1948 1988 Communist insurgency Myanmar/Burma 1960 1995 various ethnic groups; Karen rebellion 2 Namibia 1973 1989 SWAPO; SWANU; SWATF Nepal 1996 Ongoing CPN-M/UPF (Maoists) Nicaragua 1978 1979 FSLN Nigeria 1967 1970 Biafra 301 Table 32, continued Nicaragua 1981 1990 Contras & Miskitos Nigeria 1980 1985 Muslims; Maitatsine rebellion Oman 1971 1975 Dhofar rebellion Pakistan 1971 1971 Bangladesh secession Pakistan 1973 1977 Baluchistan Pakistan 1994 1999 MQM:Sindhis v. Mohajirs Papua New Guinea 1988 1998 BRA (Bougainville) Peru 1980 1996 Sendero Luminoso, Tupac Amaru Philippines 1972 1992 NPA Philippines 1971 Ongoing MNLF, MILF Russia 1994 1996 Chechnya 1 Russia 1999 Ongoing Chechnya 2 Rwanda 1990 1993 Hutu vs. Tutsi groups Rwanda 1994 1994 RPF; genocide Senegal 1989 1999 MFDC (Casamance) Sierra Leone 1991 1996 RUF, AFRC, etc. Sierra Leone 1997 2001 post-Koroma coup violence Somalia 1988 1991 SSDF, SNM (Isaaqs) Somalia 1991 Ongoing post-Barre war South Africa 1976 1994 ANC, PAC, Azapo Sri Lanka 1971 1971 JVP Sri Lanka 1983 2002 LTTE, etc. Sri Lanka 1987 1989 JVP II Sudan 1963 1972 Anya Nya Sudan 1983 2002 SPLM, SPLA, NDA, AnyanyaII Syria 1979 1982 Muslim Brotherhood Tajikistan 1992 1997 Popular Democratic Army; UTO Thailand 1966 1982 Communists (CPT) Turkey 1984 1999 PKK (Kurds) Uganda 1978 1979 Tanzanian war Uganda 1981 1987 NRA/Museveni, etc Uganda 1990 1992 Kony (pre-LRA) Uganda 1995 Ongoing LRA, West Nile, ADF, etc. United Kingdom 1971 1998 Northern Ireland Vietnam 1960 1975 NLF Yemen 1994 1994 South Yemen Yemen AR 1962 1970 Royalists Yemen PR 1986 1986 Faction of Socialist Party 302 Table 32, continued Yugoslavia 1991 1991 Croatia/Krajina Yugoslavia 1998 1999 Kosovo Zimbabwe 1972 1979 ZANU, ZAPU Zimbabwe 1983 1987 Ndebele guerillas Doyle and Sambanis (2000) 303 Table 33: UCDP/PRIO Armed Conflict Dataset (Version 4-2008) Country Estimate of annual number of civil wars Afghanistan 0.73 Albania 0.00 Algeria 0.30 Angola 1.12 Argentina 0.13 Armenia 0.00 Australia 0.00 Austria 0.00 Azerbaijan 0.56 Bahamas 0.00 Bahrain 0.00 Bangladesh 0.62 Barbados 0.00 Belarus (Byelorussia) 0.00 Belgium 0.00 Belize 0.00 Benin 0.00 Bhutan 0.00 Bolivia 0.00 Bosnia-Herzegovina 1.13 Botswana 0.00 Brazil 0.00 Brunei 0.00 Bulgaria 0.00 Burkina Faso (Upper Volta) 0.03 Burundi 0.27 Cambodia (Kampuchea) 0.90 Cameroon 0.03 Canada 0.00 Cape Verde 0.00 Central African Republic 0.00 Chad 0.77 Chile 0.03 China 0.00 Colombia 1.00 Comoros 0.08 304 Table 33, continued Congo 0.17 Congo, Democratic Republic of 0.20 Costa Rica 0.00 Cote D‘Ivoire 0.00 Croatia 0.33 Cuba 0.00 Cyprus 0.00 Czech Republic 0.00 Czechoslovakia 0.00 Denmark 0.00 Djibouti 0.22 Dominican Republic 0.00 Ecuador 0.00 Egypt 0.20 El Salvador 0.47 Equatorial Guinea 0.03 Eritrea 0.29 Estonia 0.00 Ethiopia 2.37 Fiji 0.00 Finland 0.00 France 0.00 Gabon 0.00 Gambia 0.03 Georgia 0.67 German Democratic Republic 0.00 German Federal Republic 0.00 Ghana 0.07 Greece 0.00 Guatemala 0.87 Guinea 0.00 Guinea-Bissau 0.08 Guyana 0.00 Haiti 0.07 Honduras 0.00 Hungary 0.00 Iceland 0.00 305 Table 33, continued India 2.93 Indonesia 0.80 Iran (Persia) 0.90 Iraq 1.13 Ireland 0.00 Israel 1.23 Italy/Sardinia 0.00 Jamaica 0.00 Japan 0.00 Jordan 0.00 Kazakhstan 0.00 Kenya 0.03 Korea, People's Republic of 0.00 Korea, Republic of 0.00 Kuwait 0.00 Kyrgyz Republic 0.00 Laos 0.20 Latvia 0.00 Lebanon 0.30 Lesotho 0.03 Liberia 0.27 Libya 0.00 Lithuania 0.00 Luxembourg 0.00 Macedonia (Former Yugoslav Republic of) 0.00 Madagascar (Malagasy) 0.03 Malawi 0.00 Malaysia 0.10 Maldives 0.00 Mali 0.07 Malta 0.00 Mauritania 0.13 Mauritius 0.00 Mexico 0.07 Moldova 0.11 Mongolia 0.00 306 Table 33, continued Morocco 0.53 Mozambique 0.64 Myanmar (Burma) 4.10 Namibia 0.00 Nepal 0.13 Netherlands 0.00 New Zealand 0.00 Nicaragua 0.37 Niger 0.17 Nigeria 0.03 Norway 0.00 Oman 0.13 Pakistan 0.27 Panama 0.03 Papua New Guinea 0.28 Paraguay 0.03 Peru 0.63 Philippines 1.87 Poland 0.00 Portugal 0.00 Qatar 0.00 Rumania 0.03 Russia (Soviet Union) 0.30 Rwanda 0.27 Saudi Arabia 0.03 Senegal 0.23 Sierra Leone 0.30 Singapore 0.00 Slovakia 0.00 Slovenia 0.00 Solomon Islands 0.00 Somalia 0.50 South Africa 0.87 Spain 0.17 Sri Lanka (Ceylon) 0.63 Sudan 0.73 Surinam 0.12 307 Table 33, continued Swaziland 0.00 Sweden 0.00 Switzerland 0.00 Syria 0.13 Taiwan 0.00 Tajikistan 0.67 Tanzania/Tanganyika 0.00 Thailand 0.30 Togo 0.07 Trinidad and Tobago 0.03 Tunisia 0.03 Turkey/Ottoman Empire 0.60 Turkmenistan 0.00 Uganda 0.77 Ukraine 0.00 United Arab Emirates 0.00 United Kingdom 0.73 United States of America 0.00 Uruguay 0.03 Uzbekistan 0.00 Venezuela 0.03 Vietnam, Democratic Republic of 0.00 Vietnam, Republic of 0.00 Yemen (Arab Republic of Yemen) 0.17 Yemen, People's Republic of 0.05 Yugoslavia (Serbia) 0.13 Zambia 0.00 Zimbabwe (Rhodesia) 0.23 308 Table 34: Political Instability Task Force coding of Political Failure Country No. of Years with adverse regime change No. of Years with revolutionary wars No. of Years ethnic wars No. of Years with genoPoliticides Afghanistan 7 23 8 15 Albania 1 1 0 0 Algeria 1 9 0 0 Andorra 0 0 0 0 Angola 6 25 25 22 Antigua & Barbuda 0 0 0 0 Argentina 1 0 0 5 Armenia 2 0 0 0 Australia 0 0 0 0 Austria 0 0 0 0 Azerbaijan 3 0 7 0 Bahamas 0 0 0 0 Bahrain 0 0 0 0 Bangladesh 2 0 16 0 Barbados 0 0 0 0 Belarus 2 0 0 0 Belgium 0 0 0 0 Belize 0 0 0 0 Benin 1 0 0 0 Bhutan 0 0 0 0 Bolivia 0 0 0 0 Bosnia and Herzegovina 4 0 4 4 Botswana 0 0 0 0 Brazil 0 0 0 0 Brunei 0 0 0 0 Bulgaria 0 0 0 0 Burkina Faso 1 0 0 0 Burma 0 2 30 1 Burundi 4 0 13 6 Cambodia 3 19 0 5 Cameroon 0 0 0 0 Canada 0 0 0 0 Cape Verde 0 0 0 0 Central African Republic 0 0 0 0 Chad 6 0 25 0 309 Table 34, continued Chile 1 0 0 4 China 0 1 11 6 Colombia 0 16 0 0 Comoros 4 0 0 0 Congo-Brazzaville 1 3 0 0 Congo-Kinshasa 8 4 10 3 Costa Rica 0 0 0 0 Cote d'Ivoire 0 0 0 0 Croatia 0 0 5 0 Cuba 0 0 0 0 Cyprus 1 0 1 0 Czech Republic 0 0 0 0 Denmark 0 0 0 0 Djibouti 0 0 0 0 Dominica 0 0 0 0 Dominican Republic 0 0 0 0 East Timor 0 0 0 0 Ecuador 3 0 0 0 Egypt 0 8 0 0 El Salvador 1 14 0 10 Equatorial Guinea 0 0 0 10 Eritrea 0 0 0 0 Estonia 0 0 0 0 Ethiopia 5 17 25 4 Fiji 1 0 0 0 Finland 0 0 0 0 France 0 0 0 0 Gabon 0 0 0 0 Gambia, The 1 0 0 0 Georgia 0 2 3 0 Germany 0 0 0 0 Ghana 2 0 0 0 Greece 0 0 0 0 Grenada 0 0 0 0 Guatemala 0 27 20 13 Guinea 0 0 0 0 Guinea-Bissau 2 2 0 0 Guyana 3 0 0 0 Haiti 2 0 0 0 Honduras 0 0 0 0 Hungary 0 0 0 0 310 Table 34, continued Iceland 0 0 0 0 India 0 0 23 0 Indonesia 0 2 28 18 Iran 4 6 7 12 Iraq 0 0 20 10 Ireland 0 0 0 0 Israel 0 0 13 0 Italy 0 0 0 0 Jamaica 0 0 0 0 Japan 0 0 0 0 Jordan 0 2 0 0 Kazakhstan 0 0 0 0 Kenya 0 0 3 0 Korea, North 0 0 0 0 Korea, South 1 0 0 0 Kuwait 0 0 0 0 Kyrgyzstan 0 0 0 0 Laos 6 10 10 0 Latvia 0 0 0 0 Lebanon 16 0 17 0 Lesotho 3 1 0 0 Liberia 7 6 0 0 Libya 0 0 0 0 Liechtenstein 0 0 0 0 Lithuania 0 0 0 0 Luxembourg 0 0 0 0 Macedonia 0 0 0 0 Madagascar 0 0 0 0 Malawi 0 0 0 0 Malaysia 0 0 0 0 Maldive Islands 0 0 0 0 Mali 0 0 6 0 Malta 0 0 0 0 Mauritania 0 0 0 0 Mauritius 0 0 0 0 Mexico 0 0 0 0 Moldova 0 0 1 0 Monaco 0 0 0 0 Mongolia 0 0 0 0 Morocco 0 0 15 0 Mozambique 0 17 0 0 311 Table 34, continued Namibia 0 0 0 0 Nepal 0 4 0 0 Netherlands 0 0 0 0 New Zealand 0 0 0 0 Nicaragua 3 10 4 0 Niger 1 0 0 1 Nigeria 1 6 1 0 Norway 0 0 0 0 Oman 0 7 0 0 Pakistan (1972-) 3 0 22 6 Palau 0 0 0 0 Panama 0 0 0 0 Papua New Guinea 0 0 9 0 Paraguay 0 0 0 0 Peru 1 16 0 0 Philippines 3 25 28 5 Poland 0 0 0 0 Portugal 0 0 0 0 Qatar 0 0 0 0 Romania 0 1 0 0 Russia 0 0 4 0 Rwanda 1 0 9 1 San Marino 0 0 0 0 Sao Tome-Principe 0 0 0 0 Saudi Arabia 0 0 0 0 Senegal 0 0 8 0 Seychelles 0 0 0 0 Sierra Leone 4 9 0 0 Singapore 0 0 0 0 Slovakia 0 0 0 0 Slovenia 0 0 0 0 Solomon Islands 0 0 0 0 Somalia 9 7 12 4 South Africa 0 7 10 0 Spain 0 0 0 0 Sri Lanka 0 3 17 2 St. Kitts-Nevis 0 0 0 0 St. Lucia 0 0 0 0 St. Vincent and the Grenadines 0 0 0 0 Sudan 3 0 20 20 312 Table 34, continued Suriname 0 0 0 0 Swaziland 1 0 0 0 Sweden 0 0 0 0 Switzerland 0 0 0 0 Syria 0 0 0 2 Taiwan 0 0 0 0 Tajikistan 0 7 0 0 Tanzania 0 0 0 0 Thailand 2 14 0 0 Togo 0 0 0 0 Trinidad 0 0 0 0 Tunisia 0 0 0 0 Turkey 2 0 16 0 Turkmenistan 0 0 0 0 Uganda 2 3 20 16 Ukraine 0 0 0 0 United Arab Emirates 0 0 0 0 United Kingdom 0 0 12 0 Uruguay 3 0 0 0 Uzbekistan 0 0 0 0 Vanuatu 0 0 0 0 Venezuela 0 0 0 0 Vietnam South 0 0 0 0 Vietnam 0 0 0 6 Western Samoa 0 0 0 0 Yemen 0 1 0 0 Yemen North 0 1 0 0 Yemen South 0 1 0 0 Yugoslavia 1 0 2 2 Zambia 4 0 2 0 Zimbabwe 1 8 7 0 313 Appendix D: Country List by Scarcity Group Group 1: High scarcity group Group 2: Low scarcity group Country with missing values are showed with empty cells Table 35: Country List by Mean and Median Scarcity Group Country Natural Capital _mean Natural Capital _median AFGHANISTAN ALBANIA 1.00 2.00 ALGERIA 2.00 2.00 ANGOLA 1.00 2.00 ARGENTINA 2.00 2.00 AUSTRALIA 2.00 2.00 AUSTRIA 1.00 2.00 BAHRAIN BANGLADESH 1.00 1.00 BELGIUM 1.00 1.00 BENIN 1.00 1.00 BHUTAN 1.00 2.00 BOLIVIA 1.00 2.00 BOTSWANA 1.00 1.00 BRAZIL 1.00 2.00 BULGARIA 1.00 1.00 BURKINA FASO 1.00 1.00 BURMA BURUNDI 1.00 1.00 CAMBODIA 1.00 1.00 CAMEROON 1.00 2.00 CANADA 2.00 2.00 CENTRAL AFRICAN REP. 1.00 2.00 CHAD 1.00 1.00 CHILE 2.00 2.00 CHINA 1.00 1.00 COLOMBIA 1.00 2.00 CONGO 2.00 2.00 COSTARICA 2.00 2.00 CUBA 1.00 2.00 CYPRUS 1.00 1.00 314 Table 35, continued CZECHOSLOVAKIA DEM. REP. CONGO DENMARK 2.00 2.00 DJIBOUTI DOMINICAN REP. 1.00 1.00 ECUADOR 2.00 2.00 EGYPT 1.00 1.00 EL SALVADOR 1.00 1.00 ETHIOPIA 1.00 1.00 FIJI 1.00 1.00 FINLAND 2.00 2.00 FRANCE 1.00 2.00 GABON 2.00 2.00 GAMBIA 1.00 1.00 GERMAN DEM. REP. GERMANYFED. REP. 1.00 2.00 GHANA 1.00 1.00 GREECE 1.00 2.00 GUATEMALA 1.00 1.00 GUINEA 1.00 1.00 GUINEA BISSAU 1.00 1.00 GUYANA 2.00 2.00 HAITI 1.00 1.00 HONDURAS 1.00 1.00 HUNGARY 1.00 2.00 INDIA 1.00 1.00 INDONESIA 1.00 2.00 IRAN 2.00 2.00 IRAQ IRELAND 2.00 2.00 ISRAEL 1.00 2.00 ITALY 1.00 2.00 IVORY COAST 1.00 1.00 JAMAICA 1.00 1.00 JAPAN 1.00 1.00 JORDAN 1.00 1.00 KENYA 1.00 1.00 KOREA, S. 1.00 1.00 KUWAIT 2.00 2.00 LAOS 315 Table 35, continued LEBANON 1.00 1.00 LESOTHO 1.00 1.00 LIBERIA 1.00 1.00 LIBYA 1.00 1.00 MADAGASCAR 1.00 1.00 MALAWI 1.00 1.00 MALAYSIA 2.00 2.00 MALI 1.00 1.00 MAURITANIA 1.00 1.00 MAURITIUS 1.00 1.00 MEXICO 2.00 2.00 MONGOLIA 1.00 2.00 MOROCCO 1.00 1.00 MOZAMBIQUE 1.00 1.00 N. KOREA 1.00 1.00 NEPAL 1.00 1.00 NETHERLANDS 1.00 2.00 NEW ZEALAND 2.00 2.00 NICARAGUA 1.00 1.00 NIGER 1.00 1.00 NIGERIA 1.00 2.00 NORWAY 2.00 2.00 OMAN 2.00 2.00 PAKISTAN 1.00 1.00 PANAMA 1.00 2.00 PAPUA N.G. 1.00 2.00 PARAGUAY 1.00 2.00 PERU 1.00 2.00 PHILIPPINES 1.00 1.00 POLAND 1.00 2.00 PORTUGAL 1.00 2.00 ROMANIA 1.00 2.00 RUSSIA 2.00 2.00 RWANDA 1.00 1.00 SAUDI ARABIA SENEGAL 1.00 1.00 SIERRA LEONE 1.00 1.00 SINGAPORE 1.00 1.00 SOMALIA SOUTH AFRICA 1.00 1.00 316 Table 35, continued SPAIN 1.00 2.00 SRI LANKA 1.00 1.00 SUDAN 1.00 1.00 SWAZILAND 1.00 1.00 SWEDEN 1.00 2.00 SWITZERLAND 1.00 2.00 SYRIA 2.00 2.00 TAIWAN TANZANIA 1.00 1.00 THAILAND 1.00 2.00 TOGO 1.00 1.00 TRINIDAD & TOBAGO 2.00 2.00 TUNISIA 1.00 2.00 TURKEY 1.00 2.00 U. ARAB EMIRATES UGANDA UK 1.00 2.00 URUGUAY 2.00 2.00 USA 2.00 2.00 VENEZUELA 2.00 2.00 VIETNAM 1.00 1.00 YEMEN ARAB REP. YEMEN PEOP. REP. YUGOSLAVIA ZAMBIA 1.00 1.00 ZIMBABWE 1.00 1.00

Asset Metadata

Creator Li, Wenyu (author)

Core Title Catastrophe or adaptation? Explaining the impacts of resource scarcity and adaptability on political instability

Contributor Electronically uploaded by the author (provenance)

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Politics

Publication Date 04/11/2011

Defense Date 02/24/2010

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

Tag adaptability,auto-regression,Civil War,OAI-PMH Harvest,political instability,principal component analysis,resource scarcity

Place Name Malawi (countries), Mauritania (countries), Tanzania (countries)

Language English

Advisor James, Patrick (committee chair), Rosen, Stanley (committee member), Tang, Shui Yan (committee member)

Creator Email liwenyu81@gmail.com,wenyuli@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m3732

Unique identifier UC1141934

Identifier etd-Li-4326 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-451598 (legacy record id),usctheses-m3732 (legacy record id)

Legacy Identifier etd-Li-4326.pdf

Dmrecord 451598

Document Type Dissertation

Rights Li, Wenyu

Type texts

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

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email uscdl@usc.edu

Abstract (if available)

Abstract With the rapid growth of population and human consumption, natural resource scarcity is increasingly considered as an important national security issue. A debate on the causal relationships between resource scarcity and political instability of a state, in particular civil war, has arisen in the last two decades. Although a series of case studies have been conducted by several research projects, quantitative studies on the subject still lag behind and the few existing ones have reported quite inconsistent results. This is because most of them focus on the shortage of certain natural resources and do not take into account the interaction between the natural and human system.