Growth solves war---our ev assumes every trend.
Szayna et al 17
Research department director of the Defense and Political Sciences Department and a senior political scientist at the RAND Corporation. He has over 30 years of experience in national security policy and defense analysis. From 1997 to 2011 he served as associate director of the Strategy, Doctrine, and Resources Program in RAND's Army Research Division. His research has focused on aspects of strategic planning for the U.S. armed forces, post-conflict stability and reconstruction operations, and coalition interoperability. He gave testimony for the U.S. House of Representatives and has been a keynote speaker at a number of defense conferences. Szayna received a B.A. in history and philosophy from Villanova University and an M.A. in international relations from Claremont Graduate School. Also Angela O’Mahony, Jennifer Kavanagh, Stephen Watts, Bryan Frederick, Tova C. Norlen, Phoenix Voorhies. Conflict Trends and Conflict Drivers: An Empirical Assessment of Historical Conflict Patterns and Future Conflict Projections. 2017. Download it here: https://www.rand.org/pubs/research_reports/RR1063.html
The Debate The two World Wars understandably shaped the study of world affairs for decades. Much of the American study of international relations was shaped by the overriding imperative to avoid a third such conflagration—especially since it would occur in the nuclear era. The American wars in Korea and Vietnam further reinforced the perception that peace was largely elusive and fragile when achieved. This focus, however, obscured a slowly evolving but remarkable trend in global politics: a decades-long absence of war among major powers. The Yale historian John Lewis Gaddis dubbed the postwar period “the long peace” and observed that it “now compares favorably . . . with some of the longest periods of great power stability in all of modern history.”2 In the nearly three decades since Gaddis wrote that article, the absence of war among great powers has become all the more remarkable. By the 2000s, an even broader and more remarkable picture was coming into focus, at least according to some observers. It was not merely war among the great powers that appeared to have been tamed; violent conflict in nearly all of its forms seemed to be declining. Global statistics assembled by the Uppsala Conflict Data Program (UCDP) and the Peace Research Institute Oslo appeared to show that this trend was true of wars among all types of states (interstate conflicts) and true as well—although to a lesser extent—of conflicts within states (intrastate conflicts).3 2 John Lewis Gaddis, “The Long Peace: Elements of Stability in the Postwar International System,” International Security, Vol. 10, No. 4, Spring 1986, p. 142. The political scientist Jack Levy had earlier argued that great power conflicts were declining in frequency but increasing in severity; see Jack S. Levy, “Historical Trends in Great Power War, 1495–1975,” International Studies Quarterly, Vol. 26, No. 2, June 1982, pp. 278–300. John Mueller also contributed to this general line of argument in his Retreat from Doomsday: The Obsolescence of Major War, New York: Basic Books, 1989. 3 Nils Petter Gleditsch, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Håvard Strand, “Armed Conflict 1946–2001: A New Data set,” Journal of Peace Research, Vol. 39, No. 5, 2002, pp. 615–637. For more recent research using this and related data, see for instance, Bethany Lacina and Nils Petter Gleditsch, “Monitoring Trends in Global Comb2AC---AT: A New Data set of Battle Deaths,” European Journal of Population, Vol. 21, No. 2–3, 2005, pp. 145–166; Lotta Harbom and Peter Wallensteen, “Armed Conflicts, 1946–2009,” Journal of Peace Research, Vol. 47, No. 4, July 2010, pp. 501-509; Halvard Buhaug, Scott Gates, Håvard Hegre and Håvard Strand, “Global Trends in Armed Conflict,” Centre for the Study of Civil War, Peace Research Institute Oslo (PRIO), 2007. 3 While there is a general consensus among conflict studies scholars that armed conflict has decreased over the past two decades, there is less of a consensus in explaining the decrease. Analysts and scholars have proposed many hypotheses to explain the apparent decline in the incidence of interstate conflict: • The seizure and annexation of land is less rewarding for developed societies (with a predominantly industrial or postindustrial, rather than agricultural, economic base) in an open international economy. • The enormous military predominance of the United States, a status quo power, means that a country considering an act of military aggression in a strategically important part of the world must anticipate the possibility of a decisive U.S. response. • The rise of nationalism, spread of “technologies” of violent resistance (such as guerilla warfare), and the widespread availability of small arms and light weapons has made it more difficult to pacify the populations of conquered territories. • The rise of international consensus and norms against the forceful change of boundaries, accompanied by the commensurate rise of more effective international organizations that can enforce those norms, has made military aggression less attractive.4 Similarly, observers have suggested a number of possible explanations for the decline in intrastate conflict: • The gradual strengthening of state institutions, associated with both a global increase in wealth and with the consolidation of postcolonial political orders, has made states better able to provide services to their populations and to police nonstate violence. • Rising levels of wealth give a greater proportion of populations a stake in peace. • The spread of democratic forms of government and associated practices provide institutionalized mechanisms for nonviolent conflict resolution. • The rise of effective international conflict mediation mechanisms, coupled with a credible threat of external intervention (in particular, peace operations mandated by the United Nations [UN]), has provided both channels and incentives to defuse conflict before it escalates to the level of widespread violence.5 The general consensus regarding trends since the 1990s notwithstanding,6 a number of scholars have challenged the “declinist” thesis. Their criticisms generally take one of two forms: 4 The literature on all of these points is vast. It is presented in detail in Appendix C. The appendix has been developed as a stand-alone volume and is organized so that it can serve as a quick-reference guide for those interested in understanding the current state of academic debates on a particular alleged cause of violent conflict. 5 Similarly, see stand-alone Appendix C for a review of the literature on the causes of intrastate conflict. 6 The above explanations come from scholars who have focused on the changes in conflict trends since the 1980s, which is the majority of the work in this vein. However, there is a strand of the research agenda on incidence of armed conflict that has gained popular prominence and is much broader in scope, namely Steven Pinker’s sweeping 4 either that the quantitative data have been misrepresented or misinterpreted, or that the apparent trends are in fact only a temporary anomaly and are likely to be undone by a range of new dynamics that are likely to fuel increases in conflict. Critics of the quantitative methodology underlying the “declinist” argument have challenged how conflict is defined and the timeframe over which conflict trends are observed. For example, using an extremely broad interpretation of conflict that includes nonviolent minor disputes, the economists Mark Harrison and Nikolaus Wolf claim that “the frequency of bilateral militarized conflicts among independent states has risen steadily over 131 years from 1870 to 2001.”7 Similarly, Bear Braumoeller, a political scientist at Ohio State University, argues that the apparent decline in interstate war is the result of two alleged mistakes made by proponents of the declinist thesis: The incidence of war is not observed over a sufficiently long period of time, and proponents of declinism adopt an inappropriate measure of the prevalence of armed conflict.8 Others suggest that even if the recent decline in violence is real, it is likely to be reversed by one or more of a number of future dynamics: • Globalization may lead to a quick internationalization of disruptions in one country, raising and magnifying what may be a local dispute to a systemic level. • Population pressures may lead to greater scarcity and conflict. • Climate change may induce disruptions and migrations that can overstress the capacity of states and lead to widespread unrest. • Rapidly evolving technology makes it easier for nonstate actors to organize, and it makes even small numbers of people potentially highly lethal.9 Many long-term intelligence forecasts, such as the Global Trends 2030 paper by the National Intelligence Council, refer to these new sources of conflict.10 history of the decline of violence over many millennia, positing civilizational and psychological changes at the core of the shift. Steven Pinker, The Better Angels of Our Nature: Why Violence Has Declined, New York, N.Y.: Viking Adult, 2011. See, also, the annual publications of the Human Security Report Project, available at the organization’s website. 7 Mark Harrison and Nikolaus Wolf, “The Frequency of Wars,” Economic History Review, Vol. 65, No. 3, 2012, pp. 1055–1076. Despite the title of the article, the authors are actually examining all conflict behavior, including minor disputes in which force was not used. For a forceful response, see Kristian Skrede Gleditsch and Steve Pickering, “Wars Are Becoming Less Frequent: A Response to Harrison and Wolf,” Economic History Review, Vol. 67, No. 1, 2014, pp. 214–230. 8 Bear Braumoeller, “Is War Disappearing?” paper presented at the annual conference of the American Political Science Association, Chicago, Ill., August 27, 2013. 9 The stand-alone Appendix C provides extensive documentation of all of these claims. 10 National Intelligence Council, Global Trends 2030: Alternative Worlds, NIC-2012-001, 2012. 5 Research Approach This project assessed the debate in three steps. First, the research team gathered data on conflict trends from as many sources of aggregate conflict data as could be identified. From this pool of data sources, the researchers selected those data sets that appeared to use reliable sources of data and a transparent methodology for classifying and aggregating conflicts, paying particular attention to ensure that the widest possible range of types of conflict and data sources were represented. If one data set reveals a trend, the finding is interesting, but the reliability of the underlying evidence is always open to question. If multiple data sets using different sources of data, different definitions, and different methodologies all demonstrate consistent trends, the robustness of these findings is greatly improved. Second, the research team undertook an extensive review of the scholarly literature on the causes of conflict. Without understanding why conflicts occur, it is impossible to assess if recent patterns represent a temporary anomaly or a secular trend that would require a massive shift in global politics to reverse.11 This review served as the basis for developing ten “key factors” that drive violent conflict, according to a broad consensus of social scientists. Quantitative indicators of each of these factors were also derived. Third, the research team used the quantitative data on violent conflict and drivers of conflict developed in the first two steps to assess conflict trends through the year 2040. The team first assessed conflict trends using only historical data on the recent incidence of conflict. The team then assessed how changes in any of the ten key factors might lead to changes in the future incidence of conflict. Lastly, the team developed a spreadsheet tool that allows readers of this report to “mix and match” assumptions about future conditions to see how varying these assumptions should be expected to influence the future incidence of conflict. It is important to highlight the limits of what this research approach can accomplish. The first limitation concerns the time over which we examine conflict trends. The analysis in this report uses only data from the past several decades. Going back further in history would require us to use lower-quality data and would introduce an enormous number of other variables into the analysis. Although the restricted historical range underpinning our analysis is a limitation that should be borne in mind, there are reasons to believe that these trends represent longer-term phenomena. Specifically, our research approach examines the role of a number of key factors in precipitating violent conflict. The observed decline in violence does not occur in a vacuum; it is rooted in changes in a number of critical variables. These variables, in turn, can be observed over much longer periods, and nearly all of them demonstrate long-standing changes that are consistent with our conflict trends and conflict projections. The second limitation concerns our ability to account for uncertainty. The reliability of predictions degrades rapidly the further into the future they are made. Because of inherently high 11 The results of this literature review are documented in stand-alone Appendix C. 6 levels of uncertainty, projections about what the future will look like in nearly 30 years are almost certain to be wrong in many important respects. For this reason, long-term planners have frequently made use of planning frameworks such as uncertainty-sensitive planning or robust decisionmaking. 12 Such frameworks allow for the projection of multiple possible futures, including ones in which extremely low-probability but high-impact events radically alter the future landscape. In terms of uncertainty-sensitive planning, this report provides a basis for estimating a baseline or “no surprises” future—the one that will occur if past trends continue in more or less linear fashion. It also provides a foundation for estimating a variety of “branch” scenarios—that is, alternative futures in which one or more key factors vary in ways that are not currently expected but that could be easily imagined, such as growing tensions between the United States and one or more other major powers. This report does not, however, develop so-called “shock” scenarios, such as the realization of the most dire climate change scenarios, the use of biologically engineered pathogens in a catastrophic attack, and so on. Any single such “shock” scenario is highly improbable. Over the course of 30 years, however, the aggregate probability of at least one of these possible shocks occurring is substantial. The fact that our research approach does not allow us to model these outlying scenarios does not reduce the value of the research; developing baseline and branch scenarios is absolutely critical for uncertaintysensitive planning. It is nonetheless a caveat that readers should bear in mind. Organization of the Report The remainder of this report is divided into five chapters. Chapter Two reviews trends in interstate and intrastate conflict since 1946, using a typology to classify conflict type and multiple existing data sets to examine changes in the nature and intensity of conflict, as well as how conflict may be shifting from one form to another. The chapter starts out by defining key terms, including the term “armed conflict,” then reviews the various data sets we examined in the course of our research and discusses potential measurement error and other shortcomings of available data. Most of the chapter outlines conflict trends over time, beginning with interstate conflict (looking at armed conflict as well as crises and militarized disputes), then intrastate conflict (focusing on all forms of violence, including one-sided state violence, as well as politically motivated protest that falls short of violence), and finally violence by nonstate actors (including nonstate actor conflicts against each other and terrorism).13 12 See especially Paul K. Davis, “Uncertainty-Sensitive Planning,” in Stuart E. Johnson, Martin C. Libicki, and Gregory Treverton, eds., New Challenges, New Tools for Defense Decisionmaking, Santa Monica, Calif.: RAND Corporation, MR-1576-RC, 2003; and Robert J. Lempert, Steven W. Popper, and Steven C. Bankes, Shaping the Next One Hundred Years: New Methods for Quantitative, Long-Term Policy Analysis, Santa Monica, Calif.: RAND Corporation, MR-1626-RPC, 2003. 13 Appendix A contains a more detailed assessment of the data on conflict. 7 Chapter Three addresses the potential reasons for the changes in conflict type and intensity over time, since 1946. We conducted an extensive literature review to identify the key factors having the strongest effect on the prevalence of interstate and intrastate conflict. Our review focused on literature that uses empirical and scientific approaches to conflict. We identified ten key factors, each of which is discussed in a subsection of the chapter. We operationalized each key factor using quantifiable metrics and then, using those metrics, projected the baseline trends for each factor until 2040. We also took into account expected branch scenarios by outlining the upper and lower bounds of the projections. The chapter discusses the rationale for the bounding of the projections for each factor as well as the potential for discontinuous change in the metric we use for each of the key factors. Finally, the chapter assesses the likely effect of each key factor on future conflict incidence. Appendix C contains a detailed discussion of the literature review. Chapter Four builds on the empirical analysis in Chapter Two and develops projections about incidence of conflict in the future. The chapter first discusses some of the methodological issues involved in making such projections, then outlines the baseline projections for conflict incidence and intensity for interstate and intrastate conflict. It also provides a sensitivity analysis of the conflict projections presented in Chapter Two that assesses the impact of the expected future trends for each of the ten key factors on the baseline conflict projections. The chapter then presents the revised conflict projections, which stem from both historical conflict trends and projected future trends in key factors affecting conflict. Finally, it discusses findings and projections about the incidence of conflict in the future. The starting point for Chapter Five is the assumption that all projections are bound to be wrong to some extent, and the further they are into the future, the more likely they are to be wrong. The projections presented in Chapter Four might be a good starting point but we anticipate that the value of our work is to help military planners and analysts better understand conflict drivers for experimentation and wargaming purposes. The chapter discusses our development of a range of alternative futures. We developed an exploratory analysis tool that enables analysts and planners to examine the potential impact of changes of combinations of key factors on future conflict incidence. This tool includes signposts and incorporates all the causal relationships from the work reported in earlier chapters; it also allows for a shortcut to an enormous amount of information and could be used by military planners and analysts for scenario development, wargaming, and experimentation. The chapter then discusses the main findings from our use of this Alternative Futures Tool regarding the impact that changes in several of the key factors are likely to have on the incidence of conflict in the future. We then discuss the expected effects of changes in key factors on both interstate and intrastate conflict, depending on the extent of variance from the expected baseline projections outlined in Chapter Four. We anticipate that this chapter will be of most direct use to Army planners in thinking about future operating environments. Appendix B contains additional information on the Alternative Futures Tool. 8 Chapter Six summarizes our findings and draws out implications for the Army. It also discusses caveats and some of the main unknowns that could alter some of the projections offered in the report. As noted above, several appendixes provide supporting material and additional information relevant to our study. The Alternative Futures Tool is available from the authors to authorized requesters. The project began in October 2012 and a draft report was completed in September 2013. Project team members presented the findings contained in this report to Army and DoD audiences from September to December 2013. The report was reviewed and revised in 2014 and provided to the Army in April 2015. The report was updated in July and August of 2016. 9 2. Overview of Conflict Trends Introduction The nature, intensity, and frequency of conflict have changed over the past several decades, generally shifting from wars fought directly between states to various forms of “internal” or intrastate violence, including insurgencies, guerilla wars, terrorism, organized and large-scale criminal violence, and protests. However, the timing, speed, and permanence of these shifts have varied and are not uniform for all types of political violence. This chapter reviews trends in interstate and intrastate conflict since 1946 using a typology to classify conflict type and multiple existing data sets to examine changes in the nature and intensity of conflict as well as how conflict may be shifting from one form to another. What Is Conflict? There is no single, agreed-upon definition of what constitutes armed conflict. Instead, the term may refer to civil war, ethnic war, and interstate war at high and low intensities as well as violence that falls short of war, such as militarized disputes, terrorism, and riots or strikes. Our approach in this report was to be as holistic as possible. Many studies examine particular types of violence or rely on single data sets. Although such approaches may be appropriate for the questions that other observers seek to answer, they were inadequate for our purposes. The U.S. Army must be prepared to respond to contingencies across the entire spectrum of operations, anywhere around the globe, thus, making a full consideration of all types of conflict and violence necessary for our analysis. In addition, as we identified conflict trends and made projections about the future, it was imperative that we develop as robust an empirical basis for these trends and projections as possible. By relying on multiple quantitative databases on conflict, each with its own strengths and weaknesses depending on the peculiarities of how it aggregates diverse phenomena, we were able to develop a more comprehensive and robust picture of past trends and possible futures. The fact that all of the data sets consulted in this review yielded trends that are broadly consistent despite their differences suggests that we have strong reason to believe in the results. A Typology of Conflict As a first step toward the mapping of conflict trends, we developed a typology of violence, illustrated in Table 2.1. The typology is a useful tool in the study of conflict because it identifies, describes, and categorizes different types of political violence across the conflict “concept space” and can be used to discuss changes in the most common types of conflict as well as trends in individual types of violence. We organize our typology along two dimensions: primary actors 10 and level of intensity. The first dimension focuses on the actors involved in the conflict or violence. Understanding the key actors in armed conflicts and how they change over time is an important characteristic of conflict that provides insight into its motivation, form and intensity, and potential consequences (e.g., regime change, physical destruction). The typology defines several types of conflicts based on primary actors. Table 2.1. Typology of Conflict with Examples Intrastate Interstate One-Sided State State v. Nonstate Organized Societal Spontaneous Societal High Intensity: Battle deaths of 100,000 per year or 500,000 total Wars Genocide, ethnic cleansing Civil or ethnic war, wars of independence, ethnic cleansing Ethnic or intercommunal war, ethnic cleansing Medium Intensity: Battle deaths of 1,000 per year or 100,000 total Wars Genocide, purges, coups Civil or ethnic war, wars of independence, guerilla war, insurgency Ethnic or intercommunal war, guerilla warfare Intercommunal or intracommunal violence, riots, uprisings Low Intensity: Battle deaths of 25 per year or 1,000 total Militarized disputes, armed conflict Extrajudicial killings, disappearances, coups Civil or ethnic violence, insurgency, guerilla war, extrajudicial killings, riots Terrorism, guerilla warfare, intercommunal or intracommunal violence Uprisings, intercommunal or intracommunal violence Minimum Violence: Battle deaths of 0–25 per year; nonviolent conflict/dispute Disputes/ crises Arrests, detentions, government crises Protests, strikes, demonstrations Protests, strikes, intercommunal or intracommunal disputes Strikes, protests, intercommunal or intracommunal disputes Interstate conflict involves violence between two or more states. Intrastate violence occurs within a single state and can be broken down into many forms. These include: • one-sided state violence, including genocide, “ethnic cleansing,” purges, and extrajudicial (state) killings • state versus nonstate violence, including civil wars, wars of independence, insurgencies, and some riots • organized societal violence, such as ethnic or intercommunal war or conflict between two nonstate actors in which the state is not a major participant, some types of terrorism, some riots, and organized violent crime • spontaneous societal violence, such as pogroms, uprisings, some riots, and everyday violent crime. Our second typology dimension, conflict intensity, is based on total battle deaths. We define four levels of violence: minimum violence, which includes events involving 0 to 25 deaths per 11 year; low-intensity conflict, which includes 25–999 deaths per year; medium-intensity, which includes 1,000 to 100,000 deaths per year; and high-intensity, which includes more than 100,000 battle deaths per year. Battle deaths are an imperfect measure of conflict intensity because they capture only one small piece of the cost of a conflict, namely the direct effect of a conflict in terminating individual life. For example, conflict may cause famine or disease that wipes out hundreds of thousands of lives or economic devastation that leads to widespread poverty, but neither of these costs will be included if intensity is only based on battle deaths. Furthermore, battle deaths can be difficult to count, as it is not always clear who is a combatant. There are some alternative measures of intensity being developed to address these shortcomings. Some use indexes to rate the level of violence, others look at economic costs, or disability-adjusted life years. Because most data sets still rely on battle deaths or “conflict-related deaths,” we use this measure for the current report, although we recognize its shortcomings. To provide some context, Vietnam, the most recent war in Iraq, and the conflict in Afghanistan would all be classified as medium-intensity interstate wars; the ongoing conflict in Sudan would be an example of a medium-intensity ethnic war; and the civil wars in Libya and Syria as low- to medium-intensity civil wars. The genocide in Rwanda is counted as highintensity one-sided violence by our typology, but can also be counted as high-intensity ethnic violence. Finally, ongoing sectarian violence in Iraq since the end of U.S. involvement could be counted as medium-intensity societal conflict. Our typology has a number of advantages over other definitions of conflict and approaches to the study of conflict. First, the typology disaggregates conflict along two fairly objective and usually discernible characteristics (actors and intensity), allowing us to bin individual conflicts more easily and map trends over time. Second, as already noted, we have constructed the typology and calibrated its intensity levels in a way that will allow us to move between data sets, observe and compare trends in violence according to several different definitions, and thus present a more robust picture of conflict trends. Finally, by taking a holistic view of conflict, it supports our general approach in this report. The typology also has some limitations. First, there will be overlap among the typology’s different categories. For example, the line between conflicts involving a state and a weak nonstate actor and one-sided violence may be a fine one in some instances. Certain types of conflict may fit into several different cells. For example, antigovernment protests may fall into the minimum-intensity category or they may include minimum-, low-, or high-intensity violence and may be organized or spontaneous depending on the situation. Second, some conflicts may include many types of political violence, defying an easy categorization. As a result, it will be useful to think of the typology as a guide with relatively porous borders rather than a framework with fixed internal boundaries. Finally, the conflict types and intensity thresholds defined in the typology may not always correspond to the exact conflict as defined in existing data sets. As we noted before, it may be especially hard to distinguish between organized and spontaneous 12 societal violence. Table 2.1 provides “best examples” for each type of conflict defined by the typology. The typology can be used in several different ways. First, we can use it to describe trends in conflict, considering how the distribution of ongoing or new conflicts across the matrix changes over time. Second, we can use the typology to consider how different sorts of changes or shocks to the external system have affected the form and intensity of conflict in the past and might affect the form and intensity of conflict in the future. Finally, we can use the typology as a way to integrate and combine the many different data sets that exist for the study of conflict and to discuss and use these data sets (and their definitions of conflict) together in our analysis of conflict trends. Key Data Sets To study trends in each of these different forms of armed conflict and to develop a robust understanding of these trends, we relied on a large number of major data sets on armed conflict. Table 2.2 lists these data sets, the years covered, types of conflict included, and the definition of conflict used by each. Several observations are relevant. First, coverage of interstate and intrastate violence is clearly more complete than coverage of nonstate and societal violence. Second, some data sets cover several different types of violence while others address only one form. When a data set covers several different types of violence, our ability to disaggregate the data along the boundaries of the typology will depend on how the data set reports and codes incidents of conflict. For example, although the Cross National Times Series (CNTS) Databank gives us insight into societal conflict such as riots, protests, and demonstrations, it does not provide fatality information or separate between spontaneous and organized events. In Appendix A, we present all the data and analysis summarized in this chapter. We also describe relevant trends in each data set and then compare across data sets where there is overlap. We consistently found that while there are some differences in the levels of violent and nonviolent conflict reported in each data set (largely because of differences in definitions and coding of violent events and conflict), the overall trends for each type of violence in our typology are largely the same across databases. In this chapter, we focus on one authoritative data series per conflict or violence type. Appendix A reviews additional data sources. 13 Table 2.2. Data Sets and Definitions Data Set Years Types of Conflict Thresholds for Inclusion* Correlates Of War (COW) 1816–2007 Interstate, intrastate, nonstate; extrasystemic War must have 1,000 battle deaths in 12-month period, each side must mobilize at least 1,000 persons UCDP 1946–2015 Interstate, intrastate, nonstate; extrasystemic; one-sided state; internationalized intrastate** “Armed conflict” requires 25 deaths per year; “major war” requires 1,000 Political Instability Task Force (PITF) 1946–2015 Ethnic, revolutionary conflict; genopoliticide Conflict must include 1,000 total battle deaths and at least one year with 100 battle deaths Genopoliticide: Violence must have elite sanction, endure over longer period, result in death of a substantial portion of a communal or political group Major Episodes of Political Violence (MEPV) 1946–2015 Interstate, wars of independence, intrastate (ethnic and civil) Violent episode must have at least 500 total conflictrelated deaths (base rate 100 per year) Global Terrorism Database (GTD) 1970–2015 Terrorism Incidents must have political, religious, or social goal, communicate to an audience outside victims, violate International Humanitarian Law CNTS 1916–2015 Riots, protests, antigovernment demonstrations, strikes, revolution, guerilla war Each event has specific definition, focused on the participants, motivation, and scope. Some are nonviolent, others are violent but there are no fatality thresholds.a International Military Interventions 1946–2005 Military interventions “. . . the movement of regular troops or forces (airborne, seaborne, shelling, etc.) of one country inside another, in the context of some political issue or dispute”b Militarized Interstate Disputes (MID) 1816–2010 Militarized disputes that fall short of war Involve use or display of force, but have fewer than 1,000 battle deaths International Crisis Behavior (ICB) 1946–2013 Interstate crises, most falling short of war Crises have three major characteristics: “(1) there is a threat to one or more basic values (2) an awareness of finite time for response to the value threat (3) heightened probability of involvement in military hostilities”c NOTE: *More detail and definitions for each conflict type according to each database are provided in subsequent chapters along with full citations. ** The UCDP One-Sided State Violence data set goes through 2014. a See Chapter Three for more detail. b Frederic S. Pearson and Robert A Baumann, International Military Intervention, 1946–1988, Ann Arbor, Mich.: University of Michigan, Inter-University Consortium for Political and Social Research, Data Collection No. 6035, 1993, p. 1. c Michael Brecher and Jonathan Wilkenfeld, A Study of Crisis, Ann Arbor, Mich.: University of Michigan Press, 1997. 14 Scope and Implications of Measurement Error While the data sets that we include in our review are relatively comprehensive, there are certain sources of measurement error that affect even the best of them and that will also have implications for the trends that we observe. Most significantly, while high-intensity and interstate conflicts are likely to be accurately recorded over the entire period we are considering, the same may not be true for low-intensity, intrastate, and nonstate conflicts and violence. Because most data sets rely on reporting and media coverage to count conflicts, smaller conflicts that attract less attention and such forms of violence as protests, riots, and demonstrations may be unintentionally excluded. This will be especially true in earlier decades, such as the 1940s and 1950s (the start of our window of observation), when media coverage was less consistent than it has been since 1990, particularly in developing countries. Even now, it is likely that certain instances of low-intensity violence in remote locations or authoritarian regimes go unreported or are underreported. Improvement in the quality and frequency of reporting, therefore, may lead to an apparent increase in the number of conflicts reported even if their true incidence has remained the same or even decreased. This systematic bias could contribute incorrect inferences about recent trends and future directions of certain forms of violence if it is not properly included in any analysis of trends. We ultimately cannot entirely correct for this bias, but we will consider its implications for our assessment of recent trends and future directions of certain types of conflict. What Is Excluded from the Review? Although we have attempted to build a typology and collection of data sets that includes as many forms of armed conflict and political violence as possible, there are still certain types of violence that are excluded, primarily because there are insufficient data to reliably map trends over time. Important among these excluded categories will be drug-, gang-, and crime-related violence, forms of “conflict” that are particularly severe in Central American, Caribbean, and some African countries. In recent years, this type of criminal violence has been responsible for as many fatalities as more organized forms of conflict elsewhere. In fact, homicide rates in a handful of Central and South American and Sub-Saharan African states reached rates as high as 80 per 100,000 individuals in 2013.1 For states with a population of 100 million or greater, this homicide rate would produce fatalities at the “high-intensity” level according to our typology.2 1 United Nations Office on Drugs and Crime, Global Study on Homicide 2013: Trends, Contexts, Data, Vienna: United Nations, 2014. 2 For more on transnational criminal networks and associated violence see: Phil Williams, “Transnational Criminal Organisations and International Security,” Survival, Vol. 36, No. 1, 1994, pp. 96–113; Phil Williams, “Transnational Criminal Networks,” in John Arquilla and David Ronfeldt, eds., Networks and Netwars: The Future of Terror, Crime, and Militancy, Santa Monica, Calif.: RAND Corporation, MR-1382-OSD, 2001, pp. 61–97; Fiona B. Adamson, “Globalisation, Transnational Political Mobilisation, and Networks of Violence,” Cambridge Review of International Affairs, Vol. 18, No.1, 2005, pp. 31–49; and John Arquilla and David Ronfeldt, eds. Networks and Netwars: The Future of Terror, Crime, and Militancy, Santa Monica, Calif.: RAND Corporation, MR-1382-OSD, 2001. 15 Certain data sources can give us a window into this type of violence—revealing, for instance, that it is often linked with drug and human trafficking, criminal gangs, and sexual assaults, and that it appears to be increasing in many of the areas where it is most problematic. However, in general, we simply do not have the types of data sets needed to analyze criminal-, drug-, and gang-related violence in more detail. A brief summary of data sources that do exist and their limitations will make this point more clearly. Probably the best source of data on criminal violence, including that related to drug trafficking and gang warfare, is the United Nations Office of Drugs and Crime (UNODC). The UNODC data include information on homicide, sexual assault, and other violent crime. However, these data are limited and unreliable in several ways. First, they cover a limited set of countries and years, preventing a time series analysis of international trends. Second, they rely on one of two sources of information, self-reports or victimization surveys, neither of which is completely trustworthy. Most of this information comes from country self-reports and so reveals only what the individual reporting country chooses to submit and/or is capable of reporting. This makes it unlikely that the data are complete and accurate. In addition to the UNODC, the World Health Organization (WHO) uses these data when tracking fatalities because of “violent causes,” making its fatality data similarly insufficient for our analyses. Additionally, data for some countries are collected from “victimization” surveys, which ask individuals to report whether they or someone they know have been victims of certain crimes. However, this method assumes that individuals will accurately report their experiences, is subject to sampling concerns, and exists only for certain countries (mostly European countries where rates of crime and violence are lower than elsewhere). There is more detailed material on specific issues and specific countries, such as poppy cultivation in Afghanistan, human trafficking in Sub-Saharan Africa, or drug seizures in Central and South America. However, these data are still insufficient for the description or analysis of broader global trends, which is the main focus of this report. Since we cannot provide a complete and systematic treatment of crime-, drug-, and gangrelated violence, we emphasize that societal violence of this type is still a significant source of destabilization, death, and destruction that should be taken into account when considering or assessing the total aggregate amount of violence and conflict in the international system. While it is not always captured by existing data sets, it does figure prominently in the typology and may be an area for future study and analysis. Another source of conflict that is not captured in our review or in existing data sets is cyberconflict. However, as cyberwarfare becomes an increasing national security concern, conflict researchers are going to have to find a way to measure the global level of cyberconflict in order to assess the level of stability in the international system. In fact, some researchers argue the majority of future conflict will be carried out in the cyber realm. This is an extreme view, but highlights the growing importance of this form of conflict and the fact that cyberattacks and warfare eventually will need to be incorporated into studies of conflict trends. This will mean not 16 only qualitative description of cyberattacks but also the development of cyberconflict data sets that can measure trends more quantitatively.3 Trends in Interstate Violence Trends in interstate violence suggest a significant decrease in the incidence and intensity of armed conflict. This result is true regardless of the database we consider, and appears to apply to violence at all intensities. The trend for disputes and crises that fall short of major war is somewhat more ambiguous. For these events, the downward trend is less definitive, but there is little evidence of any upward trend, especially when we focus on events that involve violence. Armed Conflict Our study of interstate conflict used three key data sets, the COW, MEPV, and UCDP. As highlighted in Table 2.2, each uses a slightly different threshold in its definition of conflict and each captures slightly different information on the conflicts included in the database, but all followed similar trends. In this section, we describe trends in interstate violence using the MEPV data set, which is also used for the baseline projections in Chapter Four. The MEPV uses a ten-point index to classify conflicts according to their total associated fatalities (including civilian deaths) but includes only those incidents with at least 500 battlerelated deaths over the course of the conflict. If there are several ongoing conflicts within a country in a given year, then the total “score” received for that country-year will equal the sum of the intensity scores assigned to each episode.4 Compared with the other data sets that measure interstate conflict, the MEPV applies a slightly more comprehensive definition of conflict and thus reports a slightly higher number of conflicts than other data sets. For example, the MEPV set for interstate violence includes such events as the September 11, 2001, al Qaeda attacks, the Bay of Pigs crisis, border disputes between Honduras and Nicaragua, clashes in India and Pakistan, and the Israeli-Hezbollah conflict, none of which are counted in the other data sets. This difference has several implications. On the one hand, the MEPV data give us a much more complete picture of the total amount of interstate violence in the system at any given time. On the other hand, the MEPV approach may overestimate the amount of violence associated with interstate war because it also includes other forms of interstate violence. For our analysis, we have roughly translated this coding scheme into our three levels of intensity and counted the number of conflicts of each type at each level of intensity ongoing in 3 Jason Andress and Steve Winterfeld, Cyber Warfare: Techniques, Tactics and Tools for Security Practitioners,Waltham, Mass.: Syngress, 2011; Kenneth Knapp and William Boulton, “Ten Information Warfare Trends,” in Lech Janczewski and Andrew M. Colarik, eds., Cyber Warfare and Cyber Terrorism, Hershey, Pa.: IGI Global, 2007, pp. 17–25. 4 Monty G. Marshall, Major Episodes of Political Violence (MEPV) and Conflict Regions 1946–2008, Center for Systemic Peace, 2010; Monty G. Marshall, Major Episodes of Political Violence (MEPV) and Conflict Regions 1946–2012, Center for Systemic Peace, 2013. 17 each year. Figure 2.1 shows the total number of interstate conflicts coded by the MEPV data set while Figure 2.2 shows the number of conflicts by intensity. As we have noted, the wars in Vietnam (1964–1973), Iraq (2003–2010), and Afghanistan (2001–2012) are all counted as medium-intensity interstate violence according to the data set we used. As alluded to several times, interstate conflict has followed a clear downward trend since its peak in the late 1970s and early 1980s. Interstate conflict at all intensity levels has declined significantly since the end of the Cold War. It has reached levels consistently near zero in recent years, but it has not disappeared, with several conflicts remaining. Figure 2.2 shows that high-intensity interstate violence has been extremely infrequent since World War II, while low- and medium-intensity interstate violence has tended to decline over time. Medium-intensity violence was most significant in the 1980s but fell sharply in the 1990s, while low-intensity interstate violence was most significant in the 1960s and late 1970s, plateaued at a lower level in the 1980s, and then declined in the 1990s. As of 2015, MEPV codes no remaining ongoing interstate conflicts. However, trends for all three levels of violence also suggest that even as it has been virtually eliminated, interstate violence may still have the potential to bubble up periodically. Figure 2.1. Number of Countries with Interstate Conflict, MEPV Data, 1946–2015 SOURCE: Marshall, 2016. 18 Figure 2.2. Countries with Interstate Conflict, By Intensity, MEPV Data, 1946–2015 SOURCE: Marshall, 2016. Crises and Militarized Disputes Not all interstate uses of military force escalate to armed conflict or war. Some involve only a show of force or an exchange of threats. For example, militarized disputes are common between Pakistan and India, where border disagreements or political discord can lead to troop mobilization and sometimes border skirmishes. Data sets that capture these types of disputes are useful supplements for a study of conflict trends because they capture a type of low-intensity violence that might otherwise be excluded. They are also useful for the perspective they provide on the types of disputes and crises that do escalate to major armed conflict and those that can be effectively de-escalated and controlled. Trends in militarized disputes and interstate crises show that, just like interstate war, the incidence and intensity of these conflict events have fallen significantly since their peak and since the end of the Cold War. However, this decline has been nowhere near as sharp or consistent as that observed for more intense forms of armed conflict. Some of the decline in armed conflict observed in recent years may result from better crisis mediation processes (including mediation by international organizations) that limit conflict escalation before the outbreak of major hostilities, but that make crises and militarized disputes somewhat more likely. We use two primary sources of data on conflicts that fall short of war, the MID data set and the ICB database. Because the MID data are used more consistently in existing literature to 19 describe trends in interstate conflict and violence, we focus on this measure of international crises and disputes.5 The data set defines a MID as a “united historical [case] of conflict in which the threat, display or use of military force short of war by one member state is explicitly directed toward the government, official representatives, official forces, property, or territory of another state.”6 Because the MID data are part of the larger COW project, “short of war” means fewer than 1,000 battle-related fatalities. In the MID data, these disputes are coded according to their levels of violence or hostility: “no violence,” “display of force,” “use of force,” or “war.” Because MIDs by definition fall short of the 1,000-death threshold, they may fall into our “minimum” or “low-violence” categories. Importantly, there may be MID incidents that correspond with “armed conflicts” in the UCDP or PITF data because of the different ways each data set codes conflicts. Figure 2.3 shows MID onsets from 1946 through 2010, the last date included in the MID data. It is important to note that while many of the figures in this report show cumulative numbers of conflicts, this particular one focuses on incident onset. Onsets of low-intensity MIDs have fallen since their peak in the 1980s at the end of the Cold War, but despite a number of temporary fluctuations, the number of low-intensity MIDs has remained at an overall stable level since the early 1990s. In contrast, minimum-violence MIDs have been at higher levels since 1990 than previously, reaching a peak in the early 2000s. Since 2006, the onset of these minimum-violence MIDs has fallen slightly. Notably, over the entire period and continuing today, low-intensity MIDs that do involve some violence appear more common than those with little or no violence. The trends in both types of MIDs suggest that militarized disputes have not disappeared, even as interstate war and more-intense conflict have fallen close to zero. This suggests two observations. First, the fact that these incidents do not escalate to full-scale war may stem from a combination of the effectiveness of existing dispute resolution and mediation procedures at the international level and the efficacy of deterrence, especially since much of the decrease in conflict occurred during a period of U.S. preeminence. Second, as long as disputes involving military forces continue, so will the risk of a flare-up into more-significant conflict, and even war. 5 Daniel M. Jones, Stuart A Bremer, and J. David Singer, “Militarized Interstate Disputes, 1816–1992: Rationale, Coding Rules, and Empirical Patterns,” Conflict Management and Peace Science, Vol. 15, No. 2, 1996, pp. 163– 213. 6 Glenn Palmer, Vito D’Orazio, Michael Kenwick, and Matthew Lane, “The MID4 Data Set: Procedures, Coding Rules, and Description,” Conflict Management and Peace Science, Vol. 32, No. 2, April 2015. 20 Figure 2.3. New Militarized Interstate Disputes, by Year of Onset, 1946–2010, by Level of Violence SOURCE: Jones, Bremer, and Singer, 1996; Palmer et al., 2015. Trends in Intrastate Violence Intrastate conflict includes all forms of violence that occur within a state, ranging from war to riots and anti-government demonstrations. Because it encompasses so many different types of violence, intrastate conflict is measured by a large number of different data sets and incorporates many different trends. Aggregate trends in intrastate conflict appear to follow the same patterns observed for interstate violence, specifically a general decline in frequency of conflict and intensity of conflict since the end of the Cold War, although there has been an uptick in some types of intrastate armed conflict since 2012. This trend applies equally to civil and ethnic conflict, is most dramatic for high-intensity violence, and is once again consistent across data sets. However, trends in other forms of low to medium intrastate violence—particularly involving nonstate and societal actors, including riots, protests, demonstrations, and terrorism— show more limited evidence of this consistent downward trend and in some cases have trended upward in recent years. Armed Conflict Our review of intrastate conflict relied on four major databases: UCDP, MEPV, COW, and PITF, which tracks the number of revolutionary and ethnic wars per year. Once again, we focus our discussion here on the MEPV data set and refer readers to Appendix A for additional detail. 21 Trends in intrastate conflict show a downward trend overall and at most levels of intensity, but this trend is more gradual and less complete than that for interstate conflict, leaving more conflict remaining within the system. Figure 2.4 shows the trend for total number of conflicts, and Figure 2.5 shows the number of conflicts by intensity level. Examples of ongoing intrastate violence include medium violence in Sudan (Darfur) and Pakistan (in the federally administered tribal areas) as well as low-intensity violence in Yemen and in Mexico (between the government and the drug cartels). The overall trend shows that the incidence of intrastate conflict increases steadily over the period from 1946 to about 1991, falls consistently until about 2001, appears to plateau until 2012, then rises slightly through 2015. Looking at violence by intensity, it seems that this recent uptick in conflict has been driven by low- and medium-intensity intrastate conflict, rather than high-intensity conflicts, notwithstanding the major conflict in Syria. Intrastate conflict incidence across intensities rises to a peak in the late 1980s and early 1990s, then declines. The decline in violence is steepest for higher-intensity conflicts and subtler at lower intensities. High-intensity intrastate conflicts have declined to very low levels since 2001. In contrast, medium-intensity intrastate conflicts plateaued through the late 1990s and 2000s, but have risen again since 2012. Finally, low-intensity intrastate conflict fell drastically throughout the 1990s, increased starting in 2000, fell again after 2005, and has been rising since 2012. This increase in low- and medium-intensity intrastate conflict is interesting and potentially significant as it diverges from the aggregate decrease observed for most other forms of conflict and suggests some persistence of low- and medium-intensity intrastate conflict, even as the other forms of conflict decrease. That said, a significant portion of remaining intrastate violence consists of long-running or repeated conflicts, rather than frequent outbreaks of brand new conflicts. However, new conflicts in places like Syria have also contributed to this increase. It is important to note that, even with recent increases, intrastate violence remains well below peak levels and well within the historic average. An important caveat on this trend is the potential bias introduced by measurement error, as discussed in the introduction. Specifically, reporting on low-intensity intrastate violence has almost certainly increased over the 1946–2015 period considered here. This improved reporting could increase the number of conflicts included in conflict data sets, even if the actual number of conflicts has not increased or has even fallen, and will affect all data sets. Since we cannot fully separate the trend we observe in violence and this measurement error, we will need to be somewhat cautious when basing inferences off of observed trends. 22 Figure 2.4. Number of Countries with Ongoing Intrastate Wars, MEPV, 1946–2015 SOURCE: Marshall, 2016. Figure 2.5. Number of Countries with Ongoing Intrastate Wars, by Intensity, MEPV, 1946–2015 SOURCE: Marshall, 2016. 23 Ethnic and Civil Conflicts The general trend reported for intrastate conflict appears to apply to both civil or revolutionary conflicts (with a political motivation) and ethnic conflicts (with an ethnic or identity group basis). Importantly, although ethnic conflict has been more common than civil conflict, the two types of conflict appear to have followed similar trends when considered in the aggregate (Figure 2.6). Both ethnic and civil conflict rose to a peak in the late 1970s and 1980s, before falling in the mid-1990s. This decline continued through about 2008, and since that time both ethnic and civil conflict appear to have increased slightly. These trends indicate that both civil and ethnic conflict contribute to the recent uptick in intrastate violence discussed above. Figures 2.7 and 2.8 show the total number of civil and ethnic conflicts recorded by the MEPV data set from 1946 to 2015 by intensity of conflict. The conflict in Libya is counted as a low-intensity civil conflict, while that in Syria is an example of a medium-intensity ethnic war within the MEPV data set.7 Figure 2.7 shows civil violence by intensity level. Low-intensity civil violence has been most common over most of the period under consideration, with the exception of a surge in medium intensity civil violence in the mid-1980s. Low-intensity civil violence reached a peak in the 1970s before declining. In recent years, high-intensity civil conflict appears to have been substantially reduced, but there has been an increase in low- and medium-intensity violence. Figure 2.8 shows trends for ethnic conflict by intensity. Ethnic conflict at high intensities has been virtually eliminated, falling consistently since the 1980s. In contrast, levels of ethnic conflict at low and medium intensities do not follow the downward trend evidenced in most other forms of violence. Medium-intensity ethnic conflict reached a peak in the 1980s and has remained at this level since then, with some fluctuations. Low-intensity ethnic conflict, which has been the most prevalent form of ethnic conflict since 1970, reached a peak in the early 1990s and fell sharply until about 2002 and then began to rise again after 2004. These trends suggest that low- and medium-intensity ethnic violence present a continued security concern and may do so for the foreseeable future. Notably, the caveat about possible bias introduced by advances in the scope and quality of reporting should be taken into account in our interpretation of these trends. 7 It is worth noting that in other data sets, such as the UCDP data set, Syria would be considered a high-intensity civil war. 24 Figure 2.6. Number of Countries with Ongoing Civil and Ethnic Wars, MEPV, 1946–2015 SOURCE: Marshall, 2016. Figure 2.7. Number of Conflicts with Ongoing Civil/Revolutionary Wars, by Intensity, MEPV, 1946–2015 SOURCE: Marshall, 2016. 25 Figure 2.8. Number of Countries with Ongoing Ethnic Conflict, by Intensity, MEPV, 1946–2008 SOURCE: Marshall, 2010. Intrastate Violence That Falls Short of War Intrastate violence may also come in forms that fall short of armed conflict or that do not involve the use of violence—for example, protests, riots, demonstrations, and strikes. These events may be significant for many reasons, including the fact that they can lead to additional violence and sometimes escalate into civil or ethnic wars. Furthermore, if intrastate violence causes instability in a country or a region that has implications for demands on U.S. Army forces, this could be just as important as armed conflict. The CNTS Databank is our primary source of information on these events.8 It includes information on a number of different types of societal violence, including assassinations, protests, riots, demonstrations, strikes, and purges over the period 1919–2015. The data set has some limitations: It does not include information on fatalities (and thus cannot be precisely disaggregated by intensity) and it does not separate organized societal violence from spontaneous violence. This is an important distinction when considering overall trends. In some cases, it is possible to infer the intensity or spontaneity of an event from the definition or type of violence, but this may not be true in all cases. For example, most assassinations have relatively low death counts and are planned. Peaceful demonstrations would also have minimum violence levels, but may be organized or spontaneous. Riots are most likely to be low intensity and are often unplanned, but purges may involve death rates at medium intensity or even higher. The quality 8 CNTS Data Archive, homepage, 2016. 26 of the CNTS data is also affected by reporting and media coverage. Events such as protests and demonstrations might not always be covered by local or national media, especially in developing countries and especially in the early decades of our analysis. As for previous analyses, improved reporting may lead to an apparent increase in number of these types of incidents over time, even if the true global incidence is the same. Each event type collected in the CNTS data has its own trajectory and trend over the period under consideration. However, one notable similarity shared by each event type is the fact that, with the exception of revolutions, they don’t clearly follow the same downward trend observed for most types of intrastate conflict over this same period. Instead, there is considerable fluctuation in all types of events from 1946 through today, with several (including antigovernment demonstrations, guerilla warfare, and riots) becoming more frequent in recent years, driven by events in the Middle East and elsewhere. This suggests that while there has been a general decline in more significant forms of conflict, low-level political violence remains a more serious concern, particularly in specific regions and countries. A brief summary of some of the more specific trends may offer insight into the drivers of these types of violence. Readers will note that we do not include all CNTS variables in this discussion, only a few of particular interest. Assassinations are likely to be planned events and to have a small number of fatalities, probably falling in the minimum-violence or low-intensity categories of our typology. Figure 2.9 shows that assassinations became increasingly frequent after 1970, reaching peaks in the late 1970s and early 1990s before declining. The number of assassinations increased between 2000 and 2005, but fell between 2005 and 2009 before rising again. Some of the peaks can be explained in the context of ongoing events, including long-simmering terror campaigns, political instability, or insurgencies. For instance, the peaks in assassinations in the 1970s and 1980s were driven by violence in such countries as Guatemala, Argentina, El Salvador, Turkey, and Spain. The smaller peak in 2004 is driven almost entirely by violence in Colombia. Although the total number of assassinations per year has fallen since its peak, this trend has consistently sloped upward since 2009, with the exception of a slight drop in 2015. This increase has been driven by an increase in assassinations in a number of countries, including Somalia, Syria, Turkey, Afghanistan, and Colombia. Assassinations may be a particularly important measure of low-level intrastate conflict because they may be associated with the crime-, gang-, and drug-related violence that remain highly pervasive and damaging but difficult to measure in many regions. Locations and time periods with high criminal violence may also have high rates of assassination, sometimes executed by similar groups of people. 27 Figure 2.9. Assassinations, CNTS, 1919–2015 SOURCE: CNTS Data Archive, 2015. NOTE: The CNTS data set does not include data for the years 1939–1946. We have chosen to present a continuous line, but have not included the omitted years on the x-axis. 28 Trends in nonviolent antigovernment demonstrations (the full series is shown in Figure 2.10; Figure 2.11 truncates the number of instances shown on the graph to provide a better sense of the variation in instances at lower levels) show a somewhat volatile trend, with a gradual increase from the early 1970s through about 1993 and then a gradual decrease until 2010, when the Arab Spring led to a dramatic increase. The slow increase during the 1970s and 1980s may have been driven by a number of causes, including growing unrest in Soviet-bloc countries (such as Poland), as well as events in Southeast Asia and in countries such as South Korea and the Philippines, whereas more-recent increases have been driven largely by the Arab Spring and subsequent unrest in the Middle East, as well as by unrelated turmoil in Southeast Asia, Africa, and Latin America. However, as noted elsewhere, some of the apparent increase in the number of demonstations is likely due to underreporting of these demonstrations in past years because of general trends in media coverage of conflict and the underlying sources upon which CNTS relies. Future trends in this form of nonviolent conflict may depend partly on political events in the Middle East, Africa, and elsewhere—and more specifically, on whether political unrest observed over the past several years continues or abates. Figure 2.10. New Antigovernment Demonstrations by Year of Onset, CNTS, Full Series SOURCE: CNTS Data Archive, 2015. NOTE: The CNTS data set does not include data for the years 1939–1946. We have chosen to present a continuous line, but have not included the omitted years on the x-axis. 29 Figure 2.11. New Antigovernment Demonstrations by Year of Onset, CNTS, Truncated Series SOURCE: CNTS Data Archive, 2015. NOTE: The CNTS data set does not include data for the years 1939–1946. We have chosen to present a continuous line, but have not included the omitted years on the x-axis. As defined in the CNTS data set, guerilla warfare (Figures 2.12, and 2.13) includes “any armed activity, sabotage, or bombings carried on by independent bands of citizens or irregular forces and aimed at the overthrow of the present regime.”9 This type of violence may involve conflict between two nonstate groups or a nonstate group and a state, even qualifying as “armed conflict” at sufficiently high levels. In our typology, this type of violence is likely to be organized and to be categorized as low-intensity violence. Figure 2.12 shows the full trend in guerilla war (again, we use the CNTS definition) while Figure 2.13 provides a close-up of the same trend line at points well below its sharp 1970 peak. The rate of guerilla warfare has been relatively stable except for the sharp peak around 1970, driven almost entirely by the Vietnam War, events in Southeast Asia (Cambodia, Laos), and possible measurement error. Outside of this spike, there is some evidence of a slight increase from about 1940 to 1969 and then a slight decrease from about 1989 to 2009. There has also been a sharp upward trend in instances of guerilla warfare in recent years. This trend is driven in part by an increase in the number of countries experiencing at least one incident classified as guerilla warfare according to the CNTS 9 CNTS Data Archive, 2015. 30 definition, but also by a number of countries that have a particularly large number of events in recent years. These countries include Syria, Egypt, Afghanistan, Turkey, Nigeria, and Sudan, among others. This suggests both a spread in scope and a change in frequency or intensity in certain specific locations. Guerilla warfare does seem to be one of the few forms of violence that is enduring and increasing even as other types of conflict begin to disappear, although (as noted elsewhere) the recent increase may be the result of trends in media coverage and awareness of these types of incidents. Our review of various types of intrastate conflict that fall short of war suggests that unlike more intense forms of intrastate violence, these lower-intensity and nonviolent events do not seem to follow an unambiguous downward trend. Instead, while several have declined since their peak, these trends appear to fluctuate much more over time and in recent years seem to be increasing in frequency—as well as scope, in some cases. These forms of conflict may continue to be relevant security concerns, especially when they escalate or trigger instability elsewhere. Figure 2.12. Guerilla Warfare, CNTS, 1919–2015 SOURCE: CNTS Data Archive, 2015. NOTE: The CNTS data set does not include data for the years 1939–1946. We present a continuous line, but have not included the omitted years on the x-axis. 31 Figure 2.13. Guerilla Warfare, Close-Up, CNTS, 1919–2015 SOURCE: CNTS Data Archive, 2015. NOTE: The CNTS data set does not include data for the years 1939–1946. We present a continuous line, but have not included the omitted years on the x-axis. One-Sided State Violence In addition to fighting against rival groups as one of two or more actors in a conflict, states also use violence unilaterally against civilian populations. One-sided state violence can range in intensity from isolated assassinations to larger campaigns of genocide or politicide (systematic destruction of a political entity and its members). Because it does often include only one actor, one-sided state violence may be excluded from standard conflict data sets (the MEPV is a noted exception). Both the UCDP and PITF data sets code one-sided state violence in separate databases. We focus on the PITF data on genocide and politicide because they cover a longer period. Appendix A provides details on the specific trends in one-sided state violence in other databases. As is the case for our other analyses, the data sets differ in what they measure but are similar in the trends that they report. As with aggregate levels of interstate and intrastate conflict, one-sided state violence is becoming less frequent—and less severe when it does happen. It has not disappeared, however. The PITF codes incidents of genopoliticide over the period 1948–2015, rating intensity on a five- 32 point scale based on the number of fatalities. But in this case, fatalities are not strictly “battle deaths,” they include the deaths of civilians targeted by the mass violence. As a result, we may see more genopoliticides at high-intensity levels than was observed for other forms of conflict. Genopoliticides are defined as involving “promotion, execution, and/or implied consent of sustained policies by governing elites or their agents—or in the case of civil war, either of the contending authorities—that result in the deaths of a substantial portion of a communal group or politicized noncommunal group.” In genocides, the victimized groups are identified by their ethnic and religious characteristics. In politicides, targeted groups are identified by political affiliation or opposition group. Authorities are complicit in the killing associated with genopoliticide and the incident is likely to endure over a longer period of time.10 According to the PITF data, the majority of genopoliticide events occurred between the mid1960s and mid-1990s and fall into the medium intensity violence category (see Figure 2.14). As noted above, violence in Rwanda in 1994 is coded as high-intensity genopoliticide. Other recent instances include Sri Lanka (government versus Tamil), where violence is classified as medium intensity; Sudan, where the violence began at high intensity and then lessened; Angola, also medium intensity, and Kosovo, with low intensity. Medium-intensity genopoliticide rises consistently in frequency over the period from 1955 to 1975, reaching a peak between 1975 and 1980 before falling throughout much of the 1980s. There is another spike in the number of these violent episodes around 1990, after which point genopoliticide becomes increasingly infrequent—but, importantly, not nonexistent. Low-intensity genopoliticide appears to have been most common in the mid- to late 1980s and early 1990s, while high-intensity conflict is generally infrequent over the entire period. Since the early 2000s, genopoliticide has been infrequent and occurred mostly at medium intensity. Both intensity levels of genopoliticide reached a low point in 2012 but have risen slightly since then. International norms against state violence and the greater willingness of third parties to intervene if evidence of genopoliticide emerges may explain both the reduction in frequency and intensity since 2000. Another explanation is the spread of democracy, which also tends to limit attacks of this type by the state on nonstate groups and civilian populations. 10 Monty G. Marshall, Ted Robert Gurr, and Barbara Harff, PITF State Failure Problem Set, Data set and Coding Guidelines, 2012. 33 Figure 2.14. Number of Countries with Genopoliticide, by Intensity, PITF, 1955–2015 SOURCE: Marshall, Gurr, and Harff, 2016. Violence by Nonstate Actors Armed Conflict Studying conflict that involves nonstate actors exclusively is complicated by the fact that while some data sets include these conflicts along with those involving states as a form of civil or ethnic war, others separate out those conflicts that involve only nonstate participants. Furthermore, armed conflict between nonstate actors can fall across the boundary between interstate and intrastate conflict, depending on the identity of the groups and the location of the conflict (and thus may span categories in our typology). The PITF and MEPV data sets include violence involving only nonstate groups in their databases, but do not identify this violence with an independent code, so we cannot assess its level or trend apart from the aggregate data. The UCDP and COW data sets also record nonstate conflict, but do so in separate databases. Trends in both data sets suggest a persistent level of nonstate violence in recent years and suggest that armed conflict between nonstate groups will continue to be relevant to conflict in the future. Although the UCDP data set covers only 1989–2014, which is a much shorter time period than the COW data set (1816-2007), the UCDP data are more current and the coding of nonstate wars more consistent and comprehensive. As a result, we focus our analysis on the UCDP data for our discussion of nonstate conflict and refer readers to Appendix A for additional details. 34 The UCDP nonstate conflict data set includes “communal and armed conflict between two organized groups, neither of which is the government of a state, which results in at least 25 battle deaths in a year.”11 Figures 2.15 and 2.16 show trends in low- and medium-intensity nonstate conflict, where the majority of nonstate violence falls. Medium-intensity nonstate conflict has been relatively episodic and infrequent. In contrast, low-intensity violence has occurred more frequently and has shown significant variation over the period under consideration. Lowintensity nonstate conflict reached peaks in the early and late 1990s, before falling consistently between 1999 and 2007. However, recent years (since 2009) reveal another significant increase in this type of violence, with the number of ongoing low-intensity nonstate actor conflicts more than doubling since 2010. This increase represents a diffusion of nonstate actor conflict and an increase in the number of specific instances of nonstate actor conflict in certain countries, including Somalia, Kenya, Mexico, Sudan, Syria, and Nigeria, as well as a proliferation of conflicts involving ISIS. Neither medium- nor low-intensity nonstate conflict appears to be decreasing according to the UCDP data, suggesting possible persistence in conflict at the nonstate or societal level even as conflicts involving state actors begin to decline. Importantly, because of its reduced time frame, this trend line is unlikely to be affected by measurement error in the same way that other data on low-intensity conflict might be. 11 Ralph Sundberg, Kristine Eck, and Joakim Kreutz, “Introducing the UCDP Non-State Conflict Data set,” Journal of Peace Research, Vol. 49, No. 2, 2012, pp. 351–362. Also see Uppsala Conflict Data Project, Armed Conflict Data set Codebook, 4-2012, Centre for the Study of Civil Wars, International Peace Research Institute, Oslo (PRIO), 2012. 35 Figure 2.15. Ongoing Nonstate Conflicts, Low Intensity, UCDP, 1989–2014 SOURCE: Sundberg, Eck, and Kreutz, 2012; UCDP Non-State Conflict Dataset v. 2.5-2015, 1989-2014. Figure 2.16. Ongoing Nonstate Conflicts, Medium Intensity, UCDP, 1989–2015 SOURCE: Sundberg, Eck, and Kreutz, 2012; UCDP Non-State Conflict Dataset v. 2.5-2015, 1989-2014. 36 Terrorism Terrorism is another form of violence also perpetrated largely by nonstate actors. (Although there is state-sponsored terrorism, we do not explicitly consider that type of violence in this subsection.) Terrorism can be interstate, transnational, or intrastate (also known as domestic). The intensity of a terrorist attack can range from low (and even no fatalities) to medium, according to our typology. Although terrorism is similar to other forms of political violence in many ways, it has a number of unique characteristics. The GTD, one of the major databases of terrorist events, covers domestic and transnational terrorism since 1970. It defines terrorism as an “intentional act of violence or threat of violence by a nonstate actor.” It also requires that the incident meet two of the following three criteria: (1) The violent act was aimed at attaining a political, economic, religious, or social goal; (2) The violent act included evidence of an intention to coerce, intimidate, or convey some other message to a larger audience (or audiences) other than the immediate victims; and (3) The violent act was outside the precepts of International Humanitarian Law. 12 This final requirement essentially means that the violence will target civilians, rather than combatants. Attacks are coded by the number of fatalities, their location, target, tactic, and (where possible) the perpetrating group. Like other forms of nonstate conflict, terrorist attacks do not seem to be following the general downward trend observed for more-traditional forms of armed conflict. An analysis of the GTD data reveals a number of key observations. First, the majority of terror attacks actually cause no deaths. Although this may seem counterintuitive, terrorists often use violence to call attention to themselves or their cause, and would rather have lots of people watching the attack than lots of people killed by the attack. Those that do cause deaths typically remain in the low-intensity category, simply because few terror attacks cause the 1,000 deaths needed to make it to the “medium” intensity level of violence (although the total deaths from a terrorist campaign may reach this level easily). Figures 2.17 through 2.19 show the number of terror attacks for each level of intensity. Figure 2.17 identifies the four attacks since 1970 that have had more than 1,000 deaths. These include a 1994 attack on Rwandan Tutsi refugees by Hutus; the 9/11 attacks in New York City; and a 2014 ISIS-claimed abduction of more than 1,500 Iraqi soldiers, the majority of whom were later killed. It is worth noting that there are a number of problems with the GTD data set, largely in the way events are collected and coded. First, the data set has changed hands many times, and this has reduced the consistency with which coding rules are applied over time. Second, the sources included in the data set change often, meaning that a perceived spike in the number of attacks could instead reflect access to a new data source. Third, the data set has recently included attacks on soldiers during the course of a war by militant 12 National Consortium for the Study of Terrorism and Responses to Terrorism (START), Global Terrorism Database Codebook, October 2012. 37 groups as terrorism. In most conceptions of terrorism, these attacks would be considered part of the ongoing war and not as terrorism. In fact, the 2014 ISIS abduction and killing of Iraqi soldiers noted above would fall into this category—ISIS and Iraq are at war and the attack was conducted as part of this continuing conflict. The inclusion of this and similar attacks in the GTD may artificially inflate the number of terrorist attacks reported in any given year. Figure 2.18 shows that low-intensity terror attacks increased sharply in the late 1970s before falling in the mid-1980s and rising again in the 1990s. The number of low-intensity attacks also has increased since 2001, declining slightly between 2009 and 2012 before rising exponentially in 2013 and 2014, with only a small decline in 2015. Minimum-violence attacks have also risen dramatically since 2007, after a decline that began in the early 1990s (Figure 2.19). The increases in the trend lines for low-intensity and minimum-violence terrorist attacks are driven by a number of factors. First, there has been a real increase in these types of attacks in places like Iraq, Syria, and Somalia. Second, as noted above, the criteria that the GTD has used to identify terrorist incidents have also changed over time and now include certain types of attacks that had formerly been excluded. Notably, many actions that would previously have been considered part of an armed conflict in places such as Iraq are now being classified as terrorist attacks. Finally, as also noted above, the sources included in GTD’s analysis also change over time. As the project has progressed, researchers have gained access to new sources that have provided better data and have allowed researchers to identify a greater number of attacks. At the same time, media attention to terrorist attacks has increased following the 9/11 attacks. However, while some of the observed increase may be due to reporting and definition issues, low-intensity and minimum-violence terror attacks appear to have become more common, even as other types of conflict have become more infrequent. Trends in terror attacks that do not cause fatalities or that cause very few fatalities remain relevant because even if they cause no death, they can still contribute to instability, retaliation, future conflict, and escalation of ongoing conflicts. In addition, the sheer number of these attacks and the uncertainty that they cause for military and political actors reinforce their continued relevance. 38 Figure 2.17. Number of Terror Attacks, Medium Intensity (Greater Than 1,000 Deaths), GTD, 1970– 2015 SOURCE: START, 2016. 39 Figure 2.18. Number of Terror Attacks, Low Intensity (Deaths Between 25 and 1,000), GTD, 1970– 2015 SOURCE: START, 2016. Figure 2.19. Number of Terror Attacks, Minimum Violence (Fewer Than 25 Deaths), GTD, 1970– 2015 SOURCE: START, 2015. 0 1000 2000 3000 4000 5000 6000 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1995 1997 1999 2001 2003 2005 2007 2009 2011 Number of Attacks Minimum Violence 40 Summary This chapter has summarized trends in conflict across major data sets, conflict types, and intensities, focusing on the period between 1946 and the present. Although each of our data sets uses its own definitions of conflict (or interventions or terrorisms or disputes) and includes slightly different sets of events, the general trends revealed by each of the data sets are largely the same. At the global level and in aggregate terms, conflict has declined in frequency and intensity. This downward trend is clearest and strongest for interstate conflicts and high-intensity conflict. The downward trend in intrastate conflict and low-intensity violence has been somewhat slower and less decisive, including an uptick in 2012–2015. This pattern generally holds, whether we consider ethnic conflict, other forms of low-intensity violence that involve nonstate actors or that occur at the societal level (such as riots, assassinations, guerilla warfare, and terrorism), or more-formal armed conflict between nonstate groups. The decline in interstate crises and disputes that fall short of war and do not involve violence also has been more muted (those that become violent have become less likely). This suggests that international dispute resolution mechanisms might be able to prevent escalation to war, but also that there is a potential for renewed conflict, perhaps just below the surface. The review has touched on all cells in the conflict typology (see Table 2.1), highlighting major trends, patterns that extend across data sets and conflict types, and observations that emerge from specific data sources and that apply most directly to particular conflict types. It is also possible to describe overall trends in conflict by considering how the distribution of conflict has moved across the typology matrix (see Table 2.3). Specifically, while medium- to highintensity interstate violence used to be the predominant form of conflict, lower-intensity forms of violence involving nonstate actors and occurring at the societal level have become more common over time. In other words, we can imagine the concentration of conflict shifting from the upper left to lower right squares in the matrix. From this perspective, conflict has not been eliminated, but rather has changed in nature and intensity. 41 Table 2.3. Conflict Trends and Typology Intrastate Interstate One-Sided State State v. Nonstate Organized Societal Spontaneous Societal High Intensity: Battle deaths of 100,000 per year or 500,000 total Medium Intensity: Battle deaths of 1,000 per year or 100,000 total Low Intensity: Battle deaths of 25 per year or 1,000 total Minimum Violence: Battle deaths of 0–25 per year; nonviolent conflict/dispute As noted earlier, the typology will exclude certain forms of criminal violence because of data limitations, and the potentially biasing effect of changes in media reporting could artificially inflate low-intensity conflict incidence. The observed shift in the frequency and intensity of conflict naturally raises the question of what might drive this change and whether these trends might be reversible. There are many domestic, international, and systemic factors that may contribute to the trends we have described in this chapter and that might also contribute to trends in future conflict, including both further reduction and a possible increase in the amount of violence. These key factors and how they might affect conflict trends is the subject of Chapter Three. High-intensity conflict was more common in the 1980s and early 1990s Since ~2001, conflict has been increasingly lowintensity and intrastate 42 3. Operationalizing and Projecting Key Factors That Will Influence the Future Incidence of Conflict Introduction If conflict type and intensity has changed over the past few decades, what are some of the reasons for the shifts, and what can we expect in the future? To address these questions, we conducted an extensive review of the massive scholarly literature explaining conflict or the lack of it. We considered a large number of factors that might have affected the prevalence of both interstate and intrastate conflict. Appendix C presents the results of our review. In this chapter, we offer a summary of the factors we identified as most important from our literature review. In subsequent chapters, we integrate data from these factors into a more nuanced set of baseline projections for future conflict incidence and intensity. Although some evidence exists to support a wide range of potential factors, our review identified ten key factors whose effects on future conflict trends are likely to be most significant. Culling this list to ten was a difficult process. Violent conflict is a complex phenomenon, with any single instance usually attributable to the interaction of several different causes. Explaining the global incidence of conflict is even more complex; the literature investigating violent conflict is therefore correspondingly complex. Social scientists have proposed dozens of variables to explain differing patterns of conflict. Each of these variables, in turn, can be measured in multiple ways.1 Nonetheless, our review of the literature indicated that some variables have a much better established relationship with conflict than others. The research team identified ten such variables where the link with conflict levels was clear, even if the relative scale of the effects remains a subject of debate. Many variables investigated in the literature on conflict were not included in our list. The reasons for their exclusion varied, but fell generally into three categories. First, for factors such as the effects of prior conflicts and state repression, there is scholarly debate about whether these factors are themselves causes of conflict (independent variables) or are expressions of deeper, underlying causes. Second, many factors (such as the presence of alliances or rivalries), while likely to affect conflict at the dyadic or state level, were judged to be relatively consistent features of the international system, and therefore unlikely to greatly affect overall future conflict trends. Third, there were several factors (such as the empowerment of women or aging populations) that, while they were potentially promising in explaining conflict trends, had not yet 1 For example, one survey of the literature on intrastate conflict alone published seven years ago assessed 89 distinct variables; the number of variables proposed by social scientists has grown since then. See Håvard Hegre and Nicholas Sambanis, “Sensitivity Analysis of Empirical Results on Civil War Onset,” Journal of Conflict Resolution, Vol. 50, No. 4, August 2006, pp. 508–535. 43 been assessed robustly in the literature. A complete discussion of all of the factors we reviewed is included in Appendix C. The ten key factors we selected are listed here with a short definition. • The capacity of state institutions. Institutional capacity affects a state’s ability to provide public goods, such as infrastructure or security, to their populations and to maintain effective and disciplined security services. The capacity of state institutions is often conceptualized as three-dimensional, composed of military capacity, administrative/bureaucratic capacity, and political institutional coherence and quality. • The prevalence of consolidated democracies. The distribution and nature of political regime types around the world has varied over time. Consolidated democracies are those that have successfully implemented a range of effective institutional rules and legal procedures constraining the executive, mandating popular election of political leaders, and ensuring civil, political, and minority rights. • The degree of ethnic and sectarian polarization. Almost all of the world’s states are composed of multiple ethnic and sectarian groups, but ethnic and religious identities do not always act as societal cleavages. Societies become polarized along ethnic or sectarian lines as ethnicity and/or religion becomes an important factor for group identification and forms the basis for political organization and the lens through which societal grievances are framed. • The rate of economic growth. Economies grow at different rates. Expressing the rate of growth in percentage terms from one period to another allows for a calculation of the change in the state’s economy in an overall sense and allows for a comparison of rate of growth to other states. High growth rates lead to faster attainment of a developed economy. Highly developed states are less prone to internal conflict. Low economic growth or differential growth rates across groups in a society can fuel grievances that may lead to conflict. • The extent of economic interdependence. Economic interdependence refers to how interrelated states’ economies are with each other in particular and with the global economy in general. There are two key characteristics of economic interdependence that influence countries’ likelihood of conflict. First, countries’ economic outcomes are affected by external conditions, such as demand in another country or a global economic shock. Second, economic interdependence means that disrupting a country’s ties to the international economy would hurt its domestic economy. • The degree of U.S. preeminence. The international state system is characterized by hierarchy in terms of power. The share of power of the most powerful state within the international state system and vis-à-vis the potential challenger to that state’s power position expresses the degree of preeminence of that state. U.S. preeminence is built on the preponderance of U.S. military power and its large share of the international 44 economy, the U.S. central role in international governance, and the U.S. position as a supporter and enforcer of international norms. • The capabilities of international organizations. International organizations can undertake key tasks in the international system, including developing solutions to crossborder problems, mediating disputes, shaping and enforcing of international norms, disseminating information, and generating shared interests and potentially aligning states’ preferences. The capabilities of international organizations to undertake these tasks depend on the amount of authority and resources that states delegate to them. • The strength of international norms. International norms represent collective expectations for the proper behavior of state actors. Strong international norms are ones that are relatively universally held and for which there is relatively universal consensus for the need for norm enforcement. Weaker norms may only be held by (or applied to) some states, or may be held more broadly but for which there is little support when it comes to punishing states that violate the norm. • The diffusion of lethal technology. Diffusion of lethal technology refers to greater access by states and nonstate actors to the technologies necessary to build and deploy lethal weapons, including nuclear, chemical, biological, radiological weapons, precision munitions, or disruptive cybertechnology. • The extent of resource stress because of population pressures. Resource stress arises from the scarcity of renewable resources, such as water and arable land, to support the population living in the area. Resource scarcity increases environmental insecurity. The main types of resource scarcity include: (1) supply-induced scarcity, in which resources are consumed at a faster rate than they can be regenerated; (2) demand-induced scarcity, in which previously stable resource consumption increases through an increase in population or increased consumption per capita; and (3) structural scarcity, in which the distribution of resources is uneven and some groups have limited access to resources. The sections of this chapter examine each of these factors in turn. The discussion of each factor includes: • an explanation of why the social science literature identifies the factor as a critical determinant of violent conflict • potential ways to translate each of these factors into a quantifiable variable that can be used to project the future incidence of violent conflict • an explanation of the specific metrics used for each factor in the remainder of this report • an exploration of “branch” scenarios, in which the key factor takes on values above or below the baseline projection • discussion of potential discontinuities (or “shocks”) that may lead to much greater changes in the key factor than those that are generally anticipated. 45 Our operationalizations of the key factors presented should be treated with caution, as the metrics selected provide only a rough proxy measurement for the factors identified by the literature review. In some cases, the factor was difficult to operationalize in one or two metrics; in other cases, limited publicly available data meant that we needed to employ more indirect metrics. We encourage analysts interested in the changing effects of a given factor on conflict to think broadly about potential signs of the factor’s increase or decrease, rather than focusing only on the metrics illustrated herein. These caveats notwithstanding, in most cases, the metrics discussed here show clear trends that allow us to project them into the future. We stress that these are projections, based on extrapolations of historical trends, and they are not predictions about the future. They do, however, begin to help us understand the factors that make some future levels of conflict more likely to occur than others. While useful, a single projected baseline trend by itself can be misleading, especially from a perspective beyond the immediate future. Consequently, each section in this chapter examines three different projections for the key factor under review: a baseline scenario that represents the “continuation of current trends” (or “expected future”) projection for that factor, and two branch scenarios, one above and one below the baseline scenario. Several different methods were used to project the baseline scenarios, and the method used is specified in the discussion of each key factor. Overall, these methods can be divided into two categories. First, baseline projections for several of the key factors were calculated using the International Futures project, a sophisticated modeling tool that incorporates a range of economic, demographic, and social variables into its analysis.2 Second, baseline projections were also calculated through the application of a trend line to historical data. The pattern and nature of the historical data determined the type of trend line that could be used; in instances where multiple trend lines were plausible, the one that had the greatest degree of fit with the historical data (as determined by the R2 value) was used.3 The branch scenarios were then typically 2 The International Futures project, along with a detailed description of its methodology, can be found at its website.The International Futures (IFs) tool models the interaction of a series of dynamic, thematic modules, incorporating economic, demographic, sociopolitical, educational, health, international political, agricultural, energy, and infrastructure data. This interactive analysis is performed for 86 countries, drawing on up to 40 years of historical data. As the IFs tool models conditions based on current and historical data, it is not ideal for predicting highly unlikely, “black swan” events that represent sharp discontinuities with the past, although it is possible to modify the parameters of the IFs tool to conduct sensitivity or scenario analyses. In our work, we employed the IFs tool to produce estimates of trends in economic or demographic factors for baseline, “no surprises” futures. 3 We attempted to calculate these trend lines based on historical data from the entire post–1945 period. There were, however, numerous exceptions. Many sources of data were not available for a wide range of countries before 1960, while in other cases, data were not widely available prior to 1970 or even 1990. In addition, for one metric, the prevalence of established democracy in independent states, data from the 1946–1959 period were judged to be misleading because they did not account for the characteristics of still widely prevalent colonial regimes. For this reason, a later time period of post-1960 was used for this metric. It is also important to note that near-term projections may diverge noticeably from the most recent historical data in several cases, and appear to “jump” up or down because they are based on a longer-term historical trend that might or might not closely fit the most recent 46 calculated as one standard deviation above and below the baseline, with certain exceptions to account for greater potential variation in some factors as we will note. Taken together, these projections mark the range of the most-likely future values of the key factor. The branch projections effectively function as a confidence interval for our baseline projection, establishing how far the baseline scenario might deviate in conditions of gradual change. Nonlinear, discontinuous changes in key factors are certainly possible, however, and there could be reasons to think they are more likely to occur with some factors than others. For that reason, we include in the overview of each factor a discussion of the dynamics that might lead to such discontinuous changes in the future. Capacity of State Institutions Our literature review identified the capacity of state institutions as one of the primary factors affecting the likelihood of intrastate conflict.4 However, this factor did not appear to exercise a significant effect on interstate conflict. States with high levels of capacity are better able to avoid intrastate conflict for two primary reasons. First, their greater capacity allows them to provide public goods, such as infrastructure or security, to their populations; this, in turn reduces the frequency of grievances against the government.5 Second, such states are better able to maintain effective and disciplined security services that can deter or defeat internal armed challenges.6 There have been many attempts to operationalize state capacity and both the quality and extent of governance. The metrics include the ratio between tax revenues and gross domestic product (GDP), GDP per capita, government spending, and military spending. All of these metrics have potential merit. However, reliable public data on such metrics as taxation are not readily available, and measures such as GDP per capita are not clearly distinct from other key factors, such as economic growth. We selected a measure of the change in the share of the population with access to public goods, as that would capture the extent to which the capacity of observations. The value of these projections rests in their direction and slope rather than their specific values— which, in any event, represent only proxy indicators for the key factors. 4 For example, see James D. Fearon and David D. Laitin, “Ethnicity, Insurgency, and Civil War,” American Political Science Review, Vol. 97, No. 1, 2003, pp. 75–90; Paul Collier, V. L. Elliott, Havard Hegre, Anke Hoeffler, et al., Breaking the Conflict Trap: Civil War and Development Policy, Washington, D.C.: World Bank and Oxford University Press, 2003; and Halvard Buhaug, “Relative Capability and Rebel Objective in Civil War,” Journal of Peace Research, Vol. 43, No. 6, 2006, pp. 691–708. 5 Ted Robert Gurr, Why Men Rebel, Princeton, N.J.: Princeton University Press, 1970; Ted Robert Gurr, “Why Minorities Rebel: A Global Analysis of Communal Mobilization and Conflict since 1945,” International Political Science Review, Vol. 14, No. 2, 1993, pp. 161–201; Fearon, James D., “Self-Enforcing Democracy,” Quarterly Journal of Economics, Vol. 126, No. 4, November 2011, pp. 1661–1708. 6 Nicole Ball, “Strengthening Democratic Governance of the Security Sector in Conflict-Affected Countries,” Public Administration and Development, Vol. 25, No. 1, 2005, pp. 25–38; Jeffrey Herbst, States and Power in Africa: Comparative Lessons in Authority and Control, Princeton, N.J.: Princeton University Press, 2000; Hegre and Sambanis, 2006. 47 the state is increasing or decreasing over time, while minimizing the concern that changes in the metric were, in fact, driven by a different key factor, such as the rate of economic growth. For our specific metric, we calculated the percentage of states where less than 50 percent of the rural population has access to an improved water source, using data available from the World Bank. The rationale for selecting this metric was to identify the prevalence of states that are likely to have extremely low levels of access and therefore be more likely to become involved in intrastate conflict. Figure 3.1 shows the historical data for this metric in black. Figure 3.1. Poor Access to Improved Water Sources SOURCE: Historical data: World Bank, World Development Indicators 2013, Washington, D.C., 2013; projections calculated by authors. To project the baseline, or most likely, scenario, we fit an exponential trend line to the available historical data and calculated the future values.7 This baseline scenario is shown in the figure in gray. To project the two branch scenarios, we calculated values of one standard deviation above and below the baseline trend. Those scenarios are displayed in the figure in red and blue, respectively. The analysis therefore suggests that the percentage of states with extremely low access is likely to range between 0 and 13 percent by 2040, with the “expected future” scenario at less than 5 percent. 7 The exponential trend line fit to the data has the equation y = 0.2679e-0.044x. This trend line had a high level of fit to the data, with an R2 of 0.92. 48 There are a number of factors that could lead to discontinuous changes in this metric. For example, a global economic downturn could lead to sharp reductions in development aid that could threaten any improvements in governance that had been made with the benefit of that assistance. Alternatively, a major jump in global energy costs could induce widespread fiscal crises similar to those of the 1970s and 1980s, ultimately leading to reduced state capacity. The branch scenario in red projects essentially no improvements in state capacity over the period in question, so larger changes would entail the rapid erosion of the capacity of states that are already relatively capable. Such changes are certainly possible, but they would constitute an example of nonlinear disruptive change. On balance, our projection is that state capacity is likely to continue to improve, which will tend to exert a downward pressure on the likelihood of intrastate conflict. Prevalence of Consolidated Democracies Consolidated democracies are less likely to fight one another and to be involved in internal conflict. While this correlation is clear, the mechanism by which democracies reduce conflict is more contested. The literature on interstate conflict has focused on the greater transparency and consistency of democratic regimes that allow them to credibly commit to peaceful solutions to disputes and the possibility that domestic norms and greater political accountability may make democratic states less likely to pursue violent conflicts.8 There are fewer arguments that the greater ability of consolidated democracies to resolve grievances within the political system leads to less intrastate conflict.9 We note, however, that partial democracies or the process of democratization itself may not be particularly peaceful and may even be associated with an increase in conflict.10 Given the importance of consolidated democracy in the literature on conflict, there is already a great deal of work in the academic literature on measurement of democracy. Several aggregate measures of democracy have been developed that include the competitiveness of elections; the state’s respect for civil, political, and minority rights; and freedoms of the press and religion. The most widely used measure of consolidated democracy, and the one we employ, comes from the Polity project. By coding a wide range of regime characteristics, such as political 8 Arend Lijphart, Democracies: Patterns of Majoritarian and Consensus Government in Twenty-One Countries, New Haven, Conn., and London: Yale University Press, 1984; Peter Liberman, Does Conquest Pay? The Exploitation of Occupied Industrial Societies, Princeton, N.J.: Princeton University Press, 1996; Charles Lipson, Reliable Partners: How Democracies Have Made a Separate Peace, Princeton, N.J.: Princeton University Press, 2003. 9 Håvard Hegre, Tanja Ellingsen, Scott Gates, and Nils Petter Gleditsch, “Toward a Democratic Civil Peace? Democracy, Political Change, and Civil War, 1816–1992,” American Political Science Review, Vol. 95, No. 1, March 2001, pp. 33–48; Christian Davenport, State Repression and the Domestic Democratic Peace, New York, N.Y.: Cambridge University Press, 2007. 10 Hegre et al., 2001; Lars-Erik Cederman, Kristian Skrede Gleditsch, and Simon Hug, “Elections and Ethnic Civil War,” Comparative Political Studies, Vol. 46, No. 3, 2012, pp. 387–417. 49 competition and constraints on the executive, an aggregate “Polity score” is produced, ranging from –10 to 10. Values of 6 or higher are typically used to identify the presence of a democracy, with a more conservative measurement of 8 or higher often used to identify consolidated democracies. We use this metric to calculate the percentage of all states that are consolidated democracies, with the historical values denoted by the black line in Figure 3.2. Figure 3.2. Prevalence of Consolidated Democracies SOURCE: Historical data: Monty G. Marshall and Keith Jaggers, Polity IV Data Set [Computer file; version p4v2012], College Park, Md.: Center for International Development and Conflict Management, University of Maryland, 2002; projections calculated by authors. We projected the baseline scenario by fitting a trend line to the historical data and calculating the future values.11 This projection is represented in the figure by the gray line. We calculated the two branch scenarios as one standard deviation above and below the baseline projection; they are shown in the figure by the blue and red lines, respectively. Discontinuous growth in the prevalence of democracies could result from various tippingpoint effects. If a high percentage of the world’s population were governed through democracy, other forms of government may come to be seen as illegitimate, and greater international pressure may be brought to bear to remove them. Within the time frame of our study, a dramatic reversal in the prevalence of consolidated democracies appears to be less likely. The reversion of 11 The trend line was fit using a generalized linear model linked to a binomial logit function. The resulting projections are therefore bounded between 0 and 1 (in this case, 0 and 100 percent). The model used in Figure 3.2 has a Pearson statistic (1/df) of .0037, suggesting a high degree of fit with the data. 50 consolidated democracies to autocracies historically has been extremely rare and is unlikely in the absence of extreme economic decline, the conquest of democracies by more powerful autocracies, or both. Either of these potential paths is likely to lead directly to increases in future conflict levels as well. Degree of Ethnic and Sectarian Polarization The academic literature generally agrees that a high level of ethnic and sectarian polarization is not sufficient by itself to cause conflict either within or between states. However, there is also agreement that in the intrastate context, where group mobilization occurs along ethnic lines, identity can become a significant contributing factor for violence, especially when strengthened by socioeconomic and sociopolitical grievances. Consequently, we identified the degree of ethnic and sectarian polarization as one of the primary factors likely to affect the level of intrastate conflict in the future. Evidence also shows that while ethnicity may not lead to conflict by itself, it may work to prolong conflicts and increase the intensity of violence in those conflicts that are already occurring.12 Such effects will most likely be strengthened if ethnic groups are deliberately disadvantaged by the state or if they are territorially based and have secessionist or separatist demands.13 Scholars tend to agree that ethnic and sectarian polarization, while a strong predictor for increased levels of intrastate conflict, is not a strong driver for conflicts between states. However, if regional and international actors become involved in intrastate conflicts, or if conflicts spread across borders, such polarization could also affect levels of interstate conflict. Such a scenario is especially likely where ethnic kin-groups in neighboring states become involved with secessionist movements.14 Quantifying ethnic and sectarian polarization is inherently difficult. While various measures have been tried, such as linguistic differences (e.g., ethno-linguistic fractionalization) or religious preferences, they are often criticized for not capturing the cleavage that gives rise to political mobilization. For example, different ethnic groups may share the same religion, and one ethnic group may speak multiple languages. It can also be difficult to determine when certain identities in a society are increasing in salience, and when they are becoming less relevant. One prominent 12 Rajat Ganguly and Raymond Taras, Understanding Ethnic Conflict: The International Dimension, Longman Publishers, 2002; Fearon and Laitin, 2003; and Daniel Bar-Tal, “Sociopsychological Foundations of Intractable Conflicts,” American Behavioral Scientist, Vol. 50, No. 11, 2007. 13 Gurr, 1970; Stephen M. Saideman, and William R. Ayres, “Determining the Causes of Irredentism: Logit Analyses of Minorities at Risk Data from the 1980s and 1990s,” Journal of Politics, Vol. 62, No. 4, November 2000, pp. 1126–1144; Monica Duffy Toft, The Geography of Ethnic Violence: Identity, Interests, and the Indivisibility of Territory, Princeton, N.J.: Princeton University Press, 2003. 14 John A. Vasquez, and Brendan Valeriano, “Territory as a Source of Conflict and a Road to Peace,” in Jacob Bercovitch, Viktor Kremenyuk, and I. William Zartman, eds., The Sage Handbook of Conflict Resolution, Los Angeles, Calif.: SAGE, 2009, pp. 193–209. 51 attempt to quantify relevant ethnic identities is the Minorities at Risk data set at the University of Maryland, which identifies minority groups by their “at risk” status—that is, by the extent to which they are disadvantaged in their relationships with other groups in the state in which they reside. However, the Minorities at Risk data identify such “at risk” groups somewhat subjectively, and the project does not claim to be comprehensive. An alternative, objective measure is to look for the degree of formal discrimination against ethnic, religious, or linguistic groups. The creation or removal of official laws providing for formal discrimination can help to identify states where identity-based grievances may become more or less salient. For capturing the degree of ethnic and sectarian polarization, therefore, we looked at the percentage of states with formal discrimination against minorities, where such minority groups make up at least 5 percent of the state’s population. We used the Ethnic Power Relations data set (EPR), which tracks the extent of access to state power for all politically relevant ethnic groups in every country of the world from 1946 to 2013. It includes annual data on more than 733 groups and codes the degree to which their representatives held executive-level state power—from total control of the government to being formally barred from holding political office. While the disadvantage of such a proxy may be that it potentially fails to capture some of the unofficial social discrimination that can lead to group mobilization, the advantage is that it allows for a more objective measure of sectarian tension. The overall levels of ethnic or sectarian polarization in the figure below may therefore be understated, but we can have more confidence in the general trend line than we could with more subjective data sources. We projected the baseline scenario by fitting an exponential trend line to the available historical data and calculating the future values.15 The projection is shown by a gray line in Figure 3.3. We calculated the two branch scenarios as one standard deviation above and below the baseline projection; they are shown by a red and blue line, respectively. 15 The exponential trend line fit to the data has the equation: y = 0.3098e-0.012x. The trend line has a high degree of fit with the data, with an R² of 0.85. 52 Figure 3.3. Percentage of States with Discriminated Minorities SOURCE: Historical data: Andreas Wimmer, Lars-Erik Cederman, and Brian Min, “Ethnic Politics and Armed Conflict: A Configurational Analysis of a New Global Data set,” American Sociological Review, Vol. 74, No. 2, 2009, pp. 316–337; projections calculated by authors. Discontinuous change in this variable may occur as a result of several factors. Historically, ethnic and sectarian factors often have increased in relevance after the breakup of larger states and empires, including the breakup of the Soviet Union in the early 1990s, or the end of colonialism in the 1960s. The breakup of other large, multiethnic states in the future could result in a similar outcome. Extremely high levels of resource stress because of population pressures also could prompt increased ethnically based conflict within states. Ethnic and sectarian polarization and grievances are latent in many societies, and may become politically important in order to mobilize groups to violence under conditions of severe resource or economic privation. Rate of Economic Growth Economic growth affects the prevalence of conflict in several ways. While territorial expansion traditionally has been a major cause of interstate war, states with higher levels of economic development may be less motivated to pursue such expansion
because of the lower relative value of land inputs in an industrialized economy. Moreover, their reliance on international capital markets may increase the potential costs of disruptions from serious 53 international crises.16 At the intrastate level, economic growth (if broadly shared) reduces grievances, bolsters the capacity of the state to handle security challenges, and increases the population’s opportunities for licit employment, thus raising the opportunity costs of participating in rebellions or insurgencies.17 Growth benefits that accrue along ethnic or sectarian lines, however, might increase the potential for intrastate conflict, as discussed in the previous section, and sharp declines in the rate of economic growth could be associated with an increased risk of internal conflict as well.18 Therefore, there are at least two different concepts that any operationalization of this factor should attempt to capture: the overall level of economic development and changes in the rate of economic growth. Over the short term, wealthy countries tend to remain wealthy and poor countries tend to remain poor, and their degree of wealth may have a strong effect on their overall likelihood of being involved in conflict. In addition, sharp declines in the rate of growth for a range of states may increase their likelihood of intrastate conflict in particular. We employed two different metrics for this key factor: the rate of global GDP growth and the percentage of states with a GDP per capita greater than USD$10,000.19 The former shows the aggregate level of economic instability on a year-to-year basis, while the latter identifies the percentage of states that are relatively wealthy and therefore less likely to become involved in conflict. Both metrics are derived from data available through the World Bank, and their historical values are shown in black in Figures 3.4 and 3.5. 16 Robert Gilpin, War and Change in International Politics, Cambridge, UK: Cambridge University Press, 1981; Erik Gartzke, “The Capitalist Peace,” American Journal of Political Science, Vol. 51, No. 1, 2007, pp. 166–191; Michael Mousseau, “The Social Market Roots of Democratic Peace,” International Security, Vol. 33, No. 4, 2009, pp. 52–86. 17 Collier et al., 2003; Halvard Buhaug, Kristian Skrede Gleditsch, Helge Holtermann, Gudrun Østby, and Andreas Forø Tollefsen, “It’s the Local Economy, Stupid! Geographic Wealth Dispersion and Conflict Outbreak Location,” Journal of Conflict Resolution, Vol. 55, No. 5, 2011, pp. 814–840; Zeynep Taydas and Dursun Peksen, “Can States Buy Peace? Social Welfare Spending and Civil Conflicts,” Journal of Peace Research, Vol. 49, No. 2, 2012, pp. 273–287. 18 Gurr, 1970; Theda Skocpol, States and Social Revolutions, New York, N.Y.: Cambridge University Press, 1979; Edward Miguel, Shanker Satyanath, and Ernest Sergenti, “Economic Shocks and Civil Conflict: An Instrumental Variables Approach,” Journal of Political Economy, Vol. 112, No. 4, August 2004, pp. 725–753. 19 Theoretically, the indicators we use are not as desirable as more-nuanced metrics that more directly identify types of economic growth that could be linked with levels of conflict, such as those that reflect the distribution of wealth within states. Unfortunately, data for such metrics were not available for a sufficiently wide range of states and over a sufficiently long time period to be used for our study. 54 Figure 3.4. Global GDP Growth Rate SOURCE: Historical data: World Bank, 2015; projections calculated by authors. To calculate the baseline projections for each metric, we relied on two different methodologies. We calculated the baseline projected global rate of GDP growth using the International Futures project.20 As evidenced by the pattern in Figure 3.4, GDP growth rates historically have experienced a great deal of year-to-year variation, and this volatility is likely to continue. However, projecting the pattern of this volatility is extraordinarily difficult, so future projections can at best suggest the direction of the long-term trend. Futures with higher longterm growth rates are likely to experience fewer sharp declines in annual growth rates, while futures with lower long-term growth rates are likely to experience such declines more frequently. Our projections for GDP growth rates should be interpreted in this light. 20 The operation of the International Futures model is discussed in the introduction to this chapter. 55 Figure 3.5. Wealthy States as a Percentage of the World’s States SOURCE: Historical data: World Bank, 2013; projections calculated by authors. We arrived at the baseline projection for the number of wealthy states by again using the International Futures model. Baseline projections are displayed in Figures 3.4 and 3.5 in gray. For both metrics, we projected the branch scenarios by calculating one standard deviation above and below the baseline scenario, shown on the two figures in blue and red, respectively. Numerous factors could lead to greatly increased economic volatility and lower growth over the period in question. Ongoing policy concerns, such as the cohesion of the Euro area, have the potential to provide negative shocks to global growth rates that may not be properly captured by projections based on historical experiences. However, greatly increased economic volatility is a much more plausible future than one in which a large number of already wealthy states see their overall GDP levels decline significantly. Such a decline would likely need to be accompanied by a more severe shock, such as a global pandemic or ecological disaster. Extent of Economic Interdependence Economic interdependence may reduce the likelihood of interstate conflict in two primary ways. Increased economic integration raises the costs that states will bear if conflict disrupts trade.21 Increased trade also provides a mechanism to share vital natural resources that a state 21 Solomon William Polachek, “Conflict and Trade,” Journal of Conflict Resolution, Vol. 24, No. 1, 1980, pp. 55–78; Bruce M. Russett and John R. Oneal, Triangulating Peace: Democracy, Interdependence, and International Organizations, New York, N.Y.: W. W. Norton & Company, 1999; Alberto Alesina, Enrico Spolaore, and Romain Wacziarg, “Trade, Growth and the Size of Countries,” Handbook of Economic Growth, Vol. 1, 2005, pp. 1499–1542. 56 might otherwise seek to control and exploit on its own.22 Both of these mechanisms are likely to reduce levels of interstate conflict. There is no evidence that suggests any significant effects on intrastate conflict levels directly because of increasing economic interdependence. The extent of economic interdependence can be operationalized in a number of different ways. While measures of exports and imports, both overall and with specific partners, are the most commonly used metric, scholars have also explored measures that involve openness to trade, trade in specific important commodities, and capital flows and financial integration. For each metric, the literature emphasizes the importance of comparing the scope of the interdependence with the size of the state’s economy. The metric used in our project is that most commonly used in the literature, the ratio of exports and imports to GDP. We calculated our measure at the global level, but data at the state or dyadic level could also be used. Our data were drawn from the International Futures project, although the original source for the historical data is the World Bank’s World Development Indicators. The baseline projection for the extent of global economic interdependence also comes from the International Futures project, which (as already discussed) relies on a complex model incorporating a range of economic, social, and political variables. This projection is shown in Figure 3.6 in gray. We calculated the branch scenarios using one standard deviation above and below the baseline projection; they are shown in blue and red, respectively. 22 Richard N. Rosecrance, The Rise of the Trading State: Commerce and Conquest in the Modern World, Vol. 386, New York, N.Y.: Basic Books, 1986; Tony Allen, “Watersheds and Problemsheds: Explaining the Absence of Armed Conflict over Water in the Middle East,” Middle East 2, No. 1, 1998, p. 50. 57 Figure 3.6. Economic Interdependence SOURCE: Historical data: World Bank, 2013; projections calculated by authors. Economic interdependence is likely to continue to increase, absent significant shocks to the international system. The last major contraction in global levels of interdependence began as a result of the First World War, and major power conflict remains the most likely mechanism for reducing trade flows. A serious global pandemic would also likely have the effect of greatly reducing international trade. Absent such a system-level shock, however, increases in economic interdependence are likely to be robust and persistent. Degree of U.S. Preeminence The high relative power of the United States vis-à-vis other major powers in the international state system since the end of the Second World War may have reduced interstate conflict in several ways. The relative dominance may have reduced the incidence of great power conflicts, including large-scale global conflicts, as the U.S. position makes a direct military challenge to it or its close allies unlikely to succeed.23 U.S. dominance also may have enabled the creation of a security community among U.S. allies in Europe (and, to a lesser extent, East Asia), as the United States could effectively deter any conflicts by credibly promising to counter the 23 A. F. K. Organski and Jacek Kugler, The War Ledger, Chicago, Ill.: University of Chicago Press, 1981; Paul K. Huth, “Extended Deterrence and the Outbreak of War,” American Political Science Review, 1988, pp. 423–443; Thomas J. Volgy and Lawrence E. Imwalle, “Hegemonic and Bipolar Perspectives on the New World Order,” American Journal of Political Science, Vol. 39, No. 4, 1995, pp. 819–834. 0 5 10 15 20 25 30 35 40 45 50 1960 1970 1980 1990 2000 2010 2020 2030 2040 Global trade flows as % of global GDP Year Historical Baseline One standard deviation above baseline One standard deviation below baseline 58 aggressor.24 The United States may also have reduced the incidence of interstate conflict during its period of dominance by actively mediating disputes between lesser powers, though the evidence for this argument is more contested.25 At the level of intrastate conflict, greater power parity tends to fuel “proxy conflicts.” One of the clearest patterns in the historical data on intrastate conflict is the influence of the Cold War. Globally, the number of internal conflicts peaked in the late Cold War era, when both the United States and the Soviet Union were committed to indirect support for partner regimes and insurgent groups around the world.26 The withdrawal of superpower support from these actors following the end of the Cold War led to the durable termination of many of these conflicts, and the overall number of intrastate conflicts has fallen in the decades since in the absence of a global challenger to the United States. The relative power of the United States can be measured in several ways. Comparisons of military and power-projection capabilities may be helpful in certain circumstances, and diplomatic and cultural interactions may be more relevant in others. Over the long term, however, the economic base of the United States will largely determine its ability to sustain and project power. The comparison of the economic capacity of the United States and other states is therefore likely to provide the best long-term method for identifying changes in the relative preeminence of the United States. To this end, this project utilizes the ratio of U.S. to global GDP to estimate the relative power of the United States over time.27 The historical values of this metric can be seen in black in Figure 3.7.