1
Studying Postconflict Violence
Approaches and Methods
Studying postconflict violence is a recent enterprise. After an earlier focus on understanding why countries go to war with each other, researchers turned their attention to forms of violence internal to states. The second half of the twentieth century saw the production of vast shelves of books, articles, and policy papers seeking to explain why civil wars, rebellions, and more localized episodic violence such as ethnic riots occur. But it was only as that most deadly of centuries drew to a close that scholars started to pay serious attention to the violence that can emerge after the formal end of extended intrastate violent conflicts.
This new focus was driven primarily by policy concerns. From the end of the Cold War, international groups—United Nations agencies, regional bodies such as the European Union, international financial institutions, international NGOs, and foreign militaries and diplomats—started to play an ever-expanding role mediating peace agreements and providing financial and technical support to their implementation. Yet this growth in policy ambition was matched by a growing realization that many peace settlements fail and that even where wars are brought to an end violence can continue and even increase in their aftermath. New puzzles started to capture researchers’ imaginations: Why do some civil wars recur while others do not? Why are some war-to-peace transitions largely peaceful while others are marked by violence? What types of policies are most likely to lead to the consolidation of peace in places scarred by previous extended violence?
This chapter and the next assess this new and growing literature on the causes of postconflict violence. In chapter 2, I consider the findings of various studies and the arguments and theoretical claims they make. But before doing so I will examine the methodological and analytic approaches that have been employed to date. Unfortunately, these have often not been well suited for providing a deeper understanding of why postconflict violence occurs with frequency in some places and not in others.
Previous Approaches to Studying Postconflict Violence and Their Limitations
Researchers of postconflict violence have tended to follow one of two paths. The first group of explorers, made up of economists and quantitatively oriented political scientists, has focused on explaining why civil wars recur in some countries and not in others after peace settlements or military victories. Large-n cross-country datasets, which code countries as being either in civil war or at peace, are used with a range of environmental factors (such as the type of war; how it ended; political, economic, and social structures; and geographic characteristics) employed as explanatory variables to test. Econometric analyses isolate the correlates of renewed civil war, and case material is sometimes used to show how such factors are indeed causing renewed civil war, although this is not always deemed necessary. The focus has been to develop and test parsimonious explanations that account for a significant share of the variance in the likelihood of civil war recidivism.1
The second group, a diverse mix of social scientists (qualitative political scientists, sociologists, historians, anthropologists) and lay researchers (journalists, policy analysts, development practitioners), have taken a different approach. Arguing that each country needs to be understood on its own terms, researchers have provided in-depth examinations of peace processes and postconflict transitions, seeking to identify what went wrong or, less frequently, why the war-to-peace transition was successful. Where such pieces are gathered together in edited volumes, introductory or concluding chapters draw out commonalities and differences, although the ambition to use the cases to develop or test any theory is usually limited.2 Postconflict contexts are seen to be exceptionally complex and the search for regularities may be fruitless. As Berdal, in his concluding reflections to one recent volume of case studies, writes:
Post-war environments do display common features and this has naturally stimulated work on more general studies or “approaches” aimed at explaining post-war violence. While these have undeniably provided valuable insights, bringing to the fore new and potentially significant points of analytical emphasis, the case material collected here makes it clear that the heuristic value of such general theories is often limited and problematic.3
Despite the differences in approach, most researchers in both traditions have tended to share a number of problematic assumptions.
The Unit of Analysis
Almost five decades ago, Przeworski and Teune noted that comparative analysis does not require comparisons between countries. Rather, the choice of the unit of analysis (the state, the region, the village, the household) should depend on “an a priori assumption about the level of social systems at which the important factors operate.”4
Scholars of violence have taken heed. Snyder’s assertion that “subnational units of analysis play an increasingly important role in comparative politics”5 certainly applies to the study of violence. Blattman and Miguel in their summary of advances in understanding civil war argue that “the most promising avenue for empirical work is on the subnational scale.”6 Others have reached similar conclusions when studying different forms of violence. Varshney and Wilkinson, for example, look at how local factors lead to variations in ethnic rioting in India, while Straus assesses sources of local variation in genocidal violence in Rwanda.7 Econometric papers have also increasingly employed subnational data to assess sources of within-country variation in violence.8
However, the study of postconflict areas has not followed this trend in violence research or comparative political science. Case-based and large-n studies have overwhelmingly used countries as the unit of analysis with both the outcome to be explained (the presence of postconflict violence) and the factors or mechanisms that explain it measured at the national level. Sorpong Peou thus talks of Cambodia’s postwar violence and situates his explanation of it in the fragility of national democratic institutions;9 Collier, Hoeffler, and Soderbom identify issues such as the presence or absence of national elections and national levels of economic development as the reasons why some countries see renewed civil war while others do not.10
This is problematic for two reasons: first, extended violent conflicts tend only to affect some areas of countries; and, second, postconflict violence usually varies in incidence and form within these areas. This variation means that factors or mechanisms at the subnational level must be important.
The majority of civil wars and other extended violent conflicts do not affect all areas of a country. Even in such national civil wars, where groups fight over control of the central state, many areas of a country will be only minimally affected. Until recent years, northern areas of Afghanistan were relatively untouched by violence.11 Providing a catalogue of examples across countries as diverse as Spain in the 1930s, Vietnam in the 1960s, and Yugoslavia in the 1990s, Kalyvas persuasively shows that civil war violence, as opposed to civil war presence, is a localized phenomenon.12
Furthermore, subnational extended violent conflicts, which include secessionist insurgencies and localized extended intercommunal violence, are at least as frequent as civil wars where combatants aim to control the central organs of the state. Weinstein argues that more than half of all civil wars since 1945 have been national in scope with rebel groups seeking to control the state accounting for 56 percent of all belligerent groups, but this overstates the amount of wars that are truly national.13 In South and Southeast Asia alone, there have been twenty-six active subnational conflicts affecting ten countries since 1992. Since 1946, such conflicts have killed more than 1.35 million people with almost 100,000 dying between 1999 and 2008. Between 1999 and 2008, subnational conflicts killed more people in Asia than every other form of conflict combined.14
In nations affected by such conflicts, violence has little direct impact in large swathes of the country. In Uganda, the Lord’s Resistance Army operated in provinces covering 44 percent of the country while other provinces remained untouched by war.15 Mindanao, the location of a series of secessionist civil wars, accounts for just one-quarter of the Filipino population, but these conflicts have only affected around 6 percent of the national population.16 The conflict-affected Deep South of Thailand houses less than 3 percent of the country’s population,17 a similar level to the proportion of the UK population that lives in Northern Ireland, which was wracked by decades of unrest.18
Similarly, Indonesia’s bouts of extended violence have affected just some locales. The province of Aceh, which saw three decades of civil war, includes less than 2 percent of Indonesia’s population. Aside from a single terrorist attack on the Jakarta stock exchange building in September 2000, Indonesians in other parts of the country saw little impact from the civil war, one reason why it took so long for a conflict resolution process to emerge. Indonesia’s intercommunal extended violent conflicts were also concentrated in a relatively small number of areas, with fifteen districts and cities, containing just 6.5 percent of Indonesia’s population, experiencing 85.5 percent of all deaths from communal violence between 1990 and 2003 according to one count.19 Describing Uganda, the Philippines, Thailand, the UK, or Indonesia as being either “in conflict” or “at peace” is not particularly illuminating, for it masks the extreme variation in levels of violence within each country.
Within countries that have experienced extended violent conflicts that have come to an end, there also tends to be marked differences in the incidence of postconflict violence across areas. As I show in chapter 3, there has been significant variation in levels, impacts, and forms of postconflict violence between areas of Indonesia affected by different violent conflicts and, indeed, between areas affected by the same conflict. But this is by no means only the case for Indonesia. One may compare levels of postconflict violence in, say, Bosnia with other postconflict countries and conclude that postconflict Bosnia presents “an encouraging exception” in that it has been largely violence free.20 Yet such cross-national comparisons draw attention away from the fact that some areas of the country have seen continuing violence while others have not.21
The presence of subnational variation in both extended and postconflict violence means that theories of postconflict violence that are built solely on national-level factors are at best insufficient. One issue is methodological and affects all of the large-n cross-country studies. These studies have developed explanations using variables measured at the national level: the strength and characteristics of central state institutions, nationwide levels of poverty, the ethnic makeup of a country, and so on. Yet national measures of mountainous terrain, GDP per capita, or ethnic fractionalization, which are deemed to predict civil war recurrence, may be poor proxies for conditions in the areas that actually experienced civil war.22 Even if ethnic makeup at the national level co-varies with the likelihood of a country experiencing civil war recurrence, if such diversity is similar in areas of a country that saw renewed violence and those that did not then it cannot be deemed to be causing recurrence. If a large proportion of a country that sees renewed violence is mountainous, but the actual areas that experience new violence are not, then terrain is not a convincing explanation.23
Beyond these methodological problems, which can be addressed through the use of better measures,24 there is a broader point to be made. Logically, national factors cannot be the source of differences in violence within countries. The presence of a set of environmental factors might explain the propensity of a given country to experience extended violence or high levels of postconflict violence within its territory. But it cannot explain why some areas of the country are affected while others are not, for, by definition, national-level factors do not vary across regions of a polity.25 To understand such local variation, we need to build theories based on fact...
