where do we go from here? conceptual, theoretical, and methodological gaps in the large-n civil war...

21
Where Do We Go from Here? Conceptual, Theoretical, and Methodological Gaps in the Large-N Civil War Research Program 1 Adrian Florea Indiana University, Bloomington Since the early 1990s, the large-N civil war research program has been a vibrant one but has reached an analytical plateau because methodologi- cal issues have obscured fundamental conceptual and theoretical con- cerns. The existing ground-clearing work has shed light on macro- and micro-level determinants of internal conflict onset, duration, and termi- nation, but has left the field theoretically impoverished. This article dis- cusses remaining gaps in the quantitative civil war literature, and proposes conceptual, theoretical, and methodological remedies. Five main lacunae are identified. First, civil war may be a longer process, with several waves of escalation and de-escalation, than generally described in the extant literature. Second, existing theories conflate civil conflict with violence—war onset and violence may be conceptually distinct. Third, we have few systemic-level theories of civil war onset and duration. Fourth, we need better theories that allow us to determine whether civil wars are different from other types of wars. Finally, we need to go beyond state-centric monadic approaches, and be more rigorous when building our models for explaining civil war onset, duration, or termination. Warfare has shifted away from great powers, away from conflict to civil war, insur- gency, and terrorism. (Levy 2007:19) The decline of interstate conflict in the post–World War II period and the end of the Cold War ushered in a paradigm shift in the study of war with students of con- flict gradually turning their analytical gaze away from international conflict and toward civil wars. According to Kalyvas (2007), scholars’ renewed interest in the underdeveloped and under-theorized field of civil war in the early 1990s was pre- mised on three main factors: first, World Bank experts specializing in the econo- mies of the war-torn African states began to investigate the economic causes of civil conflict; second, with the virtual disappearance of interstate wars, many Interna- tional Relations (IR) scholars turned to the study of internal conflict; and, third, the resurgence of ethnic conflict after the break-up of the Soviet Union and Yugo- slavia led analysts to focus on various forms of violence exhibited in internal war. Since the early 1990s, the civil war field has been a vibrant one—theories, hypothe- ses, and mechanisms have been proposed and tested regarding internal conflict onset, duration, escalation, diffusion, or termination. In the early 2000s, the 1 I would like to thank Karen Rasler, Jason Sorens, Hicham Bou-Nassif, and the anonymous reviewers for their comments and suggestions. All errors and omissions are my sole responsibility. Florea, Adrian. (2012) Where Do We Go from Here? Conceptual, Theoretical, and Methodological Gaps in the Large-N Civil War Research Program. International Studies Review, doi: 10.1111/j.1468-2486.2012.01102.x Ó 2012 International Studies Association International Studies Review (2012) 14, 78–98

Upload: adrian-florea

Post on 05-Oct-2016

213 views

Category:

Documents


1 download

TRANSCRIPT

Where Do We Go from Here? Conceptual,Theoretical, and Methodological Gaps in the

Large-N Civil War Research Program1

Adrian Florea

Indiana University, Bloomington

Since the early 1990s, the large-N civil war research program has been avibrant one but has reached an analytical plateau because methodologi-cal issues have obscured fundamental conceptual and theoretical con-cerns. The existing ground-clearing work has shed light on macro- andmicro-level determinants of internal conflict onset, duration, and termi-nation, but has left the field theoretically impoverished. This article dis-cusses remaining gaps in the quantitative civil war literature, andproposes conceptual, theoretical, and methodological remedies. Fivemain lacunae are identified. First, civil war may be a longer process,with several waves of escalation and de-escalation, than generallydescribed in the extant literature. Second, existing theories conflatecivil conflict with violence—war onset and violence may be conceptuallydistinct. Third, we have few systemic-level theories of civil war onset andduration. Fourth, we need better theories that allow us to determinewhether civil wars are different from other types of wars. Finally, weneed to go beyond state-centric monadic approaches, and be morerigorous when building our models for explaining civil war onset,duration, or termination.

Warfare has shifted away from great powers, away from conflict to civil war, insur-gency, and terrorism. (Levy 2007:19)

The decline of interstate conflict in the post–World War II period and the end ofthe Cold War ushered in a paradigm shift in the study of war with students of con-flict gradually turning their analytical gaze away from international conflict andtoward civil wars. According to Kalyvas (2007), scholars’ renewed interest in theunderdeveloped and under-theorized field of civil war in the early 1990s was pre-mised on three main factors: first, World Bank experts specializing in the econo-mies of the war-torn African states began to investigate the economic causes of civilconflict; second, with the virtual disappearance of interstate wars, many Interna-tional Relations (IR) scholars turned to the study of internal conflict; and, third,the resurgence of ethnic conflict after the break-up of the Soviet Union and Yugo-slavia led analysts to focus on various forms of violence exhibited in internal war.Since the early 1990s, the civil war field has been a vibrant one—theories, hypothe-ses, and mechanisms have been proposed and tested regarding internal conflictonset, duration, escalation, diffusion, or termination. In the early 2000s, the

1I would like to thank Karen Rasler, Jason Sorens, Hicham Bou-Nassif, and the anonymous reviewers for theircomments and suggestions. All errors and omissions are my sole responsibility.

Florea, Adrian. (2012) Where Do We Go from Here? Conceptual, Theoretical, and Methodological Gaps in the Large-N Civil WarResearch Program. International Studies Review, doi: 10.1111/j.1468-2486.2012.01102.x� 2012 International Studies Association

International Studies Review (2012) 14, 78–98

research program became dominated by large-N studies that shed light on socio-economic and political correlates of civil war onset, duration, and termination, butthat suffered from at least two limitations: (i) they were primarily state-centric inthat they focused on state characteristics and ignored the strategic interactionsbetween actors or the microdynamics of violence; and (ii) they over-privilegeddomestic variables over subnational and international ones.

My goal in this article is not to offer a comprehensive review of the findings inthe vast civil war literature2; nor is it to pinpoint past or current theoretical andmethodological advances in the field. Rather, I attempt to take the pulse of thelarge-N civil war research program, and identify remaining conceptual, theoreti-cal, and methodological gaps that have prevented us from having a fullerunderstanding of civil war as a process composed of waves of escalation andde-escalation of violence rather than a series of one-shot independent events.The argument that I advance herein is that, despite the impressive methodologi-cal ground-clearing work3 of the late 1990s and 2000s, the quantitative literatureon civil war still suffers conceptually, theoretically, and methodologically. As itwill become clearer in the discussion below, my criticism of the large-N literatureon civil wars is not epistemological, but theoretical and methodological. In otherwords, my goal is not to pinpoint the limitations of the quantitative method ingeneral, but of the application of the method in the civil war research program.

This article proceeds as follows. The first section considers two conceptualproblems that emanate from the casualty-threshold approach (the use of a spe-cific number of battle-related deaths per year) for determining civil war onset,duration, and termination in the large-N literature. I argue that the casualty-threshold approach yields two inter-related problems: arbitrary and questionablecodings of starting and ending dates for civil wars, and a conflation of civil con-flict with violence. Better conceptualization is needed: on the one hand, civilwars may be longer processes, with several waves of escalation and de-escalation,than generally described in the extant literature; on the other hand, war onsetand violence may be conceptually distinct. The second section looks at the fewsystemic-level explanations of civil war onset and duration and pinpoints theirlimitations. The third section focuses on the more recent debates of whetherinter- and intrastate conflicts are comparable phenomena. The fourth sectiondiscusses methodological problems regarding the appropriate unit of analysis;corrections for temporal and spatial dependence; models for capturing multilevelcausality; and models for accounting for multiple civil war outcomes. Finally, thefifth section draws the conclusions.

Civil War as a Conceptual Variable

The difficulty of characterizing civil wars is a conceptual problem rather thanone of measurement. (Kalyvas 2003:476)

How do we know a civil war when we see one? At first sight, the concept may seemfairly unambiguous. A civil war is generally construed as a conflict: (i) confinedwithin the territorial boundaries of a sovereign state; (ii) where the government is afull participant; (iii) violence is sustained and reciprocated; and (iv) a certain deaththreshold is reached (yearly and throughout the duration of the entire conflict).4

2Others have done so beautifully. See Blattman and Miguel (2010); Kalyvas (2007); Sambanis (2004); Tarrow(2007).

3This phrase is attributed to Karen Rasler.4See Sambanis (2004). Many definitions of civil war overlap. Where there is disagreement, it’s usually in the

operationalization of the concept.

79Adrian Florea

Upon closer scrutiny and against the empirical evidence, it soon becomes obviousthat each of these arbitrary definitional criteria is questionable at best.

First, there is little theoretical rationale for imposing on the concept the inter-nality of war to the territory of a single sovereign state. While most internal con-flicts may indeed originate within the boundaries of a single country, many civilwars actually become transnational in character5 (for example, the Bosnian civilwar of the early 1990s). So, by restricting the conceptual domain to a sovereignstate’s boundaries, we are implicitly saying that civil war as a process (that is, withall its phases—escalation, de-escalation, diffusion, termination) is unequivocallyan intra-national phenomenon—a statement that any astute observer from withinor outside the academia would certainly frown at. Second, there does not seemto be any overwhelming rationale for limiting civil wars to confrontationsbetween the government and one or more claim-making collective actor(s).6 Insome civil war situations, the government may actually not be a combatant: actorsfighting each other in countries that lack structures of governance (for example,Somalia) or in countries where the government is yet another player (for exam-ple, Eritrean People’s Liberation Front-EPLF vs. Eritrean Liberation Front-ELFin the 1970s) could be as consequential to civil war outcomes as the govern-ment–collective actor interaction.

Third, if violence is indeed the central discriminating factor, then we need tobe more specific about what level of violence is needed to observe a civil warrather than a different form of conflict. Fourth, if we rely on the death thresholdto characterize civil war onset, we may be actually conflating violence with con-flict and may thus be ignoring the ample empirical evidence which points to theendogeneity of violence to war.7

Coming up with a conceptual and operational definition of civil wars is a diffi-cult, but not intractable, endeavor. Scholars have tended to treat civil war as alargely self-evident concept and have assumed unit homogeneity for the casesincluded in the large number of available data sets on internal conflict.8 Notonly is there a tendency to assume that everyone knows what a civil war is, butmany researchers believe ‘‘that available data on conflict will accurately reflectthe theoretical concepts’’ (Gleditsch 2002:73). In many ways, and againstSartori’s (1970) warning more than 40 years ago, data collection has continuedunabated despite the lack of proper conceptualization. Not only we don’t clearlyknow what civil wars are, but we also cannot really tell what civil wars are not.9 Asa result, without theoretically driven conceptualization, civil war may not meanthe same thing to all scholars at all times. The consequences of this state ofaffairs are quite serious for our theory-building and hypothesis-testing enter-prises. It would not be a gross exaggeration to state that a conceptual morass hasengulfed the field whereby concepts remain abstract and highly aggregated; con-ceptualization of civil conflict has been insufficiently integrated with empiricalresearch; conceptual fuzziness has led to potential incomparability of cases inlarge-N research; and civil war as a variable is not quite pitched at the right level

5On the transnational dimensions of civil war, see Salehyan (2009). The PRIO ⁄ UPPSALA data set classifiesintrastate conflicts as either intrastate or internationalized intrastate wars.

6Note that I purposefully refrain from using the notion of group and instead use collective actor or organiza-tion. I follow Brubaker (2004:16) who argues that collective actors or ‘‘organizations, not [g]roups as such, are thechief protagonists of [c]onflict and violence.’’ See also Sinno (2008:3), Vasquez (2008:25), and Wimmer (2002:97)for the view that organizations rather than groups are chief protagonists in both inter- and intrastate wars.

7See the related discussion in Kalyvas (2006).8Based on a survey of the articles published in mainstream journals during the last decade, the most frequently

used data sets are PRIO ⁄ UPPSALA; Fearon and Laitin (2003); Collier and Hoeffler (2004); Minorities at Risk; Cor-relates of War intra-state wars; Heidelberg KOSIMO; Sambanis (2004); Gleditsch (2004); Cederman, Wimmer, andMin (2010). See Sambanis (2004) for a discussion of civil war codings in many of these data sets.

9In addition to counseling scholars to focus on conceptualization prior to operationalization, Sartori(1970:1042) also urges researchers to clearly state what specific concepts are not.

80 Where Do We Go from Here?

so as to connect both upwards (toward theory) and downwards (toward empiricaldata).

Because of improper conceptualization, the gap between concepts and data isgetting wider. Existing data are still treated ‘‘with excessive reverence as if thepristine raw data were a fully unbiased and accurate arbiter of the analyticalvalue of our theories’’ (Gleditsch 2002:197). An unambiguous conceptual defini-tion significantly limits any schism that may exist between theory and data. More-over, how we define our concepts has implications for how we select the keyelements of the civil war contexts—careful conceptualization ensures that thesecontexts are analytically equivalent.10 Taking the civil war concept for granted isdangerous because we have no real sense of where the conceptual boundarieslie. Thus, we are left with a familiar label for some form of conflict and littleunderstanding about its dynamics.

To be fair, it’s not easy to come up with a ‘‘perfect’’ conceptual definition andoperationalization of civil war.11 We are dealing here with a macro-level conceptthat cannot easily be factored down into specific, micro-level representations.12

At first blush, the conceptualization and operationalization problems may seemunavoidable. I disagree. While it may be difficult to come up with the ‘‘perfect’’definition of civil war that exhaustively captures all phenomena that can be clas-sified under the civil war rubric, we can nevertheless set unambiguous bound-aries that determine the appropriateness of our indicators. At least twointerlinked aspects need to be addressed in order to come up with a better speci-fication of civil war’s conceptual boundaries: (i) when exactly a civil war beginsand ends and (ii) whether civil war and violence are theoretically equivalent.I now turn to a discussion of each of them.

When Does a Civil War Really Begin or End?

Contemporary civil wars tend to have an intermittent quality, dying down forsome years before coming to life again. Conflicts may simmer for years at lowintensity, then erupt or re-erupt into full-scale civil war. (Hironaka 2005:150)

Too many cases are sufficiently ambiguous to make coding the start and end ofthe [civil] war problematic and to question the strict categorization of an eventof political violence as a civil war as opposed to an act of terrorism, a coup, geno-cide, organized crime, or international war. (Sambanis 2004:815)

Current scholarship embraces a casualty-threshold approach to determine civilwar onset, duration, and termination.13 For example, the PRIO ⁄ UPPSALA dataset requires a minimum of 25 battle-related deaths to code for the onset andduration of civil wars.14 I argue that, while existing coding rules may facilitate

10See Falleti and Lynch (2009) for a discussion of how careful conceptualization facilitates unit homogeneity(analytical equivalence).

11The reader will have observed by now that I shunned from providing a definition of civil war that addresses atleast some of the problems pointed out above. Definitional ambiguity may be ineluctable with a concept pitched atsuch a high level of abstraction. Nevertheless, I suggest that civil wars are (i) conflicts between a government and(a) collective actor(s) or between collective actors about incompatibilities in attaining various material and nonma-terial ends; (ii) which are generally fought within the territorial confines of a sovereign state but which have thepotential to be regionalized ⁄ internationalized; and (iii) which exhibit various forms of violent and non-violent col-lective action from one episode to another (onset, escalation, de-escalation, termination).

12Cf. Kalyvas (2003).13See Sambanis (2004) for a detailed discussion of coding problems in various civil war data sets.14There are actually two codings for civil war onset in this data set: one (Startdate) that marks the date of the

first battle-related death and another (Startdate2) that marks the date when the 25 casualty-threshold is reached.

81Adrian Florea

quantification, they actually create confusion about a civil war’s duration. If welook closer at the empirical evidence and regard civil wars as processes rather thanone-shot independent events, then internal wars’ start and end dates are nottrivial.15 These problems are not limited to the study of civil wars. For example,the rivalry research program is more or less confronted with similar issues. Onthe one hand, the ‘‘conflict-density approach’’ (Diehl and Goertz 2000; Klein,Goertz, and Diehl 2006) identifies rivalries by looking at a specific number ofmilitarized interstate disputes (MIDs) for dyadic pairs within specific time peri-ods.16 On the other hand, the ‘‘subjective approach’’ (Thompson 2001; Colaresi,Rasler, and Thompson 2008) focuses on decision makers’ perceptions about theexistence of competitive and threatening enemies. Both are imperfect: the for-mer excludes cases that do not register a sufficient number of MIDs and is circu-lar in the sense that it uses a conflict record to explain a conflict record; thelatter cannot identify with exact precision the start or duration of rivalries sinceit is based on subjective beliefs of enmity.17

Much as is the case with interstate rivalries,18 it may be more useful to think ofcivil wars as longer processes involving escalation and de-escalation rather thanonset, termination, and reoccurrence. Civil wars are not just isolated, one-shotevents but peaks in more continuous chains of interaction (Gleditsch 2002:85).Put differently, single disputes may be part of a broader escalation–de-escalationpattern rather than independent events. Thus, without clear-cut observable out-comes—such as enforceable peace agreements, complete military victory fromone side, unambiguous institutional outcomes (for example, secession, auton-omy)—we cannot say (and code accordingly) that a civil war has really ended.19

What is currently coded as several civil conflict onsets can actually representmere variation along the escalation–de-escalation continuum within the sameconflict.

To illustrate the plausibility of this alternative conceptualization, let’s take atime series view of a country X exhibiting several episodes of violence along acertain timespan.20 Implicitly, we are assuming that violent events are not inde-pendent of one another. Thus, to accurately explain the extent of violence attime t in country X, we cannot ignore the effects of a prior violent episode attime t-1. To facilitate comprehension, a simple first-order autoregressive (AR1)equation will suffice to illustrate the dynamic interaction between current andpast episodes of violence in country X. Hence, we have

yt ¼ u1yt�1 þ et ð1Þwhere yt is the level of violence at time t, u is a parameter of the model, and et isa white noise component with zero mean and constant variance. More generally,

yt ¼ ut1e0 þ ut�1

1 e1 þ ut�21 e2:::u1et�1 þ et ð2Þ

or

15Obviously, start and end dates matter a lot for analyses of civil war onset, duration, and termination.16Diehl and Goertz (2000) classify enduring rivalries as those dyads that experience six or more MIDs during a

20-year period.17Because it relies on perceptions of enmity rather than on directly observable antagonistic behavior, the ‘‘sub-

jective approach’’ needs a theory of both rivalry onset and rivalry termination. In the ‘‘conflict-density approach,’’rivalry onset and termination are by definition.

18See DeRouen and Bercovitch (2008) for tests of many rivalry hypotheses applied to enduring internal rivalries(EIRs).

19In addition to an observable conflict outcome, the continued existence of armed opposition movements dur-ing peace years would also be a clear indicator that war has not really ended. I thank Jason Sorens for this point.

20For the sake of exposition, I assume that we have sufficient time points to observe variation in the level of vio-lence for the given country X.

82 Where Do We Go from Here?

yt ¼X1

i¼1

ui1et�1 ð3Þ

This simple first-order autoregressive equation approximates many politicalphenomena whereby the effect of a past event persists immediately after itsoccurrence but exponentially diminishes over time. Consider the situation of asudden drop in economic growth due to an exogenous shock such as a globaleconomic recession, for instance. Its greatest impact is likely to be felt immedi-ately after it occurs, but its effects are expected to vanish exponentially with eachperiod t. For the equation above to capture the idea of exponentially diminish-ing impact, the parameter u must be restricted within the bounds of stationarity:)1 < u < 1. If u = 1, then we are simply stating that the impact of past eventsdoes not diminish over time but remains constant. If u > 1, an escalatory orgrowth process is implied—that is, as time goes by the impact of a past event isactually greater.21

What does this have to do with civil wars? The main point that I want toemphasize with the aid of this simple example is that, while most events do exhi-bit an exponentially decaying function, it may be the case that at least a subclassof civil wars does not follow this pattern and displays an escalatory function.Hence, under certain conditions—such as the lack of a clear victory from oneside, or a clear-cut institutional outcome—each additional period of peace afterthe violence episode at time t-1 may actually increase the probability of violence attime t.22 In other words, it would not be totally unreasonable to state that, atleast in some cases, peace times between episodes of violence could actually cap-ture a ‘‘boiling container effect’’ of past events23 whereby with each peacefultime point the likelihood of recurrent violence is higher.24

Current codings artificially slice civil wars and fail to capture the heterogeneityof peace years between civil war onsets. Not all peace years are similar concern-ing the effect of actor interactions during any peace year on the posterior proba-bility of war. In other words, contrary to the assumption underlying currentcoding practices, various peace years do not have similar effects on the likelihoodof war recurrence: peace years that capture a lull in hostilities or a military stale-mate may have a different impact than peace years after a peace agreement ormilitary victory. As Galtung (1969) argued a while ago, peace is an aggregateconcept: it may connote both the absence of conflict between antagonists—nega-tive peace—and cooperation and trust between former opponents—positivepeace. Most of the conflict literature—inter- or intrastate—conflates negativepeace with positive peace.25 This unfortunate, but persistent, conflation leads toomitted variable bias (King 2001) as the variation in the level of latent hostilityduring peace years remains unmeasured.26 Rarely are peaceful years—within thesame conflict or across conflicts—analytically equivalent. Therefore, negativepeace years in civil war situations—that is, peace years following the cessation ofhostilities without a clear victory from one side, credible peace agreement, or

21Technically, in the latter two cases, we are dealing with non-stationary processes, and we have to differencethe series to achieve stationarity.

22Another substantive implication is that we need to ‘‘take time more seriously’’ in civil war research and bettermodel duration dependence. See the discussion below.

23A period of peace between violent episodes may actually be a time for reorganization and mobilization fromeither or all parties involved in conflict.

24Existing codings could actually indicate this possibility: if an immediately posterior war onset is significantlymore violent than a prior episode, then this may give credence to the ‘‘boiling container effect’’ of peace times onthe probability of conflict recurrence.

25A recent exception is Klein, Goertz, and Diehl (2008).26See Gibler and Tir (2010:951) for an account of similar problems in the rivalry literature.

83Adrian Florea

mutually agreed outcome—may actually increase the likelihood of war. If weunpack the heterogeneity within peace years, we may observe fewer civil waronsets and terminations, and longer civil war durations with larger fluctuationsin violence from one episode to another.

The heterogeneity in peace years has important methodological implications:it suggests that we are, in fact, dealing with two populations of cases—positivepeace cases or those cases where civil war recurrence is unlikely because the civilwar has really ended; and negative peace cases that merely capture a lull in hos-tilities rather than termination. A duration analysis of civil war implies that, ifgiven enough time, all observations will fail—that peace years observations willeventually terminate and we will observe conflict recurrence.27 This is premisedon the assumption that, ceteris paribus, all peace years observations exert similareffects on conflict reoccurrence. If negative peace is fundamentally differentfrom positive peace in affecting the likelihood of conflict, we then have differentrisks of conflict and different populations of cases: one population composed ofimmune cases or those cases of positive peace that will never experience civil warand one population of exposed cases or those cases of negative peace that arelikely to experience renewed violence. In this case, a split-population estimator,which takes into account the two types of populations, would be more appropri-ate than traditional estimators.28

To briefly illustrate the plausibility of this alternative conceptualization andoperationalization, consider two examples of civil war processes within a relativelyshort timespan: violence in Zaire (DRC) from the 1960s to the 1970s and in SriLanka from the 1980s to the 2000s. In the case of Zaire,29 the PRIO ⁄ UPPSALAdata set codes three civil war onsets for the 1964–1978 period (1964, 1967, 1977),while in the case of Sri Lanka,30 the same data set also codes three civil war onsetsfor the 1984–2008 period (1984, 2003, 2005). A feasible alternative would be tocode for a single civil war in each case: in Zaire (DRC) lasting from 1964 to 197831

and in Sri Lanka unfolding from 1984 to 2009.32 To elaborate on the Sri Lankancase, what are currently coded as civil war terminations in 2003 and 2005 are in factmere interregna between violent episodes within the same conflict. Following a Nor-wegian-mediated ceasefire (thus, not a comprehensive peace agreement whichwould have been an indubitable indicator of conflict termination) in 2002, the Lib-eration Tigers of Tamil Eelam (LTTE) withdrew to safe havens within Tamil Eelamto regroup and prepare its escalatory attacks in 2003. The conflict de-escalatedagain in 2004 when the island was struck by a powerful tsunami which forced bothsides to concentrate on socioeconomic rather than military activities. In 2005, theconflict escalated when the Sri Lankan foreign minister was killed in a rebel attack.Finally, in 2009 the conflict terminated with the military defeat of LTTE. The SriLankan example illustrates a simple, but largely overlooked, fact: because a con-flict does not register the minimum casualty threshold in a given year, it does notnecessarily mean that it has truly ended. Dyads which have experienced severalcivil wars within a short timespan are particularly susceptible to too much arbitrari-ness in coding rules. What is currently seen and coded as the end of a civil war mayactually represent a mere interlude between episodes of violence within the sameconflict. As Hironaka (2005:151–2) puts it,

27On duration models, see Box-Steffensmeier and Jones (2004).28For an application of this estimator, see Metternich and Wucherpfennig (2011).29On the Zaire (DRC) conflicts, see Autesserre (2010).30On the origins of the Sri Lankan conflict, see Rotberg (1999).31With the following episodes: 1964—onset; 1965–1966—de-escalation; 1967—escalation; 1968–1976—de-escala-

tion; 1977—escalation; 1978—termination.32With the following episodes: 1984—onset; 2001–2002—de-escalation; 2003—escalation; 2004—de-escalation;

2005—escalation; 2009—termination (with the observable military defeat of the Tamil Tigers).

84 Where Do We Go from Here?

The ‘‘end’’ of a civil war might merely mean that the insurgents were pushedinto hills, mountains, or rural areas without actually surrendering…If theopposition can neither be fully destroyed nor fully placated, how can [civil] warstruly end? A number of countries experience perpetual low-intensity conflict. Insome years, perhaps owing to a temporary military stalemate or a fragile cease-fire, casualties may be low enough to consider a war to be over. When fightingreturns, we may say that war has recurred. But it might be more accurate to saythat such wars never really ended.33

Civil War and Violence

[V]iolence is employed by both those who wish to upend an existing order andby those who want to sustain it. (Kalyvas, Shapiro, and Masoud 2008:1)

A second problem underlying the casualty-threshold approach to internal waronset, duration, and termination is the fusing of conflict with violence. Civilwar research treats violence as a starting point, ‘‘either examining the pres-ence of violence compared to its absence, or comparing degrees of violentbehavior once a conflict already exists’’ (Chenoweth and Stephan 2010:250).In other words, civil war processes (onset, duration, termination) are con-flated with violence.34 At first sight, this may look undisputable—after all, abloody phenomenon such as civil war cannot be fully captured by a bloodlessreferent. I submit that there are at least two reasons why this conflation maybe problematic for arriving at a more fine-tuned understanding of internalconflict dynamics.

First, recent work (Kalyvas 2006; Weinstein 2007) on the microdynamics of vio-lence provides ample evidence that points to the endogeneity of violence to con-flict. If violence is typically endogenous to various civil war processes, then it maybe the case that a civil war actually starts (for example, with war declarations, casu-alty-free military skirmishes, or de facto territorial control) before violence ensues.I suggest that we need to distinguish civil war from violence beyond current direc-tions. For example, Kalyvas (2006, 2010) argues that only through disaggregationcan we fully account for the endogeneity of violence to conflict by focusing, forinstance, on private motives (for example, eliminating a rival; personal revenge)for violence that have little or nothing to do with the macro-dynamics of conflict.In fact, disaggregation (of violence, space, actors, and levels of analysis) lies at thecore of the recent turn in the civil war literature (Kalyvas 2006; Weinstein 2007;Kalyvas et al. 2008; Chenoweth and Lawrence 2010) away from focusing on recur-ring conditions that increase the likelihood of internal conflict and toward micro-level dynamics. This theoretical and analytical shift yields clear payoffs: the richnessof case study detail is no longer sacrificed in favor of sweeping generalizations gen-erated from quantitative investigations; theory and data are more closely matched;we have a better grasp of both agency (actor preferences and behavior) and contin-gency (subnational and national environmental conditions conducive to violence).While the disaggregation approach is definitely attractive, it nevertheless suffersfrom a key drawback: too much disaggregation ineluctably leads to a shift inexplanandum—from civil war as a process which encompasses both violent and

33Emphases mine.34By requiring an arbitrary minimum number of casualties to code for civil war onset, all existing quantitative

civil war data sets conflate onset with violence.

85Adrian Florea

nonviolent forms of collective action,35 to simply the variation in the level of vio-lence.36 Therefore, disaggregation is no panacea for the lingering problem of civilwars’ start and end dates.

I argue that, with more elastic indicators of conflict onset, such as war declara-tions, casualty-free military skirmishes, or de facto territorial control, disaggregationmay not even be necessary. Critics may counter that indicator elasticity will blurthe civil war concept beyond recognition. I contend that good theory can be fruit-fully informative when coming up with good indicators of conflict onset. In inter-state war research, a war declaration signals war onset even though violence is notcoterminous with it. Similarly, a war declaration in civil conflicts (or any observableaction equivalent to war declaration) can also signal war onset with violence suc-ceeding it. Consider briefly, for the sake of exposition, internal conflicts over theso-called de facto states—state-like polities that lack international recognition (forexample, Abkhazia; Nagorno-Karabakh; Republika Srpska; Somaliland; Transnistri-a; Turkish Republic of Northern Cyprus; Western Sahara). A cursory look at thehistorical evidence suggests that de facto control over a territory that belongs to asovereign state usually predates the onset of hostilities between the titulargovernment and the collective actor in control of the de facto state.37

Another objection to relying on more elastic indicators of civil war onset may bethat events such as war declarations, casualty-free military skirmishes, or de factoterritorial control are not at all common in internal conflicts. How else can weobserve onset then? Can onset be visible independently of the level of violence?To address this conundrum requires that we look at civil wars as larger processesof contention where various forms of violent and nonviolent collective actionco-evolve. A political contention perspective (Tilly 1978; Rule 1988; McAdam, Tarrow,and Tilly 2001; Tarrow 2011) strongly calls for a separation of ‘‘the determinantsof collective action from the determinants of violent outcomes to collectiveaction’’ (Tilly 1978:183). In other words, we need to distinguish between determi-nants of civil war onset and determinants of violence in civil war situations. If onsetand violence are indeed etiologically distinct, the start date for a civil war needs tobe theorized a priori rather than established a posteriori based on recorded levelsof violence. When events such as war declarations, casualty-free military skirmishes,or de facto territorial control are not clearly observable, we can, alternatively, lookfor indicators of a contested sovereignty situation which signals the emergence of (a)collective actor(s) threatening the displacement of existing power holders (Spruyt2005).

In many respects, a contested sovereignty situation is not unlike a multiple sov-ereignty situation that in the social movement literature (Gamson 1975; Tilly1978; Rule 1988; Goldstone 2001; Goodwin 2001; Tarrow 2011) marks the onsetof revolution. Multiple sovereignty in revolutions obtains when a ‘‘populationfind themselves confronted with strictly incompatible demands from the govern-ment and from an alternative body claiming control over the government—andobey the alternative body. They pay taxes to it, provide men for its armies, feedits functionaries, honor its symbols, give time to its service, yield other resources,

35See my proposed definition for civil war above.36The recent trend toward methodologically rigorous geospatial work on local, national, and transnational

dynamics of conflict (Buhaug and Gleditsch 2008; Cederman, Buhaug, and Rød 2009; Raleigh and Hegre 2009;Buhaug 2010) suffers from the same ‘‘disaggregation bias’’: these studies may purport to be about civil war, butthey are really about violence. I thank an anonymous reviewer for bringing this point to my attention.

37Contrary to the dominant view in the field (Chapman and Roeder 2007; Roeder 2007), I argue that the initialformation of many de facto states is not driven by war. In other words, many de facto states are not post-conflict insti-tutional arrangements; rather, they are endogenous products of center-periphery interactions. If violence eventuallybreaks out, de facto statehood usually predates it and is typically the main casus belli. For example, the 1992 violencein Moldova over Transnistria occurred after pro-Russian actors established a state-like entity on the left bank of theNistru river.

86 Where Do We Go from Here?

despite the prohibition of the still-existing government that they formerlyobeyed. Multiple sovereignty has begun’’ (Tilly 1975b:520–1). Thus, under multi-ple sovereignty, we observe fragmentation of state power (Tarrow 2011:210) suchthat no single legitimate authority dominates the entire government apparatus:‘‘different elements of state power are controlled by various members of the pol-ity (government and challengers)’’ (Rule 1988:178). Multiple sovereignty in revo-lutionary environments ceases to exist when a collective actor establishes controlover the government (Tilly 1978:191).

If a contested sovereignty situation could indicate civil war onset, when wouldit obtain? We would recognize such a situation when a government, partsthereof, or a given territory under its jurisdiction becomes the object of mutuallyexclusive claims from one or more mobilized collective actors.38 Several indicatorscould point to such circumstances: displacement of a power holder by another;government’s loss of control over the army and other instruments of violence; acollective actor’s control over some portion of the government (for example, ter-ritory; functional subdivision); fragmentation of the polity into two or moremobilized collective actors, each claiming exclusive control of the government.39

Granted, this alternative method for determining civil war onset is quite labor-intensive as it forces the analyst to look harder at historical facts. Yet, the payoffsare well worth the effort—by examining processes in-depth, the investigator willend up having a better theoretical handle of the complexity of civil war, and theanalytical gains will accrue exponentially with the unveiling of each episode inthe conflict.

If violence is retained as the most reliable criterion by which we identify civilwar onset, duration, and termination, we’ll end up stating that we have more orless of a civil war rather than saying that we have a civil war with more or less vio-lence. In internal conflict, collective actors may resort to both violent and nonvi-olent forms of collective action40 (for example, assassinations, riots, strikes,demonstrations, looting, ethnic cleansing, politically motivated arrests, indiscrim-inate killing). Therefore, to paraphrase Clausewitz’s famous aphorism, violence isthe mere continuation of collective action by other means—it is yet another ‘‘form ofcollective action, oriented to the same purposes that contending [actors] pursuein ‘normal’ conditions’’ (Rule 1988:175). Violence, as an extreme but ‘‘normal’’process of collective action, is but an outgrowth of the strategic interactionamong actors (Tilly 1978:183).

Second, if violence is indeed endogenous to war rather than a preconditionthereof, then we need to be more rigorous about theorizing on the role of theparticipants in civil war—particularly the state—in perpetuating or limiting theextent of violence. Unfortunately, the causal role of the state—as an actor and asa set of institutions that enable or constrain behavior—remains under-theorized.Many cross-sectional accounts of civil war occurrence, duration, and terminationfocus on country-year structural characteristics that may facilitate the break-outor escalation of violence rather than on actors’ strategic interactions. By doingso, they are only able to conclude that civil war has a greater likelihood of ‘‘hap-pening’’ in a given country-year. This is utterly unsatisfactory because we are leftwith an incomplete view of the dynamic interplay between the sides involved inthe outbreak and perpetuation of violence. Any conflict—global, regional, intra-state, or otherwise—is fundamentally centered on dynamic actor behaviors rather

38Bell (1973:118) situates more clearly this idea of a contested sovereignty situation when he argues that ‘‘civilwars tend rapidly to resemble wars of secession as the rival elites stake out areas of strength to which their support-ers can drift.’’

39A contested sovereignty situation would offer a more reliable estimate of the exact start date for the 2011 Lib-yan civil war, for example. It is more realistic to consider rebels’ control over Eastern Libya as marking the begin-ning of the conflict rather than the moment when 25 battle-related deaths were registered.

40Or, better put, to more or less violent forms of collective action.

87Adrian Florea

than on static (or slow-moving) structural elements.41 As Fearon (1995) usefullyreminded us a while ago, our understanding of war in general will remainincomplete without a theory of strategic interaction.42

Thus, to get a clearer picture of actors’ interplay in civil war—in effect, to fullycomprehend the ebb and flow of violent episodes throughout the entire span ofan internal conflict—a minimal set of assumptions about the state (strong, weak,non-present, full participant) is necessary from the very outset. For example, if webuild our explanation for the variation in violence on a state weakness assump-tion, we are able to shed clearer light on the bottom-up processes that are condu-cive to violence in the sense that state weakness offers ample political space andstructural opportunities for social actors to direct violence against the state ortoward one another. Conversely, if we construct our accounts for the fluctuationsin violence on a state strength assumption, we can better illuminate the top-downprocesses through which the state engineers or directs violence toward socialactors.43 Furthermore, if we are more precise in our assumptions and envision apoorly institutionalized state, then our analytical framework for unveiling the vari-ation in violence may focus exclusively on the strategic interactions that takeplace within the state apparatus. For example, Acemoglu, Ticchi, and Vindigni(2009) have argued that in poorly institutionalized states, the political moralhazard caused by a strong military is the main catalyst for the perpetuation ofinternal violence. Their argument is elegant in its simplicity:

[In poorly-institutionalized states], the checks that would prevent a strong mili-tary from intervening in domestic politics are absent. This makes the building ofa strong army a double-edge sword for many civilian governments, even if suchan army is necessary for defeating rebels and establishing the monopoly of vio-lence over their territory…The civilian government-military interaction is compli-cated by the fact that the elite cannot credibly commit to not reforming anddownsizing the military once the civil war is over. Consequently, a strongermilitary, which is necessary for defeating the rebels, may also attempt a coup.(Acemoglu et al. 2009:1)

If violence is indeed endogenous to conflict, then a promising starting pointfor analyzing the variation in violence would be to consider the nature of thestate and its behavior as variable rather than fixed. It is high time we ‘‘broughtthe state back in’’ our theories of civil war: ‘‘changes in state behavior ratherthan particular classes of states, may better explain episodes of [v]iolence [incivil wars]…The state, rather than being absent and unable to serve as a securityguarantor, is often an active participant in violence’’ (Chenoweth and Lawrence2010:7,12).44

To conclude this section, how can we find our way out of the conceptual andtheoretical morass that has engulfed the large-N civil war literature? One solu-tion to the problems induced by the casualty-threshold approach would be to goback to the drawing board and recode the end date of civil wars based on clear-cutobservable outcomes, such as enforceable peace agreements, complete militaryvictory from one side, unambiguous institutional outcomes (secession, auton-omy), rather than on arbitrary battle-related deaths. This is the hard road. Alter-natively, a palliative solution to the problematic casualty-threshold approach

41Bargaining, information asymmetries, credible commitment theories of war are built around this core assump-tion about conflict writ large.

42See Licklider (1993), Mason and Fett (1996), and Rosenau (1964) for other earlier efforts aimed at uncover-ing the strategic nature of internal conflicts.

43The state can do that through direct repression (Davenport 2007) or third-party-sponsored violence (Reganand Norton 2005), for instance.

44See Cederman and Girardin (2007:175) for an earlier call to ‘‘bring the state back in’’ civil war research.

88 Where Do We Go from Here?

would be to employ a split-population estimator that takes into account two pop-ulations of cases: one population composed of immune cases that will neverexperience civil war and one population of exposed cases that are likely to expe-rience renewed violence. Finally, a substitute coding for the start date of a civilwar is to focus on more elastic indicators of conflict onset (war declarations,casualty-free military skirmishes, de facto territorial control, contested sovereigntysituation) which may indicate that conflict onset actually predates violence.

Systemic Causes of Civil War

It is inappropriate to treat civil war as a fully domestic phenomenon. (Gleditsch2007:294)

I now turn to another insufficiently explored problematique in the civil warresearch program—systemic causes of internal conflict. The current civil war lit-erature focuses chiefly on national-level attributes for unveiling the causes ofinternal conflict.45 This closed-polity approach whereby countries are treated inisolation is incomplete for having a clearer image of civil conflict.46 As Skocpol(1979:30) observed some time ago, the larger geopolitical environment ‘‘createstasks and opportunities for states and places limits on their capacities to copewith either external or internal tasks or crises.’’47 States are not isolated polities,but are embedded in both regional and international structures; therefore, wemust pay closer attention to those regional and international processes thataffect civil war onset or duration.

Existing research on the external dimensions of civil war overwhelmingly con-centrates on two classes of effects: neighborhood and systemic effects. The first isby far the most popular. Scholars have proposed a myriad of mechanismsthrough which neighborhood affects the likelihood of conflict or alters civil wardynamics. For instance, we know that the transnational presence of ethnic kinincreases the likelihood (Woodwell 2004; Cederman et al. 2010) and duration(Salehyan 2008) of civil war; civil wars tend to be spatially interdependent (Gled-itsch 2007); refugee flows constitute a negative externality of internal conflictand increase the probability of violence by facilitating the transnational spreadof arms, combatants, and ideologies and by altering existing ethnic compositions(Gleditsch and Salehyan 2006); diaspora in neighboring countries provide sup-port for the rebellion, especially when fatalities are involved (Woodwell 2004);and regional rivalries enable access to external bases and contribute to longerduration of conflict (Salehyan 2008).

If the neighborhood or sub-systemic effects have so far received a great dealof attention, the systemic effects have been addressed less frequently. Trulysystemic explanations are supposed to zoom out micro- and meso-level dynam-ics; that is, they should locate explanatory variables at a higher level of aggre-gation than the regional, the national, or the subnational. Such theorieswould privilege explanations at a higher level of abstraction than cross-borderlinkages, refugee flows, or support from external actors that are so prevalentin the ‘‘neighborhood effect’’ strand of the literature on the external dimen-sions of civil wars. Systemic frameworks would concentrate, for instance, onmacro-level factors such as patterns of decolonization and state formation

45Most of the covariates employed for testing hypotheses on civil war onset and duration are domestic (forexample, regime type; state strength; level of development) rather than systemic.

46Buhaug and Gleditsch (2008) embrace a similar view.47Emphasis mine. Gourevitch (1978), Regan (2000), and Akcinaroglu and Radziszewski (2005) also seek to grap-

ple with the larger geopolitical determinants of internal violence.

89Adrian Florea

accompanying global wars, changes in international requirements for state-hood that may encourage various forms of internal violence, or changes inthe technology of warfare and rebellion associated with a global shock such asthe end of the Cold War.

A few studies from without and within the field look rigorously at systemicdrivers of internal conflict. One macro-level explanation for the prevalence ofcivil wars comes from Wimmer and Min (2006) who turn Tilly’s (1975a, 1985,1990) war-making-state-making argument on its head, and posit that both inter-and intrastate conflicts are outcomes of similar state formation processes. Ineffect, Wimmer and Min suggest that it is state formation processes which inducea combustible co-evolution of both inter- and intrastate war rather than the otherway around:

Both civil and inter-state wars are fought over the institutional structure of thestate, and thus are most likely to break out when these institutional principlesare contested… The mechanisms relating nation-state formation to war are simi-lar: wars over territory inhabited by co-nationals on the other side of a state bor-der (commonly called irredentist or revanchist wars) follow the same logic ofnationalist politics that may drive civil wars, as majorities and ethnic minoritiescompete for control over the state. (Wimmer and Min 2006:869)

Another systemic account (at least for a subset of civil wars) comes from Hiro-naka (2005) who contends that the post–World War II refusal of the interna-tional community to countenance secession outside the decolonization processhas encouraged the proliferation of civil wars by protecting weak states with arti-ficial and contested borders from fragmenting. Kalyvas and Balcells (2010) pro-pose a third macro-level explanation which suggests that the end of the ColdWar has fundamentally altered the technology of rebellion and has led to theproliferation of long-lasting symmetric-unconventional internal wars. Finally, Ras-ler (1983) argues that it is misleading to characterize internal wars as havingessentially domestic origins. Using the case studies of Lebanon in 1958 and1975, she insists that external forces create the conditions for internal war. Bothcivil wars, she posits, were inspired by regional and international politics.

These macro-level studies of internal conflict do indeed offer plausibleaccounts for the prevalence of civil wars but share one major limitation: theyleave underspecified the exact processes through which larger systemic forcesaffect the likelihood or duration of internal war. For instance, Kalyvas andBalcells’ argument about the effect of the Cold War on the technology of rebel-lion necessitates a proper specification of impact. Should we expect a step func-tion whereby following the abrupt collapse of the bipolar system rebels areconfronted with different but permanent incentive structures? Or should weanticipate a pulse function whereby the effect of the end of the Cold War isstrongly felt in the proximate years but then diminishes with time? How muchdoes it last before all actors adapt their strategies to new systemic conditions?More importantly, if both inter- and intrastate wars tend to co-evolve during stateformation waves, as Wimmer and Min (2006) suggest, can we really differentiatebetween these two types of conflict? Are they analytically equivalent? I discussthat possibility in the next section.

Are Civil Wars Fundamentally Different from Interstate Wars?

General theories of conflict should be applicable to both civil and inter-state con-flicts. (Cunningham, Gleditsch, and Salehyan 2009:595)

90 Where Do We Go from Here?

Might interstate and civil wars be caused by a similar constellation of factors?This is both a theoretical and empirical question. The conflict literature main-tains a strict distinction between intra- and interstate wars that may not be as the-oretically clear-cut as the terms suggest and that may not justify separate study.The type of unit involved in warfare is the standard discriminating feature that dis-tinguishes interstate conflicts from intrastate conflicts where an interstate dyadcontains two sovereign states and an intrastate dyad contains one or none. Thiscategorization rests on two key assumptions: (i) that sovereignty is effectivewithin sovereign states and (ii) that state behavior is fundamentally differentfrom nonstate behavior. As a result of this dichotomization, there has initiallybeen little collusion between the two types of literatures with interstate conflictresearchers being skeptical about applying theories of international conflict towars within countries. However, as the civil war research program became popu-lated with IR scholars in the early 1990s, it soon became clear that ‘‘sovereigntyis clearly less than fully effective within many existing states, and similar prob-lems of enforcement and contracting under anarchy obtain within states as well’’(Gleditsch 2002:75). As a consequence of the influx of IR students into the tradi-tionally marginal field of civil wars, we learned that there are important similari-ties in bargaining and conflict processes between interstate and civil wars:48

Posen (1993) pointed out that, under conditions of the breakdown of domesticorder, an internal security dilemma is the principal catalyst for prolonged vio-lence; Fearon (1995) and Walter (1997, 2002) argued that the same commitmentproblems found at the international level are present among internal conflictactors as well; Toft (2002, 2003, 2006) posited that just like sovereign states failto reach agreements because some issues are indivisible,49 issue indivisibility alsoprevents civil war participants from striking long-lasting bargains.

So, if anarchic situations and bargaining issues affect state and nonstate actorsin similar ways, what justification is there left for separate study? Why not testour theories on both types of conflict? Scholars’ reluctance to do so may begrounded in the realization that, despite a few similarities, there still remainimportant distinctions between inter- and intrastate wars.50 Civil wars are longer,more frequent, and more deadly than interstate wars (Regan 2002; Collier andHoeffler 2004; Lacina 2006); are fought by actors with asymmetric capabilities(Arreguın-Toft 2005); exhibit more salient commitment problems that interna-tional wars due to the post-conflict environment where internal actors, unlikestates, have to interact frequently (Fearon 2004); unfold differently from inter-state conflicts with civilians being more likely to be disproportionately victimized(Kalyvas 2006); and display a wider range of outcomes than international wars(Chapman and Roeder 2007). However, upon closer inspection, it becomes obvi-ous that the above distinctions pertain to how wars unfold and terminate (thatis, to war processes) rather than to their underlying causes.

So, are we looking at the right set of denominators that allow us to differenti-ate between the two types of wars? I submit that the main reason for our diffi-culty in adjudicating between the competing views on the similarities ⁄ differencesof interstate and civil wars has something to do with a unit type bias in conflictresearch. As long as the literature remains dominated by a focus on the sover-eign state as the sole unit worthy of investigation, we’ll always find similaritiesand differences between the two types of wars and the debate will not be put tofinal rest.

48See the special International Interactions issue [2009, Vol. 35 (3)] dedicated to the idea of applying IR theoriesto civil war. The potential analytical equivalence between inter- and intrastate wars is discussed at length in Cunn-ingham and Lemke (n.d.).

49While Toft (2003, 2006, 2010) looks at issue indivisibility as a constant that makes many wars irresoluble, forGoddard (2009) indivisibility is a variable.

50I build on Cunningham et al. (2009) and Cunningham and Lemke (n.d.) in this paragraph.

91Adrian Florea

Mainstream IR research privileges the nation-state as the unit of analysis. Whileexisting works pay due attention to the variation in the behavior of states as stan-dard units in international politics, the nature of these units is generally treatedas a constant. Most of the literature is replete with approaches that take the insti-tutional form of the modern state as given and exclude alternative polities fromanalysis. Thus, nonstate actors are often dismissed as mere anomalies to aninternational system dominated by officially recognized states. This is quite sur-prising given that ‘‘the nature of the units among which relations occur hasimportant implications for the structure of international relations’’ (Herz1957:473). Unit type matters at least as much as the distribution of capabilitiesamong units. As Spruyt (1996:5–17) puts it, structure is at least partially deter-mined by the prevalent type of unit. Hence, to fully understand the structure ofthe international system, we have to first explain the synchronic and diachronicvariation in the type of units that exist in the system.

At any given point in history, there has been little systemic stasis in terms ofthe types of unit in the system—alternative units have constantly formed. A polit-ical map of pre-Westphalia Europe shows a plethora of independent or semi-independent principalities, duchies, and states-kingdoms. Post-Westphalia, themodern state coexisted with city-states, empires, and city-leagues. Even inthe post–World War II geopolitical landscape, nonstate actors have been a con-stant feature of a world dominated by recognized sovereign states. From this per-spective, systems are rarely stagnant or stable. This shows that, when it comes tounit type, the international system is much more kinetic than mainstream IR the-ories portray it (Fazal 2007:40).51

One solution to the debate over the interstate–intrastate war dichotomy isto allow for variation in the type of units operating in international politics.This way we can better address the question of whether the interstate–intra-state conflict distinction is theoretically justifiable. This approach would entailshifting our analysis from a legalistic logic to a territorial logic—that is, frominternationally recognized sovereign states and unrecognized nonstate actorsto geographic entities that exert de facto control over a given space (regardlessof their ‘‘official’’ status) for varying amounts of time.52 A territorial logicwould enable us to shed clearer light, for instance, on those conflicts betweena sovereign state and a de facto state (for example Georgia vs. Abkhazia) thatare traditionally categorized as civil wars but that look very much like conven-tional interstate wars. This approach would also require redefining war in amore encompassing way. Because it is devoid of any unit type bias, Levy andThompson’s (2011) suggestion may be a promising start in that direction:war, in general, can be simply understood as sustained, coordinated violencebetween political organizations. Hence, to resolve the lingering debate over theinterstate–intrastate war dichotomy, we need stronger theoretical foundations,not data.

A Closing Word on Methodology

No single factor and no single level of analysis provides a complete explanationfor the causes of war. As a result, theories of war must necessarily combine causalvariables from different levels of analysis. (Levy 2007:22)

51The deficit of attention to nonstate actors in the extant IR literature is quite surprising because IR theories‘‘are not logically restricted to the analysis of officially-recognized states’’ (Lemke 2008:774).

52To my knowledge, only Wimmer and Min (2006) rigorously embrace this approach and focus on geographicterritories rather than states as the units of analysis. This allows them to explain how the interactions among unitsare conducive to global or local wars or how different types of wars tend to co-evolve.

92 Where Do We Go from Here?

With a few exceptions,53 quantitative studies of civil war focus on the country-year as the main unit of analysis. The modal strategy is to take an off-the-shelfdata set on civil wars for a given time period (usually, 1945–2008) and constructlogit or probit models to probe whether certain country-level structural covariates(for example, regime type; level of development; resource endowments) affectthe likelihood of civil war onset, duration, or termination. Obviously, thediscussion above on the conceptual and theoretical lacunae suggests that thisapproach is utterly inappropriate for fully capturing the dynamics of such a com-plex phenomenon as internal conflict. The conceptual and theoretical gaps mustbe translated into clear-cut methodological remedies. In other words, we need todo much better than merely stating that civil war has a greater likelihood of‘‘happening’’ or ‘‘lasting’’ in a given country-year.

A first remedial step is to abandon the country-year focus and recast civil warin a dyadic perspective with the dyad-year as the main unit of analysis.54 Theoriesof war are dyadic: it is the strategic interactions between actors rather thannational-level characteristics that shape war processes and outcomes. Cunning-ham et al. (2009) observe that civil war studies suffer from a failure to produceresearch at the dyadic level. Because conflict onset and duration are oftenendogenous to the bargaining between actors (who operate under certain struc-tural constraints), a dyadic approach allows us to come up with more realistic,dynamic theories of internal conflict. The authors point out not only the analyti-cal gains accrued with a dyadic approach, but also the difficulties of producingsuch dyadic data. For instance, it would not be feasible to come up with a list ofall possible dyads. Hence—for now at least—we must be satisfied looking at alimited set of dyads.55

While the dyadic perspective is definitely a step forward, it needs to rest on afirmer theoretical and methodological foothold. The multiplicity of players inmany civil wars may even suggest moving beyond a dyadic path. Some civil warsmay be better explained in regional rather than dyadic terms. For example,would a dyadic approach be appropriate for explaining civil war onset and dura-tion in Zaire (DRC) where so many internal and external players were involvedin the break-out and perpetuation of hostilities? Moreover, our estimation proce-dures with the dyadic approach need to take into account that dyad-year observa-tions may not independent of one another—subsequent observations for thesame dyad over time and space may violate the independence assumption (Beckand Katz 1996; Beck 2007).

Another methodological remedy is to model the spatial and temporal depen-dence of conflicts more rigorously. Currently, there is a big mismatch betweenthe dynamic civil war processes and the static, cross-sectional models beingemployed to capture them. If civil wars are temporally and spatially dependent,then we need to incorporate that empirical reality into our modeling strategies.This means going beyond standard corrections for temporal and spatial correla-tion in cross-sectional models56 and constructing properly specified duration andspatial regression models.57 The standard method for modeling temporal depen-dence has been to use time dummies or splines in logit models. Carter and Sign-orino (2010) show that time dummies can lead to estimation problems, andrecommend using smoothed functions of time via a cubic polynomial or spline.As for spatial dependence, it is generally treated as a statistical nuisance ratherthan estimated properly. Ignoring spatial dependence in civil war research will

53Cf. Cunningham et al. (2009); DeRouen and Bercovitch (2008).54See note above for the few recent studies that rely on the dyad-year as the main unit of analysis.55The authors suggest looking at armed insurgents who have shown a willingness to fight.56See Beck, Katz, and Tucker (1998); DeBoeuf and Keele (2008).57See Box-Steffensmeier and Jones (2004), and Ward and Gleditsch (2008), respectively.

93Adrian Florea

tend to underestimate the real variance in the data. If a country’s level of vio-lence is associated with a neighbor’s propensity for violence, this tells us some-thing important about the regional distribution of violence. Hence, spatialassociation may be a substantive feature of violence rather than a simple statisticalnuisance. If that’s the case, then the values of the dependent variable (violence)in country i are directly influenced by the values of the dependent variable in i’sneighbors, and a spatially lagged dependent variable model (SLDV model) maybe appropriate.58

Another methodological remedy relates to where the covariates of civil war,onset, and duration are located. If theory tells us that they are to be found atboth the macro-(or systemic) and meso-(or country) levels, then this needs to beclearly reflected in our model-building efforts. Hierarchical or multilevel modelsare well suited for capturing the interplay of causal factors located at differentlevels of analysis—these models incorporate the effect of explanatory factorsfrom nested structures (subnational, national, systemic).59 Finally, if civil warsexhibit a wider range of possible outcomes than other types of wars as shown byChapman and Roeder (2007), we could illuminate a war’s possible trajectories byusing competing risks hazards models where the hazards of various civil war out-comes are modeled jointly rather than separately.60

Conclusion

Where do we go from here? I have pointed out throughout this article thatbefore we are likely to make headway in reducing our collective ignorance aboutcivil wars, the problem of what civil wars are and how best to measure them mustbe directly confronted. So, first and foremost, we must go back to better concep-tualization of civil war in terms of its starting and ending points. Civil war maybe a longer process than generally described in the existing literature. It may bethe case that civil wars are longer-lasting phenomena that exhibit waves of escala-tion and de-escalation of violence occurring throughout a longer time frame.

Second, we have to turn back to theory. Specifically, we need theories of civilconflict that would be different from theories of civil violence. All quantitativecivil war data sets conflate onset with violence. Violence is typically endogenousto civil war processes—therefore, it may be the case that a civil war actually starts(with group organization for collective action; arming; war declarations; de factocontrol over a territory) well before violence ensues.

Third, we have few systemic explanations of internal conflict. Systemic explana-tions zoom out lower-level dynamics; they locate the explanatory variables at ahigher level of aggregation than current research does. A truly systemic-level the-ory of civil war would look for explanations at a higher level of abstraction thantransborder effects or the presence of diaspora communities that are so preva-lent in the extant literature on the external dimensions of civil wars. Such a the-ory would plausibly concentrate on macro-level factors such as patterns of stateformation, or changes in international requirements for statehood that mayencourage various forms of internal dissent.

58Ward and Gleditsch (2008) discuss this model in detail. The SLDV model resembles the more familiar autore-gressive time-series model where temporal serial correlation is addressed by including a lagged dependent variableas a regressor. The SLDV model fits a parameter which captures the spatial dependence of observations on thecharacteristic of interest.

59Hierarchical models can elucidate cross-level mechanisms that may affect the likelihood of civil war. One plau-sible mechanism would unfold as follows: protracted inter-state rivalry > (leads to) centralization of domestic powerfor extractive purposes in at least one of the rivals > (which leads to) resistance from certain segments of the popu-lation > (which leads to) > repression > internal war > regionalization ⁄ internationalization of war. On hierarchical ⁄multilevel models, see Gelman and Hill (2007).

60For an application of such models to conflict, see Bennett (n.d.).

94 Where Do We Go from Here?

Fourth, we need better theories that would enable us to determine whethercivil wars are really different from international wars. Scholars (Posen 1993;Fearon 1995; Toft 2003; Walter 2004, 2009) have pointed to mechanisms thatare common to both forms of conflict: credible commitment problems; securitydilemma; indivisible issues. Nonetheless, existing theories do not take intoaccount that warfare (interstate or intrastate) may be epiphenomenal to largersystemic characteristics that are related to the prevalent type of unit in interna-tional politics. Finally, we need to be more methodologically rigorous and paycloser attention to our unit of analysis, spatial and temporal dependence, andmultilevel causality when building our models for civil war onset and duration.

All the above suggest that a lot more work needs to be done. Further progressrequires fundamental changes in the way we think about and study such a com-plex phenomenon as civil war. No research program is truly progressive basedsolely on data collection efforts. The field does not suffer from a dearth ofappropriate methodologies, but from a paucity of conceptual and theoreticalinsights. For the large-N civil war research program to advance in fruitful direc-tions, small steps back to sounder conceptualizations, to better linkages betweentheory and data will actually represent great leaps forward.

References

Acemoglu, Daron, Davide Ticchi, and Andrea Vindigni. (2009) Persistence of Civil Wars. Work-ing Paper. Institute for the Study of Labor, Bonn.

Akcinaroglu, Seden, and Elizabeth Radziszewski. (2005) Expectations, Rivalries, and Civil WarDuration. International Interactions 31: 349–374.

Arreguın-Toft, Ivan. (2005) How the Weak Win Wars: A Theory of Asymmetric Conflict. New York: Cam-bridge University Press.

Autesserre, Severine. (2010) The Trouble with the Congo: Local Violence and the Failure of InternationalPeacebuilding. New York: Cambridge University Press.

Beck, Nathaniel. (2007) From Statistical Nuisances to Serious Modeling: Changing How We Thinkabout the Analysis of Time-Series–Cross-Section Data. Political Analysis 15: 97–100.

Beck, Nathaniel, and Jonathan N. Katz. (1996) Specifying and Estimating Time-Series–Cross-Sec-tion Models. Political Analysis 6: 1–36.

Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. (1998) Taking Time Seriously: Time-Series–Cross-Section Analysis with a Binary Dependent Variable. American Journal of PoliticalScience 42: 1260–1288.

Bell, David V.J. (1973) Resistance and Revolution. Boston: Houghton-Mifflin.Bennett, Scott D. (n.d.) Balance of Power Dynamics and War: A Competing Risks Model of Escala-

tion. Working Paper. Pennsylvania State University, University Park.Blattman, Christopher, and Edward Miguel. (2010) Civil War. Journal of Economic Literature

48: 3–57.Box-Steffensmeier, Janet M., and Bradford S. Jones. (2004) Timing and Political Change: Event His-

tory Modeling in Political Science. Ann Arbor: University of Michigan Press.Brubaker, Rogers. (2004) Ethnicity without Groups. Cambridge, MA: Harvard University Press.Buhaug, Halvard. (2010) Dude, Where’s My Conflict? LSG, Relative Strength, and the Location of

Civil War. Conflict Management and Peace Science 27: 107–128.Buhaug, Halvard, and Kristian Skrede Gleditsch. (2008) Contagion or Confusion? Why Con-

flicts Cluster in Space. International Studies Quarterly 52: 215–233.Carter, David, and Curtis Signorino. (2010) Back to the Future: Modeling Temporal Depen-

dence in Binary Data. Political Analysis 18: 271–292.Cederman, Lars-Erik, and Luc Girardin. (2007) Beyond Fractionalization: Mapping Ethnicity onto

Nationalist Insurgencies. American Political Science Review 101: 173–185.Cederman, Lars-Erik, Halvard Buhaug, and Jan Ketil Rød. (2009) Ethno-Nationalist Dyads and

Civil War: A GIS-Based Analysis. Journal of Conflict Resolution 53: 496–525.Cederman, Lars-Erik, Andreas Wimmer, and Brian Min. (2010) Why Do Ethnic Groups Rebel?

New Data and Analysis. World Politics 62: 1–30.Chapman, Thomas, and Philip G. Roeder. (2007) Partition as a Solution to Wars of Nationalism:

The Importance of Institutions. American Political Science Review 101: 677–691.

95Adrian Florea

Chenoweth, Erica, and Adria Lawrence. (2010) Rethinking Violence: State and Nonstate Actors in Con-flict. Cambridge, MA: MIT Press.

Chenoweth, Erica, and Maria J. Stephan. (2010) Mobilization and Resistance: A Framework forAnalysis. In Rethinking Violence: State and Nonstate Actors in Conflict, edited by Erica Chenowethand Adria Lawrence. Cambridge, MA: MIT Press.

Colaresi, Michael P., Karen Rasler, and William R. Thompson. (2008) Strategic Rivalries in WorldPolitics: Position, Space, and Conflict Escalation. New York: Cambridge University Press.

Collier, Paul, and Anke Hoeffler. (2004) Greed and Grievance in Civil War. Oxford EconomicPapers 56: 563–595.

Cunningham, David E., and Douglas Lemke. (n.d.) Combining Civil and Interstate Wars. WorkingPaper. Iowa State University, Ames.

Cunningham, David E., Kristian Skrede Gleditsch, and Idean Salehyan. (2009) It Takes Two:Dyadic Interactions and Civil War Duration. Journal of Conflict Resolution 53: 570–597.

Davenport, Christian. (2007) State Repression and the Domestic Democratic Peace. New York: CambridgeUniversity Press.

DeBoef, Suzanna, and Luke J. Keele. (2008) Taking Time Seriously: Dynamic Regression. AmericanJournal of Political Science 52: 184–200.

DeRouen, Karl, and Jacob Bercovitch. (2008) Enduring Internal Rivalries: A New Framework forthe Study of Civil War. Journal of Peace Research 45: 55–74.

Diehl, Paul F., and Gary Goertz. (2000) War and Peace in International Rivalry. Ann Arbor: Univer-sity of Michigan Press.

Falleti, Tulia G., and Julia F. Lynch. (2009) Context and Causal Mechanisms in Political Analysis.Comparative Political Studies 42: 1143–1166.

Fazal, Tanisha M. (2007) State Death: The Politics and Geography of Conquest, Occupation, and Annexa-tion. Princeton: Princeton University Press.

Fearon, James D. (1995) Rationalist Explanations for War. International Organization 49: 379–414.Fearon, James D. (2004) Why Do Some Civil Wars Last So Much Longer Than Others? Journal of

Peace Research 41: 275–301.Fearon, James D., and David Laitin. (2003) Ethnicity, Insurgency, and Civil War. American Political

Science Review 97: 75–90.Galtung, Johan. (1969) Violence, Peace, and Peace Research. Journal of Peace Research 6: 167–191.Gamson, William. (1975) The Strategy of Social Protest. Belmont, CA: Wadsworth Publishing Company.Gelman, Andrew, and Jennifer Hill. (2007) Data Analysis Using Regression and Multilevel ⁄ Hierarchical

Models. New York: Cambridge University Press.Gibler, Douglas M., and Jaroslav Tir. (2010) Settled Borders and Regime Type: Democratic Tran-

sitions as Consequences of Peaceful Territorial Transfers. American Journal of Political Science 54:951–968.

Gleditsch, Kristian Skrede. (2002) All International Politics Is Local: The Diffusion of Conflict, Integra-tion, and Democratization. Ann Arbor: University of Michigan Press.

Gleditsch, Kristian Skrede. (2004) A Revised List of Wars between and within Independent States,1816–2002. International Interactions 30: 231–262.

Gleditsch, Kristian Skrede. (2007) Transnational Dimensions of Civil War. Journal of Peace Research44: 293–309.

Gleditsch, Kristian Skrede, and Idean Salehyan. (2006) Refugees and the Spread of Civil War.International Organization 60: 335–366.

Goddard, Stacie E. (2009) Indivisible Territory and the Politics of Legitimacy. New York: Cambridge Uni-versity Press.

Goldstone, Jack. (2001) Toward a Fourth Generation of Revolutionary Theory. Annual Review ofPolitical Science 4: 139–187.

Goodwin, Jeff. (2001) No Way Out: States and Revolutionary Movements, 1945–1991. New York: Cam-bridge University Press.

Gourevitch, Peter. (1978) The International Sources of Domestic Politics. International Organization32: 881–912.

Herz, John. (1957) Rise and Demise of the Territorial State. World Politics 9: 473–493.Hironaka, Ann. (2005) Neverending Wars: The International Community, Weak States, and the Perpetuation

of Civil War. Cambridge, MA: Harvard University Press.Kalyvas, Stathis N. (2003) The Ontology of ‘‘Political Violence’’: Action and Identity in Civil Wars.

Perspectives on Politics 1: 475–494.Kalyvas, Stathis N. (2006) The Logic of Violence in Civil Wars. New York: Cambridge University Press.

96 Where Do We Go from Here?

Kalyvas, Stathis N. (2007) Civil Wars. In Oxford Handbook of Political Science, edited by Carles Boixand Susan Stokes. New York: Oxford University Press.

Kalyvas, Stathis N. (2010) Internal Conflict and Political Violence: New Developments in Research.In Rethinking Violence: State and Nonstate Actors in Conflict, edited by Erica Chenoweth and AdriaLawrence. Cambridge, MA: MIT Press.

Kalyvas, Stathis N., and Laia Balcells. (2010) International System and Technologies of Rebel-lion: How the Cold War Shaped Internal Conflict. American Political Science Review 104: 415–429.

Kalyvas, Stathis N., Ian Shapiro, and Tarek Masoud. (2008) Order, Conflict, and Violence. NewYork: Cambridge University Press.

King, Gary. (2001) Proper Nouns and Methodological Propriety: Pooling Dyads in InternationalRelations Data. International Organization 55: 497–507.

Klein, James P., Gary Goertz, and Paul F. Diehl. (2006) The New Rivalry Dataset: Procedures andPatterns. Journal of Peace Research 43: 331–348.

Klein, James P., Gary Goertz, and Paul F. Diehl. (2008) The Peace Scale: Conceptualizing andOperationalizing Non-Rivalry and Peace. Conflict Management and Peace Science 25: 67–80.

Lacina, Bethany. (2006) Explaining the Severity of Civil Wars. Journal of Conflict Resolution 50: 276–289.

Lemke, Douglas. (2008) Power Politics and Wars without States. American Journal of Political Science52: 774–786.

Levy, Jack S. (2007) International Sources of Interstate and Intrastate War. In Leashing the Dogs ofWar: Conflict Management in a Divided World, edited by Chester A. Crocker, Fen Osler Hampsonand Pamela Aall. Washington, DC: United States Institute of Peace Press.

Levy, Jack S., and William R. Thompson. (2011) The Arc of War: Origins, Escalation, and Transforma-tion. Chicago: University of Chicago Press.

Licklider, Roy. ed. (1993) Stopping the Killing: How Civil Wars End. New York: New York UniversityPress.

Mason, David T., and Patrick J. Fett. (1996) How Civil Wars End: A Rational Choice Approach.Journal of Conflict Resolution 40: 546–568.

McAdam, Doug, Sidney Tarrow, and Charles Tilly. (2001) Dynamics of Contention. New York:Cambridge University Press.

Metternich, Nils, and Julian Wucherpfennig. (2011) Civil War Onset and Recurrence: Imple-menting a Split-Population Estimation Framework. Paper presented at the 2011 annual conven-tion of the International Studies Association (Montreal, Canada).

Posen, Barry. (1993) The Security Dilemma and Ethnic Conflict. Survival 35: 27–47.Raleigh, Clionadh, and Havard Hegre. (2009) Population Size, Concentration, and Civil War: A

Geographically Disaggregated Analysis. Political Geography 28: 224–238.Rasler, Karen. (1983) Internationalized Civil War: The Impact of the Syrian Military Intervention

on the Lebanese Civil War. Journal of Conflict Resolution 27: 421–456.Regan, Patrick M. (2000) Civil Wars and Foreign Powers: Interventions and Intrastate Conflict. Ann

Arbor: University of Michigan Press.Regan, Patrick M. (2002) Third-Party Interventions and the Duration of Intrastate Conflicts. Journal

of Conflict Resolution 46: 55–73.Regan, Patrick M., and Daniel Norton. (2005) Greed, Grievance, and Mobilization in Civil Wars.

Journal of Conflict Resolution 49: 319–336.Roeder, Philip G. (2007) Where Nation-States Come From: Institutional Change in the Age of Nationalism.

Princeton: Princeton University Press.Rosenau, James. ed. (1964) International Aspects of Civil Strife. Princeton: Princeton University Press.Rotberg, Robert. (1999) Creating Peace in Sri Lanka: Civil War and Reconciliation. Washington DC:

Brookings Institution Press.Rule, James B. (1988) Theories of Civil Violence. Berkeley: University of California Press.Salehyan, Idean. (2008) No Shelter Here: Rebel Sanctuaries and International Conflict. Journal of

Politics 70: 54–66.Salehyan, Idean. (2009) Rebels without Borders: Transnational Insurgencies in World Politics. Ithaca, NY:

Cornell University Press.Sambanis, Nicholas. (2004) What Is a Civil War? Conceptual and Empirical Complexities of an

Operational Definition. Journal of Conflict Resolution 48: 814–858.Sartori, Giovanni. (1970) Concept Misformation in Comparative Politics. American Political Science

Review 64: 1033–1053.Sinno, Abdulkader. (2008) Organizations at War in Afghanistan and Beyond. Ithaca, NY: Cornell

University Press.

97Adrian Florea

Skocpol, Theda. (1979) States and Social Revolutions. New York: Cambridge University Press.Spruyt, Hendrik. (1996) The Sovereign State and Its Competitors. Princeton: Princeton University Press.Spruyt, Hendrik. (2005) Ending Empire: Contested Sovereignty and Territorial Partition. Ithaca, NY: Cor-

nell University Press.Tarrow, Sidney. (2007) Inside Insurgencies: Politics and Violence in an Age of Civil War. Perspectives

on Politics 5: 587–600.Tarrow, Sidney. (2011) Power in Movement, 3rd edition. New York: Cambridge University Press.Thompson, William R. (2001) Identifying Rivals and Rivalries in World Politics. International Studies

Quarterly 45: 557–586.Tilly, Charles. (1975a) Reflections on the History of European State-Making. In The Formation of

National States in Western Europe, edited by Charles Tilly. Princeton: Princeton University Press.Tilly, Charles. (1975b) Revolutions and Collective Action. In Handbook of Political Science. Vol. 3:

Macropolitical Theory, edited by Fred I. Greenstein and Nelson W. Polsby. Reading, MA: Addison-Wesley.

Tilly, Charles. (1978) From Mobilization to Revolution. Reading, MA: Addison-Wesley.Tilly, Charles. (1985) War Making and State Making as Organized Crime. In Bringing the State Back

In, edited by Peter B. Evans, Dietrich Rueschemeyer, and Theda Skocpol. New York: CambridgeUniversity Press.

Tilly, Charles. (1990) Coercion, Capital, and European States, AD 990–1992. Cambridge, MA: Black-well.

Toft, Monica. (2002) Indivisible Territory, Geographic Concentration, and Ethnic War. Security Studies12: 82–119.

Toft, Monica. (2003) The Geography of Ethnic Violence: Identity, Interests, and the Indivisibility of Territory.Princeton: Princeton University Press.

Toft, Monica. (2006) Issue Indivisibility and Time Horizons as Rationalist Explanations for War.Security Studies 15: 34–69.

Toft, Monica. (2010) Securing the Peace: The Durable Settlement of Civil Wars. Princeton: Princeton Uni-versity Press.

Vasquez, John A. (2008) The War Puzzle Revisited. New York: Cambridge University Press.Walter, Barbara F. (1997) The Critical Barrier to Civil War Settlement. International Organization

51: 335–364.Walter, Barbara F. (2002) Commiting to Peace: The Successful Settlement of Civil Wars. Princeton:

Princeton University Press.Walter, Barbara F. (2004) Building Reputation: Why Governments Fight Some Separatists but Not

Others. American Journal of Political Science 60: 105–135.Walter, Barbara F. (2009) Reputation and Civil War: Why Separatist Conflicts Are So Violent. New York:

Cambridge University Press.Ward, Michael D., and Kristian S. Gleditsch. (2008) Spatial Regression Models. Los Angeles: Sage

Publications.Weinstein, Jeremy M. (2007) Inside Rebellion: The Politics of Insurgent Violence. New York: Cambridge

University Press.Wimmer, Andreas. (2002) Nationalist Exclusion and Ethnic Conflict: Shadows of Modernity. New York:

Cambridge University Press.Wimmer, Andreas, and Brian Min. (2006) From Empire to Nation-State: Explaining Wars in the

Modern World, 1816–2001. American Sociological Review 71: 867–897.Woodwell, Douglas. (2004) Unwelcome Neighbors: Shared Ethnicity and International Conflict

during the Cold War. International Studies Quarterly 48: 197–223.

98 Where Do We Go from Here?