Why does armed conflict begin again? A new analytic approach

Abstract

An increasing majority of armed conflicts worldwide are episodes of recurrence (i.e. armed conflicts that occur after another armed conflict ends in the same country). This trend motivates a growing collection of scholarship on the causes of recurrence, but empirical research on this topic supports contradictory conclusions. Engaging these contradictions, this article introduces a novel approach to retroductive comparative analysis, which adapts the logic of semi-structured interviews for use in historical research, to examine the features and causes of 58 cases of armed conflict recurrence that took place between 1946 and 2005. Through this analysis, I identify five varieties of recurrence that are substantively and causally distinct. My analysis identifies scope conditions for existing theories of recurrence and, more broadly, demonstrates how distinct configurations of power inequalities, political opportunities, and the pursuit of legitimacy shape the proximate causes of each variety of recurrence.

Keywords

Armed conflict comparative historical methods configurational contentious politics political sociology recurrence

Armed conflict is a transformative force in society, reshaping individuals’ lives and social institutions alike (Goodwin, 2001; Tilly, 1978; Wimmer, 2014). In some cases, these transformations have resulted in stability and the diversification of political representation (Hughes, 2009; Hughes and Tripp, 2015). Yet, often, countries become trapped in ongoing cycles of armed conflict (Collier and Sambanis, 2002). Countries that have experienced one episode of armed conflict are significantly more likely to experience repeated episodes (Walter, 2004, 2015) and a growing majority of armed conflicts worldwide are episodes of recurrence (i.e. armed conflicts that begin after another armed conflict ends in the same country). In the 1990s, 67 percent of all armed conflicts occurred in countries with recent histories of conflict, with that figure rising to 90 percent in the first decade of the 21st century (The World Bank, 2011).

This trend motivates a growing collection of research on the causes of recurrence. Existing theories of recurrence offer varying explanations of this phenomenon (e.g. Doyle and Sambanis, 2000; Fortna, 2004; Hegre and Nygard, 2015; Kreutz, 2010; Licklider, 1995; Quinn et al., 2007; Sambanis, 2000; Walter, 2004, 2015). Yet, rather than simply emphasizing one cause over another, empirical research on this topic supports contradictory conclusions. Summarizing research on the causes of armed conflict recurrence, Call (2012) writes that there is “tremendous disparity among scholars about whether certain factors are important or not, and about the degree to which they are important” (p. 30). In extreme cases, different analyses show opposite effects for the same key variables (e.g. Kreutz, 2010; Sambanis, 2000; Walter, 2004).

In this article, I contend that these contradictions are due to a misalignment between how armed conflict recurrence is studied and the characteristics of recurrences as events. Research on the causes of armed conflict widely shows that substantively distinct varieties of armed conflict have divergent causes and implications (e.g. Kalyvas, 2007; Krcmaric, 2018; Sambanis, 2001; Wimmer et al., 2009). Recurrence is typically treated as one of these distinct varieties, and research on this topic examines the differences between recurrences and other conflict types. However, the defining characteristics of recurrence encompass a broad and heterogeneous set of armed conflicts. Consequently, there is reason to believe that there may be multiple varieties of recurrence with distinct causes requiring different explanations. Yet, efforts to distinguish between varieties of recurrence are notably absent.

To address this lacuna, I examine what differentiates episodes of armed conflict recurrence from one another. Using a novel approach to retroductive comparative analysis (Ragin and Amoroso, 2010) that adapts the logic of semi-structured interviews to historical research, I analyze the characteristics and causes of 58 cases of armed conflict recurrence that took place between 1946 and 2005. Through this analysis, I identify five distinct varieties among these cases of recurrence: delayed continuation, elite insurrection, opposition factionalization, government resumption, and external instigation. I detail both substantive and causal features of these varieties, showing how these recurrences are (or are not) different from one another and from other forms of armed conflict onset.

This research advances scholarship on armed conflict recurrence in several ways. The analysis identifies the important scope conditions for existing theories of recurrence. Every case analyzed in this article was included in a prior quantitative analysis of armed conflict recurrence, and thus was treated as empirically and theoretically equivalent to the other cases. Yet, there is wide variation in the extent to which these cases align with established theories of recurrence. Some cases conform to existing theories, and the conditions hypothesized as influencing recurrence are causally relevant. Other cases exhibit conditions that are theorized as influencing the probability of recurrence, but the historical analysis shows that those conditions were causally irrelevant. Still other cases fall entirely outside the scope of existing theories. By mapping these differences, this research complements and extends existing scholarship on armed conflict recurrence and provides a foundation for further theoretical development and empirical research. More broadly, this research advances a configurational approach to the study of armed conflict recurrence. A growing collection of scholarship shows how distinct configurations of political power produce different types of conflict with different causes (Buhaug, 2006; Cederman et al., 2010; Wimmer, 2013; Wimmer et al., 2009). Research on armed conflict recurrence typically examines how broad categories of motivation, opportunity, and constraint influence the likelihood that non-state actors will renew violence (e.g. Call, 2012; Loyle and Appel, 2017; Quinn et al., 2007; Walter, 2004). However, consistent with configurational approaches to explaining armed conflict, my research shows that the effects and relevance of individual conditions (i.e. ethnic grievances, or the way the preceding conflict ended) are shaped by power inequalities, political opportunities, and the pursuit of legitimacy within state institutions. These conditions influence both state and non-state actors’ efforts to secure power through renewed conflict, resulting in different proximate causes for each variety of recurrence.

Armed conflict recurrence

Armed conflict recurrence can be defined as an episode of armed conflict that takes place in a country that previously experienced armed conflict within its borders (Call, 2012; Gates et al., 2016; Quinn et al., 2007). Thus, like all types of armed conflict, recurrences are events. Yet, recurrences are distinguished from other conflict events because they are defined in relation to (1) the location where the event occurs and (2) the history of armed conflict in that location. Consistent with research on non-recurrent armed conflicts, theories of recurrence emphasize the motivations and structural constraints that influence the probability of recurrence. However, due to the geographically and historically contingent nature of recurrence, theories of recurrence typically focus either on the relationship between an episode of recurrence and the conflict that preceded it, or on social and economic conditions within the relevant territory.

This dual focus results in competing theories regarding the causes of recurrence (e.g. Fortna, 2004; Hoddie and Hartzell, 2005; Kreutz, 2010; Roeder and Rothschild, 2005; Sambanis, 2000; Walter, 2004). Among theories emphasizing the relationship between recurrence and the preceding conflict, explanations typically focus on the motivations for the preceding conflict or the way that the preceding conflict ended (see Karlén, 2017). For example, Licklider (1995) argues that the way a conflict ends will influence the probability of another conflict subsequently beginning in the same country. He contends that peace agreements result in redistributions of power that allow segments of the political elite to retain a base of popular support, thereby facilitating conflict renewal. Conversely, he contends that conflicts ending in total victory meaningfully reduce this possibility (see also Toft, 2010). However, Licklider (1995) specifies that the stability of peace agreements depends on why the preceding conflict began, with agreements being less stable if the preceding war was an “identity” (i.e. ethnic or religious) war. Doyle and Sambanis (2000) similarly argue that efforts to establish peace are significantly less likely to succeed in the wake of an ethnic or religious conflict (see also Kreutz, 2010).

Other scholars argue that who is involved in ending the preceding conflict is important for explaining the causes of recurrence. For example, Fortna (2004) shows that the presence of peacekeepers significantly contributes to the stability of conflict resolution, reducing the probability of recurrence. Some scholars note that actors who feel that peace would undermine their power and interests often act as spoilers, using violence to undermine negotiations (see Stedman, 1997). Similarly, others argue that peace will be harder to achieve as the number of actors with independent demands increases (e.g. Cunningham, 2006; Karlén, 2017; Kreutz, 2018).

In contrast, theories emphasizing territorial conditions typically focus on how the quality of the economy and the quality of governance in a country motivate recurrence or constrain opportunities for renewed violence. For example, Walter (2004) contends that poor economic and social conditions following one armed conflict significantly increase the likelihood of another conflict because average citizens are more likely to take up arms in the face of severe personal hardship and the absence of non-violent means for bringing about change (see also Keels and Nichols, 2018). Similarly, Hegre and Nygard (2015) argue the key to preventing recurrence is establishing effective law enforcement and providing social services after a war ends (also Walter, 2015). More generally, Call (2012) attributes recurrence to governmental illegitimacy rooted in political exclusion.

Some scholars contend that both post-conflict economic conditions and the conditions of the preceding conflict jointly influence the onset of recurrence. Quinn et al. (2007) contend that post-war economic growth reduces the probability of recurrence. However, they also draw on the concept of dual sovereignty (Tilly, 1978: 191; see also multiple sovereignty, p. 194) as an explanation for war onset, arguing that certain types of conflict resolution reduce the probability of recurrence. Dual sovereignty is a condition where a political contender advances an exclusive alternate claim to power that is supported by a segment of the population. Quinn and colleagues argue that both decisive rebel victories and peace agreements are effective at eliminating dual sovereignty because it is harder for government officials and members of the military to simply blend back into society after a rebel victory, whereas peace agreements depend on the mediation of competing claims to sovereignty.

While these explanations for recurrence are supported by empirical research, analyses have produced notably divergent findings. In extreme cases, different analyses report directly contradictory results. For example, while some studies show that ethnic mobilization has a significant positive effect on the probability of recurrence (Karlén, 2017; Kreutz, 2010), others find a significant negative effect, or no effect at all (Call, 2012; Sambanis, 2000; Walter, 2004). Similarly, efforts toward total takeover (i.e. when rebels seek a complete overhaul of society rather than partial reforms) range from a significant positive effect on armed conflict recurrence (Kreutz, 2010) to a non-significant but negative association (Walter, 2004). Different analyses also support divergent conclusions regarding the effects of conflict intensity (measured in the number of battle deaths) and the duration of the initial conflict (Kreutz, 2010; Quinn et al., 2007; Walter, 2004).

Not all conditions have such inconsistent effects. Several studies find that gross domestic product (GDP) and the presence of peacekeepers have a negative effect, while higher rates of post-war infant mortality have a positive effect on the probability of recurrence (e.g. Fortna, 2004; Mukherjee, 2006; Quinn et al., 2007; Walter, 2004). However, except for the negative effect of peacekeepers on recurrence, these findings are only supported by analyses of civil wars, whereas analyses that also include armed conflicts with fewer casualties—such as coups or smaller rebellions—find no effects (Kreutz, 2010).¹ Consequently, there is little agreement regarding the causes of armed conflict recurrence (Call, 2012: 30).

Rethinking recurrence as a unitary phenomenon

Each of the studies highlighted above present independently compelling theories rooted in thoughtful empirical analyses. Yet, the fact that both the significance and the direction of various conditions change across analyses suggests that there may be meaningful substantive differences between the cases of recurrence themselves. Armed conflicts—as social events—can exhibit distinctive features across space and time, and thus erupt for diverse reasons (Kalyvas, 2007). A growing collection of research identifies distinct varieties of armed conflict, highlighting substantive differences between guerilla wars versus non-guerilla wars (Krcmaric, 2018), identity (i.e. religious or ethnic) wars versus non-identity wars (Sambanis, 2001), and different forms of rebellion and infighting (Wimmer et al., 2009). Moreover, analyses show that the causes of armed conflict differ from one variety to the next (e.g. Sambanis, 2001; Wimmer et al., 2009).

Despite the prevailing wisdom that the causes (and effects; see Krcmaric, 2018) of armed conflict vary across varieties of armed conflict, efforts to disentangle varieties of recurrence are notably absent. While some studies differentiate between recurrences that simply continue the preceding conflict versus recurrences that were initiated by new actors (e.g. Walter, 2004, 2015), comparisons between these types show few differences (see Walter, 2004). Moreover, because there is no temporal limit defining when an armed conflict should be counted as a recurrence versus being considered an entirely new conflict (i.e. conflicts may be coded as recurrences even when the preceding conflict ended more than 30 years before; see Kreutz, 2010), it is not entirely clear when or how recurrences are distinct from other forms of armed conflict onset. Thus, while there is little dispute that understanding the causes of recurrence is of great importance, efforts to develop such an understanding may be hampered by substantively meaningful heterogeneity in the dependent variable. In the sections that follow, I analyze 58 cases of armed conflict recurrence to identify which sources of heterogeneity distinguish varieties of recurrence.

Methods and data

Analytic strategy: “interviewing” history

To identify different varieties of recurrence, I use an analytic strategy characterized by what Ragin and Amoroso (2010) term retroduction. Retroduction refers to the iterative incorporation of both induction and deduction. Retroduction is widely evident in comparative research, which intentionally engages assumptions and ideas regarding pertinent factors influencing social phenomena (typically based on existing theory) in dialogue with empirically derived patterns among cases (see Mahoney and Rueschemeyer, (2003)). While retroduction is common in comparative research, different strategies for comparative analysis privilege either induction or deduction, with the division typically corresponding to the number of cases analyzed. Small-N comparative research (i.e. fewer than a “handful” of cases; see Ragin, 2014 (1987)) tends to privilege induction for the purposes generating hypotheses, pursuing conceptual innovation, and identifying homogeneous subpopulations of cases (Mahoney, 2007). Here, theory is used primarily to determine and delimit scope conditions and, relatedly, for case selection. Conversely, medium-N comparative analyses tend to privilege deduction for the purposes of theory testing (see Ragin, 2014 (1987) and use induction primarily for elaborating causal processes and resolving empirical contradictions (Ragin, 2014). There is an intuitive logic to these tendencies: generating hypotheses, pursuing conceptual innovation, and differentiating between subpopulations of cases all benefit from case-intensive knowledge (Mahoney, 2007), which becomes increasingly challenging to achieve as the number of cases increases (Ragin, 2014(1987)

The purpose of this research—to identify varieties of recurrence—aligns with strengths of small-N comparative research. Yet, in the context of research on armed conflict recurrence, the empirical and theoretical ambiguity that characterizes current knowledge on this topic provide many reasons to question the logic of research that analyzes only a limited subset of cases. Thus, accounting for as many episodes of recurrence as possible is preferable. However, the analytic techniques commonly used for medium-N comparative analysis (e.g. qualitative comparative analysis (QCA)) privilege prior theory and research—which, in the context of conflict recurrence, is problematic—by relying on established classifications and analytic categories (see Ragin, 2008: 82).

To preserve the privileged position of induction found in small-N comparative analyses (Mahoney and Rueschemeyer, (2003)) while accounting for more than a handful of cases (i.e. Ragin, 2014 (1987)). I adapt the logic of semi-structured interviewing for analyzing historical cases. Like small-N comparative research, semi-structured interviewing explicitly incorporates retroduction (see Ragin and Amoroso, 2010: 50), but privileges inductive analysis and commonly includes tens of cases (Lamont and Swidler, 2014). Data collection through semi-structured interviewing begins by establishing a framework of themes to be explored, then using open-ended questions to allow new ideas and information to be introduced and explored during an interview. While the themes explored during data collection are typically determined based on prior theory and research, analysis of the data relies on the inductive identification of themes and patterns within the data. In this research, I build on well-established efforts to leverage the similarities between interview/observational research and historical/documentary analysis (i.e. Glaser and Strauss, 1967: 161; see also Vaughan, 2004 on historical ethnography) by adapting the logic of semi-structured interviewing to compile data on historical accounts of 58 cases of armed conflict recurrence, and then inductively analyze that data.

I began my research by developing the equivalent of an interview guide that consisted of three questions: Who initiated the recurrence? Why did was the recurrence initiated? How was the recurrence related to the preceding conflict? These questions guided the data collection process. I compiled existing research by historians, area studies specialists, social scientists, research institutes, government agencies, and non-governmental organizations.² I then reviewed these sources to develop narrative histories of each case, constructing open-ended answers to the questions from my interview guide.³

After these data were collected, I used qualitative data analysis software to inductively code each of the case histories for major themes and issues pertaining to the nature of each conflict and its causes. I then compared across codes to identify initial substantive and causal patterns among the cases. Once I identified common substantive and causal features among cases, I used the qualitative data analysis software to construct codes corresponding to the variables associated with each case in previous quantitative analyses (in this process, I accounted for all variables in the statistical models published by Kreutz, 2010; Quinn et al., 2007; Walter, 2004). I then examined how the qualitatively identified patterns informed the substantive interpretation of variables typically included in quantitative analyses of recurrence, and examined whether the presence of those conditions informed my interpretation of the qualitatively identified patterns among cases.

Case selection

To identify cases of armed conflict recurrence, I used the replication data provided by Kreutz (2010) for his analysis of the causes of armed conflict recurrence to identify a baseline set of recurrences. Kreutz’s data build on the widely used UCDP-PRIO Armed Conflict Dataset (Gleditsch et al., 2002; see also Walter, 2015; Wimmer et al., 2009). The UCDP-PRIO Conflict Termination dataset defines a conflict episode as a continuous period of fighting between a state and a non-state actor where there is a minimum of 25 casualties per year. Once a conflict fails to meet these criteria, an episode is coded as having ended. Kreutz coded a recurrence for all conflicts that were followed by an armed conflict in the same territory.

I used Kruetz’s data for case selection because of its comprehensiveness. Analyses of armed conflict recurrence often focus on sub-populations of armed conflicts, limiting the conflicts included in analyses based primarily on conflict intensity (measured in battle deaths). Kreutz uses the least restrictive criteria for inclusion, thereby capturing the largest population of armed conflicts during the time period. While some scholars question the logic of including conflicts with as few as 25 battle deaths in the context of quantitative analyses due to issues of comparability between such small episodes of armed conflict and much larger civil wars, the qualitative nature of my research allows me to account for more subtle variation in conflict intensity as one dimension for comparison in the analysis of cases. Thus, using the most comprehensive data on armed conflicts to identify cases for my analysis reduced the extent to which arbitrary (see Sambanis, 2004) distinctions based on conflict intensity would influence my findings, and allowed me to follow best practices by qualitatively accounting for the importance of conflict intensity without reifying quantifiable cutoff points as theoretically (rather than logistically) important (Staniland, 2017).

Kreutz’s dataset includes five different types of conflict termination that can be followed by a recurrence: rebel victory, government victory, peace agreement, ceasefire, and other. The latter category accounts for all conflicts that never formally ended, but diminished to very low levels of activity. In his analysis of armed conflict recurrence, Kreutz (2010) analyzes recurrences of intrastate conflicts (omitting interstate conflicts) and uses a binary dependent variable where a recurrence is coded as present (1) for each episode of intrastate conflict that ended in any of those five forms of termination and was subsequently followed by an armed conflict in the same country. I used this variable to establish a baseline set of recurrences, then removed all cases that were coded as recurrences following conflict terminations coded as other. Terminations coded as other typically represent increased activity in ongoing conflicts rather than the recurrence of a conflict that formally concluded, making them substantively distinct from recurrences following the other four types of termination (see Staniland, 2017).

Using Kreutz’s data, I identified 60 episodes of recurrence that followed a rebel victory, government victory, peace agreement, or ceasefire for which there was complete data available in Kreutz’s (2010) quantitative analysis. Of these 60 episodes, two conflicts were later omitted because I was not able to locate sufficient historical data to account for the causes of recurrence. In a total, I analyzed 58 recurrences in 41 countries that occurred between 1946 and 2005.

Results

In the sections that follow, I detail each variety of recurrence. I discuss the characteristics and causes of each variety, highlight the features that distinguish each variety from the others, and provide substantive illustrations. I then discuss how my findings align with or inform existing theories, and illustrate how various conditions interact to produce different proximate causes for each variety of recurrence.

Delayed continuation

Accounting for 12 of the 58 cases analyzed, this variety consists of recurrences where a large-scale armed conflict erupted following a related, but much smaller episode of armed conflict. These recurrences effectively continued the preceding conflict after a multi-year delay. Table 1 lists each episode of recurrence, and the conflict that preceded it.

Table 1.

Cases of delayed continuation.

Country	Recurrence start year	Recurrence	Preceding conflict start year	Preceding conflict
Argentina	1973	ERP and Monteneros	1955	Military faction led by Doucet
China	1956	Tibetan military	1950	People’s Revolution
Congo/Zaire	1978	FNLC (Shaba II)	1977	FNLC (Shaba I)
Cuba	1957	26th of July Movement	1953	26th of July Movement
El Salvador	1979	ERP-FAL-FARN-FPL-PRTC	1972	Military faction led by Colonel Benjamin Mejia
Ethiopia	1974	1974 Popular Revolution/Derg Coup d’Etat	1960	Attempted coup against Emperor Haile Selassie
Indonesia	1999	Free Aceh Movement (GAM)	1990	Free Aceh Movement (GAM)
Iran	1966	Democratic Party of Iranian Kurdistan (KDPI)	1946	KDPI
Sri Lanka	1989	Janatha Vimukthi Peramuna (JVP)	1971	JVP
Sudan	1976	National Front	1971	Sudanese Communist Party
Sudan	1983	Sudan People’s Liberation Movement-Army	1963	Anyanya-Southern Sudan Liberation Movement
Uganda	1977	Uganda National Liberation Army	1971	Military forces of Idi Amin-Kikosi Maalum Militia

ERP: Peoples’ Revolutionary Army.

Consistent with the historically contingent nature of recurrence, the defining features of this variety are rooted in the relationships between the episodes of recurrence and the preceding conflicts. In each case, the same basic grievances that motivated the preceding conflict also motivated the recurrence. The groups opposing the governments in these conflicts all sought dramatic political transformations either in the form of a complete overthrow of the existing political system (9 cases) or through secession (3 cases). However, the initial attempts to secure these changes all failed. In every case, the initial conflict ended in a military victory by the government, after which there were no meaningful efforts to address the grievances that motivated the initial conflicts. Instead, each country returned to the same arrangement of political power that the preceding conflicts sought to upend, thereby setting the stage for recurrence.

The dynamics of this variety are illustrated by the conflict between the Janatha Vimukthi Peramuna (JVP) and the Sri Lankan Government. In 1971, the communist JVP initiated an armed revolt to overthrow the Sri Lankan government (Horowitz and Jayamaha, 2007). The conflict lasted for less than 3 months, and the government regained control over the area. Following the conflict, there was a brief period of political liberalism in Sri Lanka during which the JVP entered legal politics. However, the ruling United National Party (UNP) soon implemented constitutional changes that allowed for the suppression of dissenting political institutions and strengthened the executive power of the presidency (Abeyratne, 2004). In 1983, the JVP was banned from legal politics, leaving them effectively in the same position they occupied institutionally and politically prior to the 1971 insurrection. Efforts to jump-start the failing Sri Lankan economy benefited only a limited subset of society, failing to address the grievances that motivated the JVP to begin with (Abeyratne, 2004; Horowitz and Jayamaha, 2007). With the conditions driving the initial insurrection effectively reinstated, the JVP mobilized and renewed the conflict in 1987. As Moore (1993) writes, “The insurrection of 1987-1989 was better-prepared and more deeply-rooted than that of 1971; the human costs and societal consequences of its extirpation were correspondingly greater” (p. 593).

Causally, the recurrences in this variety do not meaningfully differ from the conflicts that preceded them. Consequently, the same conditions that might be used to explain armed conflict onset would likely predict these recurrences equally well. Theoretically, this is notable because the way that the preceding conflict ended (one of the few causal conditions that is unique to theories of recurrence) was consistent across all cases, but had little bearing on the resumption of violence. Multiple scholars argue that a decisive victory will reduce the likelihood of recurrence because governments consolidate power or simply eliminate the opposition altogether (Quinn et al., 2007; Toft, 2010; Walter, 2004). Within this variety, governments’ victories did not reconfigure the existing power inequalities, alter the political opportunities available to the opposition, or improve the government’s legitimacy. The fact that conflict renewal was delayed—and in several cases, initiated by new actors—indicates that a decisive victory may have delayed opposition mobilization, but did not prevent conflict renewal in the face of enduring grievances.

Importantly for theories of recurrence, the grievances that motivated the conflicts in this variety were heterogeneous. Theories emphasizing grievances as causally relevant typically contend that some types of grievances (i.e. ethnic or religious) are more intractable than others (e.g. Doyle and Sambanis, 2000; Licklider, 1995). In this variety, the conflicts were all intended to fundamentally alter the distribution of power among politically organized segments of society—for example, opposition to an imperial system (Ethiopia 1974), advancement of a communist system (e.g. JVP in Sri Lanka), ethnic separatist movements (e.g. Free Aceh Movement (GAM) in Indonesia, Democratic Party of Iranian Kurdistan (KDPI) in Iran)—but the contested political inequalities do not map neatly on to the categories of motivation identified in previous research.⁴

Elite insurrection

Accounting for nine cases, this variety is defined by recurrences that were initiated by factions of each country’s military. This single feature distinguishes these cases from other recurrences in several causally and theoretically important ways. An implicit assumption in much of the literature on armed conflict recurrence is that non-state actors initiate recurrences. This is reflected in the theories discussed above, which uniformly focus on conditions that would either constrain or motivate non-state actors and the general populace to revolt against the government. The stated motivations for the insurrections in this variety align with the motivations for many other types of armed conflicts (see, for example, Sambanis, 2001; Wimmer et al., 2009), with some insurrections (such as the Colorados revolt in Paraguay led by General Andres Rodriguez, or the coup against King Faisal II in Iraq) aimed at overthrowing repressive regimes, and others (such as the Navy Revolt of 1963 in Argentina and the 1985 rebellion in Uganda led by General Tito Okello) seeking to establish strong military rule to the explicit exclusion of key groups in society. Yet, the constraints commonly faced by non-state actors—such as limited access to resources, limited training, and overt efforts to crush rebellion—are far less relevant for members of the military, who can leverage institutionally established positions of power to organize an insurrection (Table 2).

Table 2.

Cases of elite insurrection.

Country	Recurrence start year	Recurrence	Preceding conflict start year	Preceding conflict
Argentina	1963	Navy Revolt	1955	Military faction led by Doucet
Ghana	1981	Jerry John Rawlings and military factions	1966	National Liberation Council
Ghana	1983	Attempted coup by military factions	1981	Military faction led by Jerry John Rawlings
Iraq	1963	Military forces of Abd as-Salam Arif, NCRC	1958	Free Officers Movement
Paraguay	1954	Military forces of Alfredo Stroessner	1947	Opposition coalition between Febreristas and Communists
Paraguay	1989	Colorados coup	1954	Opposition rebellion led by Alfredo Stroessner
Togo	1991	Military factions	1986	Togolese Movement for Democracy (MTD)
Uganda	1981	Military faction Led by General Tito Okello	1977	National Resistance Movement
Venezuela	1992	Military faction led by Hugo Chavez	1962	Naval faction

NCRC: National Council of the Revolutionary Command.

Poor economic conditions attributed to bad governance were consistently used to justify these insurrections. This is illustrated by the 1981 coup led by Jerry John Rawlings in Ghana. From 1966 through the end of the 1970s, the political system in Ghana was highly unstable, transitioning power between the military and democratically elected leaders. During this time, Ghana suffered from a weak economy, which was exacerbated by three severe draughts in 1971, 1977, and 1980 (Biswal, 1992). Citing the failures of the regime to stabilize the economy, Rawlings led a faction of the military that overthrew the government and delivered power to the Armed Revolutionary Forces Council (Adedeji, 2001). Except for Paraguay after the 1954 conflict and the 1992 attempted coup in Venezuela, the countries in this variety had below-average GDP per capita relative to other countries that experienced recurrence. However, in both Paraguay and Venezuela, the coups were preceded by periods of economic instability and recession (Coppedge, 1994; Sacks, 1988).

Theoretically, these cases generally align with assertions that poor economic conditions and poor governance increase the likelihood of conflict onset (e.g. Hegre and Nygard, 2015; Walter, 2004, 2015). However, rather than poor economic conditions and weak governance providing a basis for popular revolts, these conditions were instead incorporated into the justifications that groups of elites used to legitimate political agendas (i.e. Brubaker, 2004). Yet, in instances where the coups succeeded, the new governments generally failed to rectify the problems cited to justify the insurrections.

Opposition factionalization

Accounting for nine cases, this variety is defined by cases where divisions among the opposition resulted in the formation of factions that re-initiated armed conflict after a negotiated settlement was established. Much like the cases of delayed continuation, the defining features of this variety are rooted in the relationships between the episodes of recurrence and the preceding conflicts. The preceding conflicts were initiated with the general aim of altering the institutional structure of the state (either politically or territorially), and the factions that instigated the recurrences retained these aims as a motivation for conflict renewal. Yet, this variety is distinct in several important ways, each of which is rooted in the fact that they followed conflicts ending via peace agreement (Table 3).

Table 3.

Cases of opposition factionalization.

Country	Recurrence start year	Recurrence	Preceding conflict start year	Preceding conflict
Djibouti	1999	FRUD-Dini faction	1991	FRUD
India	1969	Communist Party of India-Marxist	1948	Communist Party of India
India	1992	All Tripura Tiger Force (ATTF)	1978	Tripura National Volunteers
India	1997	ATTF (faction)	1992	ATTF and National Liberation Front of Tripura
India	1964	Naga National Council (faction)	1962	Naga National Council
Iraq	1973	Kurdistan Democratic Party (KDP)	1961	KDP
Mali	1994	Islamic Arab Front of Azawad	1990	Azawad People’s Movement (MPA)
Niger	1995	Coordinated Armed Resistance (CRA)	1992	Air and Azawad Liberation Front (FLAA)
Niger	1997	Faction of the CRA	1995	CRA

In each of these cases, the configurations of political power that resulted from the peace agreements shaped both the motivations and opportunities for recurrence. Regarding motivations, while each of the conflicts that preceded these recurrences cohered around the general aim of achieving large-scale institutional change, the fact that some parties were willing to accept a peace agreement while others would not suggests that the opposition was not unified in their specific aims. Eight of these recurrences were motivated by ethnic grievances, while the other was initiated by the Communist Party of India (CPI), which was riven by ideological divisions. Consistent with the observation that revolutions often emerge through the formation of negative coalitions (Dix, 1984), ethnicity and ideology provided a basis of shared opposition. However, shared opposition is not synonymous with shared political positions (Brubaker, 2004), and divisions within the opposition coalitions undermined efforts to negotiate an end to violence (Kreutz, 2018).

Regarding the distinct constraints associated with recurrences following peace agreements, once a sufficient majority of the opposition is willing to end the conflict and agrees to terms for peace, the size of the opposition coalition willing to fight on behalf of a new faction, the new faction’s access to resources, and the faction’s legitimacy within the community they purport to represent are greatly diminished. Thus, there are more barriers to mounting a meaningful challenge against the government.

The features typical of this variety are illustrated by the conflict in Djibouti. Following its independence from France in 1977, Djibouti was ruled by a coalition government composed of the three major ethnic groups: the Gadaboursi, the Issa and the Afar (Kadamy, 1996). Within a few years, the president implemented single party rule that favored members of the Issa ethnic group, and by the 1980s the national coalition broke down. In response to these changes, in 1991 the Front for the Restoration of Unity and Democracy (FRUD), a group dominated by ethnic Afars, initiated a rebellion against the government to restore multi-party rule (Uppsala Conflict Data Program (UCDP), 2015). The conflict lasted for 3 years until the FRUD agreed to peace in 1994, laying down their weapons in exchange for the appointment of members of the Afar ethnic group to several government positions (Bollee, 2003). However, the group split, and a subset led by Ahmed Dini refused to accept the terms of the agreement (Schraeder, 1993). While their violent activities were generally minimal, the rebellion by the FRUD-Dini Faction reached the level of a recognized armed conflict in 1999 (UCDP, 2015).

Theoretically, these findings inform our understanding of how ethnic cleavages affect recurrence, and the factors influencing the stability of peace agreements. While ethnic conflicts recur on average more often than non-ethnic conflicts (Kreutz, 2010; see also Doyle and Sambanis, 2000), this analysis suggests that ethnicity is most salient as a cause of recurrence when it provides a basis for coalition formation without ideological cohesion. Moreover, these findings suggest that when (1) the opposition is a coalition rather than a single organization and (2) the coalition is based on shared social position (e.g. minority status) or broad ideological categories (e.g. the division between communism and capitalism) rather than shared agendas, the likelihood that veto-players (i.e. Cunningham, 2006) will emerge is greater.

Government resumption

Representing eight cases, this variety is defined by recurrences where the government renewed the conflict after a ceasefire was established (see Table 4). Like delayed continuation and opposition factionalization, this variety is defined in no small part via the relationships between the episodes of recurrence and the preceding conflicts. In these cases, the recurrences continue the preceding conflicts, with the same actors involved. Like cases of elite insurrection, who initiated the recurrence is both substantively and causally significant. As noted above, theories of armed conflict recurrence widely assume that recurrences will be initiated by non-state actors. This is important because accounting for what motivates recurrence is central to many of these theories. The implicit assumption that only non-state actors initiate recurrences overlooks the fact that state actors have a vested interest in maintaining control and protecting the institutional and territorial legitimacy of the state (i.e. Wimmer et al., 2009; Wimmer and Min, 2006). As a result, conflict can be more advantageous for a regime than peace, as illustrated by this variety.

Table 4.

Cases of government resumption.

Country	Recurrence start year	Recurrence	Preceding conflict start year	preceding conflict
Croatia	1995	Serbian Republic of Krajina	1993	Serbian Republic of Krajina
Myanmar	1994	Karen National Union (KNU)	1992	KNU
Papua New Guinea	1992	Bougainville Revolutionary Army (BRA)	1989	BRA
Philippines	1997	Communist Party of the Philippines (CPP)	1969	CPP
Philippines	1998	Communist Party of the Philippines (CPP)	1997	CPP
Republic of Congo	1999	Supporters of President Lissouba	1993	Ninjas-Cocoyes-Cobras
Russia	1999	Chechnya	1994	Chechnya
Senegal	1995	MFDC	1993	MFDC

MFDC: Movement of the Democratic Forces of Casamance.

In each of the eight cases, the regimes in power either controlled newly formed countries or faced threats from multiple distinct insurgencies. Consequently, not only were the governments vulnerable, but the legitimacy of the states themselves was threatened. While ceasefires sometimes signal a genuine intention to end the violence, they also provide a reprieve that allows actors to regroup. When an insurgency is more resilient than expected, the opportunity to recalibrate strategies can be essential for a government to ultimately defeat their opposition (see Paul et al., 2010). Not all governments necessarily want to continue the violence. However, in these eight cases, the rebels established dual sovereignty (Tilly, 1978), enjoying popular legitimacy and thereby threatening the coherence and stability of the polity. In each of these cases, moving from a ceasefire to a peace agreement would threaten the integrity of their territory, compromise the regime’s control over the government, or signal to other insurgents that concessions could be extracted using violence (i.e. Walter, 2004). In conjunction with the political vulnerabilities associated with the newness of the regimes or threats from multiple insurgencies, this created conditions where the governments could gain more by using the ceasefires as an opportunity to regroup and resume the conflict.

These dynamics are illustrated by the conflict between Myanmar and the Karen National Union (KNU). Since Myanmar’s independence from the British in 1948, there has been persistent conflict between the central government of Myanmar and the Karen people. In 1947, the KNU was formed, and in 1949 the affiliated Karen National Liberation Army initiated an armed conflict to secure Karen independence from Myanmar. The conflict continued until 1992 when the military declared a unilateral ceasefire. However, far from an effort toward reconciliation, “[m]ilitary analysts suggested that, since declaring a unilateral ceasefire in April 1992, Burmese soldiers were simply waiting for an opportunity to strike when the Karens were weak” (Harriden, 2002: 85).

The conflict between the government of the Philippines and the Communist Party of the Philippines (CPP) followed a similar trajectory. Beginning in 1969, the CPP initiated a protracted “Peoples War” to gain control of the government. While efforts to reach a peace agreement failed in 1995, the two sides agreed to a ceasefire in 1997. However, less than 2 years later, the ceasefire was broken. According to a 2009 Armed Conflicts Report by Project Ploughshares (2009), “After a short respite for peace talks, government forces were ordered to resume full hostilities against the communist rebels in April.”

External instigation

This variety is defined by 13 recurrences that were made possible through support to the opposition by an external state. These 13 recurrences do not cohere around common classifications from previous research and substantively vary quite widely. What these recurrences do have in common is that external state support was instrumental to the onset of recurrence (Table 5).⁵

Table 5.

Cases of external instigation.

Country	Recurrence start year	Recurrence	Preceding conflict start year	Preceding conflict
Bolivia	1967	ELN led by Che Guevera	1952	Revolutionary Nationalist Movement (MNR)
Cuba	1962	Bay of Pigs Invasion	1956	26th of July Movement
Guatemala	1954	Forces of Carlos Castilla	1949	Military factions
Iraq	1982	Islamic Supreme Council of Iraq (SCIRI)	1963	Nasserite Officers Coup
Iraq	1973	Kurdistan Democratic Party (KDP)	1961	KDP
Laos	1963	Pathet Lao	1959	Pathet Lao
Laos	1989	Pathet Lao	1973	Pathet Lao
Lebanon	1975	Lebanese Civil War	1958	Independent Nasserite Movement-Mourabitoun Militia
Liberia	1989	Liberian Civil War	1980	Military coup Led by Samuel Doe
Liberia	2000	Liberians United for Reconciliation and Democracy	1989	National Patriotic Front of Liberia/Independent National Patriotic Front of Liberia
Nicaragua	1982	Contras-FDN Conflict	1977	Sandinista Revolution
Rwanda	1996	Democratic Forces for the Liberation of Rwanda	1990	FPR
Thailand	1974	Communist Party of Thailand	1951	Naval Officers Coup

ELN: National Liberation Army; FDN: Nicaraguan Democratic Force; FPR: Rwandan Patriotic Front.

In 11 of these 13 cases, the rebels received direct material aid from external states. This is illustrated by the recurrence following the Sandinista revolution in Nicaragua that overthrew the Somoza regime in 1979. Following the conclusion of the Sandinista revolution, supporters of the ousted Somoza regime and the remnants of the military were living in exile in Honduras, Costa Rica, and the United States (Simon, 2007). These supporters of the ousted regime would form the basis of the Contra rebels who, in 1980, reinitiated the conflict and sought to overthrow the Sandinista regime. To undermine what President Ronald Reagan referred to as the Marxist Sandinistas of Nicaragua, the United States supplied them with the weapons and resources necessary for re-initiating armed conflict (see Paul et al., 2010).

In the other two cases (the conflict reinitiated by the Democratic Forces for the Liberation of Rwanda (FDLR) and the Liberian Civil War that was led by Charles Taylor), the recurrences were facilitated by states providing the rebels with safe-haven. These states may not have provided direct military aid. Yet, they played a critical role in facilitating the rebels’ mobilization after their governments decisively won the preceding conflicts.

As Karlén (2017) details, while research on recurrence accounts for the role of state intervention in maintaining peace agreements (e.g. Fortna, 2004), “biased third-party intervention” (p. 499) is quite common but largely treated as anomalous in empirical analyses (e.g. Call, 2012: ch. 6). External aid to rebels facilitates the resumption of armed conflict, and even the anticipation of aid appears to increase the likelihood of recurrence (Karlén, 2017). In all 13 cases, the historical literature indicates that involvement by external states was contingent on the states’ own interests and agendas. Cold War rivalries motivated some interventions. Others were rooted in a combination of economic and political interests, such as the US-backed coup against Guatemalan President Jacobo Arbenz, which was ostensibly motivated by Cold War rivalries but was heavily influenced by the political lobbying of the United Fruit Company (Chapman, 2007). Still other conflicts—such as the insurgency by the Liberians United for Reconciliation and Democracy (LURD) that was funded by Guinea in an effort to overthrow Charles Taylor’s regime in Liberia—were driven by regional political rivalries (Smith, 2006). Regardless of the external states’ motivations, their support allowed disaffected militants to overcome many of the most intractable barriers to pursuing armed insurrection.

Non-conforming cases

While the five varieties of recurrence outlined above account for approximately 86 percent of the conflicts analyzed here, eight cases (Table 6) were substantively inconsistent with these varieties. Multiple ethnic conflicts in Ethiopia followed the 1991 victory of the Eritrean Liberation Front/Eritrean People’s Liberation Front that established the nation of Eritrea in a pattern that is consistent with Walter’s (2004) assertion that partition will produce copy-cat movements hoping to establish their own regime. The conflict between the Khmer Rouge in Cambodia and Vietnam was initiated by the victors of the previous conflict and was motivated by fears over Vietnam’s regional ambitions. The 1952 Bolivian National Revolution was initiated by supporters of Estenssoro after he won the 1951 presidential elections but was denied his victory by the military Junta. While many characteristics of this case are consistent with the cases of elite insurrection, the rebels were not a part of the government at the time of the rebellion. The insurgency led by the Communist Party of Nepal-Maoist (CPN-M) in 1990, the Burundian Civil War, and the 2000 conflict in Guinea occurred at least 30 years after the preceding conflicts. While these conflicts technically fit the definition of recurrence, substantively they are better understood as new conflicts. The armed conflict fought by the Muslim Brotherhood beginning in 1979 is like cases of external instigation because the Muslim Brotherhood—founded in Egypt—supported this rebellion to advance their ideological goals. However, support coming from an international organization rather than an external government was unique to this case.

Table 6.

Non-conforming cases.

Country	Recurrence start year	Recurrence	Preceding conflict start year	Preceding conflict
Bolivia	1952	Bolivian National Revolution	1949	Popular Revolutionary Movement
Burundi	1993	Burundian Civil War	1965	Military faction Loyal to Gervais Nyangoma
Cambodia	1978	Vietnam War	1967	Khmer Rouge-FUNK
Ethiopia	1990	Multiple Ethnic Conflicts	1962	Eritrean Liberation Front/Eritrean People’s Liberation Front
Guinea	2000	Rally of the Democratic Forces of Guinea	1970	Operation Green Sea
Mexico	1996	People’s Revolutionary Army (EPR)	1994	Zapatistas National Liberation Army (EZLN)
Nepal	1990	Communist Party of Nepal-Maoist (CPN-M)	1961	Nepali Congress Insurgency
Syria	1979	Muslim Brotherhood	1966	Military Factions Loyal to Neruddin Atassi and Youssef Zeayan

FUNK: National United Front of Kampuchea.

Discussion

Recent scholarship on armed conflict emphasizes the importance of understanding how power is distributed in a political order for identifying causes of armed conflict (see Wimmer et al., 2009). Systematic power inequalities can motivate various forms of political mobilization and afford some (but not other) actors the opportunity to initiate an armed conflict (Buhaug, 2006; Cederman et al., 2010; Sambanis, 2001; Wimmer et al., 2009; Wimmer and Min, 2006). This collection of scholarship demonstrates how different configurations of conditions cause distinct varieties of armed conflict, and the analysis presented here suggests that this configurational logic is equally relevant for the study of recurrence.

Every episode of armed conflict analyzed here meets the definition of armed conflict recurrence. Yet, the characteristics of the episodes and their causes systematically differ in ways that both challenge and inform widely held assumptions regarding the nature and causes of recurrence. The first variety—delayed continuation—broadly reflects the prototypical recurrence, as it is discussed and conceptualized in prior research. In this variety, the initial episodes of armed conflict ended, but the issues motivating them went unresolved and the conflicts were later renewed. Consistent with prior theories that emphasize the conditions that would motivate citizens to pick up arms to resolve grievances (e.g. Walter, 2004), the post-conflict social and economic conditions were causally relevant. However, the cases in this variety also exhibit some shared characteristics—namely, the preceding conflicts all ended in a decisive government victory—which prior theories highlight as causally relevant, but ultimately had little bearing on the onset of recurrences in this variety. This mismatch between the features of the cases and their causal relevance pointedly illustrates how political context shapes the causes of recurrence. Despite all the recurrences following a government victory, the victories did not alter social and economic inequalities or reform the distribution of power in meaningful or lasting ways. As a result, the theorized mechanisms linking a decisive victory to the absence of recurrence were inoperative in these cases.

The elite insurrection variety was distinguished by the fact that the conflict episodes were coups initiated by factions of the military. These conflicts were causally unrelated to the preceding armed conflicts, and did not include members of the general population or other non-state actors. While the instigators consistently cited popular grievances and poor governance as motivations, these issues did not motivate popular revolts in the way that prior theories emphasizing the importance governance and economics to explain recurrence assume (Call, 2012; Hegre and Nygard, 2014; Walter, 2015). Because theories of recurrence focus on motivations for the general population to rebel or emphasize the importance of the preceding conflict, the defining characteristics of this variety place the cases outside the scope of existing theories of recurrence and indicate that they are better understood as independent conflict episodes. Within this context, these findings align with Wimmer et al. (2009) finding that the presence of multiple competing groups within the government will increase the likelihood of infighting. While Wimmer and colleagues focus on ethnically segmented central governments, the logic of their explanation appears to apply in this variety as well.

Episodes of opposition factionalization are broadly consistent with prior scholarship on spoilers (Stedman, 1997) and highlight the challenges associated with maintaining peace agreements (Call, 2012; Fortna, 2004). Like delayed continuation, these cases align well with how recurrences are conceptualized in prior research. Yet, in this variety, the conclusion of the preceding conflict critically shaped the causes of recurrence. This is illustrated by the way that peace agreements undermined the unifying effect of the oppositions’ motivations. In this variety, rebels were unified around shared opposition to the government, but this opposition did not provide the basis for a broader shared agenda. The peace agreements shifted the distribution of political power, undermining the basis of the broad negative coalition (e.g. Dix, 1984) that had initiated the preceding conflict. Consequently, the factions that renewed the conflicts lacked widespread support and, in almost all cases, were unable to meaningfully challenge the government. Thus, consistent with a configurational conceptualization of causality, the type of motivation was not causally relevant for the recurrences in isolation. Rather, motivations influenced the onset of recurrence because of how they contributed to the formation of opposition coalitions in conjunction with the way peace agreements affected the unity of those coalitions.

Government resumption was unique because of who initiated the recurrence. Like elite insurrection, the fact that the regimes in power initiated each renewal places this variety outside the scope of existing theories of recurrence, which treat the state as a reactive (rather than pro-active) agent. In these episodes, the states were either recently formed or their territorial and political sovereignty was widely challenged. Concessions by the regime posed a fundamental threat to the legitimacy of basic political institutions. Consequently, there was a strong motivation for the regimes to avoid concessions and pursue complete victory. In this context, the ceasefires provided a respite, allowing the regimes time to adjust their strategies and improve their chances of victory. Consistent with research showing that recurrence becomes less likely the longer that conflict is absent (Kreutz, 2010; Quinn et al., 2007), most of these ceasefires were short-lived.

The final variety—external instigation—is characterized by recurrences that were made possible by biased third-party intervention (Karlén, 2017). The importance of external support in these cases highlights a long-standing gap in the literature on recurrence, which treats intra-national armed conflicts as being driven exclusively by intra-national processes. Consistent with the tendency to treat states as reactive agents in the context of armed conflict recurrence, theories and research on recurrence have accounted for external intervention only insofar as it contributes to sustaining peace (e.g. Fortna, 2004), but not as a contributor to continued violence. Yet, consistent with Karlén’s (2017) research showing that third-party intervention matters a great deal for recurrence, I find that foreign governments have the capacity to assist rebels in overcoming many of the most intractable barriers to mounting an armed challenge against the state by providing money, weapons, intelligence, and safe-havens (see also Paul et al., 2010). In this way, external intervention provides a mechanism for circumventing constraints imposed by the various power configurations, allowing motivated actors to reinitiate the conflict

Taken together, these findings show how power inequalities, political opportunities, and the pursuit of legitimacy can variably shape the effect and relevance of individual conditions as causes of recurrence. An implicit assumption underlying much of the research on recurrence is that various characteristics of the preceding war, the territory, or the actors involved will additively affect the likelihood of recurrence, and the effect will be equivalent from one episode to another. This has resulted in the “tremendous disparity” about whether certain factors are important and about the degree to which they are causally important for recurrence that Call (2012: 30) observes. My findings show how thinking about causality in configurational terms and accounting for how political and institutional factors shape these configurations can help reconcile these disparities. Comparing how different types of motivation (i.e. ethnic, religious, and political) matter across varieties illustrates this point. For example, among cases of delayed continuation, specific classes of motivation (e.g. ethnic or religious) did not matter. Rather, the recurrences were motivated by structural conditions embedded in political institutions, resulting in grievances that endured well beyond the conclusion of the initial conflict. Similarly, in each case of government resumption, the substantive issues were ultimately inconsequential compared to the risk of losing legitimacy. In contrast, cases of opposition factionalization variety illustrates how a broad class of motivation (i.e. ethnic grievances) provided the foundation for a negative coalition (i.e. Dix, 1984) among actors with competing agendas. In these cases, the motivations mattered a great deal, but their effects were contingent on systematic exclusion, which changed with the establishment of the peace agreement.

The same configurational dynamics are reflected in the variable relevance of conflict resolution. While some types of resolution—such as peace agreements—limit support for conflict renewal, others—such as ceasefires—provide opportunities for actors to pursue their objectives, regardless of whether their objective is total victory or continuing toward peaceful resolution. In contrast, every case of elite insurrection was preceded by a decisive victory, yet the preceding conflicts were unrelated to the episodes of recurrence, and therefore imposed none of the constraints commonly associated with decisive victories (e.g. Quinn et al., 2007; Toft, 2010). Ultimately, when and how previously theorized causal conditions mattered depended on the presence or absence of other conditions.

The conceptual mapping of recurrences provided in this article stands to complement and inform statistical analyses of armed conflict recurrence. Focusing only on cases where a recurrence (broadly defined) occurred, this research does not address what distinguishes each variety of recurrence from non-recurrent conflicts, or whether the conditions identified as substantively important here can predict the probability of recurrence in other cases. Examining these possibilities would undoubtedly offer important new insights into the causes and dynamics of recurrence. By mapping the different varieties of recurrence, this research identifies new avenues for further exploration and provides a foundation for future research on recurrence to explore new mechanisms driving recurrence onset.

Footnotes

Appendix 1

Sources for the historical analysis were selected by searching the names, dates, and/or relevant actors for each conflict in GoogleScholar, then using the lists of references provided in initially identified studies to identify additional sources. I compiled sources until I was able to answer each of the three questions (detailed in the manuscript) that guided the construction of the stylized case studies. In cases where there were clear discrepancies across historical accounts, I worked to identify multiple sources then erred on the side of the majority of the evidence.

A list of sources for the historical analysis is provided below. Sources are organized by country rather than by conflict because the same sources were sometimes used for more than one episode of conflict. In addition to sources listed for each individual country, the UCDP Conflict Encyclopedia (Uppsala Conflict Data Program), Civil Wars of the World (DeRouen and Heo 2007), and Victory Has a Thousand Fathers (Paul, Clarke and Grill 2010) all provided data on multiple conflicts, and are therefore not listed with any single country. References for these sources are listed first.

Acknowledgements

I would like to thank Tim Bartley, Colin Beck, Bart Bonikowski, Ronald Breiger, Joseph Galaskiewicz, Steve Lopez, David Melamed, Susan Olzak, Charles Ragin, Sebastian Schutte, Robin Stryker, Robert VandenBerg and Michael Vuolo for their feedback and advice during the various stages of this research. Any errors in this manuscript are the sole responsibility of the author. I would also like to thank the editor and anonymous reviewers for their comments and suggestions.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

ORCID iD

Eric W Schoon

References

Abeyratne

(2004) Economic roots of political conflict: The case of Sri Lanka. The World Economy 8(1): 1295–1314.

Adedeji

(2001) The legacy of J.J. Rawlings in Ghanaian politics, 1979–2000. African Studies Quarterly 5(2): 1–27.

Biswal

(1992) Ghana, Political and Constitutional Developments. New Delhi, India: Northern Book Centre.

Bollee

(2003) Djibouti: From French outpost to US base. Review of African Political Economy 30(97): 481–484.

Brubaker

(2004) Ethnicity without Groups. Cambridge, MA: Harvard University Press.

Buhaug

(2006) Relative capability and rebel objective in civil war. Journal of Peace Research 43(6): 691–708.

Call

(2012) Why Peace Fails. Washington, DC: Georgetown University Press.

Cederman

L-E

Wimmer

Min

(2010) Why do ethnic groups rebel? New data and analysis. World Politics 62(1): 87–119.

Chapman

(2007) Bananas: How the United Fruit Company Shaped the World. Edinburgh: Canongate.

10.

Collier

Sambanis

(2002) Understanding civil war: A new agenda. Journal of Conflict Resolution 46(1): 3–12.

11.

Coppedge

(1994) Prospects for democratic governability in Venezuela. Journal of Interamerican Studies and World Affairs 36(2): 39–46.

12.

Cunningham

(2006) Veto players and civil war duration. American Journal of Political Science 50(4): 875–892.

13.

Dix

(1984) Why revolutions succeed & fail. Polity 16(3): 423–446.

14.

Doyle

Sambanis

(2000) International peacebuilding: A theoretical and quantitative analysis. American Political Science Review 94(4): 779–801.

15.

Fortna

(2004) Does peacekeeping keep peace? International intervention and the duration of peace after civil war. International Studies Quarterly 48(2): 269–292.

16.

Gates

Nygård

Trappeniers

(2016) Conflict recurrence. Conflict Trends 2: 1–4.

17.

Glaser

Strauss

(1967) The Discovery of Grounded Theory. Chicago, IL: Aldine Publishing Company.

18.

Gleditsch

Wallensteen

Eriksson

et al . (2002) Armed conflict 1946–2001: A new dataset. Journal of Peace Research 39(5): 615–637.

19.

Goodwin

(2001) No Other Way Out: States and Revolutionary Movements, 1945–1991. Cambridge: Cambridge University Press.

20.

Harriden

(2002) “Making a name for themselves”: Karen identity and the politicization of ethnicity in Burma. Journal of Burma Studies 7: 84–144.

21.

Hegre

Nygard

(2015) Governance and conflict relapse. Journal of Conflict Resolution 59(6): 984–1016.

22.

Hegre

Sambanis

(2006) Sensitivity analysis of empirical results on civil war onset. Journal of Conflict Resolution 50(4): 508–535.

23.

Hoddie

Hartzell

(2005) Signals of reconciliation: Institution-building and the resolution of civil wars. International Studies Review 7(1): 21–40.

24.

Horowitz

Jayamaha

(2007) Sri Lanka (1972-present). In: DeRouen

Heo

(eds) Civil Wars of the World: Major Conflicts since World War II. Santa Barbara, CA: ABC-CLIO, pp. 715–734.

25.

Hughes

(2009) Armed conflict, international linkages, and women’s parliamentary representation in developing nations. Social Problems 56(1): 174–204.

26.

Hughes

Tripp

(2015) Civil war and trajectories of change in women’s political representation in Africa, 1985–2010. Social Forces 93(4): 1513–1540.

27.

Kadamy

(1996) Djibouti: Between war and peace. Review of African Political Economy 23(70): 511–521.

28.

Kalyvas

(2007) Civil wars. In: Boix

Stokes

(eds) The Oxford Handbook of Comparative Politics. Oxford: Oxford University Press, pp. 416–434.

29.

Karlén

(2017) The legacy of foreign patrons: External state support and conflict recurrence. Journal of Peace Research 54(4): 499–512.

30.

Keels

Nichols

(2018) State repression and post-conflict peace failure. Conflict, Security & Development 18(1): 17–37.

31.

Krcmaric

(2018) Varieties of civil war and mass killing: Reassessing the relationship between guerrilla warfare and civilian victimization. Journal of Peace Research 55(1): 18–31.

32.

Kreutz

(2010) How and when armed conflicts end: Introducing the UCDP conflict termination dataset. Journal of Peace Research 47(2): 243–250.

33.

Kreutz

(2018) New rebels in postconflict settings: The principal–agent dilemma of peacebuilding. Peace & Change 43(2): 218–247.

34.

Lamont

Swidler

(2014) Methodological pluralism and the possibilities and limits of interviewing. Qualitative Sociology 37(2): 153–171.

35.

Licklider

(1995) The consequences of negotiated settlements in civil wars, 1945–1993. American Political Science Review 89(3): 681–690.

36.

Loyle

Appel

(2017) Conflict recurrence and postconflict justice: Addressing motivations and opportunities for sustainable peace. International Studies Quarterly 61(3): 690–703.

37.

Mahoney

(2007) Qualitative methodology and comparative politics. Comparative Political Studies 40(2): 122–144.

38.

Mahoney

Rueschemeyer

(2003) Comparative historical analysis: Achievements and agendas. In: Mahoney

ueschemeyer

(eds) Comparative Historical Analysis in the Social Sciences. Cambridge: Cambridge University Press, pp. 3–40.

39.

Millington

(1964) President Arturo Illia and the Argentine military. Journal of Inter-American Studies 6(3): 405–424.

40.

Moore

(1993) Thoroughly modern revolutionaries: The JVP in Sri Lanka. Modern Asian Studies 27(3): 593–642.

41.

Mukherjee

(2006) Why Political Power-Sharing Agreements Lead to Enduring Peaceful Resolution of Some Civil Wars, But Not Others? International Studies Quarterly 50(2): 479–504.

42.

Paul

Clarke

Grill

(2010) Victory Has a Thousand Fathers: Sources of Success in Counterinsurgency. Santa Monica, CA: RAND Corporation.

43.

Project Ploughshares (2009) Armed conflicts report—Philippines—CPP/NPA. Available at: http://www.justice.gov/eoir/vll/country/armed_conflict_report/Philippines.pdf (accessed 19 March 2015).

44.

Quinn

Mason

Gurses

(2007) Sustaining the peace: Determinants of civil war recurrence. International Interactions 33: 167–193.

45.

Ragin

(2014[1987]) The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. Oakland, CA: University of California Press.

46.

Ragin

(2014) Comment: Lucas and Szatrowski in critical perspective. Sociological Methodology 44(1): 80–94.

47.

Ragin

Amoroso

(2010) Constructing Social Research: The Unity and Diversity of Method. Newbury Park, CA: Pine Forge Press.

48.

Rock

(1977) Revolt and repression in Argentina. World Today 33(6): 215–222.

49.

Roeder

Rothchild

(2005) Sustainable Peace: Power and Democracy after Civil Wars. Ithaca, NY: Cornell University Press.

50.

Sambanis

(2000) Partition as a solution to ethnic war: An empirical critique of the theoretical literature. World Politics 52(4): 437–483.

51.

Sambanis

(2001) Do ethnic and nonethnic civil wars have the same causes? Journal of Conflict Resolution 45(3): 259–282.

52.

Sambanis

(2004) What is civil war? Conceptual and empirical complexities of an operational definition. Journal of Conflict Resolution 48(6): 814–858.

53.

Schraeder

(1993) Ethnic politics in Djibouti: From “eye of the hurricane” to “boiling cauldron.” African Affairs 92(367): 203–221.

54.

Simon

(2007) Nicaragua (1978–1979 and 1980–1989). In: DeRouen

Heo

(eds) Civil Wars of the World: Major Conflicts since World War II. Santa Barbara, CA: ABC-CLIO, pp. 551–566.

55.

Smith

(2006) US-Guinea relations during the rise and fall of Charles Taylor. Journal of Modern African Studies 44(3): 415–439.

56.

Staniland

(2017) Armed politics and the study of intrastate conflict. Journal of Peace Research 54(4): 459–476.

57.

Stedman

(1997) Spoiler problems in peace processes. International Security 22(2): 5–53.

58.

The World Bank (2011) World Development Report. Washington, DC: The World Bank.

59.

Tilly

(1978) From Mobilization to Revolution. Reading, MA: Addison Wesley.

60.

Toft

(2010) Ending civil wars: A case for rebel victory? International Security 34(4): 7–36.

61.

Vaughan

(2004) Theorizing disaster: Analogy, historical ethnography, and the Challenger accident. Ethnography 5(3): 315–347.

62.

Walter

(2004) Does conflict beget conflict? Explaining recurring civil war. Journal of Peace Research 41(3): 371–388.

63.

Walter

(2015) Why bad governance leads to repeat civil war. Journal of Conflict Resolution 59(7): 1242–1272.

64.

Wimmer

(2013) Waves of War: Nationalism, State Formation, and Ethnic Exclusion in the Modern World. New York: Cambridge University Press.

65.

Wimmer

(2014) War. Annual Review of Sociology 40: 173–179.

66.

Wimmer

Min

(2006) From empire to nation-state: Explaining wars in the modern world, 1816–2001. American Sociological Review 71(6): 867–897.

67.

Wimmer

Cederman

L-E

Min

(2009) Ethnic politics and armed conflict: A configurational analysis of a new global data set. American Sociological Review 74(2): 316–337.