Bridging Micro and Macro Approaches on Civil Wars and Political Violence

Bridging Micro and Macro Levels of Analysis in Conflict Research

The recent move in the literature on civil wars and political violence from aggregate cross-national analyses to research on micro-level dynamics has led to new significant insights into the behavioral and relational foundations of conflict and violence. It has also resulted in important new knowledge about particular conflict-affected countries and/or areas within countries. A major concern with existing micro-level studies is, quite unsurprisingly, that of their external validity. Indeed, a large number of these studies are based on a subset of countries in which either fieldwork is feasible or fine-grained data are already available to researchers. So far countries such as Colombia, India, Peru, or Indonesia have been extensively studied, ³ while others such as Chad, Congo-Brazzaville, or Central African Republic have been largely overlooked. ⁴ This may raise important issues related to the ability to generalize results and conclusions. In addition, we do not know whether the results of different cases aggregate in a consistent way. Put otherwise, one of the risks of micro-level research is leading to an accumulation of sui generis knowledge, which might not make coherent sense as a whole.

A second major drawback of the micro-level approach to conflict analysis is theoretical. We do not know the way in which the dynamics of conflict observed at the local level may explain wider political, economic, or social processes. Are local dynamics of concern only for understanding local processes or do they provide the foundations for explanation of wider phenomena? In contexts of violent conflict, the ways in which individuals and communities cope with and adapt to violence often develop independently of how the conflict progresses at the macro level (Justino 2009). Individuals and groups act in response to localized forms of control, information flows, and institutional change (Kalyvas 2006; Balcells 2010). And yet the choices and behaviors of local populations and their relations with local networks and institutions are not purely local events. They depend to a large extent on how fighting and negotiations between different factions unfold in the wider political arena. Moreover, these local dynamics of conflict also have wider consequences (Justino, Brück, and Verwimp 2013). For instance, strong organizations at the local level may attract violent actions because local leaders may resist taking sides or because they are being abusive toward civilians. ⁵ Other local organizations may instead repel violence from rebel or state factions due to the types of alliances that are formed between local and national leaders. Local social, economic, and political dynamics may have important consequences for how a war and peace negotiations progress at the national level. Notably, economic, social, and political choices made locally may have significant effects on the strength and levels of authority exercised by state and nonstate groups, the levels of support they can expect from local populations, and the abilities of different actors to operate and intervene locally (Justino, Brück, and Verwimp 2013). Establishing a theoretical link between macro-level political processes and local conflict dynamics is therefore crucial for gaining a better understanding of key outcomes such as the outbreak, duration, and eventual cessation of violent conflict. Needless to say, it is also important for the design of policies to break vicious cycles of violence, war, and underdevelopment.

Existing Literature

There are some relevant works within the literature on social movements and political violence that have dealt with the relationship between the different levels of analysis (these types of works are surprisingly more absent in the civil wars literature ⁶ ). Three books come to mind: The first one is Donatella Della Porta’s (1995) Social Movements, Political Violence, and the State: A Comparative Analysis of Italy and Germany, which is a major study of social movements that escalate from (leftist) social protest into terrorism. Della Porta explores three different levels of analysis, the interaction of which explains the emergence and dynamics of political violence: a macro level of analysis, which refers to what type of society is political violence most likely to develop in (the political opportunity structure or POS); a meso level of analysis, which refers to organizational aspects of collective mobilization and social movements (i.e., what groups are most likely to use violent repertoires); and a micro level of analysis, which refers to the motivational and ideological orientations of the actors themselves (i.e., which individuals are more likely to resort to political violence). Despite POS approaches being highly structuralist and unstrategic, as Trejo (2012) points out, Della Porta’s combination of the three levels of analyses is quite sound. Her theoretical approach, combined with fine-grained ethnographic research of terrorist movements in Germany and Italy, makes the work highly valuable and her explanation of radicalization quite convincing. ⁷

A second book is Roger Gould’s (1995) Insurgent Identities, which is an exemplary work of how major macro political changes and policy decisions taken at the level of the state and the city (i.e., Paris’ urban renovation projects) had differentiated effects on the collective action capacity of working-class groups depending on residential location and neighborhood network structures. According to Gould, collective action mobilizations emerge from the interplay of collective identities with major events, ⁸ social interactions, and organizational leadership.

A third book is Guillermo Trejo’s (2012) Popular Movements in Autocracies, which shows that the impact of land liberalization policies on the escalation of protest into rebellion in rural Mexico was mediated by the governance strategies that state-level governors adopted to implement reforms and on the village-level network structures that facilitated rebel recruitment. In Trejo’s approach, land-related grievances (due to macro-level policies) do not translate into collective action (at the individual level) in the absence of organizational resources and networks for mobilization (at the meso level). In the Mexican case, Trejo explains that these networks of mobilization were created by Catholic clerks, who promoted an organizational infrastructure once they faced serious competition by Protestant missionaries (after the 1930s).

Challenges of Multilevel Approaches to Conflict

Linking micro outcomes and macro processes in conflict analysis, like in any other areas of social science research, involves considerable theoretical and methodological challenges. Schelling (1978) famously demonstrated how individual behavior may result in unintended aggregate behavior through incremental steps. Similarly, local dynamics of conflict (e.g., individual and community-level exposure to violence, individual participation in and support for armed groups, local forms of resistance, and other aspects of everyday life in conflict areas) may not be easily connected to specific macro processes.

Another fundamental challenge in bridging different levels of analysis is the identification and isolation of the precise mechanisms, whereby macro-level processes (e.g., peace negotiations, the employment of different warring technologies, and recruitment patterns) may affect actors, institutions, and processes at the micro level. A large literature in economics has attempted to theorize linkages between macro-level dynamics and the micro level that take place, for instance, between agricultural growth and household consumption expenditure (Delgado et al. 1998), direct foreign investment and individual firm productivity (Haddad and Harrison 1993), and trade liberalization and household welfare (Winters, McCulloch, and McKay 2004). Yet, these approaches are still quite absent in the study of civil war and political violence.

At the empirical level, the explicit test of hypotheses that move across different levels is also a challenge. Micro-level theories cannot be tested with macro-level data, ⁹ macro-level theories cannot be tested with micro-level data, ¹⁰ and theories bridging the micro and macro levels cannot be tested without a combination of data at different levels. One way of addressing the complexity of micro–macro theories is to use a multi-method approach, with a combination of case studies and large-N datasets. One example of this type of combination is Christia’s (2012) study of multiparty civil wars, alliance formation, and within armed group fractionalization. This author finds that the key actors are often local elites “operating at a meso level that links the national-level cleavages with individual-level motivations” (Christia 2012, 5). She explores these micro–macro-level linkages with two case studies, Afghanistan and Bosnia, along with a large-N dataset on multiparty civil wars. Another example of a successful multi-method exercise is Kalyvas’s (2006) work on violence against civilians in civil war, which combines a fine-grained case study of the civil war in Greece with extensive secondary evidence on a large variety of additional cases.

However, using different methods is not always necessary. Another possible approach consists of using large-N datasets combining data at different levels of analysis (e.g., the individual, the locality, the state, or the country) and exploring them with multilevel econometric models that link them (some authors in this issue successfully use this strategy). Another alternative is testing the multilevel hypotheses with a (theoretically) selected number of cases and using micro-level quantitative data for each of them. ¹¹

This issue introduces several articles that make use of a variety of methodological approaches in original ways, allowing them to tackle the complexities of bridging different levels of analysis.

Structure of the Special Issue

This special issue features original research on different dimensions of political violence and civil wars, such as the determinants of communal violence; the causes and consequences of electoral violence; taxation by armed groups; the formation of wartime political orders; and civil war duration, severity, and outcomes. A first set of articles addresses the micro–macro-level link mostly from a theoretical point of view, even if they all test their hypotheses with either cross-national or subnational empirical data. These articles have the commonality of theoretically articulating the connection between different levels of analysis; the linkage is made via meso-level institutions, via specific variables, or via the analysis of the organization of warfare during conflict. ¹² A second set of articles addresses the micro–macro link from an empirical perspective, using multilevel empirical models focused on single countries and deriving implications for additional cases. These works allow us to empirically understand how patterns and dynamics of conflict at the level of the individual or the household may affect broader processes of conflict onset, resolution, and duration at the national level.

The first set of articles in this issue includes works by Ana Arjona, Laia Balcells and Stathis Kalyvas, and Theodore McLauchlin. Arjona’s “Wartime Institutions: A Research Agenda” presents wartime institutions as a crucial link between micro- and macro-level processes. Arjona explain that during wartime, there are different types of institutions that originate from the interactions between armed groups and civilians and that are highly conditional on the characteristics of prewar local institutions. The different forms of governance materialize in different “contracts” (what she calls disorder, rebelocracy, or aliocracy). These contracts are located at the meso level, and they impact a myriad of civil war aspects, including violence, displacement, recruitment, and postwar dynamics. The author presents original empirical evidence from Colombia showing that these different forms of social contracts exist in wartime areas and that they have implications on civilians’ lives and decision-making processes. ¹³ The article has a local institutionalist approach that allows for the establishment of a clear link between behavior at the micro-level (of civilians and armed organizations) and macro-level patterns. This link has strong micro foundations and allows the generation of clearly observable implications. Overall, this article is an excellent example of a theory that successfully travels across different levels of analysis. In addition, despite being based on the case of Colombia, the author uses evidence to show that her typology captures forms of wartime governance that are common beyond this particular case.

Balcells and Kalyvas’s “Does Warfare Matter? Severity, Duration, and Outcomes of Civil Wars” addresses the relationship between technologies of rebellion, which are determined by the absolute and relative capabilities of participating groups, and civil war duration, severity, and forms of termination. The article uses the different categories of warfare, which are defined by micro-level factors, to explain important macro-level outcomes. Technologies of rebellion constitute a meso-level variable allowing them to link micro-level behavior (e.g., recruitment) and macro-level outcomes (e.g., civil war duration). This approach is innovative. There is a large literature on civil war duration that has ignored the way civil wars are fought on the ground; instead, the focus has been on the nature of armed organizations and on the number of factions within these organizations (e.g., Cunningham, Gleditsch, and Salehyan 2009; Christia 2012). ¹⁴ By focusing on the technologies of rebellion, Balcells and Kalyvas provide clear explanations for patterns of duration: civil wars that last longer are those fought asymmetrically between a weak but robust insurgency and a strong state. These wars are characterized by attrition and protraction. In contrast, civil wars fought symmetrically, implying frontal clashes and greater destruction, are more likely to end sooner. Technologies of rebellion can also explain patterns of civil war severity in the battlefield, with symmetric wars being more deadly than asymmetric ones, as well as patterns of civilian victimization, with irregular wars being more deathly against civilians. Finally, they can explain why some civil wars (i.e., conventional and symmetric nonconventional ones) tend to end up in settlements, while others (i.e., irregular ones) tend to end up in incumbent victories. ¹⁵

Balcells and Kalyvas’s article ties well with Arjona’s theoretical framework. For example, Arjona (2014, 32) argues the following: “If transforming local institutions and establishing rebelocracy is essential for holding territorial control over the long run, only certain rebel groups may be able to fight long wars”. This is coherent with Balcells and Kalyvas’s finding that irregular wars are significantly longer than other types of civil wars: irregular wars are fought by robust insurgencies capable of establishing Arjona’s “rebelocracies,” which in turn allow the waging of protracted civil wars against strong states.

McLauchlin’s (2014) “Desertion, Terrain, and Rearguard Control in Civil Wars” incorporates a frequently used macro-level variable (rough terrain) in a micro-level analysis of wartime dynamics. Rough terrain constitutes the meso-level variable permitting the connection of individual behavior and general patterns in this framework. In this case, the dependent variable is not violent tactics but a generally less explored variable: desertion from armed groups. The empirical test is undertaken using individual-level data from soldiers fighting in the province of Santander in northern Spain during the 1936–1939 civil war. McLauchlin takes advantage of the availability of fine-grained data in this historical civil war; the combination of archival and secondary sources allows the construction of rich datasets. Desertion is an unexplored facet of conflict, but it is an important one. Desertion shapes civil war outcomes because an armed group cannot keep fighting if it falls apart due to desertion; moreover, desertion affects politics and society in wartime because in order to prevent desertion, an armed group will often try to tighten its control over both soldiers and civilians. McLauchlin qualifies the notion that desertion depends on micro-level characteristics of individuals, as has been assumed in the existing literature. Rather, the decision to desert has much to do with the nature of terrain at the local level, which provides greater or lesser opportunities to hide and avoid sanctioning by the well-organized forces fighting conventional civil wars. A multilevel statistical test allows the authors to establish sound empirical connections between this variable (measured at the level of the locality) and micro-level behavior. The article makes an important contribution to the literature by giving the “rough terrain” variable a different theoretical and empirical significance than the one it has been given in previous macro-level studies (e.g., Fearon and Laitin 2003). It also contributes to our understanding of control over territory in civil war. Its analysis suggests that some areas are persistently hard to control because of structural characteristics like rough terrain, not just because they are being fought over in wartime (as in Kalyvas 2006).

The second set of articles in the special issue includes works by Gupte, Justino, and Tranchant (2014); Rachel Sabates-Wheeler and Philip Verwimp (2014); and Roxana Gutierrez, Gupte, Justino and Tranchant’s (2014) “Households amid Urban Riots: The Economic Consequences of Civil Violence in India” examines the determinants of riot victimization in India. This article makes use of a multilevel framework to identify how conflict dynamics play out for different units of analysis. In addition, this article represents a classical example of how to navigate across different levels of analysis (micro and meso in this case) through the use of multilevel econometric modeling techniques. The article shows how riot victimization works through channels operating at both the micro (individual) and meso (neighborhood and district) levels. This approach allows the authors to observe how civil violence dynamics play out at the neighborhood and district levels and to understand how wider patterns of victimization are related to individual-level dynamics. This is an original departure from the literature on communal violence in India that has mostly focused on explaining the emergence of riots at the state or city levels (Varshney 2002; Wilkinson 2004). However, as shown by these authors, communal riots operate in different (and sometimes contradictory) ways at different levels of analysis. For instance, as discussed in the article, economic vulnerability increases victimization through an individual-level channel. But the link between poverty and riot victimization is less clear at the aggregate level; in fact, riot victimization seems to be higher among less disenfranchised neighborhoods.

The analysis in this article is based on a unique household- and neighborhood-level survey collected by the authors in the Indian state of Maharashtra in 2010. These data were collected using clustering methods that have allowed the authors to explain how people experience riots in different ways within the same communities in Maharashtra and how those local patterns of victimization in turn map onto the persistence of civil violence in some areas. This is a unique study of low-intensity violence at different levels of analysis. Also, contrary to the well-established literature on the causes of communal violence in India, this is one of the first studies to analyze empirically the consequences of violence at the micro level and their association with dynamics of violence at higher levels of aggregation.

Sabates-Wheeler and Verwimp’s (2014) “Extortion with Protection: Understanding the Differential Effects of Rebel Taxation Systems on Civilian Welfare in Burundi” takes on a similar approach to that of Gupte, Justino, and Tranchant (2014) in trying to account for a macro pattern (i.e., government taxation capacity) using micro and meso dynamics. The article analyzes the effects of forced contributions (cash and forced labor) to rebel movements during Burundi’s civil war on household welfare levels. Forced taxation can be seen as an example of the type of meso-level institutions that originates from the interaction between armed groups and civilians, like in Arjona’s article. Interestingly, Sabates-Wheeler and Verwimp (2014) find that cash payments to rebel groups increase the economic welfare of households that pay them. The welfare increase is between 16 percent and 25 percent. Forced labor has no significant effect on household welfare. Regular cash payments to armed groups therefore appear to act as an insurance mechanism against negative outcomes for individuals with a certain economic profile. This positive welfare effect is in turn linked to the persistence of weak state capacity in conflict-affected countries, as payments to armed groups may reinforce their power relative to that of states.

Gutierrez’s (2014) “An Inquiry into the Use of Illegal Electoral Practices and Effects of Political Violence and Vote-Buying” delves into the causes and consequences of electoral violence and vote buying. The author focuses on Kenya, where communal violence was widespread after the 2007 elections. She makes use of an original panel research design that allows her to measure the causes of electoral violence at the local level, on one hand, and the impact of this violence on citizens’ preferences, on the other. In the panel, the respondents who were interviewed two weeks before the elections were reinterviewed several months after the elections—and the associated violence. This allows for a rigorous measurement of the impact of violence on postwar identities and behavior, as it permits the author to control for previolence identities and preferences using propensity score matching. ¹⁶ At the same time, the author matches individual-level survey indicators to meso-level district variables in a multivariate analysis that allows her to measure the impact of violence across individuals within areas affected by violence as well as between violent and nonviolent areas.

Regarding the cause of violence, unlike most of the literature on voting in new democracies that focuses exclusively on clientelism, this article analyzes both the use of “carrots and sticks” by incumbent parties and their competitors. The article finds that political parties use vote buying and intimidation strategically. Wherever the main political parties use intimidation tactics (targeting rival’s supporters), they also use vote buying (targeting swing voters and own voters). Consistent with previous literature (Collier and Vicente 2012), the incumbent party was more frequently reported threatening people in areas with higher electoral competition.

With regard to the consequences of violence, the author finds that, after the conflict, victims of electoral violence do not significantly reduce their support for elections or democracy, as compared to nonvictims of violence. Gutierrez also finds that victims of violence are more likely to identify themselves in ethnic terms. Also, the victims of two very different malpractices, violence and vote buying, are more likely to support the use of violence to resolve disputes. The latter implies a risk of conflict recurrence and provides support for the hypotheses that violence breeds violence (Fridja 1994) and that violence contributes social polarization (Kalyvas 2006; Balcells 2012).

Like Gupte, Justino, and Tranchant’s piece, this article focuses on low-intensity forms of violence and not on insurgency or armed conflict. In Kenya, violence had a communal character, with violence perpetrated by citizens organized in gangs. Communal violence is generally less destructive than civil war violence, but it is still extremely lethal and disruptive at the social and political levels. The results in the article suggest that the commonalities between this and other forms of political violence are striking and that we should be able to explore them through a similar theoretical lens.

Conclusion

The literature on civil wars and political violence has tended to focus on two extreme poles, the individual level (i.e., the micro level) and the societal level (i.e., the macro level), and it has generally overlooked the connections between them (Christia 2012). ¹⁷ In this special issue, we have put together a set of articles that manage to bridge the micro, meso, and macro levels of analysis in the study of civil conflict and political violence. They do it in original ways both at the theoretical and empirical level and using diverse methodologies. In addition, they all are strong empirical articles.

As Kalyvas (2012) has noted, micro-level insights can be scaled up and tested at the macro and meso levels, and in this way illuminate macro-level patterns. At the same time, insights coming from macro-level theoretical and empirical models can lead to better understandings of micro-level dynamics. The articles in this issue go in one direction (i.e., from the micro to the macro), the other (i.e., from the macro to the micro), or both. Regardless of their specific design, all the articles in the issue manage to bridge different levels of analysis and in this way provide us with important insights on processes of domestic armed conflict and political violence.

We hope that the approach in this issue will soon be more common in the realm of conflict studies. We believe that new research at the macro level should pay more attention to micro-level foundations and delve into micro-level implications and tests. At the same time, new micro-level research should look more into its wider implications. Discerning on how local-level institutions operate at the meso level or on how different forms of rebellion and warfare aggregate to different outcomes are some examples of how to establish this micro–macro connection.

Overall, this issue is a step forward toward a more comprehensive research agenda on conflict that reconciles the insights and methodologies of two approaches that have thus far been largely divorced: those that look at processes taking place at the national and cross-national levels and those that look at processes happening at the local and household levels.