Youth Cohorts and The Risk of Conflict Recurrence:
A Global Quantitative Analysis
EVA FRONEBERG Master's Thesis
Spring 2019
Department of Peace and Conflict Research, Uppsala University Supervisor: Ralph Sundberg
Word Count: 15,204
2
Abstract
While the role of youth in post-conflict settings has increasingly gained policy attention, systematic academic studies on this topic remain scarce. This research adds to existing literature by the youth bulge theory of conflict onset to a post-conflict setting. It hypothesizes that large youth bulges at the time conflict episodes end increase the feasibility for rebel groups to relaunch their insurgence and are therefore associated with a higher risk of conflict recurrence. A global quantitative analysis revealed a statistically significant positive correlation between youth bulges and conflict recurrence. Three conditional hypotheses which focus on elements of economic, political and social exclusion that are argued to interact with youth bulges to influence the risk of conflict recurrence found no empirical support. However, the combination of findings for Hypothesis 1 and non-findings for Hypotheses 2 may indicate that the underlying cause for the relationship between youth bulges and conflict recurrence may not be based on young people’s individual grievances and motives to join a rebel group. Instead, other factors may be at play which could not be tested within the scope of this study.
Despite its shortcomings, this study therefore confirms the necessity of continuing to investigate the relationship between the age distribution of a population and the risk of conflict recurrence.
Keywords: youth, youth bulge, demography, post-conflict, conflict recurrence
3
Acknowledgements
There are several people without whom I would not have been able to complete this thesis. Firstly, I would like to thank my advisor, Ralph Sundberg, for his constant guidance and support throughout this process. Thank you to my all of my incredible friends who always provided the right words of encouragement and perfect distraction whenever they were needed. A special thank you goes to Johanna – there is no way I would have survived the last two years with my sanity semi-intact without you.
Thank you also to my grandparents whose financial support allowed me the luxury of pursuing a graduate degree in the first place and to Heather for opening her home to me so I could fulfil my dream of working in Washington D.C. for a semester.
Thank you to my parents who have never been anything but ridiculously supportive and encouraging of everything I have attempted to do. And finally, to Maddie for always entertaining me, letting me complain and just being my favorite person in the world.
4
List of Figures
Figure 1: Causal Diagram for Hypothesis 1 ... 15
Figure 2: Causal Diagram for Hypothesis 2 ... 17
Figure 3: Predicted Probabilities of Conflict Recurrence ... 29
Figure 4: Predicted Probabilities for Model 2 (Hypothesis 2a) ... 32
Figure 5: Predicted Probabilities for Model 4 (Hypothesis 2a) ... 32
Figure 6: Predicted Probabilities for Model 2 (Hypothesis 2b) ... 35
Figure 7: Predicted Probabilities for Model 4 (Hypothesis 2b) ... 35
Figure 8: Predicted Probabilities for Model 4 (Hypothesis 2c) ... 37
Figure 9: Predicted Probabilities for Model 2 (Hypothesis 2c) ... 37
List of Tables
Table 1: Descriptive Statistics for Conditions ... 26Table 2: Descriptive Statistics for DV and IV by region ... 27
Table 3: Descriptive Statistics for Control Variables ... 27
Table 4: Logit regression results for Hypothesis 1 ... 28
Table 5: Descriptive Statistics for Youth Unemployment ... 31
Table 6: Logit Regression Results for Hypothesis 2a ... 31
Table 8: Descriptive Statistics for Level of Democracy ... 34
Table 7: Logit Regression Results for Hypothesis 2b ... 34
Table 9: Descriptive Statistics for Secondary Education Enrolment ... 36
Table 10: Logit Regression Results for Hypothesis 2c ... 36
5 Table of Contents
1. Introduction ... 6
2. Previous Research ... 9
2.1 Conflict Recurrence ... 9
2.2 Youth Bulges ... 12
3. Theory ... 14
3.1 Youth Bulges and Feasibility ... 14
3.2 Conditions Affecting the Hypothesized Relationship ... 16
3.2.1 Economic Exclusion ... 17
3.2.2 Political Exclusion ... 18
3.2.3 Social Exclusion ... 19
4. Research Design ... 21
4.1 Data Sources and Operationalization of Variables ... 21
4.1.1 Dependent Variable ... 21
4.1.2 Independent and Interaction Variables ... 22
4.1.3 Control Variables ... 24
4.2 Statistical Models ... 25
5. Empirics and Results ... 27
5.1 Descriptive Statistics ... 27
5.2 Hypothesis 1: Youth Bulges and Conflict Recurrence ... 28
5.3 Conditions Affecting the Relationship ... 31
5.3.1 Hypothesis 2a: Economic Exclusion ... 31
5.4 Hypothesis 2b: Political Exclusion ... 34
5.5 Hypothesis 2c: Social Exclusion ... 36
6. Discussion ... 38
6.1 Implications and Limitations ... 38
6.2 Alternative Explanations and Avenues for Further Research ... 41
7. Conclusion ... 44
References ... 46
Appendix ... 49
6
1. Introduction
A large proportion of armed conflicts today are in fact recurrences of previous wars. The notion of a so called ‘conflict trap’, in which violence is fueled by the devasting consequences of previous episodes of war, is well established within the field of peace and conflict (Hegre, Nygård, and Ræder 2017; Gates, Nygård, and Trappeniers 2016; Kreutz 2012). Estimates on the likelihood of renewed violence range from 40% within ten years of the cessation of hostilities (Collier, Hoeffler, and Söderbom 2008), to 60 % of all conflicts (Gates, Nygård, and Trappeniers 2016). Thus, a growing literature has sought to understand which factors are most likely to increase the risk of conflict recurrence. Most studies focus on features of the previous conflict (Kreutz 2012; David Mason et al.
2011; Quinn, Mason, and Gurses 2007) or the conditions of post-conflict societies as drivers of renewed violence (Walter 2004; Themnér 2013). Others, evaluate the effect of conflict resolution and peacebuilding tools by examining the effect they have on the duration of peace, most often conceptualized in quantitative studies as an absence of violence. Despite the number of studies focusing on various drivers of conflict recurrence, there is no consensus on which factors are of the greatest relevance.
At the same time, the complex role of youth as victims, perpetrators of armed conflict and as a resource for peace has gained increased policy and intelligence attention. In 2015, Resolution 2250 was adopted by the United Nations Security Council as “the first resolution fully dedicated to the important and positive role young women and men play in the maintenance and promotion of international peace and security” (UNFPA and PBSO 2018, ix). The Resolution highlights the fact that “today’s generation of youth is the largest the world has ever known, and young people often form the majority of the population of countries affected by armed conflict” (Resolution 2250 2015).
Academic circles have also increasingly sought to investigate the causes and consequences of this matter.
One prominent field of thought linking youth and the onset of organized violence is the youth bulge theory. According to this theory, the risk of armed conflict, social unrest and terrorism is heightened significantly when youth make up a disproportionately large percentage of a population (Urdal 2006; Wagschal and Metz 2017; Weber 2019). No general consensus has been reached on the strength of this proposed relationship, though differing results can primarily be explained by varying definitions and operationalizations. More recent studies have sought to add nuance to the theory by focusing on the underlying causal mechanisms or by investigating possible interaction effects. While youth bulges are at times mentioned as a possible factor of interest in studies on peace duration and
7
conflict recurrence (Urdal 2011; Schwartz 2010; Haer and Böhmelt 2015), no study to my knowledge has thus far focused explicitly on the role the age distribution of a population may have on conflict recurrence.
This study will attempt to bridge this gap by investigating whether there is a relationship between youth bulges and conflict recurrence by asking the following question: Under what conditions does a youth bulge in a post-conflict setting increase the likelihood of conflict recurrence?
In order to do so, I first investigate whether the presence of a youth bulge in and of itself has an effect on the likelihood of conflict recurrence. Based on previous research on youth bulges, I argue that a large youth cohort relative to the total adult population in a post-conflict setting will affect primarily the feasibility for rebel groups to renew their fighting. A youth bulge in the simplest terms therefore represents a large number of potential recruits for rebel groups. Following Collier, Hoeffler and Rohner’s argument that “where a rebellion is financially and militarily feasible it will occur” (2009), the presence of a youth bulge should therefore increase the likelihood of conflict recurrence. Thus, I arrive at the following hypothesis:
H1: The larger the size of a youth bulge when a conflict episode ends, the higher the risk of conflict recurrence.
I further argue that certain conditions will affect the proposed relationship between youth cohorts and conflict recurrence. Building on the argument that youth bulges affect feasibility, which in turn rests on individuals’ motives to join a rebel group, I zero in on certain elements that could further increase a young person’s willingness to (re-)engage in violence. More specifically, I argue that grievances based on economic, political and social exclusion will interact with the size of a youth cohort to further increase the risk of conflict recurrence. Again, more detail on the underlying causal mechanism can be found in Section 3. This reasoning leads me to the following hypotheses:
H2a: The larger the size of a youth bulge when a conflict episode ends, the higher the risk of conflict recurrence under the condition of high youth unemployment.
H2b: The larger the size of a youth bulge when a conflict episode ends, the higher the risk of conflict recurrence under the condition of low levels of democracy.
H2c: The larger the size of a youth bulge when a conflict episode ends, the higher the risk of conflict recurrence under conditions of low levels of education.
To test these hypotheses, I conduct a quantitative study on a global dataset merged from various sources covering conflict episodes between 1981 and 2007. The dependent variable conflict recurrence is operationalized as a binary variable indicating whether another conflict broke out between
8
the same actors within five years of the cessation of previous hostilities. To improve the robustness of my results, I run logistic regressions using three different measures of youth bulges, which are most commonly used in previous research. Due to the conditional nature of Hypotheses 2, I run logistic regressions which include multiplicative interaction terms. The results of this study suggest that there is in fact a positive statistically significant correlation between the size of a population’s youth cohort at the end of a conflict episode and the risk of conflict recurrence. However, there is no empirical support for the conditional hypotheses. This may be due to the limitations the research design for this thesis suffers from, which are described in greater detail in Section 6. However, the combination of statistically significant findings for Hypothesis 1 and non-findings for Hypotheses 2 may also indicate that the underlying cause for this relationship may in fact not be based on young people’s individual grievances and motives to join a rebel group. Instead, other factors may be at play which could not be tested within the scope of this study.
As such, this research adds to the growing body of research on both conflict recurrence and youth bulges. As systematic studies of the role of youth in post-conflict settings are lacking (Schwartz 2010), this thesis can serve as a building block for further research. A greater understanding of the conditions under which a youth bulge may fuel conflict recurrence will also be important in order for policy-makers to address possible grievances and harness the positive transformative power youth can have.
The following section will introduce relevant concepts in greater detail and outline some of the previous research on conflict recurrence, as well as on youth and armed conflict. Based on these, I will outline my proposed theoretical framework, as well as the causal mechanisms for my four hypotheses. Section 4 will describe the research design in greater detail while Section 5 presents the results of the empirical analyses. Section 6 will discuss the theoretical and policy implications of these findings, as well as limitations and avenues for further research.
9
2. Previous Research
2.1 Conflict Recurrence
Before going into greater detail on previous literature dealing with conflict recurrence, it is important to clearly define this concept. Due to the cyclical nature of violence during most armed conflicts, distinguishing between breaks in fighting and conflict recurrence is not always easy. This study follows the approach taken by the UCDP Conflict Termination dataset and Joakim Kreutz in defining both the termination of an armed conflict and its recurrence. When a clear victory or a signed agreement between two actors is present, it is easier to identity the end of a conflict. However, a majority of armed conflicts end in less clear circumstances (Kreutz 2012, 26). To solve this problem, the UCDP Conflict Termination dataset considers a conflict to have ended when “an active year is followed by a year in which there are fewer than 25 battle-related deaths(Kreutz 2010, 244). If fighting in the same country resumes, a new conflict episode is entered into the dataset. This relapse in violence is categorized into three different types by Kreutz: 1) conflict recurrence (involving previous belligerents), 2) splinter conflict (involving splinter factions made up by ex-combatants), and 3) new conflict (involving new rebel organizations (2012, 29). Thus, in order for the resumption of hostilities to be considered a conflict recurrence, which is the focus of this study, the fighting in the new conflict episode must involve the same actors as the previous conflict1.
An expanding literature has sought to understand the causes of conflict recurrence. Most studies take a similar approach to seminal studies on the causes of armed conflict by testing the explanatory power of key variables thought to increase the likelihood of an armed conflict recurring.
These aspects can roughly be divided into four strands: the causes of the original war, how the original war was fought, the termination of the original war and the post-conflict environment.
The first strand builds on the argument that some incompatibilities are harder to solve than others. Armed conflicts that are fought along ethnic lines, for example, are thought to be more likely to recur than ideology-based wars. Ethnic identities are far more rigid than ideological commitments and often become hardened during armed conflict, making coexistence with other groups more difficult if fighting ends, thus increasing the likelihood of recurrence over time (Kaufmann 1996). In
1 The terms “recurrence”, “relapse”, etc. will henceforth be used synonymously to refer to this definition of conflict recurrence. Any reference to “post-conflict peace” refers to the absence of violence in the sense that a conflict has not recurred, unless specified otherwise.
10
addition, the type of demands a rebel group makes has also been thought to affect the risk of hostilities resuming (Kreutz 2012, 26; Walter 2004).
The second strand, focusing on the way the previous conflict was fought, highlights the impact which the duration and intensity of fighting may have on the likelihood of conflict recurrence. Lengthy conflicts, for example, are argued to be less likely to recur given the fact that resources and energies on both sides are increasingly depleted. In addition, ongoing fighting sends clear signals to combatants that the cost of achieving victory if the fighting resumes would be high, thus reducing their incentives to restart hostilities (Kreutz 2012). There is less agreement on how the intensity of a previous conflict affects the risk of conflict recurrence. Some authors argue that “wars that inflict high costs on combatants and their supporters could exacerbate animosity between them and create a strong desire for retribution even after the war ends” (Walter 2004, 374). Others find an opposite effect and argue that this is due to the fact that “supplies have been exhausted, soldiers fatigued, or popular support used up” (Walter 2004, 374).
The third strand highlights the outcome or the way the previous conflict episode was terminated as an important explanatory factor for the durability of peace. A clear military victory by either side is found to have a stabilizing effect and lower the risk of conflict recurrence (Kreutz 2012;
David Mason et al. 2011). While peace agreements in general are more likely to be followed by conflict recurrence than victories, the stipulations of the agreement can also impact this risk. Some argue that if rebel grievances are addressed after hostilities have ended, there should be no reason for them to resume fighting and the peace should therefore last. Other arguments highlight the role of power- sharing agreements and partition in predicting the risk of conflict recurrence (Quinn, Mason, and Gurses 2007).
The underlying assumption for most of the above-mentioned explanations is that a rebel group will resume fighting if they have the motivation, based either on grievances or greed, to return to fighting. However, few of them explicitly take into account factors of opportunity, which build on the greed or grievance framework (Collier and Hoeffler 2004), by adding feasibility as another layer of explanation for the onset of armed conflict. Collier, Hoeffler, and Rohner test this explanation and find that there is “little evidence that motivation can account for civil war risk but […] there is evidence to support [the] feasibility hypothesis: that where rebellion is financially and militarily feasible it will occur” (2009, 1). Thus, according to this framework, it is primarily factors of opportunity which can explain the onset or in this case the recurrence of conflict.
11
The fourth strand of conflict recurrence explanations which focus on the post-conflict environment falls more closely into this category of feasibility. Some studies focus on conditions that affect the post-conflict security dilemma and commitment problems. The deployment of peacekeeping troops, for example, is often found to reduce the risk of conflict recurrence (Fortna 2004) as it raises the opportunity costs for returning to fighting. Others highlight financial and material aspects of opportunity by looking at, for example, the role natural resources could play in destabilizing a post-conflict environment (Rustad and Binningsbø 2012; Roy 2018) or the legacy of foreign patrons (Karlén 2017).
A third group of studies draws on rebel recruitment literature to highlight the importance of rebel groups having an abundant supply of potential fighters to employ in order to relaunch their campaigns. As rebels do not have standing armies to draw upon, they must rely on ordinary citizens being willing to fight for their cause. Authors like Thémner focus on the demographic which may most immediately be considered appropriate recruits – former fighters of the original war. He argues that ex-combatants are rarely triggered to become violent merely through the presence of arms, lack of opportunities or experiences of insecurity. Instead, the interaction of entrepreneurs of violence, military affinities, intermediaries and selective incentives increases the risk of recurrence (Themnér 2013). Walter (2004) argues that it is primarily the economic and political situation in the aftermath of a conflict that will influence a person’s decision to (re)join a rebel group seeking restart a conflict. She argues that without such popular support, rebel groups would be unable to relaunch their insurgence no matter their motives and reasoning, making this the primary determining factor of whether a conflict will recur.
The theoretical framework for this study builds on this approach to understanding conflict recurrence by testing whether or not demographic factors of a post-conflict population further influence the opportunity for a rebel group to relaunch their armed conflict. The question of whether demography can help to explain conflict onset has only recently regained popularity among researchers of peace and conflict (Wagschal and Metz 2017, 55). Though scholars have, for example, found some robust support for the fact that the size of a population can influence the risk of conflict onset (Wagschal and Metz 2017; Collier and Hoeffler 2004; Fearon and Laitin 2003), this approach has not been the focus of studies on conflict recurrence. One particularly prominent theory within demography and conflict onset links youth bulges – large cohorts of young people within a population – with the onset of social unrest, armed conflict and terrorism (Urdal 2011, 2006; Weber 2019; Yair
12
and Miodownik 2016; Wagschal and Metz 2017). The following section will outline previous research on this theory before highlighting how it could help further our understanding of conflict recurrence.
2.2 Youth Bulges
As mentioned, the youth bulge theory links the presence of a disproportionately large youth cohort relative to the rest of the population to an increased risk in the onset of armed conflict, social unrest, violent crime and terrorism. The question of whether age distribution of a population can increase our understanding of armed conflict has received increasing scholarly attention since roughly the 1970s (Wagschal and Metz 2017, 57). It has since been further popularized by being used to explain the rise of the Arab Spring and violence in Egypt, Syria and other Middle Eastern and North African countries (Weber 2019, 81). It is also frequently mentioned in policy and intelligence discussions on the risks for conflict onset (Hvistendahl 2011) and as a challenge for countries looking to escape the conflict trap. Academic literature, however, has only in the last couple of decades looked into this relationship more thoroughly and systematically.
Evidence for the proposed relationship has been mixed. Earlier work by Laitin and Fearon found no support for the hypothesis that the number of young people within a population can lead to conflict. Subsequent work by Urdal (2006), Fuller and Pitts (1990), Mesquida and Wiener (1999), amongst others, have found a clear and positive correlation between youth bulges and intrastate conflict, terrorism and rioting. This discrepancy in findings is caused in large part by the different definitions and operationalizations of youth bulges that authors employ. Firstly, the term youth is not easy to define in and of itself. While international legal definitions clearly make the distinction between
“child” and “adult”, there is less clarity surrounding the concept of a youth demographic situated between the two. The age within which this transitionary stage between childhood and adulthood is reached can vary both at an individual and a cultural level. Nonetheless, the most commonly applied age brackets to refer to youth are 15-242 and 18-293.
The greater disagreement between youth bulge scholars lays in precisely which segments of the population to compare to each other. Originally, authors testing this theory primarily measured the size of a youth cohort relative to the total population (Fearon and Laitin 2003; Collier and Hoeffler
2 “The United Nations defines ‘youth’ as individuals between the ages of 15 and 24, for ‘statistical purposes’ and ‘without prejudice to other definitions by Member States’ (UNGA 1981). However, diverse United Nations entities use different age definitions” (UNFPA and PBSO 2018, 9).
3 This is the age bracket United Nations Security Council Resolution 2250 refers to, while noting that variations exist at national and international levels.
13
2004). However, this approach has been criticized particularly by Henrik Urdal for not adequately capturing the “demographic bottleneck” a youth bulge creates. According to Urdal, a youth bulge leads to unrest when the labor market is no longer able to absorb the number of young people entering the labor market (2006). Therefore, a more appropriate measure of the youth bulge compares the number of youths to the total adult population. Staveteig (2005) and Wagschal and Metz (2017) go even further and apply a “relative cohort size” measure proposed by Richard Easterlin, which compares the number of young people (ages 15-24, sometimes 29) to that of older working age adults (ages 30-54). They argue that this measurement best captures the frustration and economic marginalization which causes a youth bulge to lead to violence (Staveteig 2005, 13). Another approach focuses on the risk young men in particular pose by measuring the number of male youth relative to the total population (Collier and Hoeffler 2004).
An author’s choice of measurement relies predominately on their proposed theoretical framework underlying the relationship between youth and armed conflict. These range from socio- biological explanations for the volatility of young people, and young men in particular (Mesquida and Wiener 1999; Collier and Hoeffler 2004) and a lack of civic knowledge stemming from growing up in a child saturated context (Hart et al. 2004) to feasibility and motive frameworks (Urdal 2006).
However, none of these studies apply the youth bulge theory in a post-conflict context and therefore do not take into account how the experience of war may affect the relationship between youth cohorts and violence. This is surprising, given the previously mentioned growing policy interest both in the role of youth, not just as victims and perpetrators, but as potential agents for peace, and the importance of preventing conflict recurrence. Instead, most research on the role of youth post conflict has been conducted in the form of case studies (Brett and Specht 2004; Peters, Vlassenroot, and Richards 2003), work on and with former child soldiers (Haer and Böhmelt 2015) or micro-level studies on the psycho-social impact the experience of armed conflict has on children and adolescents (Newnham et al. 2015). These studies can help elucidate micro-level needs, motivations and attitudes towards peace or renewed conflict. However, less work has been done to gain a better macro-level understanding of the impact a large youth population can have in a post-conflict setting.
In order to bridge this gap, I argue that the youth bulge theory has the potential to capture the way in which a large youth cohort may affect both the feasibility and motive structures of a post- conflict society and can thus be applied to gain a better understanding of the effects of a population’s age distribution on the likelihood of conflict recurrence. In the following section, I will elaborate on the causal mechanisms underlying this proposed relationship.
14
3. Theory
As mentioned, I posit the presence of a youth bulge as an additional mechanism within the feasibility and motives framework leading to an increased likelihood of conflict recurrence.4 The mere presence of a youth bulge is theorized to primarily affect the feasibility structure for rebel groups to renew their fighting by affecting their ability to recruit potential fighters5. To then assess the ways in which the presence of a youth bulge may affect individual level motives to rebel, I also hypothesize that there are certain conditions which will interact with the population’s age distribution to further increase the risk of conflict recurrence. The following sections will lay out each of these arguments in greater detail to present the hypotheses which will be tested in this study.
3.1 Youth Bulges and Feasibility
Some authors suggest that large youth cohorts are linked to the onset of organized violence directly through the development of a ‘generational consciousness’, “especially so out of awareness of belonging to a generation of extraordinary size and strength, enabling them to act collectively” (quoted in Urdal 2011, 118). However, Urdal highlights the need for identity groups to form in order to overcome the collective action problem, and identity groups along purely generational lines are rare (Ibid). Thus, the relationship between youth bulges and conflict recurrence necessitates a further causal mechanism connecting the two. I argue that understanding how youth bulges affect the feasibility structure for a rebel group to (re)launch an insurgence can help bridge this missing link.
In their seminal study “Beyond Greed and Grievance: Feasibility and Civil War”, Collier, Hoeffler and Rohner find general support for their hypothesis that “factors that are important for the financial and military feasibility of rebellion but are unimportant for motivation decisively increase the risk of civil war” (Collier, Hoeffler, and Rohner 2009, 3). I apply this notion to understanding the risk of conflict recurrence by following Walter’s argument that in order for a rebel group to relaunch their
4 It is important to note that I am not claiming age distribution or in fact any demographic factors are the only necessary factor to explain conflict recurrence. As Walter explains in her work on the role of individual incentives for enlistment,
“for civil war to occur, intergroup antagonism and grievances must exist, leaders must emerge to coordinate and manage recruitment, resources and supplies must be available to support the movement over time” (Walter 2004, 375). I argue that since the age distribution of a population has thus far not been linked to conflict recurrence it could be one important explanatory factor in understanding how insurgencies are relaunched.
5 Factors of feasibility, motive, greed and grievance are closely interlinked, but are broadly conceptualized as follows in this study: Factors of feasibility capturing the supply of rebel recruits, for example, are interwoven with micro-level motivations for joining a rebel group. An individuals’ motives can then be based on greed and/or grievance, which can be affected by structural conditions. A rebel group’s macro-level motivations (based on greed and/or grievance) may also play a role in the likelihood of conflict recurrence. However, they are not the focus this study following Collier, Hoeffler and Rohner (2009, 3) and Weinstein (2005) who argue that motivations may primarily be determined by the feasibility of civil war.
15
insurgence they must be able to recruit sufficient troops to do so. This study therefore does not focus on the motives behind a rebel group’s decision to change the status quo and restart their fighting, but on their ability to do so. I argue that the size of a country’s youth cohort immediately following a conflict episode will impact this feasibility and thus have an effect on the likelihood that this conflict will recur.
It is often highlighted that most fighters fall into the youth age category (Weber 2019, 80;
Mesquida and Wiener 1999). These authors therefore point the finger at young people themselves, using socio-biological and psychological explanations to highlight their inherent ‘restlessness’ and
‘danger’ of youth. Thus, an argument could be made that a large number of young people within a population represents a large number of people who are more predisposed towards violence, therefore driving down recruitment costs for a rebel organization and making conflict onset or, in this case, recurrence more likely.
While Latin and Fearon (2003) and Collier and Hoeffler (2004) do not find support for the inclusion of a ‘youth bulge’ measure, Urdal, as mentioned, argues that the absolute number of young people compared to the size of the total population does not effectively capture the reasons a youth bulge may lead to unrest. According to him, it is not only the number of young people but the cohort’s size relative to other generations that causes issues. “Not only do youth bulges provide an unusually high supply of individuals with low opportunity costs, but an individual belonging to a relatively large youth cohort generally has a lower opportunity cost relative to a young person born into a smaller cohort” (Urdal 2011, 120). Thus, the feasibility structure is affected not only by sheer numbers, but by the socio-economic conditions a youth bulge can create. The following sections and hypothesis seek to examine this aspect further. Nonetheless, the above-mentioned arguments lead me to expect that the size of a youth bulge in a post-conflict setting on its own will already impact the feasibility for rebels to (re)recruit enough to affect the risk of conflict recurrence. This is visualized in the following causal diagram:
Figure 1: Causal Diagram for Hypothesis 1
Therefore, I will test the following hypothesis:
Hypothesis 1: The larger the size of a youth bulge when a conflict episode ends, the higher the risk of conflict recurrence.
Youth Bulge in a post-
conflict setting Increased feasibility for
rebels to (re)recruit Conflict Recurrence
16
3.2 Conditions Affecting the Hypothesized Relationship
The age distribution of a population is an inherently static condition as it normally takes decades for demographic changes to take an effect (Urdal 2006) . Thus, the size of a country’s youth cohort is likely to interact with other elements. In addition, not all countries with a youthful population descend into violence and social unrest. In fact, “in well-functioning societies, a youthful population can add to the vigor and productivity of a society” (quoted in Wagschal and Metz 2017, 64). Urdal refers to this notion of large youth cohorts as a blessing as a demographic dividend (2011, 118).
Therefore, it may not be the case that youth bulges are linked directly to the risk of conflict recurrence.
Instead, it may be that the presence of a large youth cohort must be accompanied by other conditions to actually create or increase the likelihood of relapsed violence.
Wagschal and Metz argue that “the young constitute a subset of the population, so ills and/or factors of opportunity that might induce an interest in rebellion among youthful members of a society are expected to be of prominent importance” (2017, 61). While the previous section focused on the feasibility for rebels to relaunch their insurgency, this part of my research thus focuses on a set of conditions which may increase a young person’s willingness to (re)join a rebel group. I follow Walter’s reasoning that “the fact that personal motives for joining may be connected to a previous experience with fighting does not reduce the importance of current incentives to enlist” (2004, 375). As such, I focus on immediate or structural incentives rather than war exposure or psychological factors like potential desires for revenge to explain motives to enlist6.
While there are several such structural conditions which may affect a young person’s choice to (re)join a rebel group, this study focuses elements of exclusion that create grievances based on relative deprivation and a blocked transition to adulthood7, thus reducing opportunity costs for joining a rebel group. The very notion of exclusion implies that someone is not given access to an activity or service which others are receiving. Particularly in a post-conflict setting that is almost always defined by the destruction caused during an armed conflict, the exclusion of youth from economic, political
6 The role these factors may play as alternative explanations are briefly explored in Section 6. The choice not to include them in this study is also based on findings such as those by Newnham et al. which indicate that on a micro-level
“children and adolescents affected by war are not reacting solely to traumatic exposure; rather, it appears that their engagement with economic opportunities and healthy interpersonal relationships in the aftermath of war are vital to their psychological recovery” (2015, 120). Therefore, a focus on ‘daily stressors’ rather than war exposure is warranted.
7 The notion of a blocked transition to adulthood is often termed ‘waithood’ in literature on youth in post-conflict and is explored in greater detail by authors like Sommers (2012) and McEvoy-Levy (2014).
17
and social processes8 is therefore likely to effectively block their transition to adulthood (Hilker and Fraser 2009, 4; McEvoy-Levy 2014) and thus increase their feelings of being deprived of opportunities for a better future. According to Gurr (1970), this gap between expectations and reality will increase frustrations and thus increase a person’s willingness to (re-)engage in violence. This grievance perspective is thus a relevant overarching theoretical framework for the following three hypotheses which focus on elements of exclusion that are argued to affect young people in particular and therefore their willingness to join a rebel group. The underlying causal mechanism for all three versions of Hypothesis 2 are visualized in Figure 2 below.
The next sections will outline the specific elements of exclusion included in this study and the causal mechanisms underlying these hypotheses.
3.2.1 Economic Exclusion
“The overall performance of a society is an important factor determining the income forgone by joining a rebel movement, and thus the opportunity for rebellion” (Urdal 2011, 212). Therefore, the economic grievances, which a disproportionately large youth cohort cause, are often the focus of youth bulge theoretical frameworks. Urdal quotes several studies which find that “large youth cohorts are associated with a significant increase in youth unemployment rates” (Urdal 2006, 810). Youth bulges create an ‘institutional bottleneck’, in which the size of a youth cohort is much larger than that of the preceding generation leading to a situation in which the labor market cannot cope. The role of youth unemployment is also often cited when explaining the social unrest during the Arab Spring, for example (Hvistendahl 2011). The resulting pool of unemployed youth is then likely to become frustrated with their situation, which lowers their opportunity cost for joining a rebel group further.
8 While exclusion can occur along other dimensions, such as cultural or religious lines, the focus of this study is on the economic, social and political sphere. These are the elements most often cited when discussing youth exclusion, as highlighted by the Independent Progress Study on Youth, Peace and Security, for example: “Evidence strongly suggests that a
‘cocktail’ of economic, social and political factors […] underpins the sense of justice experienced by youth” (UNFPA and PBSO 2018, 30).
Youth Bulge in a Post-Conflict
Setting
Economic, political or social
exclusion of youth
Increased motives + low
opportunity costs to join a
rebel group
Increased feasibility for rebel groups
Conflict Recurrence
Figure 2: Causal Diagram for Hypothesis 2
18
Therefore, the economic problems a youth bulge can create has, for example, been argued to
“potentially provide fertile ground for recruitment to terrorist organizations” (quoted in Urdal 2011, 121).
In a post-conflict setting, there is the added element of many young people potentially already having been members of an armed group. The prospect of rejoining civilian life is likely to be particularly unappealing when there are few economic prospects for the future. Thus, youth unemployment may also act as a mechanism leading former combatants to rejoin a rebel group. Given all of these aspects, I argue that high levels of youth unemployment will interact with the size of a youth bulge to decrease opportunity costs and thus increase a young person’s willingness to (re)join a rebel group. As a result, the likelihood of conflict recurrence will be high. Therefore, I test the following hypothesis:
Hypothesis 2a: The larger the size of a youth bulge when a conflict episode ends, the higher the risk of conflict recurrence under the condition of high youth unemployment.
3.2.2 Political Exclusion
Political exclusion is often at the forefront of discussions on the role of youth and conflict9. Young people often feel voiceless or disempowered when they mistrust their governments and feel excluded from political processes. The resulting feelings of frustration may themselves cause enough grievances to destabilize peace. However, this condition also embodies a lack of non-violent alternatives to create change within a system. This is the approach Walter takes, for example, when listing political exclusion as one of her core explanatory factors to explain the risk of conflict recurrence. She argues that individuals are likely to (re)enlist in a rebel group when they are “in a situation of individual hardship or severe dissatisfaction with one’s current situation” and/or there is
“an absence of any nonviolent means for change” (Walter 2004). When this is the case, violence is considered the only alternative to address all grievances, thus lowering opportunity costs for joining a rebel group. Wagschal and Metz therefore “expect that the less democratic a regime is, the more motives for rebellion exist [and] the less autocratic a regime becomes, the more feasible a rebellion
9 The Independent Progress Study on Youth, Peace and Security, for example, lists “meaningful political inclusion [as a] core issue at the heart of the YPS agenda” (UNFPA and PBSO 2018, 63).
19
becomes, as the possibility for violence increases” (2017, 69). Following this logic, I test the following hypothesis:
Hypothesis 2b: The larger the size of a youth bulge when a conflict episode ends, the higher the risk of conflict recurrence under the condition of low levels of democracy.
3.2.3 Social Exclusion
While the term “social exclusion” is often used in policy discourse, there is no clear consensus on how to define it. It is most often used to explain a condition in which a certain group of people is denied access to basic services and subsequently are denied opportunities for a better future. Social exclusion is thus a complex concept, involving several elements including, for example, access to health care, public amenities, social citizenship and means of communication (Peace 2001, 23).
Measuring and evaluating all of these is unfortunately outside the scope of this study.
Instead, I focus on the role of access to education in affecting a young person’s willingness to join a rebel group. The reasoning behind this choice is twofold. Firstly, access to schooling is a basic service that uniquely impacts young people, as the group directly benefitting it, and is considered a social process which enhances opportunities for an improved future. According to the Independent Progress Study on Youth, Peace and Security, “education featured universally as a core peace and security concern for young people across all consultations undertaken for the study – whether as an object of grievance and frustration, or as the embodiment of young people’s aspirations and hopes…” (UNFPA and PBSO 2018, 85). As such, a lack of education captures the particular type of grievances which are argued to arise from youth exclusion. In addition, “investment in basic education is seen as an important success factor behind the realization of the demographic dividend in East Asian countries such as South Korea and Taiwan” (Barakat and Urdal 2009, 2). Thus, access to schooling is a particularly relevant aspect of studies linking the size of a youth cohort to the risk of organized violence.
Previous research has not come to a conclusion on the specific role education plays in the onset of organized violence. Much like poverty and low alternative income opportunities, lack of schooling is often cited as a reason for joining a rebel group (Brett and Specht 2004; Barakat and Urdal 2009; Østby, Urdal, and Dupuy 2018; Urdal 2006). Educated young people are assumed to have better income-opportunities, which decreases their opportunity costs for joining a rebellion. Barakat and Urdal (2009) find support for their hypothesis that “youth bulges increase conflict risk in societies
20
where male secondary education is low”. On the other hand, high levels of education raise expectations, which often cannot be met by an unreceptive labor market and thus may in fact increase the risk of social unrest or organized violence.
For the purposes of this study, I focus on the role of education on its own in affecting the relationship between youth bulges and conflict recurrence, without initially taking into account economic prospects. Therefore, I test the following hypothesis:
Hypothesis 2c: The larger the size of a youth bulge when a conflict episode ends, the higher the risk for conflict recurrence under conditions of low levels of education.
21
4. Research Design
To empirically investigate my hypotheses, I conduct a quantitative study on a global dataset which covers conflict episodes between 1981 and 2006. The unit of analysis for this study is conflict episodes, rather than countries. Since the aim of this study is to analyze post-conflict conditions and the risk of conflict recurrence, countries which have never experienced an armed conflict within the time period covered are not included in this analysis. Thus, a comparison is made between episodes of conflicts that recurred and those that did not. The dataset also does not include every year within the chosen time-period for each country listed. Instead, measurements are included either for the year a conflict episode ended (t) or for the first five years hostilities were terminated (t+5). This ensures temporal order is controlled for as measurements of the independent and interaction variables are captured before a potential relapse occurred.
Given the fact that my theoretical framework operates primarily at the macro-level, a global quantitative approach is most appropriate to investigate the proposed relationships. In addition, available data for the phenomena I am trying to capture (namely population age distribution, unemployment, levels of democracy and education data) are generally measured at a country-level thus making a global comparison fitting.
As with any method, there are of course shortcomings to this approach. Large-N studies are, for example, unable to establish causality between the variables of interest. The implications of this and other limitations are discussed in greater detail in Section 6. Nonetheless, the statistical models within this study allow me to effectively test whether there is any empirical support for my hypotheses.
4.1 Data Sources and Operationalization of Variables
To statistically test my hypotheses, I merge data from different sources into a single dataset.
The period of analysis is determined by the availability of conflict recurrence data and is thus set to 1981-2007 following Steinert, Steinert, and Carey (2018) whose replication dataset forms the foundation of this study. More information on how the authors coded variables relevant to this study, as well as on those that have been added to the dataset, can be found below.
4.1.1 Dependent Variable
My dependent variable, conflict recurrence is taken from a replication dataset for the article
“Spoilers of peace: Pro-government militias as risk factors for conflict recurrence” by Steinert, Steinert
22
and Carey (2018). It is based on the UCDP/PRIO Armed Conflict and Conflict Termination datasets, which defines armed conflict “as any contested compatibility that concerns territory and/or government where the use of armed force between two parties leads to at least 25 battle-deaths” (Ibid.
2018, 7). The authors define conflict recurrence “as a transition from a year with fewer than 25 battle- deaths to a year with more than 25 battle-deaths” and code it as a binary variable “indicating whether a new conflict between the same conflict parties occurs within the first five years after the conclusion of the preceding conflict” (Ibid. 2018, 7). It includes only those conflicts for which information is available for the entire five-year period after a conflict that took place between 1981 and 2007 has ended. It thus excludes those which are still ongoing. This reduces the initial 228 observations included in the dataset to 196. While episodes for which there is no information on whether or not a conflict occurred are still part of the final dataset, they excluded in the regression models10. Table 2 in Section 3 shows that there are 14 observations missing, which means the maximum number of observations for any of my models is 182. Steinert et al.’s focus on the first five years after a conflict episode has ended is an effort to increase the likelihood that subsequent conflict episodes include the same actors as the previous ones, thus ensuring it is in fact a conflict recurrence as defined in the previous research section of this study.
4.1.2 Independent and Interaction Variables
The independent variable for H1, youth bulge, was calculated using UN data on population size by age groups from World Population Prospects (2017). It follows Urdal (2006) in measuring the ratio of youth (age 15-24) to the adult population (15 and older). He argues that not including the under-15 population prevents the youth bulge measure from becoming deflated by implicitly taking into account continued fertility rates (Urdal 2011, 125). To verify the robustness of my findings, I also test three alternative measures of youth cohorts described above. The first follows Laitin and Fearon and measures the ratio of youth (age 15-24) relative to the total population. The second measures the ratio of young men (age 15-24) relative to the total adult population (15 years and older). Finally, I include a measure for the relative cohort which captures the ratio of youth (ages 15-24) compared to the working age population (ages 30-54) following Easterlin and Staveteig (2005).
10 This missingness is caused either by a lack of information on the post-conflict context of the episode, which could, for example, be due to the fact that the conflict episode ended too close to when this dataset was constructed.
23
Hypotheses 2a, b, and c each have youth bulge as their independent variable, but also include interaction variables. Economic, political and social exclusion form the overarching conceptual framework for these conditional hypotheses. However, they only test the specific elements of these types of exclusion which come closest to capturing the kind of grievances argued to affect the risk of conflict recurrence.
To understand the effect of economic exclusion, the first conditional hypothesis includes youth unemployment. This variable is measured using data from the International Labor Organization’s ILOSTAT database, which measures the “share of the labor force that is without work but available for and seeking employment” (ILO 2019) in the year a conflict episode ends.
Unfortunately, this data is only available starting in the year 1991, which creates 50 missing values in the merged dataset reducing the number of observations for Hypothesis 2a to 129. A lack of alternate measures which could come as close to capturing the grievances caused by economic exclusion and the descriptive statistics for this variable in Table 5, Section 5, show why this is nonetheless an appropriate measure to use.
The next interaction variable, level of democracy, focuses on grievances caused by political exclusion. When levels of democracy are low, there is a lack of non-violent means to address grievances, thus capturing the concept of political exclusion. It is measured using the Polity IV score for each country in the year a conflict episode ends. This score captures [the] regime authority spectrum on a 21-point scale ranging from -10 (hereditary monarchy) to +10 (consolidated democracy) and is included in Steinert et al.’s dataset (2018).
Finally, secondary education enrolment is used to measure one theoretically relevant element of social exclusion. Barakat and Urdal argue that “secondary education is found to provide the most suitable discriminator in assessing the role of education […]. The transition to secondary- level qualification is a threshold for participation in the modern economic sector and is such likely to mark a significant rise in opportunity costs for participation in violent conflict” (2009, 12). Thus, I have included UNESCO data measuring a country’s ‘Gross Secondary Education Enrollment Ratio’
in the year a conflict episode ended obtained from the World Bank database (2019). Though 50 observations are missing for this variable it is the most appropriate and complete measurement available to me and is thus included. The limitations of this are discussed further below.
24 4.1.3 Control Variables
Drawing on previous research on conflict recurrence, conflict onset and youth bulges, I control for six possible confounders. While there is a whole set of conditions which could have an effect on the dependent variable, so the risk of conflict recurrence, not all of them influence the independent variable. Thus, this study only controls for the variables described below, which affect both the dependent and independent variables or vital elements of the causal mechanism.
Logged GDP/ capita is included in the models as a proxy for the level of development in a country, which has been found to have a significant relationship with conflict onset (Collier, Hoeffler, and Rohner 2009; Collier and Hoeffler 2004; Hegre and Sambanis 2006), as well as conflict recurrence (Walter 2004). A country’s level of development is also closely related to the size of its youth population. Extreme poverty and famine, for example, can affect fertility and infant mortality levels, thus altering the age distribution of a population (Nordås and Davenport 2013, 923). Therefore, I control for the possible confounding effects of a country’s level of development using the GDP/
capita measure included in Steinert et al.’s dataset.
The logged total population of a country has also been linked to the risk of conflict onset and is frequently included as a control variable in conflict recurrence literature (Steinert, Steinert, and Carey 2018). As the total population in the year a conflict episode ended captures another demographic factor, which could affect the feasibility based on sheer numbers argument, it is included in this study as a possible confounder. The data was obtained from UN Population Prospects and measured for the country and the year in which a conflict episode ended. Figures for the non-logged version of this variable are presented in thousands.
As mentioned in the previous research section, conflicts that were fought along ethnic lines are argued to be more likely to recur than non-ethnic conflicts (Walter 2004; Steinert, Steinert, and Carey 2018). In addition, Yair and Miodownik (2016) have also found that ethnicity plays a role in whether youth bulges lead to conflict onset or not. According to their findings, a “youth bulge affects only mobilization patterns leading to non-ethnic war” (Yair and Miodownik 2016) and thus do not affect the risk of ethnic conflict. Therefore, I control for the possible confounding effect of ethnicity by including the degree of ethnic fractionalization at the end of the original conflict, which is included in the Steinert et al. dataset and is measured with the Ethnic Fractionalization Index from the Ethnic Power Relations Dataset (Steinert, Steinert, and Carey 2018, 8).
Two controls are included which take into account the way the previous conflict episode was fought and could have an impact on the age distribution of a population. The age distribution of a
25
population is likely to be affected as young people die in battle and economic and social infrastructures are destroyed affecting fertility rates as well as the overall health. Therefore, intensity is operationalized using a binary variable to indicate whether the conflict episode fulfils the UCDP Conflict Dataset’s definition of civil war (at least 1000 battle-related deaths in one year). Intensity will have the greatest confounding effect when losses are so tremendous that they affect both the age distribution of a population and the cost-benefit structure for people to join a rebel group. Thus, it makes theoretical sense to include a civil war measure rather than the number of battle-related deaths during a conflict episode, which are highly skewed. The duration of the conflict episode is argued to have similar potential impacts on the demographic characteristics of a population, as well as on cost- benefit structures. It is calculated in years using the start and end date of a conflict episode as captured by the Steinert et al. dataset which itself is based on the UCDP Conflict Termination dataset.
The involvement of third-party actors has been proposed to alter the cost-benefit structure for civilians or ex-combatants to (re)join rebel groups (Schwartz 2010, 13). In particular, the deployment of peacekeeping troops is regularly found to significantly reduce the likelihood of conflict recurrence (Doyle and Sambanis 2000; David Mason et al. 2011; Fortna 2004). As the presence of peacekeeping troops alters the same feasibility structure as a youth bulge would, a binary variable indicating whether or not they were deployed within five years after the end of the conflict episode is included as a control (peacekeeping within t+5).
4.2 Statistical Models
As my dependent variable is binary, I will be running binomial logistic regressions.
For Hypothesis 1 this model will include my primary youth bulge measure, as well as all six control variables. As mentioned, I will also run the same regression using three alternative measures of my first independent variable to verify the robustness of my findings.
26 Due to the conditional nature of Hypotheses 2a, b, and c, I run logistic regressions which include multiplicative interaction terms.
Given the small size of my dataset, I have categorized the independent variables in an effort to increase their explanatory power. To improve the robustness of my findings, I include both a binary (high, low) and a three-scale version (high, medium, low) of the variables. The dummy variables are split along their mean.
Each step in the categorized versions of the variable
represent a third of the difference between their minimum and maximum values.
Youth unemployment (binary) is therefore operationalized as 0 when unemployment is below 16.4% and 1 when unemployment is higher than that. The youth unemployment (category) variable is 1 when youth unemployment is lower than 20%, 2 when it is between 20 and 39.5% and 3 when it is higher than 39.5%. Level of democracy (binary) takes on the value 0 for all scores below 0 and 1 for positive scores. Level of democracy (category) distinguishes between autocracies (-10 to 06), which are coded as 1, anocracies (-5 to 5), which are coded as 2, and democracies (6 to 10), which are coded as 3. Secondary education enrolment (binary) takes on the value 0 when the rate falls below 44.6% and 1 when it is above that rate. Finally, secondary education enrolment (category) is coded as 1 for all values below 38%, 2 for all values between 38% and 71% and 3 for all values above 71%. While regressions using the continuous versions of the variable are not included in the main analysis, regression tables and plots of these versions can be found in Appendix II for reference.
The following section will present the results of all of my models.
Table 1: Descriptive Statistics for Conditions
27
5. Empirics and Results
5.1 Descriptive Statistics
Table 2 shows descriptive statistics for both the dependent variable, conflict recurrence, and independent variable, youth bulge, grouped by World Bank region. While there are 196 conflict episodes between 1981 and 2006 included in the dataset, information on whether a conflict recurred within the first five years after the episode ended is only available for 182. The missingness for conflict recurrence is spread relatively evenly across regions, thus limiting the risk that the excluded observations would have changed the global applicability of results significantly. Of the 182 conflict episodes, 59 or approximately 32.4% relapsed within five
years. While a significant proportion of conflict episodes took place in Sub-Saharan Africa, South Asia is the only region in which the median for conflict recurrence is 1. Latin America and the Caribbean saw no conflict recurrences. Though there thus appear to be some regional variations in conflict recurrence, the same cannot be said for the youth bulge variable. There is no region for which the youth bulge standard deviations around the mean do not overlap with the global standard deviations around the mean. Table 3 presents descriptive statistics for the six confounding variables included in all of the models.
Table 2: Descriptive Statistics for DV and IV by region
Table 3: Descriptive Statistics for Control Variables
28
5.2 Hypothesis 1: Youth Bulges and Conflict Recurrence In order to test Hypothesis 1, I ran logistic
regressions where the dependent variable is a binary indicator of whether or not conflict recurred. Results are presented in Table 4. All four models include the same set of control variables described in the Research Design section. The independent variable for Model 1, youth bulge, is the primary measurement of a youth cohort chosen for this study. An initial look at the distribution of youth bulge means across whether or not a conflict recurred, as well as a bivariate regression using only the independent and dependent variable11, do not indicate support for Hypothesis 1. Once the control variables are added to the model, however, the youth bulge variable reaches statistical significance at the 95% confidence level. This suggests that there is empirical support for Hypothesis 1, meaning an increase in the size of a population’s youth cohort raises the risk of conflict recurrence as defined by this study.
To make these results more intuitive and
to assess the strength of the relationship between youth bulges and conflict recurrence, I calculate a series of predicted probabilities and visualize them in Figure 3. The calculations are based on Model 1, and as a baseline I use a case where all variables are set to their median12. The predicted probability
11 The results table for the bivariate regressions can be found in Appendix I.
12 This means that the youth bulge is 33.9% of the total adult population, GDP/ capita is 464.5 USD, ethnic
fractionalization is 0.580, the total population is 18.453 million, the duration of the previous conflict episode was 2 years, it was not a civil war and there was no peacekeeping troop deployed.
Table 4: Logit regression results for Hypothesis 1
29 of conflict recurrence for this case is 36.08%. Holding all other variables constant at the baseline, the predicted probability of recurrence increases to 50.03% when the youth bulge reaches its maximum value of 40.75% of the total adult population. When the size of a youth cohort is at its minimum value of 15.86%, the risk of recurrence decreases to 11.09%. Given the fact that this means there is a 38.94% difference in the risk of conflict recurrence depending on whether the size of a population’s youth cohort is at its minimum or maximum value, I would consider the relationship between these two variables to be quite strong.
Moreover, these results and their visualization in
Figure 3 do not support the suggestion made by some authors, such as Fuller and Pitts (1990) that there is a certain “critical threshold level” of youth ratios beyond which it causes unrest. Instead, it appears that the effect of youth bulges on conflict recurrence is relatively even.
To test the robustness of these findings, I ran logistic regressions using three alternative measures of youth bulges. The results for these can be found in Table 4. Model 3, which uses the male youth bulge measurement appears to confirm the results of Model 1. The non-relative cohort size variable, however, does not reach any standards of significance. Due to Urdal’s criticisms of this measure, which has been explained above, I do not believe that this decreases the validity of the findings for Hypothesis 1. The relative cohort size measurement does reach statistical significance at the 90%
confidence interval, indicating some empirical support for Hypothesis 1 using this definition of a youth bulge. The predicted probabilities of conflict recurrence using the relative cohort size variable are quite similar to those using the youth bulge measure. Holding all other variables constant at their median, the risk of recurrence increases from 15.74% to 52.31% when the relative cohort size is at its minimum and maximum values, respectively. Again, the difference of 36.57% between likelihoods of recurrence is high, thus lending overall empirical support for Hypothesis 1 and the results of Model 1 discussed above. An in-depth discussion of the implications of these results for all four models and the resulting robustness of my findings will follow in Section 6.
Figure 3: Predicted Probabilities of Conflict Recurrence
30
VIF scores for all four models indicate that there are no issues of multicollinearity potentially obscuring the results13. The majority of control variables do not reach any levels of statistical significance, with the exception of logged total population and intensity, at the 99% and 90% confidence level respectively. This is true for all four models. The highly statistically significant nature of the relationship between the total population and conflict recurrence supports Wagschal and Metz’s findings that “large populations are more prone to violence or (vice versa) less peaceful” (Wagschal and Metz 2017, 79). Interestingly, intensity is statistically significant with a negative coefficient, indicating that an intensely fought conflict episode (with more than 1000 battle deaths a year) decreases the risk of conflict recurrence. As mentioned in Section 2, previous research has not come to a general conclusion about the effect of intensity of fighting on conflict recurrence. To my knowledge, none of the work focusing on characteristic of fighting in a conflict episode on the risk of relapse has included youth bulge measurements. While this is interesting to note, I will not investigate this result further given the scope of the present study.
The following sections present the results for regressions testing the remaining hypotheses before Section 6 discusses the theoretical implications for all of the findings.
13 None of the VIF scores are above 2.
