The links between education and conflict are difficult to disentangle because they operate on a multitude of levels—from long-term structural influences to immediate triggering factors for conflict. They also work in different causal directions: the education system and education policy can influence the likelihood and dynamics of conflict, but conflict likewise has major impacts on and implications for education in post-conflict societies.

In this article I focus on factors that run in a broadly causal direction from education to conflict: How does education affect the likelihood and dynamics of conflict? I suggest that education is particularly important for the dynamics of conflict, but also a particularly difficult arena in which to develop conflict-sensitive policy approaches because its influence operates on many different levels. Among these levels, four are key:

  • Education has structural and socio-economic effects on the likelihood of conflict. The education system constitutes one of the most important institutions through which social inequalities along class, gender, ethnic, religious or other lines are created and maintained. These inequalities are increasingly recognized as lying at the root of many contemporary conflicts, but reducing such inequalities—including through the education system—can also increase the likelihood of conflict when advantaged groups feel they are discriminated against unfairly.

  • Education has political effects on the dynamics of conflict. Education can be a tool for both political inclusion and political exclusion. It can be used to promote a common national identity and, in some cases, a lingua franca that serves as a national language, or to reinforce ethnocentric nationalist narratives. Institutionally, segregated education systems that cater differently for different groups within society are often seen as undermining these goals and as reinforcing divisions within society, but in some circumstances they can contribute to cultural autonomy being recognized enough to dampen the potential for conflict.

  • Cultural preferences influence the education system. Different ethnoreligious groups can have preferences for different educational styles and content, preferences that can play into conflict dynamics in a particularly intractable way. On the one hand, refusal to recognize or give leeway to alternative cultural preferences—such as traditionalist Islamic education over modern curricula—may feed into minority groups’ perceptions of marginalization and “cultural status inequality” (Langer and Brown 2008). On the other hand, where cultural plurality is recognized in the education system, minority groups may find themselves ill-equipped for a labour market structured around the norms of the majority community.

  • Education is a site of mobilization. Beyond the effects that education can have on the structural causes of conflict, educational establishments can constitute sites for rebel organizations to indoctrinate, radicalize, and directly recruit minority groups, although the evidence for this is much more limited than typically assumed.

Moreover, because education systems are often an important factor in the exclusion and marginalization of particular groups in society, education policy itself is highly politicized in deeply divided societies. Changes to the education system can arouse deep suspicion and resentment, particularly if they are perceived as being imposed upon minority communities by a dominant majority. Rarely is an education system and policy directly implicated in a turn to violent conflict, but education is arguably one of the most important—and certainly most challenging from a policy perspective—contextual factors for conflict likelihood, precisely because it intersects with so many other dimensions of conflict dynamics.

Education and economic theories of violent conflict

Because violent conflict is so complex, most comparative accounts of the socio-economic causes of violent conflicts—as opposed to case studies of particular conflicts—focus on factors that appear likely to increase or decrease the likelihood of conflict, rather than asserting one or more specific “causes” for it. Broadly, these studies can be divided into two approaches that relate, respectively, to the opportunities for conflict and to the motivations for violent group behaviour.

Put simply, opportunity theories of conflict assert that conflict is more likely to occur in contexts where it is more feasible to instigate “rebellion”. This may or may not involve attributing motivations to insurgent behaviour, but the point is that even when motivations are asserted, they are thought to be less important in explaining the incidence of conflict because those motivations are presumed to be present even when conflict is not. Hence, to explain conflict, we must explain the conditions under which such motivations are translated into violent behaviour, and this is a matter of explaining opportunity conditions.Footnote 1 In one of the first econometric studies of violent internal conflict, for instance, Fearon and Laitin (1996) concluded that two factors best account for the incidence of conflict: the presence of a weak state, which makes success more likely for insurgent groups, and the nature of the country’s geographical terrain, with mountains and forest cover providing more feasible grounds for rebellion.

Fearon and Laitin are largely agnostic about what motivates insurgents, but an alternative opportunity account is offered by Paul Collier and his collaborators; they use economic theory to suggest that the main motivation for conflict is the individual desire to profit from rebellion (Collier 2000; Collier and Hoeffler 2004). Following this approach, the conditions that favour insurgency are those that lower the opportunity cost of engaging in rebellion. On the one hand, this points to a set of factors that make rebellion more profitable, notably the presence of natural resources that can be looted, including oil and diamonds. On the other hand, another set of factors increase the profitability of rebellion relative to alternatives including high levels of male youth unemployment. Assuming that young men are the primary recruits into rebellion, high unemployment among this group means a higher risk of conflict because there are more potential recruits for whom the alternative is a life of poverty. Proponents of this type of explanation have been able to marshal considerable econometric evidence as well as case studies to support their argument.

The alternative approach, focused on the motivations for violent conflict, suggests that while some degree of political and economic grievances may well be a universal characteristic of developing (and developed) societies, there are nonetheless important differences in the extent of grievances: societies with higher levels of grievance are more likely to see conflict than those with relatively low levels of grievance. Moreover, this approach suggests that certain forms of grievance are more liable to be mobilized into violent conflict than others. In particular, proponents of this approach suggest that conflict is more likely where grievances align with cultural divisions within society—primarily religious or ethnic—because cultural affinity provides a powerful framework for conflict mobilization. This approach is mostly associated with Ted Gurr’s (1970, 1993) thesis of “relative deprivation” and Frances Stewart’s (2000, 2008) notion of “horizontal inequalities”. While both theories point to largely the same set of factors, there are important distinctions between them. Gurr’s approach is based on a primarily internal group calculus which he formulates as the difference between group aspirations and group achievement. Stewart’s conceptualization of horizontal inequalities focuses more explicitly on between-group differences. The evidence associated with horizontal inequalities suggests that three particular contexts tend to increase the likelihood of conflict:

  • The intersection of horizontal socio-economic inequalities and political exclusion appears particularly incendiary. Econometric analysis shows that politically excluded ethnic groups are more likely to rebel (Cederman, Wimmer, and Min 2010) and case study evidence from Côte d’Ivoire (Langer 2005) and Sudan (Cobham 2005) has clearly demonstrated how the political exclusion of economically marginalized groups can feed into a process of ethnic mobilization and ultimately violent mobilization.

  • Where communal conflict between ethnic groups emerges, state interventions that are seen to favour one ethnic group over the others can transform communal violence into anti-state rebellion and, where ethnic groups are geographically concentrated, violent secessionism (Brown 2008). Similarly, conflicts that begin as primarily ideological can become ethnicized through the prosecution of violence that often affects marginalized ethnic groups more than other groups; the conflicts in Latin America have largely followed this trend (Caumartin 2005; Caumartin, Gray Molina, and Thorp 2008).

  • Decentralized institutions, including federal structures, have a particularly complex relationship with inequality and conflict: econometric evidence suggests that fiscal decentralization appears to mitigate conflict incidence in geographically segregated countries, but can be inflammatory in the presence of high inter-regional inequalities (Bakke and Wibbels 2006; Brown 2009; Tranchant 2008). Case studies provide relatively strong evidence that increased fiscal decentralization can appease restive provinces that want to secede (Alemán and Treisman 2005). Political devolution appears to appease secessionist pressures in relatively wealthy minority regions, but to exacerbate such pressures in relatively poor regions (Brown 2009).

Of course, these two accounts alone do not exhaust the wide and sophisticated literature on conflict dynamics, and other analyses have pointed to a range of additional factors that appear to influence the incidence of conflict. One factor that is important in virtually all accounts is GDP per capita; statistically, it is the only economic factor that is robust to different definitions of “conflict” (Sambanis 2004) and different combinations of independent variables (Hegre and Sambanis 2006). Poor countries are more at risk of conflict than richer countries. Population demographics are also important. Larger populations are certainly more susceptible to violent conflict, but recent attention has also been drawn to the particular danger of “youth bulges” in countries where the population is increasing dramatically in the lower age cohorts, although the validity of these claims is disputed (see below). Gender dynamics are also receiving increased attention, with feminist scholars suggesting that conflict can be seen as an outcome of “frustrated masculinities”: where societies place high expectations on men to fulfil functions of breadwinning and masculine behaviour, conditions that frustrate these expectations may lead men to respond in violent ways in order to reassert their masculinity.

Clearly, we have no particular reason to assume that either approach entirely invalidates the other and it seems plausible to assert that conflict dynamics are likely to involve dimensions of both opportunity and motivation. For our purposes here, however, reconciling these two approaches is less important than discerning the particular role that education can play in either fostering or mitigating them. Here, I take each approach in turn and consider what kinds of theoretical roles education might play and the evidence available to support these claims, with particular attention to potential policy implications.

First, from the opportunities perspective, education is primarily important as a determinant of later incomes and, hence, the opportunity cost of violent conflict. Particularly among men—as the main agents in rebellion—low levels of education will translate into lower normal incomes and the economic risk from engaging in violent behaviour will be concomitantly lower. Collier and Hoeffler (2004) found econometric evidence that countries with lower rates of male secondary school enrolment are indeed more susceptible to violent conflict, but they note that this finding is difficult to interpret, as rates of male secondary school enrolment are highly correlated with levels of GDP per capita. The policy implications of this approach are both straightforward and uncontentious: policy should prioritize increasing the general rates of education in the population. Still, it is important to note that, from a conflict perspective at least, higher rates of education may be important only insofar as they translate into higher incomes. Increasing education without increasing subsequent employment opportunities may be less effective at reducing the likelihood of conflict.

When we consider the motivations-based approach to understanding conflict, however, it is clear that education is critical to creating and maintaining, or ameliorating and reducing, the social inequalities that lie at the heart of many violent conflicts, particularly those that appear to fall along ethnic or religious lines. Here, however, the role of education is more complex and multi-layered, involving three main interlinked economic aspects:

  • The distribution of educational opportunities between ethno-religious groups within the population;

  • The extent to which education translates into economic opportunities, that is the general economic returns to education; and

  • Differences in returns to education between ethno-religious groups.

Low returns to education are also important from the motivations-based perspective because they can exacerbate the gap between aspirations and achievement that the relative deprivation theory sees as lying at the heart of violent conflict. Education increases aspirations—and expectations—of income and frustrating these aspirations can contribute to violent mobilization. Hence, for instance, the inability of educated youths in Sierra Leone to find gainful employment has been seen as a major driver of the widespread sense of shame among young men in the country that fuelled the civil war. David Keen (2005) cites an ethnographic example:

Many young, educated Sierra Leoneans find themselves in what can only be described as an inescapable trap. For example, the disinherited son of [a] widow […] was unable to approach his patrilineage to request farmland because he was not married. […] While some young men in similar positions do waged labour on farms to support themselves and save some money, the young man’s education and the association of superiority it carried with it made such labour an intolerable option, even though the nature of the young man’s education did not qualify him for any particular employment. (p. 72)

This example illustrates how education is important because it links together different risk factors for conflict. The centrality of shame as a motivation for conflict in Sierra Leone is clearly commensurate with the gendered perspectives on conflict outlined above. The Sierra Leone case also illustrates the way in which youth bulges per se are not necessarily problematic for violent conflict. As Marc Sommers (2006) and others (Hartmann 2009; Hendrixson 2004) have pointed out, growing populations can, in fact, be good for economic progress. When they represent a risk factor for violent conflict is when these youths have inadequate opportunities; clearly, education is central here. If education systems are inadequate and unable to equip pupils with the appropriate tools for the labour market, they can in fact feed rather than mitigate violent conflict by creating aspirations that cannot be matched in reality.

Differences in average educational achievement rates between ethno-religious groups also play an instrumental role in creating and maintaining broader socio-economic horizontal inequalities within a society because of their more general knock-on effect on earnings potential and social mobility. Across Central and South America, the educational rates of indigenous group and, in some countries, Afro-Caribbeans, are substantially lower than those of the mestizo or ladino elites.

When education is particularly problematic, it appears, is when differences in the absolute levels of educational attainment between different ethno-religious groups overlap with different returns to education. This phenomenon of double disadvantage is at the heart of the persistence of horizontal inequalities over very long time periods. Stewart and Langer (2008) have developed a formal model to explain the persistence of horizontal inequalities, modelling average group incomes as a function of three forms of capital: physical capital (assets, etc.), human capital (primarily education, but also health), and social capital (networks). Crucially, however, they argue that an individual’s returns to one form of capital are dependent upon the levels of other forms; without access to financial assets or strong social networks, for instance, one individual is not able to exploit their education (human capital) as effectively as another individual with the same level of education but higher endowments of other forms of capital. Hence, educationally deprived groups are not able to benefit from education to the same extent as other groups. Moreover, facing such a situation, members of deprived groups are less likely to invest in their children’s education, especially because poor people must often educate their children at the expense of benefitting from their labour; in turn, this perpetuates the cycle of educational deprivation and horizontal inequalities.

Language, education, and conflict

Thus far I have considered education as a structural cause of conflict in relatively narrow economic terms but, as noted above, one reason education is so critical to violent conflict is that it cuts across political, economic, and social realms. Lower returns to education for deprived groups, for instance, are often not just the product of the kind of persistent economic logic outlined above, but are also due to labour market discrimination and, in some cases, deliberate political exclusion.

Critically linking the economic and political dimensions of exclusion and marginalization in education is the issue of language and language of instruction. In many ways, language policy in education embodies and epitomizes the tensions and risks of managing diversity more broadly. On the one hand, the promotion of a single lingua franca has been seen by many governments, particularly those in transition to independence, as an important milestone on the road to “national unity”. France provides a good historical example of such a project succeeding; at the turn of the nineteenth and twentieth centuries it linked the promotion of the French language (as opposed to the many local dialects of the period) with a huge expansion in basic education in order to transform “peasants into Frenchmen” (Weber 1976). Similarly, many economists suggest that a single lingua franca is of net benefit to economic efficiency because it reduces transaction costs: it is cheaper for people to do business together if they speak the same language.

On the other hand, denial of mother-tongue educational opportunities can contribute to nationalist backlashes, and create obstacles to educational attainment for minority groups. This appears to be particularly problematic where the promoted lingua franca is associated with the domination of one particular ethnolinguistic group, but promoting a “neutral” lingua franca usually involves prioritizing the former colonial language, which in turn carries implications of vertical (i.e. class-based) rather than horizontal marginalization and exclusion. In post-colonial Africa, for instance, virtually all the former colonies decided to adopt the colonial language—whether English, French, or Portuguese—as the national language and main medium of instruction in education; this move has been seen as entrenching elite domination at the expense of “authentic nationhood” (Mansour 1993, p. 199; see also Cleghorn 2005; Myers-Scotton 1993). From an economic perspective, a single medium of instruction potentially acts as a disincentive for minority groups to enter the education system, and even when they do it can have negative impacts on their performance, particularly in the early years when the language gap is significant.

Education, then, is both a major agent in perpetuating the socio-economic inequalities that often undergird violent conflict and an important—and often symbolic—tool for the political inclusion or exclusion of minority groups. Moreover, tensions often arise between the economic and political roles of education, which highlight its conflict-inducing potential. Comparative evidence suggests that the demographic make-up of the country is important here. The imposition of a national lingua franca—whether indigenous or colonial—as the sole medium of instruction in secondary, and sometimes primary, education appears to be much less problematic in situations of extreme linguistic diversity. This may reflect a popular pragmatism, a recognition that some form of lingua franca is needed in highly multilingual societies and that education plays a fundamental role in promoting that language. Malawi, Botswana, the Philippines, and Tanzania—all very linguistically diverse societies—have all had relatively unproblematic relationships with colonial languages in education (Djité 2000; Hau and Tinio 2003; Kayambazinthu 2004; Nyati-Ramahobo 2004). Meanwhile, in Indonesia, the selection of Malay as the language of education and the national language served both to bridge linguistic divisions across the extremely diverse archipelago and, to some extent, to allay fears of ethnic Javanese domination (Bertrand 2003). As with most conflict dynamics, however, this relationship is not automatic and it depends on broader policies of inclusion and exclusion. In Nepal, for example, the imposition of Nepali as the language of instruction fed into the broader set of grievances among non-Nepali speaking castes and ethnic minorities that drove the civil war (Gates and Murshed 2005).

The third dimension along which these economic and political functions of education intersect with conflict dynamics is through the recognition—or not—of cultural preferences in educational style and content. As outlined above, recognition of minority traditions can be an important step towards multicultural toleration, but it also risks creating parallel education systems that do not sufficiently equip members of minority groups with the skills demanded by the labour market to compete with those from the mainstream schooling system. The role of more traditional madrasah in majority non-Muslim countries epitomizes the tensions between the cultural/political and economic functions of education. For many Muslim minorities, sending their children to Islamic schools is an important assertion of cultural distinctiveness and autonomy. Yet the skills children learn in madrasah are often not as closely matched to the labour markets’ demands as the skills taught in secular schools. Hence, the assertion of cultural autonomy through madrasah schooling may itself contribute towards the persistence of religious horizontal inequalities.

There is certainly clear evidence that the appeal of madrasah-style schooling among Muslim minorities is often linked to a sense of cultural distinctiveness. In India, Muslims’ expression of cultural autonomy through madrasah attendance combines with language issues and a preference for Urdu-medium teaching that the secular school system does not provide. Similarly, in Southern Thailand, two major causes of resentment among the Malay Muslim minority are, first, the intermittent attempts by the Thai state to regulate—and, some perceive, ultimately disestablish—the pondok traditional Islamic school system, and second, the under-representation of Muslims in the formal education sector, particularly in the urban centres of the South (Brown 2008). Yet clear tensions exist between these two sources of grievance.

Thus the key point here is this: not only does education intersect importantly with many of the political, economic, and cultural risk factors for violent conflict, but it does so in ways that often pull in contradictory directions not merely for between-group relations, but also within minority groups. For instance, in Suresh Canagarajah’s (1993) ethnographic study of Tamil students learning English in Jaffna, Sri Lanka, one of her respondents reported:

I am from [a rural area]. There was no English from Grade 3 to 7. I lack opportunities. But I really (extremely) desire learning English… Please don’t reveal this to anybody else in the class. (p. 610)

This ambivalence between social mobility and cultural distinctiveness lies at the heart of the educational dilemma for minority groups, and helps explain why it is such a politically sensitive contributor to conflict dynamics.

Political institutions and decentralized education

From a policy perspective, various forms of decentralization and federalism are widely debated as potential conflict-mitigating structures in diverse societies, and it is worth considering more extensively the relationship between such political structures and the education system. It is useful here to adopt Daniel Treisman’s (2007) typology of forms of decentralization; the first three of them are particularly important for us here.

  • Administrative decentralization refers to arrangements where the implementation of centrally-determined policy is undertaken by local agents.

  • Political decentralization refers to arrangements which involve a degree of policy-making authority.

  • Fiscal decentralization refers to situations where subnational tiers either have tax-raising powers or constitute a significant proportion of total government spending.

  • Appointment decentralization refers to the conduct of local elections for local leaders.

  • Constitutional decentralization refers to arrangement where local authorities have a say in national policy-making.

Educational decentralization through administrative decentralization is being promoted around the world, in part due to a broader shift as donors and international agencies see decentralization as a more transparent and efficient means of service delivery. With specific reference to education, they see it as a means of addressing the problems of marginalization by promoting decentralized community schooling systems such as the PRONADE system in Guatemala and the PROHECO schools in Honduras, which are typically geared toward the inclusion of indigenous students. Evidence on the effectiveness of these school systems varies, with some systems appearing to improve academic outcomes, others less so (see de Grauwe 2005; Nielson 2007 for broad evaluative reviews). With specific reference to the educational inclusion of marginalized groups, however, the evidence in more systematically disappointing. In Guatemala, the PRONADE schools are failing to address the indigenous groups’ performance deficit in school (Marshall 2009). More generally, the community-based design of these projects does not appear to promote inclusion as marginalized groups are often left out of, or under-represented in, school management committees (de Grauwe 2005; Nielson 2007). In this sense, decentralized educational structures are still simply reproducing the patterns of domination within society more broadly. One notable exception to this is the successful Education Centre Tatutsi Maxakwaxi which caters for the Wixarika community in the Mexican state of Jalisco, but it is worth noting that it is a community-driven project rather than the result of a formal decentralization process (Santizo Rodall and Martin 2009).

While administrative decentralization with substantial political autonomy appears not to have succeeded in promoting the educational inclusion of minorities or marginalized groups, a combination of political and fiscal decentralization, usually in the form of some kind of federal structures, has been more effective. A good historical example of this is Switzerland, where education policy—like most areas of social policy—is highly decentralized down to the canton level. Since the nineteenth century for primary education, however, and more recently for higher levels of education, the federal government has mandated minimum levels that cantons are obliged to provide and also subsidized educational provision in poorer cantons (Linder and Vatter 2001). As a result, despite its high level of linguistic and ethnic diversity, by 1870, Switzerland had achieved an estimated primary enrolment ratio of 75%, the highest estimate in the world at that time, and far above the regional average of 56% (Benavot and Riddle 1988).

The relationship between federalism, political decentralization, and education in resolving or mitigating ethnic disparities and the associated potential for conflict does, of course, depend fundamentally on the form of federalism and the geographical spread of ethnic communities. In Malaysia, for instance, the sub-national states have relatively limited powers and education is designated a federal issue. Similarly, the main ethnic groups of Malaysia are largely geographically integrated, with the exception of the predominantly Malay east coast states. Hence, the link between federalism and education in Malaysia is not likely to be of much import. In contrast, in neighbouring Indonesia, ethnic groups are highly regionally concentrated; along with other social services, education was radically decentralized in the post-Suharto era to the second-level districts (rural kabupaten and urban kota) rather than the first-level provinces (propinsi). In fact, the contemporary education system in Indonesia is similar to the Swiss one in many ways, with the central government stipulating minimum guidelines for educational standards but the districts implementing this as they see fit. As in Switzerland, this is partnered with strong fiscal decentralization that provides districts with revenue-raising powers and stipulates an equalization grant from the centre to poorer regions.

Initial evidence suggests that this has led to a significant reduction in regional inequalities in enrolment rates, with a drastic drop in the variance in enrolment gaps from 31.9 in 1998 (when the decentralization measures were passed) to 13.2 in 2004 (Arze del Granado, Fengler, Ragatz, and Yavuz 2007). Poorer districts, however, still struggle to provide education services, on average spending a higher proportion of their budget on education than richer districts, but translating into lower per student expenditure. Given the paucity of data, however, systematic evidence is lacking on the impact that decentralization has on ethnic, rather than regional, inequalities in Indonesia. There is some evidence that Muslim children are less likely to transit to secondary school than non-Muslim children of equivalent regional and socio-economic background (Suryadarma, Suryahadi, and Sumarto 2010), but survey evidence also suggests that non-Muslims, who account for around 14% of the population, are significantly less likely to be satisfied by education provision in their district (Lewis and Pattinasarany 2009).

A similar model has also been promoted and at least partially implemented in Ethiopia, where educational inequalities were a particularly inflammatory aspect of the broader ethno-regional inequalities that have stoked ethnic conflict in the country in the past (Abbink 1997). After the collapse of the Derg in the late 1980s and the final ouster of Mengistu in 1991, the new Ethiopian government promoted major decentralization, including of education, to the newly reorganized ethnically-delimited federal states. As in Indonesia and Switzerland, the federal government retained broad-policy setting powers at the primary level (although higher education remains a federal issue) but implementation was devolved to the states; it also allowed for minority language instruction, which had been illegal under Haile Selassie and discouraged by the Derg. Regional disparities in enrolment rates remain high, however, with gross enrolment rates in the eastern states of Afar and Somali stagnating at around 10% between 1996 and 2000, compared with figures increasingly in excess of 60% in western states such as Gambella and Benishangul, as well as the capital region Addis Ababa. As in Indonesia, but to an even larger extent, the main culprit here appears to be the failure to fully implement fiscal decentralization. A notable exception is the politically restive southern states of Oromiya and the Southern Nations, Nationalities, and People’s Region (SNNPR), both of which have received significant federal revenues and as a result have seen a dramatic increase in enrolment rates: from around 30% to 50% in Oromiya and from 40% to 60% in SNNPR over the same time period (Keller and Smith 2005).

A final example of federal arrangements for educational provision comes from India, where the initial state boundaries were reorganized along linguistic lines a decade after independence. Language of instruction and other aspects of educational policy are decided at the state level. Despite similar arrangements in terms of fiscal and political decentralization of education, overall educational disparities in India appear to be increasingly accounted for by regional disparities between states, in addition to the persistence of educational inequalities by ethnic and religious groups noted above. An analysis of the inequality in literacy rates in India shows that in 1981, between-state inequalities accounted for around half of the total inequality in literacy rates; by 1998, this had risen to three-quarters (Noorbakhsh 2003). Explanations of this trend suggest that it is partially caused by the simple enormity of India’s educational problems, which engulf states’ budgetary capacity, along with the discretionary allocation of federal funds between states which typically favours relatively rich states (Tilak 1989).

Case study evidence suggests, then, that an effective combination of political and fiscal decentralization in ethnically-divided countries with broad territorial segregation of ethnic groups provides a potentially effective model for ameliorating the impact of educational disparities on ethnic conflict. Educational federalism allows for effective expression of (perceived) minority language rights and can help to equalize absolute educational attainment, if not returns to education. But it is important to note that this is no panacea. First, ensuring adequate and equitable financial flows from the centre to the regions is a vital component of such a strategy. Second, educational federalism has sometimes contributed to the emergence of political disputes and sometimes violent clashes between groups within the federal units; this has been observed in Nigeria (Ukiwo 2007), Ethiopia (Keller and Smith 2005), and India.

Decentralization may provide a useful conflict-mitigation mechanism in diverse and geographically segregated societies, but it is clear that those cases where the education system is most directly implicated in violent mobilization—such as Sri Lanka, Thailand, and Algeria—are those where the ethnic demography is one of relatively small minority groups facing a dominant majority group. In these cases, however, it is clear what contributed to the escalation into conflict: particular changes in education policy that were seen as discriminating against the minority. As the comparison below of Sri Lanka and Malaysia demonstrates, similar educational policies can be received in very different ways by minority communities depending on how the state manages and legitimizes them. In Malaysia, pro-Malay policies in education, particularly at the tertiary level, were accompanied by a concerted project of political legitimation and consensus building that saw the minority communities largely accept—albeit often reluctantly—the political necessity of these programmes. In contrast, similar programmes were implemented in Sri Lanka as part of a broader project of exclusivist Sinhalese nationalism—and the Tamil minority received them as evidence of deliberate discrimination and marginalization. Hence, just as important as the design of policies is the way they are developed and implemented.

Education and rebel recruitment

We have seen, then, that education plays an important role in conflict dynamics because it crosscuts with multiple causal pathways and risk factors for violent conflict. In general, however, the education system is not a direct cause of conflict but rather an enabling contextual factor, albeit one that is particularly problematic from a policy perspective because of its often contradictory implications.

In certain contexts, however, education has been argued to play a more direct role in conflict as a realm of radicalization and recruitment into rebel networks. In the contemporary international security environment, this is particularly associated with the perceived role of Islamic madrasah in recruitment into “terrorist” networks. Other cases, however, show evidence of recruitment through school cooptation and radicalization, for instance the Free Aceh Movement in Indonesia (Schulze 2004). Similarly, the Eritrean People’s Liberation Front recruited through networks of students rather than directly through schools (Weinstein 2005).

While it is certainly clear that particular schools have served as recruitment centres for particular organizations, the more general trend is less clear. Sohail Abbas (2007), for instance, reports that the majority of jihadis incarcerated in Pakistan after capture in Afghanistan were educated in the mainstream education system rather than in madrasah. Similarly, Olivier Roy (2004) points out that the 9/11 bombers themselves were mostly the product of secular education systems; one even attended a Christian school in Lebanon. More generally, however, comparative studies and case studies of recruitment into rebel groups do not show a systematic tendency to recruit through schools or other educational institutions (Kalyvas 2006; Nillesen and Verwimp 2009; Weinstein 2005). Instead, these studies focus on the primacy of coercion and on sanctions against non-recruits, although this can also take place through schools. Thus, for instance, Lilja (2009), who studied recruitment into the LTTE in Sri Lanka, writes,

In the early 1990s LTTE conducted awareness campaigns for teenagers at schools and sport stadiums throughout the Vanni. Four accused traitors were asked to make five-minute confessions as to their alleged acts of treason […] After having been brought to all Vanni schools, the men were sentenced to death by LTTE and assassinated in public. (p. 315)

Thus the common perception that schools are ideological breeding grounds for insurgent groups, while accurate in some cases, is largely misplaced. Moreover, it seems clear that the full range of recruitment strategies that rebel groups use to mobilize—including ideological propaganda, but also intimidation and outright coercion—are as prevalent in school recruiting when it does occur as in other venues; there is no special relationship between education and radicalization.

Conclusions

To conclude, then, I have argued that the popular perceptions of the relationship between education and conflict—epitomized by the image of radical Islamic madrasah in Pakistan serving as conduits for Al Qaeda and other Islamic militant groups—do not adequately or appropriately capture the complexity of the relationship. Education influences the incidence of conflict primarily through its indirect impact on other structural causes of conflict. The reason that education is so important for conflict dynamics is not its direct impact, but the fact that it interacts crucially with so many other dimensions of conflict. Economically, education systems are instrumental in creating and maintaining social stratification and horizontal inequalities between ethnic and religious groups. Politically, education is an important tool of inclusion and exclusion, at both the mass and the elite levels. And culturally, education is a symbolic venue for the recognition of minority cultures, languages, and practices. The challenge for education policy makers seeking to minimize the potential for conflict in divided societies is to weigh all these different factors—and their often contradictory implications—in developing appropriate interventions and responses.