Introduction

The September 2013 attack at the Westgate shopping mall brought worldwide attention to Kenya, with Al-Shabaab claiming responsibility and maintaining that it was conducted in revenge against Kenyan military operations in Somalia.1 However, this particular incident represents only one of many examples of ‘violent extremism’ (VE) in Kenya over in recent years, and indeed the intensity has escalated over recent months. Home to many individuals of Somali descent, Nairobi’s district of Eastleigh is one prominent VE hotspot. Such violence also occurs with relative frequency along the Kenyan coast, both driven by and provoking elevated tensions between religious communities, particularly after the killing of prominent clerics such as Sheikhs Aboud Rogo (August 2012), Ibrahim Rogo (October 2013), Abubakar Shariff (April 2014), and Idris Mohamed (June 2014).

Within this tense context USAID’s Office of Transition Initiative (OTI) implemented what was essentially a pilot program of the new Countering Violent Extremism (CVE) concept, forming one component of the wider Kenya Transition Initiative (KTI). Between 2011 and 2013 the program was operational in Eastleigh and its environs, and in 2012 it was expanded to the coastal regions of Lamu, Kilifi, Kwale, Malindi and Mombasa. It was delivered via a flexible funding mechanism that supported individuals, networks and organisations, often with small grants implemented over a short duration. These grants funded a broad range of activities aiming to counter the drivers of VE, including livelihood training, community debates on sensitive topics, cultural events, counselling for post-traumatic stress disorder, and so on. While the CVE concept is essentially still under development (as discussed below), it now commonly forms one component of a wider response to VE that also incorporates law-enforcement, counterterrorism, development and other initiatives. While this article focuses narrowly upon CVE programming, a key determinant of the extent to which VE is countered will undoubtedly be the degree to which these diverse efforts are adequately prioritized and coordinated.

With the KTI team eager for the lessons from this pilot to be disseminated to the wider community of CVE practitioners, this essay summarises a qualitative evaluation undertaken by Integrity Research and Consultancy as part of the program close-down phase.2 With implementing organizations rarely willing and/or able to reveal their ‘lessons learned’ to a broader audience, we believe that it offers the reader a rare access ‘inside’ a highly significant US-funded security initiative.3 We also believe that its relevance is heightened both because the CVE concept continues to grow in prominence amongst US and European donors, and as many insights presented in this paper are also of substantial relevance to stabilization, counterterrorism and other related forms of programming. Although it is unfortunately not possible to provide additional details given the sensitive nature of the program, during the field research we collected considerable qualitative evidence to suggest that KTI’s efforts achieved a positive impact in dissuading certain individuals from following the VE path. We also concluded that a critical strength of the program was its flexibility, enabling rapid responses to the aftermath of specific incidents such as the Westgate attack and the killing of Sheikh Aboud Rogo. It also showed flexibility in its ability to scale-up associations with successful grantees.

This essay is structured loosely to reflect the KTI program lifecycle, sequentially focussing upon the goal statement, drivers of VE, vulnerable population subsets, the intervention logic, and negative effects. The final section delivers two overarching conclusions intended to be of relevance to the wider CVE community. Firstly, we argue that KTI programming decisions would have benefitted from a more comprehensive understanding of VE within the local context. For instance, subsets of the population more narrowly ‘at-risk’ of being attracted to VE should have been identified and targeted (potentially including teenagers, members of specific clans, ex-convicts, etc.), and a greater focus should have been placed upon comprehending the role of material incentives, fear, status-seeking, adventure-seeking, and other individual-level drivers. A second is that the KTI team would have profited from additional top-level guidance from their donors, for instance, providing direction on the extent to which efforts should have been targeted at those supportive of violence versus those directly involved in its creation, the risks associated with inappropriate donor branding, and the contexts in which the pejorative term ‘extremism’ should be pragmatically replaced by neutral terminology. As a priority donors and the wider community should also provide suitable definitions of the CVE concept, rather than leaving practitioners to construe, undoubtedly inconsistently, it’s meaning from the available definitions of VE.

Defining the program goal

Providing a platform for the subsequent discussions, and CVE objective-setting more broadly, this initial section comments upon the KTI program goal statement, stipulated as: ‘Stronger identity and self-confidence of youth to allow them to reject extremism.’ This statement is problematic for a variety of reasons. Firstly, we argue that the term ‘identity’ should have been omitted given that the relationship between this concept and VE is highly complex and nuanced (e.g. Schwartz et al 2009), and this runs counter to the need to maintain simplicity in the overarching goal.4 Indeed, there was confusion amongst KTI stakeholders regarding the basic point of whether the term was intended to refer to national, religious or clan identity, or even identity as youth (widely defined as those between 16 and 35 in Kenya). The focus upon ‘self-confidence’ is also problematic as it implies an association between this psychological factor and an attraction to VE, despite an apparent absence of supporting evidence for this link. This is of particular relevance given the growing consensus against the idea that it is possible to create suitable psychological profiles (e.g. Sageman 2008: 17). In any case, the converse relationship is also plausible, i.e. that elevated levels of self-confidence may be associated with an attraction to VE.

Perhaps of greater relevance, the phrase ‘allow them to reject extremism’ should have been adapted as it is suggestive of ‘top-down’ assumptions, implicitly removing degrees of agency from individuals who often self-select to travel the VE path. Put another way, it infers that radicalisation is a process that happens to individuals, rather than being something that is actively pursued. Seeking to counter this common perception, Marc Sageman (2004: 122) asserts that ‘joining the jihad is more akin to the process of applying to a highly selective college.’ He continues that ‘many try to get in but only a few succeed,’ and that ‘candidates are enthusiastic rather than reluctant’ (Sageman 2004: 122). While counterexamples certainly do exist of VE entities actively recruiting from the populace, the wider point is that it is mistaken to assume that enlistment is solely a ‘top-down’ process.

Of greater relevance still, the language is somewhat ambiguous, and this reflects broader issues that remain with the conceptualisation of CVE. Indeed, the concept of CVE is itself rarely defined by donors, and this undoubtedly results in inconsistent interpretations as implementers are simply left to construe its meaning from the available definitions of VE. Compounding this issue, the VE concept is extremely broad as it focuses upon both those directly involved in the production of violence, and their supporters. For instance, the USAID paper Development Response to Violent Extremism and Insurgency (from which the KTI team drew heavily throughout program design and implementation) identifies VE (USAID 2011: 2) as ‘advocating, engaging in, preparing, or otherwise supporting ideologically motivated or justified violence to further social, economic and political objectives.’ The Australian National Counter-Terrorism Committee similarly defines the concept as ‘a willingness to use or support the use of violence to further particular beliefs, including those of a political, social or ideological nature’ (cited in Nasser-Eddine et al 2011: 8).

With this breadth in mind, and pending suitable definitions of CVE from donors, we tentatively suggest the following goal statements for future initiatives:5 1) To reduce the rate of VE acts in Location X, and 2) To reduce the level of support for VE acts in Location X. In practice, however, we also argue that the comparative importance of the latter objective varies according to context. In particular, CVE initiatives should place an elevated focus on undermining support for VE in locations where the perpetrators of violence are highly dependent on local communities for material resources, shelter, information on state forces, and so on. This should also be the case in locations where support for VE is particularly widespread, and thus long-term solutions necessarily involve extensive efforts to change attitudes. This has programmatic implications given that supporters of violence are often considerably more numerous than those who are directly involved (Khalil 2014),6 and thus interventions would have to target a broader cross-section of the populace.

Identifying the drivers of violent extremism – push factors

However articulated, efforts to achieve CVE aims require an ample understanding of what drives this violence in any given environment, and in pursuit of such comprehension the KTI team adopted the language of ‘push’ and ‘pull’ factors, as outlined in USAID’s Development Response. According to this paper (2011: 3–4), the former are ‘important in creating the conditions that favor the rise or spread in appeal of violent extremism or insurgency,’ whereas the latter ‘are associated with the personal rewards which membership in a group or movement, and participation in its activities, may confer.’ Development Response does not aim to offer a definitive list of potential drivers, and the KTI team correctly sought to identify factors of specific relevance within their areas of implementation, as indicated in Figure 1. It should be briefly observed that, while the concepts of ‘push’ and ‘pull’ provide a convenient means through which to structure an understanding of VE drivers, other such systems of categorization may also be adopted, e.g. distinguishing between ‘motivators’ and ‘enablers.’

Figure 1 

Push and pull factors (summarised from project documents).

Whilst it is beyond the scope of this paper to comment upon all such drivers, the role of discrimination and profiling in ‘pushing’ individuals to VE is worth highlighting. This issue is routinely discussed within the Kenyan literature, for instance, with the International Crisis Group (2012: 10) reporting that ‘the war against Al-Shabaab has led to an increase in ethnic profiling and discrimination against Somalis in particular, and Muslims in general.’ Anneli Botha (2013: 9), from the Institute for Security Studies, similarly states that ‘members of the Somali-Kenyan and Somali communities claim to be victims of racial or ethnic profiling and to have been rounded up and arrested for little reason other than their race and ethnicity.’ Indeed, in response to escalating levels of violence, such patterns have increased in 2014 through Operation Usalama Watch.7 This initiative was launched in the aftermath of the Westgate incident and associated attacks in Nairobi, with its many critics asserting that it applies a disproportionate use of force against these same communities (e.g. Amnesty International 2014).

While poverty and unemployment were also identified as drivers by KTI, there is substantial debate in the wider literature as to whether such factors actually correlate with VE (e.g. Mercy Corps 2013; USAID 2009: v; Venhaus 2010: 5). Critically, even if a relationship between these factors and VE is revealed in a specific location, it should not be taken for granted that the former drive the latter given that there are other potential causal routes. Specifically, as shown in Figure 2, unemployment may drive individuals towards VE, or conversely VE may lead to a lack of employment opportunities (perhaps as individuals become tainted by the former), or an ‘external variable’ (clan affiliation, education levels, and so on) may drive both. Complexity is added, however, as causality may flow in multiple directions simultaneously (e.g. X ‘causes’ Y and Y ‘causes’ X). Such considerations are of critical relevance given that initiatives to enhance employment in any specific region will not serve to reduce VE if the uppermost option in Figure 2 is incorrect. Of course, this logic is potentially also applicable to all candidate drivers of VE (i.e. those identified in Figure 1), and so alternate causal directions should be considered in each instance.

Figure 2 

Potential directions of causality.

Furthermore, while the issue of causal direction is of clear relevance, it should be recognised that it represents no more than the tip of the iceberg in terms of complex effects. The idea of a simple linear relationship between causes (e.g. unemployment) and effects (i.e. VE), gives way to a reality of tipping points, disproportionate feedback loops, interaction effects between multiple variables, and other complexities. These issues are exacerbated as the analysis is commonly performed on data of questionable quality given that study informants may offer false or misleading information, (a) as they themselves are misinformed, (b) to discredit others, (c) to be viewed favourably by the interviewer (often referred to as ‘social desirability bias’), (d) out of fear of potential repercussions of divulging information, (e) to aggrandise their own role in events, or (f) as a process of unwitting self-deception.8 Thus, rather than attempt to ‘prove’ causality, the more modest aim of studies designed to inform CVE efforts should be to draw caveated findings to support or contest a predetermined list of hypotheses on the potential drivers of VE, i.e. including those listed in Figure 1. And, with such considerations in mind, we argue that CVE practitioners should certainly not treat the resultant findings as being definitive. Arguably, the KTI team placed too much faith in their commissioned research.

Identifying the drivers – pull factors

In any case, even if discrimination, unemployment, poverty and other such ‘push’ factors are shown to drive VE in specific environments, it is increasingly recognised that an exclusive focus upon these variables is insufficient. As observed in USAID’s Guide to Drivers (2009: 11) document, ‘efforts to bring about significant reductions in such alleged “root causes” as high unemployment, pervasive poverty, systematic political exclusion, endemic corruption, and a lack of political and economic opportunities (to name but a few) will require large-scale investments, carried out through hard-to-implement and expensive programs sustained over long periods of time.’ In other words, there are few ‘quick wins’ achievable on the push side. In addition, it is often observed (e.g. USAID 2009: ii) that such push factors are common to many global regions, whereas VE remains comparatively rare. Thus, at best these drivers provide insufficient explanations for VE, and it is thus necessary to additionally consider what ‘pulls’ individuals towards violence.

With many interventions focusing upon push factors almost by default, the KTI team can be credited with placing an elevated focus upon those that ‘pull.’ However, the drivers identified by KTI diverge notably from those suggested in USAID’s Development Response (see Figure 1), and in particular it can be noted that the team neglected to focus on those that motivate at the individual level, rather than collectively.9 For instance, these include ‘selective incentives’ such as Al-Shabaab’s reported monthly salary of 50–150 USD (Hassan 2012: 18).10 Status-seeking is also commonly identified as a VE driver, with a former member claiming that ‘walking in the city with a gun as a member of al-Shabaab ensured everybody feared and respected you’ and that ‘girls also liked you’ (Hassan 2012: 19). Adventure-seeking, fear and revenge are also routinely identified as motivators in a range of cases (e.g. Kilcullen 2009: 40–41; Ribetti 2007).11 Having overlooked such factors during their initial research efforts, the KTI subsequently downplayed the relevance of these drivers as they designed their program and selected grants.

Such considerations have two main programmatic implications. Firstly, individual-level drivers tend to be particularly applicable to those directly engaged in the production of violence, rather than the supporters of these activities, as they are often contingent upon behaviours (Khalil 2014). Put another way, the benefits of material gain, elevated status, a sense of adventure, and so on, are not generally gained by those who merely remain on the sidelines. And, with these somewhat divergent motivations, this underlines the point discussed above that it is necessary to treat the perpetrators and supporters of violence as distinct (albeit overlapping) populations. Secondly, programs within the scope of CVE can be designed to counter at least some of these individual-level drivers if they are determined to be of key significance. For instance, if research reveals that revenge is an important motivator in a specific location, projects can be established to bridge the gap between religious, ethnic, and clan-based communities.

Identifying ‘vulnerable’ sub-populations

While research into such push and pull factors provides a suitable framework upon which to base CVE efforts, additional nuance is required in the contextual understanding as certain individuals are more susceptible than others to the appeal of VE. Amongst practitioners there is some resistance to the idea of identifying such ‘vulnerable’ individuals as this is suggestive of stereotyping at a time when it is increasingly recognised that terrorist profiles either do not exist or may never be identified (Horgan 2009: xxii; Sageman 2008: 17). However, we believe that this is misguided as the objective should not be to precisely identify individual ‘types,’ but more modestly to narrow targeting efforts on a probabilistic basis. Indeed, this logic was in any case applied by KTI through placing an emphasis upon Somalis and youth (taken to be those between 16 and 35) as specifically ‘at-risk.’ It was also applied in a geographical sense in that the program focus was upon specific coastal regions and Eastleigh’s environs, rather than Kenya in its entirety, as these locations were deemed to be highly susceptible.

There is some anecdotal evidence that teenagers (i.e. a tightening of the concept of ‘youth’), ex-convicts, members of specific clans, and other narrow sub-populations are specifically vulnerable to VE in Kenya. We argue that the KTI program would have benefited from additional research into this matter, and that a greater proportion of KTI grants should have targeted those subpopulations positively identified through comprehensive research as being specifically ‘vulnerable.’ This conclusion applies in particular given that VE is associated with a very small minority in the Kenyan context.12 In practice, however, our investigation revealed that a number of interventions inadvertently selected against the most at-risk within the broad categories of Somalis and youths. While it is not possible to provide additional details given the sensitive nature of this programming, put simply, those individuals most likely to follow the VE path were seemingly less likely to attend a range of KTI-sponsored events.

Determining the intervention logic

In considering the topic of intervention logic it should firstly be noted that CVE programming paradoxically cannot alone serve to counteract VE. For instance, it is beyond the scope of CVE to provide protection to individuals who are coerced into VE, or to attempt to cut the funding of VE entities in order to undermine their ability to offer material incentives for such acts. And, in a striking disconnect between policy and practice, neither is it possible in certain US CVE lines to provide training to officials in the security forces or to engage with religious institutions that may play a key role in either provoking or preventing VE.13 The former is of particular relevance in contemporary Kenya as many state officials seemingly remain unaware of the apparently counterproductive effects of their repressive practices. The simple retort to these observations is that CVE forms only one component of a wider response that also often involves law-enforcement, counterterrorism, security sector reform, development, and other such initiatives. Yet, as observed above, this riposte is only valid if these efforts can be adequately coordinated and deconflicted, and this is no trivial matter given the range of (often highly bureaucratic) agencies typically involved.

As also noted above, we collected considerable qualitative evidence to suggest that KTI’s efforts achieved a positive impact, i.e. in preventing certain individuals from following the VE path. However, we also argue that KTI could have enhanced this success through applying the Theories of Change (ToC) logic in order to maximize the likelihood that each separate grant contributed to the broader program goal, and this section serves to provide a brief introduction to this topic. ToC provide a map of the pathways from inputs (e.g. funds, time, etc.) through to impacts (i.e. a reduction in VE or support for such acts), involving the articulation of assumptions upon which such efforts are based. Certain KTI grants aimed to enhance livelihood opportunities through linking local youth and financial institutions, with the intention being that the former obtain loans or grants from the latter. Such efforts rested on assumptions that (inter alia) (a) youth did not already engage with such entities, (b) youth had the motivation/time to engage with these bodies, (c) financial institutions were willing/able to support youth, (d) youth are able to channel such resources to enhance their livelihood opportunities, (e) a lack of opportunities for youth drives VE (as discussed above), and (f) individuals specifically at-risk of being attracted to VE will partake in such initiatives (also discussed above). Ideally the resources would be available for implementers to research each of these assumptions, but the process of articulating such suppositions in itself commonly reveals potential weak links. And, this enables adjustments to be made to the program design at suitably early stages in the process.

Another example that perhaps even better demonstrates the role of ToC is offered by KTI efforts to combat post-traumatic stress disorder (PTSD) amongst all community members. These initiatives rested upon assumptions that (inter alia) (a) PTSD occurs in the community, (b) individuals are willing to partake in counselling despite the potential stigma, (c) it is possible to recruit suitably qualified/experienced counsellors, (d) such treatment is able to successfully counter PTSD, (e) PTSD drives VE, and (f) individuals specifically at-risk of VE are willing and able to partake in the counselling. While it is necessary to challenge all of these assumptions as a matter of due process, the fifth is perhaps of particular concern. A key KTI document maintained that former Al-Shabaab members are likely to reenlist if unable ‘to get appropriate psychological counselling to deal with their trauma,’ but this assertion seemingly lacked supporting evidence. Indeed, focussing on a selected sample of ‘global Salafi mujahedin,’ Marc Sageman (2004: 97) maintains that ‘there was no pattern of emotional trauma in their past, nor was there any evidence of pathological hatred or paranoia when the facts are analysed.’ While counselling of this nature would undoubtedly have a positive impact upon the community, clearly it will not serve to undermine VE if this disorder does not actually drive this violence. The wider point is that entire lines of programming may wholly fail to contribute to the broader KTI goal if only one assumption is misguided.

Mitigating negative effects

Of course, it is necessary to consider not only the positive effects of programs, but also potential negative consequences. For instance, grantees and other stakeholders may be subject to threats or actual targeting as a result of their association with external donors. Indeed, one grantee highlighted that local youth distributed leaflets to denounce their KTI-funded activities, and that these same individuals subsequently threatened staff members. While KTI offered to cancel this grant in order to resolve this issue, the grantee elected to continue with the project. The obvious recommendation is that such initiatives require robust systems of risk management, assessment and mitigation. In making this suggestion, however, it is necessary to recognise that CVE programs must also avoid the temptation to become overly risk-adverse as this will undoubtedly impinge upon their ability to achieve their program objectives. Put simply, CVE efforts inevitably entail elevated levels of risk.

A more specific recommendation is for donors (and particularly those from the US given that ill-feelings towards the pivotal Western nation are commonly pronounced) to be cautious with branding. KTI’s policy was flexible, allowing grantees to operate without the USAID logo if they deemed it to be unsuitable. Advocates of branding in this context argue that the logo serves to demonstrate the good-will of the US, and in doing so it may draw individuals away from the VE path. Yet, this argument may be outweighed, firstly, by the fact that those who seek to encourage VE can utilise this branding as evidence of external ‘meddling,’ and thus it may inadvertently serve as a rallying point for violence. Secondly, with opposition to Western influence being common in many regions, such branding may actually discourage those individuals who are part-way down the VE path from participating in CVE initiatives. And, thirdly, it likely increases the risk of grantees and other associates being threatened or physically targeted through advertising their links to an external entity. Put simply, in certain contexts the net effect of branding may actually be to undermine progress towards CVE aims, or to elevate the costs of such efforts, and thus selectivity is certainly required.

It is also necessary to reflect upon the normative nature of the VE/CVE concepts. The ‘extremism’ label (alongside ‘terrorism’ and other such tags) is routinely utilized to vilify opponents in an effort to isolate them from the wider populace.14 However, this label may instead serve to polarise communities by distinguishing between ‘extremists’ and (implicitly) ‘normal’ or ‘moderate’ individuals, and in doing so it may inadvertently cause some to shift towards the former camp. Such pejorative labels are of course associated with attempts to shape perceptions and disguise agendas, and the sense of being manipulated may in itself also discourage some from associating with initiatives that carry the CVE tag. In addition, the label may attract unwanted attention from unsavoury elements, thereby elevating risk for grantees and other program stakeholders. Indeed, presumably with this threat in mind, various KTI grantees elected to ‘tone-down’ the CVE language, instead portraying their projects as being focused upon, for instance, ‘peace-messaging’ and ‘coexistence.’ In a very different sense the label may also undermine interventions through implying that it is primarily or solely nonstate actors who are at fault for the violence (i.e. as ‘extremists’), thereby potentially steering practitioners away from pursuing necessary state reforms. Each of these possible consequences remain under-research (a matter that should be remedied as a priority), but all may substantially undermine CVE initiatives. This is certainly not to suggest that the CVE terminology should be abandoned, but rather that it may be necessary to adopt neutral language in certain contexts.

Conclusion

While this brief essay has covered substantial ground, from the program goal statements through to negative effects during implementation, it is worth concluding with two overarching points. Firstly, KTI programming decisions would have benefitted from additional research in order to provide a more thorough understanding of the VE context in Kenya. For instance, while the KTI team correctly placed an elevated focus upon drivers that ‘pull’ individuals to VE, they essentially neglected the role of material incentives, fear, revenge, status-seeking and other individual-level drivers. Similarly, the program team would have benefitted from a more precise understanding of which subsets of the populace (potentially including ex-convicts, members of specific clans, teenagers, etc.) were specifically vulnerable to the appeal of VE, particularly as comparatively few individuals are drawn to this violence in Kenya. More generally, CVE practitioners should also aim to become more conversant in the limitations of research and analysis in such environments. In particular, this applies to the reliability of data collected from informants, and regarding questions of causality, e.g. to consider whether a correlation between a lack of livelihoods opportunities and VE in a specific region may result from the former driving the latter, or vice versa, or whether an external factor is responsible for both.

The second conclusion is that the KTI implementers would have benefitted from additional top-level guidance from donors. For instance, they would have gained from direction with regard to the extent to which the CVE concept should be intended to target individuals who are directly involved in the creation of violence, versus those who only support such acts in cases such as Kenya. Additional guidance would also have been beneficial, for instance, with regard to circumstances in which practitioners may potentially undermine their own programs through the inappropriate application of branding, and in which the pejorative term ‘extremism’ should be pragmatically replaced by neutral terminology. Above all, we argue that USAID and other donors should offer suitable definitions of CVE, rather than leave practitioners to construe its meaning from the limited number of available definitions of VE.