Government through clanship: Governing Ethiopia’s Somali pastoralists through a community-based social protection programme

Drawing on an analysis of the implementation of Ethiopia’s Productive Safety Net Programme (PSNP) in the Somali periphery, we consider how the programme is promoted as an ‘innovative’ social protection programme that links food security with development projects. Underpinning its ‘innovative’ role is a community-based approach, focusing upon the institutions, values and capacities of a community. Taking the case of the nomadic pastoralists in Ethiopia’s Somali region, we consider the role of clans as the ‘dominant’ grassroot socio-political organizations. Our analysis, drawing on ethnographic fieldwork shows how in the implementation of PSNP the mobilization and (re)deployment of clanship values and rules create legible and governable Somali pastoral subjects. This is in line with the Ethiopian state’s conception of ‘improvement’ and ‘modern’ way of life based on sedentary-based development and governance. We illustrate how clan leaders unwittingly (re)organize their clan (leadership) values and capacities to support this project. We argue that clan-based implementation of PSNP has become an ‘effective’ mechanism of extending state power and governing nomadic pastoralists, leading to changes in relations of authority and in forms of (inter)subjectivity between pastoralists, their clan (leaders) and the state. Towards this end, we put forward the concept of ‘government through clanship’ to reflect the assemblage of these practices, processes and changes which would offer critical analytical insights into social policies claimed to be community-based.


Introduction
In 2005, the government of Ethiopia, in collaboration with donors, launched an 'innovative' social protection programme called the Productive Safety Net Programme (PSNP) for five million chronically food insecure farming highlanders. The aim was to close the consumption gap of food insecure households in the short-term, while lifting them permanently out of poverty in the longterm by linking social protection to development projects (FDRE, 2016). PSNP was launched in pastoral areas in 2008 and the programme now claims to service ten million beneficiaries. There are two main components of PSNP: 'public work' and 'direct support'. The latter constitutes unconditional cash and food transfers for twelve months for targeted households who have no adult able-bodied labour at all to participate in public work. This component reflects a smaller proportion of beneficiaries given the productive focus of PSNP through the public work component.
Public work constitutes participatory community-based activities. In pastoral areas this includes activities such as rangeland management, water point development, improvement of livestock health, and construction of social infrastructures, in which beneficiaries participate as a source of employment (FDRE, 2012). The aim of public work is to provide conditional cash and food for six months as a wage so that beneficiaries can meet their consumption needs in the short-term while they simultaneously enhance the communal resource base on which the production of and access to food, and/or livelihood security depend in the long-term (FDRE, 2006). Indeed, the underlying rationale of PSNP is that it should focus on productive investments and integration with broader rural livelihood opportunities. As such, more emphasis in terms of PSNP budget is allocated, and more technical and administrative personnel are involved in this component.
Underpinning PSNP's 'innovativeness' is its community-based approach, focusing upon the mobilization and enrolment of the institutions and values of community in its implementation. PSNP uses community-based targeting in which community members are actively involved in the selection of programme beneficiaries (FDRE, 2016). In terms of public work activities, the community participates in the identification, planning, monitoring and evaluation of public work sub-projects to ensure they are tailored to the prevailing livelihood (FDRE, 2016). All these community roles, in the nomadic (Somali) pastoral context, are vested in clans (FDRE, 2012(FDRE, , 2016 as the most 'dominant' grassroots form of socio-political organization. In this way, PSNP embodies the rationale that improvement in pastoralists' food security and livelihoods will be better achieved through active participation of local communities. The logic is presented as a win-win, where communities make their own choices and transform them into desired actions and outcomes based on their local realities while, in doing so, they are empowered to take (even) better actions and to counteract subordination and domination. This, in line with Li (2007), constitutes the 'will to improve' local communities from below by enhancing their capacities rather than imposing interventions from abovewhich draws from the contemporary discourse and practices of community-based social development promoted, in the global South, by NGOs and donors since the 1980s in response to perceived failed top-down approaches dominated by hierarchical and unequal power relations.
In this article we ask whether community (in this case clan)-based implementation of PSNP has become Ethiopian state's 'effective' technology of governing nomadic (Somali) pastoralists. Following Li (2007), we are not suggesting this is a hidden, cynical agenda concealed by the community/clan-based rationale of PSNP that most critics (e.g. Agrawal and Gibson, 1999;Leach et al., 1999;Mosse, 2001) would argue for. The limitation, critics claim, is the lack of 'genuine' engagement with community institutions and capacities for selfmanagement. While we do not discard these critiques, by drawing on the governmentality (through community) approach, we argue that 'genuine' engagement withrather than lack ofclan values and capacities in the implementation of PSNP becomes the 'effective' mechanism of governing (Somali) nomadic pastoralists. This constitutes, we argue, the Ethiopian government's 'will to improve' nomadic pastoralists. This is done by envisaging improvement in terms of a sedentary-based development and governance logic. As such, clan leaders/elders become complicit in modern state formation and/or the extension of state power over pastoralist subjects.
In line with McKee (2009), by emphasizing the empirical reality of PSNP and exploring its concrete governmental practices/effects, our arguments and findings offer insights for critical social policy analysis, including the need to go beyond the conventional dichotomy between community-based versus top-down approach or freedom versus domination in researching impacts of social policy.
Research setting: (Local) political economy, PSNP policy rationales, and fieldwork methodology PSNP marks a policy shift from decades of ad hoc relief food aid based on annual emergency appeals to a long-term and integrated approach to food security. Financed through donors' multi-year predictable and reliable resource transfers, PSNP is designed and evolved as an integral part of the Ethiopian government's rural development strategy (for a more extensive review see Lavers, 2019). PSNP is implemented in a way that 'prevents asset depletion at the household level and creates assets at the community level…addresses immediate human needs while it simultaneously encourages households to engage in production and investment and supports the rural transformation process' (FDRE, 2006: 1). In doing so, central to PSNP's aim is 'resilience to shocks and livelihoods enhanced, and food security and nutrition improved, for rural households vulnerable to food insecurity [so that] all Ethiopians enjoy social and economic well-being, security and social justice' (FDRE 2016, chapter 2, p. 3).
PSNP is implemented by existing or newly established Food Security directorates, committees, task forces, etc. within the relevant offices at all levels of government (e.g. within the Ministry of Agriculture at the federal level). The most important ones at the local level to facilitate the actual implementation process include: the Woreda 1 Food Security Task Force (WFSTF) which coordinates PSNP implementation under the oversight of the Woreda Council; the Kebele 2 Food Security Task Force (KFSTF) manages the implementation of PSNP at the kebele level under the oversight of the Kebele Cabinet; and the Community Food Security Task Force (CFSTF)composed of representatives of kebele administration and various categories of community (men, women, youth and elders)is responsible for PSNP beneficiary targeting. Depending on (regional) contexts, these PSNP implementation structures may take different forms and function differently.
National selection criteria for beneficiaries have been established. Eligible beneficiaries include chronically food insecure households who have faced continuous food shortages (three months of food gap or more per year) in the last three years; households who suddenly become more food insecure because of a severe loss of assets; and households without adequate family support and other means of social protection (FDRE, 2016). Locally relevant criteria, such as livestock and land holdings, and other sources of income, are used by CFSTF to select eligible households at the village/community level. Beneficiary targeting is both annual and periodic. The latter involves full (re)targeting of PSNP beneficiaries every three to five years. PSNP transfers/payments are made at the end of each month either in cash or food or both. Both public work and direct support beneficiaries receive 15kg cereal and 4kg pulses per person per month while cash transfer is calculated based on the costs of buying these food items.
While inclusion/exclusion errors are commonly reported in relation to local misinterpretation of the programme, political interference, elite capture and nepotism (Cochrane and Tamiru, 2016;Lavers et al., 2020), research findings on impacts of PSNP are still inconclusive. Many researchers (Béné et al., 2012;Berhane et al., 2014;Desalegn and Ali, 2018;Devereux et al., 2008) claim that PSNP improves household food consumption and nutrition outcomes. Thus, PSNP prevents distress selling of household assets to cover consumption gaps during shocks (Baye et al., 2014;Desalegn and Ali, 2018). However, many (Baye et al., 2014;Béné et al., 2012;Desalegn and Ali, 2018;Devereux et al., 2008;Sabates-Wheeler et al., 2021) argue that PSNP is not yet able to build significant livelihood resilience against the impacts of shocks, such as the drought in 2021 which, according to the World Food Programme, put nearly six million Ethiopians in need for food assistance. However, while PSNP may 'fail' in terms of its stated/intended objectives, as many researchers (Alene et al., 2021;Cochrane and Tamiru, 2016;Lavers et al., 2020) observe, it has become instrumental for achieving 'unintended' objectives of extending state-party power. This suggests the need for enhanced critical-empirical perspectives for understanding the actual governing practices and effects of PSNP.
The article is based on twelve months of fieldwork from 2017/18 by the first author in three pastoralist sub-kebeles, namely Qurenjale, Dhaladu, and Gobanti, together forming one kebele administration called Gobley in the Afdem woreda of Ethiopia's Somali region. The pastoral population of the subkebeles belongs to the Issa Somali clan family and depends on nomadic pastoralism as a source of livelihood and way of life. The Issa clan is segmented into twelve major sub-clans each of which further segmented down to mag-paying (blood compensation-paying) group, 3 which is the most localized and functional one. The defining elements of mag-paying lineage solidarity include, among others, that its male members contribute to blood-debts (mag) payment for the murder of member/s of another mag-paying lineage group, and collectively develop and exploit an area of local clan resource, mainly ella (a traditional hand-dug water point) over which they have generational ownership and use rights. Different mag-paying lineage groups live together in each sub-kebele while one of them constitutes the majority, owning the ella in the sub-kebele. Collective identity based on (sub-)clan lineage is important and reciting the genealogy of one's own (sub-)clan is important for an individual as a committed (sub-)clan member. On first meeting another Issa-Somali person, next to exchanging simple greetings, the first topic of discussion is often enquiring about each other's (sub-)clan.
The lead author conducted 26 unstructured in-depth interviews with pastoralist PSNP beneficiaries composed of men and women, youth and elderly, and clan leaders/elders. Eight in-depth expert interviews were conducted with bureaucrats at (sub-)kebele, woreda and region levels, selected on the basis of their role in PSNP implementation. One intensive focus group discussion was held with a group of pastoral clan leaders and elders. Participant observation was conducted in formal and informal meetings, PSNP retargeting, public work activities, monthly PSNP cash/food transfers, and clan leaders/elders' meetings and deliberations on issues, such as magpaying (blood compensation) arbitrations between sub-clans. Research material was also collected through informal interactions and conversations, such as gossips, rumours and pastoralists' metaphors. Throughout the research processes, we adhered to ethical procedures of data collection, including explaining the research objectives and obtaining informed consent of research participants. We systematically implemented a strategy for securing confidentiality and anonymity in line with the data management plans of our science groups.

Government through community: Insights for critical social policy analysis
Community takes different forms depending on its organizing principle or purposes. Populations are reimagined as forming many 'natural' communities, such as, religious, territorial, lifestyle-based and clan-based (Li, 2007). Community is also 'a way of making collective existence intelligible and calculable' for governmental interventions (Li, 2007: 232), such as PSNP. Community is often discursively (re)constituted by outsiders, who are positioned within a specific regime of knowledge and power, in the way it fits to their project models (Leach et al., 1999). As Rose (1999: 172-73) writes, it is through 'the political objectification and instrumentalization of this com-munity…that government is to be re-invented'government through community. A wider governmental power, Foucault (1980: 72-73) notes, is transmitted faster through communities and civil societies beyond the state apparatus as it makes use of the agency and/or capacities of the governed themselves, 'yet often sustain the state more effectively than its own institutions, enlarging and maximising its effectiveness'. Li's (2007) exegesis of Rose (1999) involves the notion of 'government through community' in (social) development in the global South which is particularly relevant for our case/analysis. Governing through community in development requires that community first be rendered legible in view of the (rationalities of) specific programme of intervention; it should be identified, (re)defined, visible and formalized, and only then will its organizational and social resources and capacities be enrolled (Li, 2007). Li (2007) observes that local communities become the principal units of intervention in contemporary social development policies because it is reasoned that they have 'natural' capacities for democratic and egalitarian selfmanagement/improvement -'the secret to the good life'. However, this way of (re)constituting (government through) community is problematic because there are complex (unequal) power relations and contestations among different social groups on whom governmental interventions may have differential impacts (McKee, 2009). There is, again, a paradox here, that on the one hand local communities, such as clans, have the capacities for good life, while the good is yet to be defined and 'experts must intervene to secure that goodness' on the other hand (Li, 2007: 232). In these ways, power is operationalized 'by shaping and mobilizing particular subjectivities' through the production and deployment of particular discourse/ knowledge upon subjects (communities) who have, of course, the capacity for resistance (McKee, 2009). However, resistance is not absolute rejection or liberation. Rather, it is about bringing forward alternatives, so that there is always a reconciliation between the interest of the governed subjects and the rules of the governor (McKee, 2009). In this way, following McKee (2009), we argue that government through community could offer conceptual insights for critical analysis and understanding of the contemporary social policies which claimed to be community-based.

Clanship and Somali society
Clanshipmore than residenceis the basis for socio-political organization and mobilization of Somali society (Lewis, 1999). Xeer is defined as the norms and rules through which clan members are governed (Lewis, 1999). Customary leadership is at the centre of Somali clanship. Uggas (clan chiefdom) is predominantly a nominal authority derived based on genealogical position, perceived wisdom, religious knowledge, seniority or elderliness, and other personal attributes. That said, actual leadership power at the local level, or the lowest clan segmentary lineage, is held by (sub-)clan elders who have similar sources of authority as the Uggas (Lewis, 1999) to interpret and enforce xeer (Gaas, 2019).
Clan elders need the support and trust of their clan members to maintain their status while they, reciprocally, protect the interest of their clan members (Collins, 2004). Clans are social constructions 'enacted by people using them for the purpose of identification, making claims, dividing or competing for power [while they are not] the most dominant at all times or absent in Somali society' (Gaas, 2019: 2). Clanship is dynamic and flexible so that at least its characteristics and functions change in accordance with changing political and economic contexts (Kapteijns, 2004). (Somali) clanship is constituted by different social groups, based on gender, age and class, who have different capacities to access and exercise power (Devereux, 2010).
Clanship is constituted by rules and sanctions of collective management, access and use of rangeland resources and mutual help with which nomadic pastoralism is supported. In view of Li's (2007) framework above, all these features (of self-management) constitute Somali clanship as the relevant site for governmental interventions through community-based social development programmes. Hence, in the context of Somali society, to 'succeed', community-based social development interventions, such as PSNP, need to be framed in terms of features of clanship (values, rules and leadership) which then to be mobilized and enrolled.
Critics (Gaas, 2019;Kapteijns, 2004) suggest that other factors, beyond agnatic ties (e.g. political/economic alliance), also matter in the construction of clan identity and/or in the socio-political organization of Somali society. While we agree with these views in general, clanship is still the dominant principle of socio-political organization of Somali society as apparent in the case of our research area. (Government through) clanship is relevant because recent interventions in development (PSNP) and governance (e.g. 'decentralized' ethnic federalism) by the Ethiopian state have drawn on the discourse of clanship that, as Kapteijns (2004) observes, continues to dominate thinking about Somali society.
In this way, clan represents the ideal model of Li's (2007) community, as viewed by policymakers, as an authentic and already existing selfmanaging entity making it legible for 'effective' government. This form of government can be conceptualized as (the paradox of) government through community (Li, 2007). In the context of nomadic pastoral Somali society, the clan is viewed by outsiders, such as those who planned the interventions of PSNP, as the only or most dominant organizing principle upon which the implementation of PSNP can rely for effective government of pastoralists.
Adopting Li's (2007) conceptualization helps us to examine the actual governing practices and effects of PSNP interventions through clanship more critically and empirically beyond its discursive rationalities and intentions. This suggests the need for moving away from a linear and technical approach to social policy which depoliticizes the policy process. This requires, in our case, attention to political-economic relations between the sedentary Ethiopian state and the clan-based nomadic Somali pastoralists both historically and today under the ethnic-based federal policy which we briefly outline next.

The ethnic federal Ethiopian state -Somalis (clanship) relations
The Ethiopian state expanded from its highland core to the Somali lowland peripheries in the late 19th century to subjugate nomadic pastoralists (Markakis, 2011). State-building in the Somali periphery has always been contested and complicated up to present day. The (Imperial and Socialist) Ethiopian regimes viewed the (nomadic pastoral) Somalis as 'uncivilized', troublemakers, alien to Ethiopian nationhood and incapable of organizing and running formal state institutions (Hagmann, 2005). Hence, state-backed highlanders settled in the Somali lowlands as trustees/administrators (Hagmann and Korf, 2012) for a mission of dar-ager makinat (in Amharic, to civilize the peripheries). To this end, the state, as Hagmann and Khalif (2008: 42) write, 'employed the stick more often than the carrot…'. Sedentarization had been promoted as one of the main sticks of extending state power and turning 'trouble-making' nomadic pastoralists into governable and civilized Ethiopian subjects (Korf, Hagmann and Emmenegger, 2015). Devereux (2010) observes that the Somalis have tended to view state expansion and accompanying modes of governance as colonial occupation, worse than political and economic domination. It is experienced as an imposition of highland Ethiopian values on their way of life (Devereux, 2010). Hence, by organizing themselves through clans, Somali pastoralists have continually resisted state expansion and administration (Hagmann, 2005).
It has been argued that the introduction of ethnic-based federal policy in 1991 has improved Ethiopian state -Somalis relation (Hagmann, 2005;Markakis, 2011;Samatar, 2004). Ethnic-based federal policy instituted nine 'self-governing' regional states corresponding to major ethnic groups to grant them the right of self-rule, but critics (e.g. Markakis, 2011;Samatar, 2004) argue that this is no more than a new method of (indirect) rule from the centre. Ethiopia's Somali region is one of the regional states to incorporate Ethiopian Somalis and their clan-based governance structures into the Ethiopian state through 'self-rule' rather than coercion (Hagmann, 2005). At the regional, zonal and woreda levels (sub-)clan elders are selected to form Guurti (a council of clan leaders/elders) that 'assists/advises' government officials in daily political affairs, (Hagmann, 2005). Hagmann and Khalif (2008: 31) observe that this 'was primarily motivated by the Ethiopian state's need to extend government control into rural pastoral areas [and] government officials rightly conceived of elders as bearers of intimate knowledge of the clan lineages they represent and of being capable of mobilizing their communities'. Guurti members are incentivised by the salary they are paid by the government like any civil servant (Hagmann, 2005).
On the other hand, clan leaders/elders have now committed themselves to participate in formal government affairs and represent their kinship interests within the formal government administration and party (Hagmann, 2005). Hence, instead of viewing clanship and government as irreconcilable, the current ethnic federal Ethiopian state co-opts Somali clans to govern nomadic pastoralists more 'effectively'. However, 'effective' government of pastoralists, for the state, continues to be defined in terms of sedentary cropfarming as this, facilitates the political and administrative management of pastoralists, while nomadic pastoralism is portrayed as the other side of civilization and development (Korf, Hagmann and Emmenegger, 2015).

PSNP beneficiary targeting and rendering nomadic Somali pastoralists legible/governable through clanship
Given that the key principle of PSNP is the active participation of local communities who can plan for and have control over the implementation process, it is fundamental that locally relevant communities are identified as a unit of intervention. The Issa/Somali nomadic pastoralists were (re)imagined as forming different sub-clans, so that they could be mobilized as groups for 'participatory' PSNP beneficiary targeting and public work implementation. While very little was known by the authorities about Issa/Somali pastoral populations, they identified sub-clans with their respective elders/leaders who were called upon for consultation about launching PSNP. An elder who represented his sub-clan during the launching of PSNP recounts his experience as: Three elders including me [each representing one's own sub-clan] and subkebele officials were called by the woreda authorities for a meeting about starting off PSNP. Given the fact that both pastoralists and the government trust us and we know the details of each pastoralist household's situation better than sub-kebele officials, the woreda authorities delegated us to carry out the first targeting process. The woreda authorities did not give us criteria for targeting except telling us the quota allocated to our sub-kebele is 800 beneficiaries, of which only 60 should be targeted as direct support beneficiaries while the rest should be targeted as public work beneficiaries. We then invited three other elders and we together selected PSNP beneficiaries based on our detailed knowledge of their wealth status without their direct participation. The subkebele administrator was there with us just for approval while there was another person [PSNP officer from the woreda] just to record lists of beneficiaries we selected.
There has been competition between sub-clan units (in this case between their respective elders/leaders) for inclusion in the programme, including 'successfully' appealing to the woreda authorities when some of them were not called upon for the consultation. As a result, several sub-clan units (elders/ leaders) were identified and legitimized to organize beneficiary targeting themselves. According to the PSNP implementation manual, targeting is formally expected to be carried out by CFSTF. In our research area, however, CFSTF and other kebele level PSNP implementation structures have not yet played any role if they were formed at all, while (sub-)kebele administrations were weak and/or not functional by then. Woreda authorities granted legitimate targeting power to sub-clan elders/leaders to draw on their rich experience of organizing traditional mutual support practices, such as sedekah (alms giving). Then, elders 'successfully' targeted the neediest pastoral households while the allocated PSNP budget was not enough to target all eligible households during the first five years. Elders sometimes targeted pastoralists from their clan in other villages at the expense of eligible pastoralists from other clans in their respective villages.
In connection with this, a conflict between woreda officials and clan leaders/elders emerged around the nature of beneficiary targeting. Officials ordered clan leaders/elders to implement the 'full family targeting' principle of PSNP in which all household members must be targeted so that a 'sufficient' level of transfer is provided for the household to quickly acquire productive assets such as livestock. The intention is that this will better support beneficiaries to graduate from the programme. Clan leaders/elders resisted this because it is not acceptable, in view of their resource sharing norms, to create differences between clan members by targeting some households while completely excluding others. Yet, as one pastoralist clan leader told us, 'wealth difference among our pastoralists is not as such important; the one who is rich today may be poor tomorrow because of recurring droughts'. So, without getting into direct confrontation with the officials, clan leaders/elders employed a 'false compliance' strategya form of resistance similar to what Scott (1990) called 'hidden and open transcripts'to undertake targeting in line with their own resource sharing norms and wealth analysis. As a result, this reduced the level of transfers that could have been provided to the targeted beneficiaries so that they could reach more households. However, they documented and reported fewer households to the officials, as though they had applied 'full family targeting'.
In practice, the strategy of the clan leaders/elders brought more households under the in/direct governing role of PSNP. The budget of PSNP has, however, gradually increased and during our fieldwork all households, regardless of wealth status, were officially targeted, documented and reported to the woreda officials. We note that a household cap of targeting five persons per household has been introduced nationally during phase 4 of the PSNP (from 2015 to 2020). In doing so, the most evasive and nomadic Issa/Somali pastoralists became visible and legible to authorities at two interrelated levels: sub-clan and household/individual. Legibility at the sub-clan level is achieved as pastoralists were categorized into different sub-clan units and at the household/individual level legibility is achieved through beneficiary targeting by clan elders/leaders who applied their insider knowledge to reveal the socioeconomic characteristics of their clan members.
The governmental or in this case programmatic legibility of pastoralists is more than just making pastoralists visible for direct political control. Legibility accompanies the (re)definition of pastoralists as subjects of government-driven development (PSNP) that 'unintendedly' tangles with the sedentary-based development and governance vision of the Ethiopian state. All these were (unwittingly) facilitated by clan leaders/elders. Some of these clan leaders/elders were already co-opted, before PSNP, into the woreda administration as Guurti members. Their current co-option into the management of PSNP is related to the 'prospect' of influencing PSNP beneficiary targeting in favour of their clan interest, such as targeting more of their clan members, for which clans competed. It is also related to personal interests, such as accessing exemption from public work labour contribution as a compensation for their services. To continue accessing these benefits, clan leaders/ elders cooperate with the woreda authorities in instrumentalizing PSNP for the implementation of other coercive policies, such as sedentarization. By being co-opted into the management of PSNP, clan leaders/elders are also co-opted into the (local) state, both of which reinforce each other.

Somali clan leaders, and the enrolment and optimization of clan values and rules in public works
Public work is based on the formation of work teams. According to the implementation manual (FDRE, 2016, chapter 8, p.7), work 'teams can be formed by grouping selected beneficiaries who reside in the same neighbourhood…or on the basis of traditional (e.g. clan) groupings…to ensure cooperation within teams, such as covering each other's tasks willingly when some are unable to attend, monitoring each other's attendance…'. Public work beneficiaries, in our research area, (re)organized themselves into work teams based on their mag-paying lineage organization, constituting a group of pastoralists who collectively develop and use an ella. While members of the same mag-paying lineage group may live in different locations, it is those members who usually move and/or establish temporary hamlets together who (re)organized themselves as a public work team.
Public work teams draw on their clanship norms and sanction mechanisms in managing their clan affairs, such as labour cooperation in livestock production and in management and use of rangeland resources, for the organization and accomplishment of public work sub-projects. Overseen by their respective clan leaders/elders, absentees are punished within their clanship sanction framework, rather than within the PSNP official sanction mechanisms. Work team members view their mutual commitment to the accomplishment of some public work sub-projects, such as improving their ella, in terms of augmenting the resource base of their lineage group. An adult pastoralist man explained: We could have reported absentees to the kebele or woreda authorities for punishment [in terms of denying PSNP transfers], but we do not do that because we do not want to see our relatives get punished. They will rather compensate their absenteeism in terms of contributing more labour in the next public work activities or in terms of searching better grazing land to which our [team members'] livestock will move.
While planning and implementation of public work sub-projects are backed by implementation manuals or directives from woreda authorities, a pastoralist clan leader stated that: we mobilize pastoralists to identify both the location and types of public work activities in favour of our [pastoralists'] interest as much as we can. Even when the woreda fails to send us directives, we still plan, make the woreda authorities aware of our plan, execute our plan, report our activities to the woreda to claim our PSNP transfers while we fear that the woreda authorities might intentionally fail to send us directives and use our failure to do public works as a pretext to deny our PSNP transfers.
However, there are directives or 'public work' sub-projects, that clan leaders and/or their work teams are expected to directly implement without modification or avoidance. One example is the clearing and preparing of rangeland for collective crop-farming. Here, participating work team members would share produce (which is not officially stipulated in the implementation manual). Another example is the construction of schools to expand education. Such public work sub-projects are promoted by state authorities to push pastoralists toward a settled crop-farming and use of modern social services -(woreda) state authorities' criteria of improvement.
Clan elders/leaders are 'autonomized' to 'improve' their clan members, but in line with the woreda/state authorities' (re)definition of the official programmatic goals of PSNP. The clan elders/leaders, for instance, took up the above directive on 'collective crop-farming' and mobilized their respective public work teams to collectively clear and prepare land for crop-farming. However, clanship rules of collective/communal use and management of rangeland resources, and resource sharing norms, are not applicable in all contexts of pastoral production and ownership of resources. As such, team members resisted collective production and sharing of crops through the 'collective crop-farming' public work sub-project. This is because it is not the norm to collectively own/produce and use resources such as crops and livestock. As one pastoralist man explained: 'I can share food [including PSNP transfers] with a needy household, and I can even give a cow [to the household], but how could I cultivate crops or raise livestock together and [own and] share with such and other households who may be lazy'. Because of this, clan leaders/elders had to substitute the woreda authorities' directive on collectively cultivating the prepared land and sharing produce. They instead subdivided the prepared land among the team members for private cultivation. To enhance their capacities for 'improving' pastoralists (in state authorities' criteria), selected clan elders/leaders from the woreda are sent to the Somali region capital city twice a year for, in the words of the woreda administrator, 'awareness creation'. They are incentivized with per diems for this 'awareness creation' travel and per diems from their respective (sub-)kebele PSNP budget for (re)mobilizing PSNP for the way it will 'improve' their clan members based on the awareness they acquire.
The irony is, however, that the role of clan leaders in managing PSNP has gradually reduced in favour of (sub-)kebele authorities. This is because, first, a pastoralist clan elder argues, 'we already did the initial groundwork of mobilizing our clan members who did not trust government officials when PSNP started; this has now 'improved' and (sub-)kebele officials do not have any reason to harm our pastoralists by mismanaging PSNP, but we still intervene if they do'. Second, as a pastoralist sub-kebele deputy administrator claims, the 'problematic' clannish tendencies of clan leaders/elders in (mis)managing PSNP have lessened as (sub-) kebele administration has now been strengthened to manage PSNP 'appropriately'.
Nevertheless, selected clan leaders/elders continue to act as 'middlemen' between pastoralists and (sub-)kebele administration more broadly. Pastoralists appeal to these clan leaders/elders against mistreatment, such as exclusion from PSNP, by (sub-)kebele authorities while the latter appeal to clan elders, instead of taking actions independently, against pastoralists' failures to obey government rules. (Sub-)kebele authorities seek the support of clan leaders/elders in collecting/deducting different forms of levies or taxes from pastoralists' own PSNP transfers.
In these ways, PSNP has 'unintendedly' enhanced the expansion of the Ethiopian state in the Somali periphery, not only as social protection provider, but also as tax collector. Pastoralists have come under regular state surveillance because of their legibility and visibility already achieved through PSNP targeting and documentation. Clanship values and leadership have been (re)organized to serve broader governmental power beyond implementing PSNP for its own sake, including the use of PSNP finance to incentivize clan leaders/ elders being co-opted into (local) state bureaucracy.

Competing realities and contradictions of government through clanship
In line with Kapteijns' (2004) critique of clanship, government through clanship in the context of PSNP has been based on oversimplified representations and understanding of Somali (pastoral) society just in terms of clanship. This conceals the actual dynamics and complexity of a society based on other dimensions, such as age, class and gender. Class and gender are important elements of social stratification and sociopolitical organization of Somali society (cf. Kapteijns, 2004). Hence analysis and understanding of Somali society and policy interventions just in terms of clanship would be misleading. The case of gender is illustrative because, from the perspective of state authorities, clanship is identified as the key institution for PSNP intervention which has mainly a male-dominated ideology and practice in Somali society. As we observed on the ground, it runs contrary to the participatory principle of PSNP to actively involve and empower women Underlying the contradictions of government through clanship is the notion held by government authorities that clans are authentic pastoral community organizations with the capacity for managing their affairs in development. However, development is yet to be defined by technocrats/authorities and the capacity of clan leaders needs to be optimized as such. Development, for local state authorities, is oriented towards settled crop-farming and use of modern social servicesthe kind of 'development' aspirations clan leaders/elders (and pastoralists) are believed to be lacking. Hence, the woreda administrator explains: every six months we send ten selected clan elders to Jigjiga [i.e. Somali region capital city] for awareness creation about the importance of a transition into agropastoralism and sedentarization to lead a modern and good life, so that elders, in turn, would inculcate their respective clan members with this idea and we have gotten good results in this regard.
The contradictions of government through clanship can be also observed within the broader context of the woreda administration/governance system and practices that have a direct impact on PSNP. The woreda administrator, who is from another (non-Issa) Somali clan, from another woreda and directly appointed by the regional government under the Dib-Curaasho policy (see below), claims: We closely work with local clan elders. We have one [salaried] clan elder in our office to advise us on matters regarding the local communities…[But] we cannot fully depend on the 400 years old traditional xeer to administer the population. Previously, woreda officials [who used to be from the local Issa clan communities before Dib-Curaasho] were powerless compared to clan elders who used to sit just under shades of trees to pass verdict even on the woreda administrator just for carrying out his government duties. This is not acceptable now.
In this regard, Somali regional state administrative policy called Dib-Curaasho (meaning new approach in Somali) is instructive. Dib-Curaasho is a rotational political (re)appointment of woreda and zonal administrators and PSNP officers away from areas of their own (sub-)clan to decrease the influence of informal clan-based networks and hence nepotism/corruption. However, this led to a widely held view locally that Dib-Curaasho has brought woreda officials who have treated local communities coercively and embezzled some of the PSNP budget because they lack empathy for local communities to which they do not belong.

Discussion and conclusion
This article shows how the implementation of PSNP has (re)constituted, mobilized and enrolled clanship values and rules to enhance the practice of (self-)government of Somali nomadic pastoralists. We call this government through clanship. This has enabled the Ethiopian state to extend its modern governmental powercouched in 'a civilizing' mission based on sedentarization and expansion of modern social servicesinto the Somali periphery more than its previous top-down development programmes had ever done.
PSNP has achieved government through clanship, first, by rendering evasive Somali nomadic pastoralists legible and visible, and by (re)imagining them as forming different self-governing sub-clan units. Beneficiary targeting has further enhanced the legibility and visibility of pastoralists as their socioeconomic characteristics (used as targeting criteria) are revealed, documented and reported to government authorities. This subjected Somali pastoralists to what Dean (2010) calls 'forms of visibility' that make them legible for further interventions both within and beyond PSNP.
Second, PSNP has achieved government through clanship by mobilizing, optimizing and enrolling clanship values and norms, mainly those around social contracts between clan leaders and pastoralists, in the implementation of its public work sub-projects. In this regard, PSNP's governmental techniques include: the mobilization and application of clanship rules and sanctions to harness pastoralists' commitment to public work participation; the enhancement of the 'autonomization' and capacities of clan leaders/elders to facilitate this. In doing so, in line with Mamdani's (1996) observation on colonial administration, clan leaders/elders have consolidated 'non-customary powers' over their clansmen subjects. This has coincided with and enhanced the consolidation of non-customary state administration, (sub-)kebele, run by non-customary pastoralist authorities who have now taken over PSNP implementation roles from clan leaders/elders as the desired modern way of governing pastoralists.
Hence, there is a paradox in 'government through clanship' in PSNP: on the one hand, clanship is constituted as a 'naturally' existing authentic way of organizing and mobilizing Somali pastoralists in development and governance from below. On the other hand, clanship is viewed as traditional and hence needing to be optimized so that its roles can be (re)framed in terms of Ethiopian state's hegemonic sedentary-based (pastoral) development and governance paradigm. This hegemonic political economy ideology of the Ethiopian state itself has 'unintentionally' shaped the implementation of PSNP. Yet, at a policy level PSNP is 'depoliticized' in the sense that it claims to be responsive to local realities and needs.
However, pastoralists challenged the above trend, including clan leaders/ elders' co-option and complicity. Pastoralists resisted the instrumentalization of PSNP, for example, by openly and 'successfully' resisting collective cropfarming that local state authorities, in collaboration with clan leaders/elders, imposed as a new public work sub-project. From the perspective of pastoralists, settled crop-farming is not a viable economic practice given climatic/ ecological volatility in the area. Pastoralists also resisted PSNP's official goals and principles, such as 'full family targeting', because this contradicts their clan-based norms of resource sharing, without offending state authorities. As a result, pastoralists have become sceptical about the custodianship of co-opted clan elders/leaders. This is expressed through pastoralists' popular metaphor that 'they [clan leaders/elders] sell us just for khat'. This phrase suggests that by being co-opted by the local state and, in exchange, receiving money (i.e. per diem) from woreda authorities to spend on khat, many clan leaders have prioritized their own personal interests at the expense of the interests of their fellow pastoralists.
Yet, clanship continues to be relevant for pastoralists to mobilize collective agency to resist (or optimize their relations with) the state. For instance, both clan leaders/elders and their clansmen collaborate to ensure the political representation of their clan within government administration, (sub-)kebele. Paradoxically, in doing so, they make themselves subject to state administration at the expense of their clan governance. Hence, we concur with Odysseos's (2011) notion of 'governing dissent' that by exercising their collective clan agency in the framework of formal state governance, pastoralists unwittingly guide their conduct in ways that reinforce the very (inter)subjectivities they seek to resist This by itself is a form of resistance because resistance should not be 'conceived in terms of [absolute] liberation from an oppressor; rather as an invention of alternatives to current governing practices' (McKee, 2009: 471).
Hence, clanship is, we argue, still both a relevant, dominant organizing principle and a form of collective resistance in Somali society, though it is not the only one and it is adaptable especially in the context of political economy changes, such as the introduction of ethnic federal policy and PSNP. Clanship has become important for the Ethiopian government as well, in 'effectively' governing nomadic Somali pastoralists from below (or through resistance) via PSNP. 'Effective' government of pastoralists, in this regard, involves the expansion of sedentary-based development and noncustomary administrative structures of the state in the Somali periphery. While these are unexpected outcomes of PSNP and/or beyond its explicit policy intentions/goals, they reflect the Ethiopian government's hegemonic development and governance paradigm. In this way, PSNP has helped the government to achieve its strategic political-economic goal that many of its well-intended and planned policies or state-building projects were unable to achieve in the past. Even the actual implementation of PSNP has been shaped and instrumentalised by technocratic Somalis within the local state because they believe this to be strategically important in the interest of pastoralists themselves.
By drawing on the governmentality (through community) perspective that to govern is not to dominate/harm others, but to enhance and direct their capacity for action, and based on our findings, we conclude that government through clanship by PSNP is constituted by the Ethiopian government's 'will to improve' its nomadic pastoral subjects, defined as sedentary-based development and non-customary governance that government through clanship via PSNP 'unexpectedly' achieved. Finally, it should be noted that the Ethiopian state acknowledged clanship governance not for its own sake, rather, by optimizing/instrumentalizing it through PSNP, as a technology of state's (non-customary) power.
In this way, our conclusion offers important insights for critical social policy analysis. First, social policy, implicitly or explicitly, deploys (political) power, so that the political economy and dominant ideology (in our case, the sedentary metaphysics of the Ethiopian state) matters in shaping the actual implementation and outcomes on the ground. Second, contemporary social policies which claim to be community-based may be shaped by the dominant ideology or other factors to bring about contradictory effects that could trigger resistance at the receiving end of the policy. Third, although the governed subjects have the power/capacity for resistance against governor, different sections of the population (e.g. women) do not access and exercise this power equally, so that they are impacted differently.