Building Local Governance : Participation and Elite Capture in Slum-upgrading in Kenya

Community and democratic participation are still an essential component of current mainstream development interventions. However, elite capture seriously undermines the outcomes of development projects. This article analyses the effects on (in)equality of the implementation of policies that are technically participatory, in the context of an internationally-funded urban development programme in Nairobi, which was implemented in the aftermath of the post-election violence of 2007–2008. Ethnographic data reveal how the institutionalization of pre-existing power imbalances between landlords and tenants is accomplished through the creation of structures of community governance and ‘participatory enumeration’. The article concludes that without the resources to challenge powerful interests within the settlement, the programme is likely to worsen the condition of a large section of the residents. In the context of strong pre-existing inequalities and conflict, participation needs careful management and firm external agency to achieve genuine social transformation.


INTRODUCTION
Community and democratic participation remain essential components of mainstream development interventions. Development agencies need community partners in order to implement their programmes, and therefore 'build' communities and their governance structures. In Nairobi, informal settlements host heterogeneous and fragmented residents who are socially divided into two main groups: landlords and tenants. Focusing on an internationally-funded urban development programme, this article analyses the effects on (in)equality of the implementation of policies that are technically participatory.
Much research in the area of urban development studies has shown how state-driven regeneration programmes have resulted in displacement and gentrification (Bassett et al., 2003;Campbell, 1990). In contrast, other studies have shown how communities self-organize to demand public services, and have praised the collective action of disadvantaged urban communities in pursuit of their own development (Appadurai, 2001;D'Cruz et al., 2009;Mitlin, 2008;Weru, 2004). Despite their different approaches and conclusions, these two bodies of work share an exploration of the relationships between 'the state' (or other development agencies) and 'communities'. However, contributors to this debate seldom problematize the community itself, despite the fact that it is widely acknowledged in the literature that slums are some of the most unequal settlements in the world.
Scholars have long observed the marginality of poor urban tenants, their relationship with landlords, and their exclusion from local organizations (Amis, 1984;Gilbert and Varley, 1990;Nelson, 1979;Rakodi, 1995b). Research has shown how community leaders in low-income urban areas do not represent all residents and how the interests of women are frequently ignored (Ward and Chant, 1987). There is also significant work exploring the diversity of livelihood strategies among slum dwellers (Beall, 2002;Rakodi, 1995a;Rakodi and Lloyd-Jones, 2002). More interestingly, some recent work (Dill, 2009;Lemanski, 2008;de Wit and Berner, 2009;Zérah, 2009) has analysed the impact of residents' social organization and internal community dynamics on participatory projects in urban areas. However, there is still insufficient micro-level analysis of these internal community processes in relation to slum-upgrading outcomes. There is, of course, much significant work questioning the notion of community in development at a more general level (e.g. Mohan and Stokke, 2000), and with regard to rural participatory projects in particular (e.g. Chhotray, 2004;Mosse, 2001;Platteau, 2004). This article draws on the latter literature to examine how pre-existing inequalities affect the implementation (and, by implication, outcome) of an urban development programme embedded in a 'participatory' framework. It looks inside the 'black box' of 'community', in order to analyse, firstly, the process whereby structures of community governance are created and, secondly, the manner in which 'participatory enumeration' 1 is carried out in one slum-upgrading programme. It also explores how the wider political climate -specifically post-election violence -as well as limited government capacity affect the way participation is managed.
After a selective review of the relevant literature on the concept and practice of 'community participation' and 'elite capture', this article will introduce the history, context and social organization of the settlement under Participation and Elite Capture in Kenya 259 study. A brief presentation of the programme and the actors involved is followed by an ethnographic account of the process through which structures of community governance were created and 'participatory enumeration' was carried out. These accounts provide the basis for analysing the progressive process of exclusion of certain sections of residents, and for a critical examination of the impact of pre-existing social inequality on the participatory project. 2

COMMUNITY PARTICIPATION IN CURRENT DEVELOPMENT DISCOURSES AND PRACTICES
In the 1980s, the failure of large-scale, state-driven and top-down approaches -which ignored the priorities and needs of the poor -opened a debate on participatory development with the idea of 'putting the last first' in the planning of development interventions (Chambers, 1983). The key idea was to learn from the poor, and to understand the complexity of their social reality in order to design more appropriate programmes together with the beneficiaries. The concept of participation entered mainstream development, defining what became critically termed a 'new orthodoxy' in the 1990s (Gardner and Lewis, 1996). 'Participation' remains at the heart of development discourses and practices today.
With the passage of time, an increasing body of work started to show how these participatory approaches and techniques could easily turn into another form of domination (Woost, 1997). According to Mosse (2005: 96), these techniques have become rituals to transform people into beneficiaries of development interventions. Moreover, participatory projects have unwittingly built upon pre-existing power structures, reinforcing them (Chhotray, 2004;Mosse, 2005) to the advantage of the 'learning elites' (Wilson, 2006). These learning elites are formed by local people who have learnt how to manage the discourse of participation and its language, and are able to exploit these skills to gain (or maintain) privileged access to development resources. They present themselves as representatives of the community, and become intermediaries between project officials and beneficiaries. Therefore, several studies have argued that participatory projects are vulnerable to 'elite capture' (Platteau, 2004) and create parallel structures that detract from the democratic process and public scrutiny (Green, 2000(Green, , 2002.
A common argument is that in communities with 'serious power imbalances . . . the poor are heavily dependent on vertical links with local elites, [therefore] it is difficult to form the horizontal associations necessary for organising collective action for the common good' (Das Gupta et al., 2004: 28). When collective action does take place, elites strongly oppose it or capture the benefits (Bardhan and Mookherjee, 2000). In very unequal communities, 'the "elite capture" problem becomes more acute' (Abraham and Platteau, 2004: 229). Zérah corroborates this point by arguing that 'class remains a relevant category in understanding participation as a terrain of struggle' (2009: 859) -an important argument for the analysis of the slum-upgrading project set out here.
Another problematic aspect of participatory approaches is their oftenidealized view of harmonious 'natural' communities, a view which suffers from a lack of understanding of power structures (Guijt and Shah, 1998;Mohan and Stokke, 2000;Mosse, 2001) and of how community representatives may also be motivated by individual interests rather than the good of the collective (White, 1996). Participation is the outcome of a political process influenced by participants' inequalities in resources and power (Mayoux, 1995: 245). These misunderstandings regarding communities and personal incentives have meant that participatory projects may serve to entrench rather than reduce inequalities. As a result, their capacity to achieve more efficient and equitable results than their top-down predecessors is still seriously questioned (Dill, 2009).

THE KWA-MAJI URBAN DEVELOPMENT PROJECT AND THE POLICY CONTEXT
The first two years (2008-2010) of the Kwa-maji Urban Development Project (KUDP) 3 were aimed at improving the living conditions of Kwamaji residents through the provision of security of tenure and the building 3. Kwa-maji is a pseudonym and the names of the key actors have been anonymized. This choice was made for multiple reasons. Many institutional actors were represented by a limited number of staff who would be easily recognizable if I were to mention the actors and the settlement. This would violate research ethical guidelines as it would threaten the anonymity that I committed to grant to my informers and interviewees. Moreover, the programme deals with one of the most problematic political issues in Kenya: the distribution of land. This issue generated conflict, and two community leaders had already been violently killed; the second murder in particular reinforced my conviction of the need for unequivocal anonymity. At the same time, the programme is ongoing, and the research itself could create internal conflicts among the actors or, even worse, be used by competing government departments in future election campaigns. Moreover, there is a risk that misuse of the research, extrapolating parts out of their context, could be used to attack the professional reputation of implementing staff and institutions. These risks are very concrete. While many other researchers have anonymized their work (e.g. Platteau, 2004), sometimes even concealing the country (Rossi, 2004), without having affected the relevance of the argument, I am aware of the limits that such a level of anonymity implies, particularly around facilitating debate and comparative work.
Participation and Elite Capture in Kenya 261 of infrastructure, with an emphasis on participation and inclusion. This first phase of the KUDP was funded by an OECD donor (hereafter known as AID) as part of its most important bilateral initiative with the Government of Kenya (GoK). Initially, AID wanted a United Nations organization (UNX), considered expert and neutral, to be the main implementing body of the programme. However, the GoK wanted to maintain a prominent role itself and AID agreed to have a GoK Department (hereafter the Lead Government Agency), as the principal implementing body, thus significantly reducing the role of UNX. While the GoK, through a different department, had an established collaboration with the United Nations in the field of slum-upgrading, AID did not itself have any previous experience of slum-upgrading in the region, but nevertheless wanted to use this pilot programme to develop an approach that was possible to replicate elsewhere. Slum-upgrading in Kenya and particularly in Nairobi has a long history. Starting in the 1970s, the response to informal settlements was the introduction of 'site and services' schemes. These schemes were supposed to relocate slum dwellers to different areas and provide them with basic urban services such as roads, water and electricity. They also provided finance for housing construction that would often be started by the project, and completed by the beneficiaries. These projects succeeded in partially addressing the demand for housing; however, they generally failed to consult target groups in the planning process, and were also criticized for corruption in allocation procedures (Syagga et al., 2001). Many programmes were conceived according to middle-class standards and were not economically sustainable for the target group, leading to 'gentrification' (Campbell, 1990;Syagga et al., 2001). This process has been widely described by Mitullah (1992) in the case study of the Umoja Tenant Purchase Scheme; Huchzermeyer (2008) in relation to the Kibera High Rise project of the early 1990s; and Ochieng (2007) in Pumwani, among other examples. In line with international policies, the GoK has also recognized that, together with their residents, slums can be upgraded in situ, and has tried out different approaches, exploring diverse types of tenure (Bassett and Jacobs, 1997). An important feature of this new generation of programmes has been the emphasis on community participation that has become a central pillar of mainstream development practice. 4 It was in this policy context that the KUDP officially started in the spring of 2008. This was in the aftermath of the worst violence that Kenya had seen since independence, generated by the general election of December 2007. This violence claimed more than 1,100 victims and resulted in over 350,000 people being internally displaced, some even within Kwa-maji. The first priority of the Kenyan government officials, when called on to implement the programme with limited resources, was therefore to avoid any further violence. In other words, there was a conscious need to ensure that this project did not create additional conflict that the government was not prepared to deal with, and that would have been unacceptable to the donor. This was no easy objective to achieve, considering that the programme intended to allocate land -arguably the most delicate political issue in Kenya (Ndungu, 2004;Southall, 2005) -and at the same time, to limit government intervention and foster a community-led process.

HISTORY AND SOCIAL ORGANIZATION OF KWA-MAJI
Kwa-maji was created in the 1970s as a settlement for people who had been evicted from slums closer to the centre of Nairobi. The growth of the slum was enabled by temporary licences distributed to occupants on purely subjective criteria. The allocation of plots to new settlers was managed through these licences by a committee composed of the area Chief, 5 his assistants, local elders and local representatives of the KANU party (which governed from independence until 2002). Possessors of these licences then built structures which they rented out to newcomers arriving from rural areas. Since that time, an informal market in structures has emerged, which makes it possible for new, affluent people to buy and own structures. The allocation of plots was often made along ethnic lines, thus creating systems of political patronage. Since the first evictions in 1977, resettlement in Kwa-maji favoured the Kikuyus, who still own the majority of structures in the settlement.
The creation of a committee of elders by the local Chief to assist in local administration is normal practice in Kenya, dating back to colonial times. In Kwa-maji, the influence of the elders was such that one of the villages (the settlement is composed of eight smaller neighbourhoods called villages) was originally named after a prominent elder on the allocation committee. From the outset, village committees, working closely with the Chief, dealt with housing problems, village administration and the settlement of disputes. They also controlled the informal market in structures (buildings).
This process created the main, and lasting, social division in the settlement -that between tenants and structure owners. This kind of social division has characterized Nairobi slums for a long time: at the beginning of the 1980s, Amis (1984) found an already well-developed system of informal ownership and landlord-tenant relations. In the 1990s, there was a huge campaign to rename this category 'structure owners', and not landlords as they were previously known. This was done in part to clarify that they did not own the land, as well as being part of an attempt to push an agenda of tenants' rights (Huchzermeyer, 2008;Syagga et al., 2001). In a recent article, 5. In Kenya, the area Chief is an appointed representative of the central government in a certain location; he/she has many powers, and can implement his/her decisions through the administrative police under his/her command. Gulyani et al. (2012: 252) demonstrate that living conditions for tenants in Nairobi are considerably worse than for structure owners. In 2000, President Moi promised to give the Kwa-maji land owned by the government to its residents. A powerful group of structure owners, formalized into an 'association of structure owners', immediately mobilized and went to court to claim the land. The same year, other residents -including some structure owners -started to work with an NGO to undertake an enumeration of structures and residents in the settlement, with a view to planning a process of slum-upgrading. In 2001, some members of the community were even flown to India in order to learn from other slum communities.
Even though this enumeration was only an initial step and had no direct implications for the allocation of land, structure owners felt threatened by the process and used every available means to stop it. The enumerators had to be protected by the police while they performed their duties. The planning of the slum-upgrading came to a halt shortly after the end of the enumeration when the government withdrew its support because of a court case against government institutions, filed by the structure owners' association.
Since the first multi-party elections in 1992, episodes of ethnic violence appeared in Kwa-maji, often around election times. The clashes were mainly between the two main tribes: Kikuyu and Luo. Similar clashes followed the 2007 elections, and in one village -which is part of Kwa-maji but not included in the project area since its land is private -Luo tenants expelled their structure owners and refused to pay rent. Structure owners in the other villages became worried that their position could be undermined.
Today, while the land in Kwa-maji belongs to the government, the structures belong to 4,300 structure owners, over half of them living outside Kwa-maji. According to the figures from the enumeration carried out by the programme in 2010 (these figures are problematic in themselves, see below), there are 34,000 residents in Kwa-maji. Of these, 19 per cent are households of resident structure owners and 81 per cent are tenants' households. These figures exclude the 2,300 absentee structure owners who own 55 per cent of Kwa-maji structures.
This short account shows that Kwa-maji was a conflict-ridden settlement with an already-consolidated social structure and organization in place, and previous contested experience of slum-upgrading. Social organization was mainly built around patterns of property ownership, shaped historically through patronage politics and ethnicity. Ethnic tensions around elections were connected with national politics but also influenced by local patterns of property ownership.

THE COMMUNITY ELECTIONS OF 2008
With its emphasis on democratic community participation, it was necessary for the KUDP to establish legitimate representatives of Kwa-maji, both in order to deal with government officials in the implementation of the programme, and to secure community support. The KUDP therefore set about organizing the election of a Residents' Committee.
A first stage in this process was a two-day listening survey (June 2008) that asked people who the opinion leaders were in the community, among other things. Names were collected and these people were subsequently invited for an opinion leaders' sensitization workshop, attended by forty-two people (July 2008). This workshop drew up the methodology for elections, including the key aspect of the composition of the Committee. The official report of the workshop states that, 'the Residents' Committee should comprise six community representatives from each of the seven villages: 6 two structure owners, one tenant, one woman, one youth and one elder' (KUDP, 2008a: 7). Crucially to the eventual outcome, there was no further elaboration of how these categories would be interpreted.
Barazas (public meetings) were then held in each of the villages, where, in the words of the official report on the Committee elections, 'the structure owners met and agreed to collectively support the programme' (KUDP, 2008b: 3, italics added). In one subsequent baraza, the District Officer announced that the elections would take place at the beginning of August. Prior to this date, there was an information campaign, involving street posters and mobile loudspeaker announcements from the programme vehicle. As regards the actual voting process in August 2008, the government report states: 'The voter turnout for the Residents' Committee election was high, with some villages recording several hundreds of people who had turned out in readiness for the exercise' (KUDP, 2008b: 19).
The outcome of this process was that the KUDP now had its own fortyeight-member Residents' Committee. The elected leaders were given training and became the conduit between the government and the 'community'. During the Leadership Training Workshop that took place ten days after the Committee was elected, a government official explained that the Committee would have important functions relating to the key issue of land allocation. These included 'definition and identification of the target groups and beneficiaries; coordinating verification of the lists of beneficiaries; selection of development and planning options, including land tenure' (KUDP, 2008c: 6-7).
The first, and most basic, point to make about these elections is that the composition of the Committee itself was socially unbalanced: a minority (structure owners) were assigned twice as many representatives as the majority (tenants). This makes it hard to sustain the argument that this process was in essence democratic. A second point of criticism is that the voting method chosen was the mlolongo vote, 7 whereby voters line up in front of 6. Initially two villages were considered one by the programme. 7. The national election held in 1988 introduced the mlolongo system (Swahili word for queuing). In 1992, the multi-party elections saw a return to the secret ballot.
Participation and Elite Capture in Kenya 265 their favoured candidate. According to the government election report, the mlolongo was preferred because a secret ballot 'would require too much time and resources, and was more open to abuse' (KUDP, 2008b: 4). However, as emerged from my ethnography and as recognized in the KUDP's own Concept Paper, 'The tenants have no tenure security, as the structure owners enjoy the right to evict for non-payment of rents or any other reason' (KUDP, 2008d: 4). Therefore, it is likely that many tenants voted for their structure owners out of fear of eviction, or as a way of reinforcing their client relationship with their structure owners. In addition, the mlolongo method of queuing, without even registering voters, and the confusion regarding guidelines, facilitated the 'importation of voters', whereby candidates would bring people from nearby villages to queue in front of them. 8 The lack of clear guidelines and the lack of an extended effort to build political awareness within each different interest group led to the paradox that, at least in one case, even the tenants' representative came from a structure owner's family, as noted in the official election report (KUDP, 2008b). The whole process led to a Residents' Committee dominated by structure owners. This could probably have been predicted in advance from the Committee composition agreed during the initial sensitization workshop. However, the results turned out to be even more unbalanced than this composition would imply, since in some villages, the representative 'woman' and 'elder' elected were also structure owners. In this way, it was possible that four or even five out of the six positions assigned to each of the eight villages could be occupied by structure owners. I calculate that between thirty-two and forty of the forty-eight members were themselves structure owners. 9 The widespread official narrative that voter turnout was high also needs further scrutiny. The visual image of long queues in front of candidates was taken as evidence of high turnout. However, the report on the elections estimates that only around 3,500 to 4,000 people voted, which must be deemed a very low turnout, even if one accepts the (also very low) total figure of 34,000 residents arrived at by the programme's enumeration (see below).
Even after the elections, there was an ongoing process of establishing a strong power-block of structure owners inside the Residents' Committee. At the Leadership Training Workshop already mentioned, an Executive Residents' Committee was elected from the overall forty-eight members. The government report on this workshop states that after discussions, the following composition of the Executive Residents' Committee was proposed and 8. Importation of voters was a problem acknowledged in the official report on the election (KUDP, 2008b: 19). 9. This was a very sensitive issue to discuss; however, my calculations were confirmed by various community members, including a local missionary who raised the issue in a meeting with the government, 'In the Residents' Committee more than 60, maybe even 70 . . . 80 per cent are structure owners'. elected: eight structure owners, two tenants and a youth. The area Chief, a city council officer and the area councillor were included as ex officio members (KUDP, 2008c). Clearly, this further skewed the already-unbalanced representation in favour of structure owners.

THE PARTICIPATORY ENUMERATION OF 2010
Having in this way established a community governance structure and having started on some infrastructural work, 10 the next decisive step of the programme was to carry out a 'participatory enumeration', in effect a detailed census of all residents which would identify project beneficiaries. Whatever the political choice in relation to land allocation, a properly conducted enumeration is necessary in any slum-upgrading programme. It is normally a preliminary step, but in Kwa-maji, due to the sensitivity of the exercise, it was postponed for a long time. Of course, as UN-Habitat has pointed out, an enumeration is never a neutral exercise. Well carried out, it can be used to rectify existing inequalities. A carefully implemented participatory enumeration can counter 'the actions of powerful elites within a settlement'; and 'marginalised groups (women, tenants . . . ) can be included in the upgrading and development processes that follow' (UN-Habitat, 2010: 8). On the other hand, if such an enumeration is conducted in a superficial manner, it can 'favour only particular groups or classes of residents, at the exclusion and expense of others' (ibid.: 118). Indeed, enumerations can be said to generate conflict because they expose to public scrutiny pre-existing inequalities in resource allocation and relationships between tenants and landlords (ibid.: 140). An enumeration is normally a comprehensive and costly exercise, but if vigilantly implemented can provide all the information needed to plan a slum upgrading. Types of data and accuracy differ in every enumeration depending on the purpose for which they are collected (UN-Habitat, 2010). In some cases, the aim may be simply to ascertain the total number of residents. In others, and this was the case in Kwa-maji, complete accuracy is of the highest importance, since the enumeration would be used to identify beneficiaries of the programme. In such cases, any mistakes would have the serious consequence of excluding potential beneficiaries from the programme.
In Kwa-maji, the process of data collection constituting the enumeration was divided into three different research projects -the numbering of structures, a socio-economic survey, and the enumeration of residents. Prior to the first of these, the Lead Government Agency carried out a physical mapping of the whole area. It then assigned a number to every structure, identifying its owner/s by name, and connecting it to the GIS (Geographic Information Participation and Elite Capture in Kenya 267 System). It is important to the analysis here that this exercise identified only the structure owners.
When that had been completed, UNX commissioned a local research consultancy to carry out a socio-economic study (October-December 2009) based on a random sample of 540 households. Before the household survey began, its contents were carefully examined by the Residents' Committee in an ad hoc workshop with the consultants. At this stage, the Committee managed to eliminate the collection of any data they deemed controversial, in particular regarding the relationship between structure owners and tenants. The Residents' Committee cleverly argued that collecting such data would generate conflict and further division in the community. It is significant that neither the Lead Government Agency nor UNX had any objection to these simplifications, since they accepted at face value this threat of conflict or 'community resistance' to the research.
It was eventually established that long-term residents were also to be included among the beneficiaries of the programme, and that therefore a full enumeration of residents was needed. At this point, the Residents' Committee gave their consent to the enumeration on the condition that it was implemented, under their control, by the Lead Government Agency. As explained at the beginning of this article, UNX had lost its role as main implementer of the programme; nevertheless it still held responsibility for the politically delicate enumeration. However, the top leaders of the Residents' Committee insisted that UNX was not seen in positive terms by the community, and persuaded the Steering Committee of the KUDP to reassign the sensitive exercise to the Lead Government Agency. The Residents' Committee had viewed UNX with suspicion from the start because the United Nations was implementing another major slum-upgrading programme in neighbouring Kibera with a policy of redistribution of assets. After receiving compensation (still under negotiation), structure owners were to be equated with tenants, both being beneficiaries on equal terms.
While enumerations are generally admitted to be costly (UN-Habitat, 2010), the Kwa-maji enumeration was conducted with a budget of only 1.2 million Kenyan shillings (11,000 euro), about 0.5 per cent of the project budget. The team of enumerators was made up of one government officer who was in charge of two members of the Residents' Committee and one to three female enumerators recruited from new university graduates. The enumerator would call at the door of the dwelling and ask for the ID and voting card of the person. The enumeration form was then filled in with both parties standing outside the door of the dwelling. This was carried out on weekdays between 10.30 am and 3 pm, a time when many people are away from their homes. 11 When no one was found in a dwelling, neighbours 11. Other NGO workers who had conducted household surveys, as well as academic researchers, were keen to explain how they needed to work at night or weekends in order to find people at home.
were approached for information; however, in most cases neighbours refused to provide information on others. Moreover, no documentary details were obtained of people who were absent on the day of enumeration. At one point, I was able to ask the enumerators what they would do to get information about those who were absent. The team leader answered: 'We told people about the enumeration and we are trying our best, but we know that if we are enumerating 10,000 people we have maybe an error of 500'. I stressed the fact that mistakes in an enumeration interested in knowing the total number of people were not that important, but when inclusion in the list of beneficiaries was the aim of the exercise, accuracy was fundamental, and people could not be excluded just because they were away at work. I received the answer, 'It is unfortunate for them'. I then asked whether people would have the chance to come and register themselves at a later stage, such as one of 'public data verification' (as advised by best practices); but I was told that it was not in the plan at that time. Enumeration forms used by other organizations in analogous circumstances all had a length of about eight pages per household. However, in Kwa-maji the form was drastically simplified to just one page per structure. Given that most structures consist of eight rooms, each one occupied by a 'household', it must be concluded that the enumeration in Kwa-maji was designed to collect up to sixty-four times less information than a standard enumeration undertaken for similar purposes. In general it can be said that the more data are collected on residents, including all the tenants and their families, the greater the chance of recognizing them as project beneficiaries with specific needs. However, in this case, this specific information was not collected. Moreover, the enumeration form was too simple to record the distinction between a room that was vacant and a room where the residents were out at work. Therefore, even if the will had been there to improve the data by returning to those residents who had been out when the enumerators called, it would have been impossible to know which forms were in need of such completion.
The intention was to record the names and document numbers of the heads of household in the case of each room in the dwelling. However, as I myself witnessed, many structure owners told their tenants to tell the enumerators that the owner was living with them. Since tenants had no clear idea of how the data would be used, some of them acceded to this request. As the form only had space for one name, the enumerator had then to put the name of the structure owner as head of household and delete the name of the tenant. In this way, many tenants, without knowing it, lost any chance of being included as a beneficiary of the project. Structure owners, on the other hand, wanted to be registered as residents because they knew that in other slumupgrading programmes, absentee structure owners were not benefitted in an equivalent way to those resident in the settlements.
No verification inside the house was made, and it was not possible to know whether the person at the door was the resident, or a relative of the structure Participation and Elite Capture in Kenya 269 owner purposely sent to enact the role of resident and get registered instead of the tenant. Simple checks recommended by international best practice to avoid such errors were dismissed as unnecessary by the Lead Government Agency, on the grounds that the Residents' Committee was supervising the process. This represented a failure to consider how this committee was itself a concrete expression of the structure owners and their interests.
There is also an important issue of gender: when the only name reported on the enumeration is the head of household, this is almost always given as the man, if a couple is living in the household. This problem has been widely noted elsewhere, but the result in this case was particularly problematic, since it meant that many women were not recorded as potential beneficiaries. Only their male partners would be listed and this, in the context of dynamic and unstable relationships that characterize Nairobi slums (Flores Fernandez and Calas, 2011: 133), renders women even more dependent. For instance, if a woman leaves a violent husband she will not benefit from the slumupgrading. This is yet another case to add to the literature that shows how household surveys may end up worsening women's situations (Kandiyoti, 1999).

INSTITUTIONALIZING PRE-EXISTING INEQUALITIES
The following analysis of this ethnographic account explores the relationship between 'participation' and 'elite capture'; it is divided into five subsections, each of which deals with one aspect of the process which served to institutionalize a pre-existing inequality. The first section looks at how the perceived need to avoid further conflict led the parties to adopt a pragmatic alliance with the structure owners as 'representatives' of the community. The second addresses barriers to participation in the project process, particularly as regards gender. The third looks specifically at how capture by an elite worked to exclude those who had originally been intended as the 'target group' of the participatory programme. The fourth examines discourses of community and their relation to heterogeneity and political divisions among structure owners in Kwa-maji. The final subsection looks at how it was that policies designed to avoid 'elite capture' were not followed in this case.

Avoiding Conflict
The Residents' Committee elections took place just after general elections had led to the most severe political crisis of independent Kenya, a crisis which claimed many lives, some within the programme area. This clearly explains the government's concern with avoiding violence in the Kwa-maji elections. From this perspective, the election exercise was very successful -disregarding for the moment at what cost this was achieved. Moreover, a Residents' Committee dominated by structure owners -those with strength and power in the community -had advantages in terms of the effective implementation of the programme. Conflict was avoided by supporting and promoting those whose capacity to generate conflict had caused the failure of a similar upgrading programme in 2001.
Up until now, I have used the words 'elite' and 'structure owners' as loosely interchangeable. However, the group that became dominant in the Residents' Committee has multiple identities, and its members are hard to define with a single term. They are almost always structure owners. Property ownership is considered a prerequisite for joining the elite group because, as explained by an experienced development worker, 'when you own something you have an interest in the area, you are a shareholder of [Kwa-maji], and therefore you have the legitimacy to participate in the decision-making process'. They are also normally considered wazee, elders, which does not necessarily mean that they are 'old'; wazee is more a title of reverence. It means that they are called on to settle personal matters and small conflicts, or to help the Chief in doing so. They are also gatekeepers, who manage the access of external actors wanting to operate in Kwa-maji, and are also considered to be community representatives, legitimated and formalized in this role by the elections.
Years of dealing with community issues means that many residents are connected to this elite through relations of patronage. The loyalty of people towards their patrons allows the latter to mobilize the residents easily and make them comply with project implementation (Platteau, 2004). In this case, no other actor could so successfully manage residents' consent. The publicly visible voting system contributed to the reproduction of these relations of patronage. Many residents themselves, after seeing the disaster of national elections conducted through secret ballot, genuinely considered queuing in front of the candidates a more transparent way to display alliances and loyalties within Kwa-maji. To this extent, the Residents' Committee was a success in reconciling conflicting policy needs of the project, since it could be officially considered a democratic and legitimate form of community representation, while still maintaining pre-existing power relations unchanged.
Both the GoK and AID needed to achieve their objectives. One can surmise that they pushed towards the easier route: the alternative of dealing with the conflict more openly could have destroyed the entire process at its inception, with the structure owners boycotting the upgrading, as had happened in 2001. Mosse's (1997) work is particularly instructive in demonstrating that participatory and community-driven development is not possible without the transformation of the social equilibrium through which elites have maintained control over the management of resources. Challenging this equilibrium necessitates a slow and gradual process and requires a full understanding of the political and social dynamics. In the current case, the KUDP seems to have taken a decision to avoid disturbing the existing social equilibrium, even though it is clear that in order to genuinely achieve full participation and social inclusion, a very different and more conflictual Participation and Elite Capture in Kenya 271 intervention would have been necessary. In addition, an intervention that challenged the status quo could not have been undertaken in a short space of time.

Barriers to Participation
An important issue, severely underestimated by the KUDP but widely acknowledged in the literature, is that in order to be able to participate, it is necessary to remove minimum barriers for vulnerable groups; participation is costly (Corbridge et al., 2005: 149;Mansuri and Rao, 2004: 6). Mayoux (1995) argues that this has an important gender dimension, since the resources required, especially skills and time, lead to inequality in opportunities for participation. Lack of formal education, or of mobility to attend meetings, frequently put women at a disadvantage when it comes to participating. Moreover, participatory programmes require considerable amounts of time, and the resultant higher 'opportunity costs' also work to the disadvantage of women (ibid.: 246-8).
In Kwa-maji, involvement in public meetings and community governance structures requires a substantial time input, as well as a minimum level of education in order to understand the process and be able to participate meaningfully. In female-headed households, women cannot afford to leave their daily work in order to participate as this would imply a direct loss of livelihood. During the Leadership Training Workshop, it was specified that members 'should be committed and available for intensive involvement. Frequent meetings, sometimes on short notice; should be ready to work on voluntary basis -no pay' (KUDP, 2008c: 7). Only those who do not need to engage in paid work can afford these conditions: in this context, this means primarily the structure owners, who gain income from their rent revenues. Recent research undertaken in the informal settlements of Dar es Salaam (Hooper and Ortolano, 2012) similarly shows that property owners were more likely than tenants to actively participate in risky and timeconsuming community initiatives, whereas the advocates of those initiatives had expected higher participation among tenants. Hooper and Ortolano also found the division between structure owners and tenants to be the most important in explaining different patterns of participation (ibid.: 112).
In Kwa-maji, the selection of community leaders included a general criterion regarding 'levels of education': this, too, operated to discourage tenants (who are among the poorer people in Kenya and have a lower level of formal education) from putting themselves forward for the Executive Residents' Committee or other official positions. Moreover, for the most disadvantaged residents to begin to publicly express their social and economic needs would imply taking up a position against powerful groups. As Corbridge et al. point out (2005: 252) the disadvantaged will think carefully before making requests that would challenge an important patron, since they have realistic assumptions on the long-term character of power structures. In this regard, 272 Andrea Rigon Crook (2003: 86) argues that 'popular perceptions of the logic of patronage politics' reinforce 'elite capture of local power structures'. In their detailed review of practices of participation, Mansuri and Rao (2004) emphasize that, along with material barriers, there are psychological obstacles to participation. In line with Mayoux's (1995) conclusions in relation to India, Nicaragua and Kenya, we find that in Kwa-maji, the most vulnerable and marginalized people need government or NGO actors to be fully cognisant of their interests and to give extra support to their participation through, for instance, affirmative action strategies. As Das Gupta et al. (2004: 32) put it, empowering the urban poor requires 'considerable political commitment on the part of the state, to overcome landlord resistance as well as tenants' fears of retribution from landlords'.

Elite Capture and Participation: Excluding the Target Group
Analyses of participation elsewhere have exposed how participatory projects are often built upon pre-existing oppressive power structures (Cooke and Kothari, 2001;Mosse, 2005), particularly if there is no will to challenge the existing social equilibrium (Mosse, 1997). What is interesting in Kwamaji is how community participation, planned through the construction of local governance structures in order to facilitate the implementation of the upgrading, has further empowered dominant elites.
Right from the start, the programme's strategy of democratic community participation was paving the way for the future handover of the management and implementation of the slum-upgrading programme to the community. In practice, this meant handing over to the Residents' Committee; in order to enable this, the Committee was registered as a non-profit corporate body and a formal constitution was drawn up. This formalization process fostered the notion among the structure owners that, since they were effectively in control of these local governance structures, they would eventually become the legitimate owners of both the land and the developments of Kwa-maji. Platteau (2004) has argued that in rural community-driven projects, 'elite capture' does not prevent the project benefitting everyone in the community. In such cases, elites who deal with development agencies and attract funding tend to appropriate a much larger share of the benefits of projects. Nevertheless, the projects will generally improve the situation of the wider community and the latter will often be thankful to the elite for bringing in the project. Olowu (2003: 46) argues that local elites have 'resources, knowledge, influence and networks' and are therefore crucial to local governance structures, but should not be allowed to exclude the poor. He concludes that institutional frameworks should be designed in order to involve elites but, at the same time, to prevent them from exploiting the structures of local governance for private interests. This is difficult given that when access to external funds is limited to a small elite, the bargaining strength of non-elite groups is very limited and the latter are obliged to accept highly unequal patterns of distribution in terms of project benefits (Abraham and Platteau, 2004: 226). 12 On the other hand, Dasgupta and Beard (2007) argue that even in the context of 'elite capture', projects may continue to benefit the poor. In their urban case studies, local elites were willing and able to contribute the time and know-how needed to facilitate community-level projects and governance.
In the case of Kwa-maji, elite capture is more worrying than in some other development contexts. Here it is not a matter of simply excluding some people from the benefits of the project; rather, the programme has the potential to lead to the complete disruption of the lives and livelihoods of relevant sections of the population. This is because, in urban slum-upgrading, what is at stake is the very land on which people live. In Kwa-maji, the elected committee had to decide who would be entitled to receive land and how. The Residents' Committee was called on to make important political choices that would affect all residents, but without itself being constitutionally representative of these different groups. The Committee had to address a number of key questions. Should the project give land only to structure owners? Should land allocation include or exclude absentee structure owners? Should the project give land to tenants who had resided in Kwa-maji for over ten years? If so, what would become of the other tenants (at least 45 per cent of the residents and probably more) who would then be displaced? Would tenants who were not granted land rights be able to afford rent, once houses are improved and the area fully serviced? Besides these formal choices, the degree of control vested in the Residents' Committee allowed it to shape the outcome of activities such as the enumeration, even though that had originally been designed to bring other interest groups into the programme. All in all, the 'participatory' process gave the structure owners access to areas of decision making and influence that extended their existing power in new directions. As argued by Conning and Kevane (2002: 389), social exclusion is 'deeply rooted in local social divisions and the way the community operates and regulates access to resources. Changing these structures and breaking down social divisions often requires challenging established structures and mobilizing the disadvantaged'. They conclude that, rather than letting 'the community' choose rules and criteria to obtain programme benefits, 'carefully chosen national targeting rules, criteria and national political support can help strengthen the position of disadvantaged groups in these local contests' (ibid.).
Following Dasgupta and Beard (2007), it is certainly important to recognize that local elites not only possess know-how but also enjoy community recognition which may facilitate the implementation of community projects.
These authors argue that elites may also be willing and able to contribute their time and play a key leadership role. Therefore, they argue, it is useful to distinguish between elite control over the project and elite capture of the benefits (ibid.: 244). In complex contexts, local power structures of patronage may be utilized productively in projects. However, these elites may exploit their power to shape the project to their advantage. In these circumstances, the state has the duty to ensure that the prominent role of the elites in the project does not undermine the stakes of the other residents. The role of state implementers may actually be to ensure that the local expertise of elites is used productively and not at the expense of the poor. A project may not be able to dismantle consolidated social structures -and there are also good reasons for projects not to aim at revolutionizing the existing social order -but at the same time the state should ensure that the poor are not excluded from interventions that are likely to radically change their lives. These findings support Werlin's (1999) argument for a strong state administration of slum-upgrading projects.

Discourses of 'Community' and Internal Heterogeneity
It would be incorrect to argue that the implementers of the project were not aware of conflicts and differences in interests among the residents. However, their assumption was that, following the election process, legitimate leaders would emerge and would express the will of 'the community'. Negotiations would take place among the elected leaders, and the three top representatives, who had the right to sit on the Steering Committee, would bring community issues to the attention of the implementers. From the perspective of programme officers on the ground, to challenge choices made by community leaders would have been tantamount to a betrayal of the participatory process. It would also have implicated them in the adoption of a patronizing attitude towards the community, rather than viewing it as a partner.
In the discourse of the programme, what these leaders said was what the community said. They became the voice of the community. This discourse had important effects; an interesting example was what happened with the enumeration. Programme officials argued that in this case there was no need for the checks and balances normally employed to prevent abuses, because the community, through their leaders, was directly involved (my italics). The notion of 'community' was thus invoked in such a way as to conceal oppressive power relations within that same 'community' (Guijt and Shah, 1998;Mohan and Stokke, 2000).
Another example of the deployment of this discourse occurred when the leaders of the Residents' Committee said in the Steering Committee that the community did not trust UNX and did not want UNX to undertake the enumeration. There was no questioning of who within the community did not want UNX. The task was simply reassigned to the Lead Government Agency of the GoK, and UNX was requested to transfer the money they had Participation and Elite Capture in Kenya 275 received for this activity to the Lead Government Agency. This discourse was so strong that it allowed the obfuscation of the clear political reality that for the structure owners, a UN body that in other programmes had been assiduous in recognizing tenants' rights represented a threat, and was certainly less easy to influence than the Lead Government Agency.
However, while there is no doubt about the structure owners' dominance over the Residents' Committee, structure owners were also a very fragmented category. Some of them refused to take part in project activities at all, and remained faithful to the structure owners' association that had blocked the previous slum-upgrading attempt in 2001, despite its eventual court defeat. These structure owners continued to boycott the project, raising a stream of allegations against the Residents' Committee and the implementers. Even among the structure owners elected to the Residents' Committee, conflicts emerged. For instance, one prominent member, together with the area Chief, was managing a community hall rented out to development agencies and community groups for meetings. Other Committee members reclaimed the community hall and brought it back under the control of the Residents' Committee, so cutting off one source of income for the Chief and the committee member. Other splits arose around the divide between absentee and resident structure owners. Internal divisions such as these have actually opened up a space for more balanced decisions in the Residents' Committee. For instance, some structure owners have come to recognize that tenants with over ten years' residency in Kwa-maji have a more legitimate claim to land than do absentee structure owners. This group of structure owners has been using this claim in furthering their own struggle against absentee structure owners.

Learning Elites Capturing the Programme
UN-Habitat clearly recognizes the problems I am analysing here. Its own publication, entitled 'Count Me In: Surveying for Tenure Security and Urban Land Management', and produced in Nairobi, states that in enumerations it is difficult to ensure representation of vulnerable groups, and that 'there is a danger that the results may solidify an already unequal distribution of rights, assets and access to resources' (UN-Habitat, 2010: 141). To avoid this risk, the document argues that it is important to conduct the enumeration carefully. However, in the case of Kwa-maji, the implementation would seem to have been careless rather than careful. The research tool was oversimplified, contributing to the dumbing down of the entire exercise. The survey was conducted in a hurry, during the central hours of the day, and without verifying individual households by entering the dwellings. The results were startlingly out of line with previous quantitative estimates, but, remarkably, no one questioned them. The initial project documents had mentioned a total population of 100,000-120,000 in Kwa-maji. This figure is widely thought to be an exaggeration, but the figure of 34,000 arrived at by the enumeration is likely to have seriously underestimated the tenant population. 276

Andrea Rigon
The Kwa-maji socio-economic study, commissioned by UNX and implemented just a few months before the enumeration, states that the population inside the settlement had continued to grow in the last ten years. The enumeration conducted in 2001 had identified 18,537 households, while the new enumeration found only 10,581 households. Surprisingly, this huge discrepancy did not provoke any reaction. Had the Lead Government Agency wanted to understand how 8,000 households had disappeared over a ten-year period, while their own socio-economic study argues that population had increased in that same period, they might have asked the NGO that conducted the previous enumeration for their dataset. This never happened, despite the fact that the dataset was offered to them. Whenever I personally raised any doubts, the response was always that the Residents' Committee -hence the community -had supervised the process and therefore it must have been fair and correct: had there been irregularities, people would have raised their voices. It is not only the possibility of quantitative errors in the data collection (and therefore mistakes in the total number of residents) that is significant here, but also, as indicated above, the limited and poor quality of information collected regarding tenants. Even if tenants had been included in the total number of residents, the scant availability of data may still have worked to exclude them in practice in subsequent phases of the project.
It cannot be argued, either, that the elite in Kwa-maji were unaware of the political nature and ramifications of an enumeration in the context of slum-upgrading. Some elite members of the Residents' Committee, who had been involved in the previous slum-upgrading programme, had at that time travelled to India with the aim of learning from other enumerations and slumupgrading approaches. They were therefore fully aware of the potentially empowering effects of enumerations to challenge the status quo; it was in full knowledge of this potential that they acted to transform the exercise in their favour. In this they were supported by a government interested in finishing the exercise quickly and without generating any conflict with the local elite that was their main partner in the settlement. The instrument for the socio-economic survey was simplified each time a discussion was held about its content, and the same thing happened subsequently with the enumeration form. Again, the discourse of the presence of 'community leaders representing the interests of all residents' was used to justify a certain level of laxity in relation to these vital processes. This lack of care was undoubtedly also tolerated because of the desire to finish what was turning out to be a tiring and under-funded exercise as quickly as possible. The actual enumeration was carried out by young, female graduates on short-term contracts, with little likelihood of renewal. They therefore had little incentive to do the job well, and simply followed instructions given to them by community members. They certainly did not have the social power or the will to challenge or further verify the information given to them. In sum, we can say that the exclusion of marginalized groups in Kwa-maji was accomplished socially through a series of small, bureaucratic omissions involving land distribution and gain legitimacy as key decision makers with regard to land allocation, the likely outcomes include not only the exclusion of some beneficiaries from their fair share of the benefits: there is also a serious risk that the programme will exacerbate previous inequalities, with potentially disastrous results for the poorest. For instance, if the land is given to a minority of the residents and the state provides infrastructure and services as well as facilitating housing improvement, rents will increase and most tenants will not be able to afford to live there.
In Kwa-maji, the Lead Government Agency feared that confronting the structure owners would result in a much more time-consuming and conflictual process, and this in the immediate aftermath of post-election ethnic violence that had destabilized the country. Such an undertaking would have required a very intensive use of resources and would have been difficult to justify to other actors, not least to the international donor, who wanted to see community consensus.
Participation is not only costly for beneficiaries but also for implementers. In order to ensure that marginalized categories (tenants, single mothers, and so on) contribute to decision making in a participatory programme, the government needs to act consistently and firmly, and has to be able to challenge powerful local groups. As argued by Das Gupta et al. (2004: 28), in communities characterized by severe power imbalances, 'higher levels of government may actually be better placed to help the disadvantaged' than local actors with vested interests. One of the paradoxes and criticisms of participatory practice in general is that whilst it emerged out of a widely-felt need for a power reversal between development agencies and beneficiaries (Chambers, 1983), it nevertheless relies on some external agency to organize participation. This has led many critics to argue that participatory processes are shaped by outsiders (Cooke and Kothari, 2001;Mosse, 2005). The lesson from Kwa-maji is that a stronger display of power from the 'outsiders'government and development agencies -is required in the management of participation. Reducing poor people's dependence on the local elite 'takes much more careful design and effort', with greater effort required in more unequal contexts (Das Gupta et al., 2004: 29). Engaging communities requires 'careful planning and active management' (ibid.: 48). Participation cannot be naively considered to be an entirely people-driven process where external agency is interpreted as a failure to achieve 'genuine' participation. Participation must be managed carefully if a programme involving land distribution is to avoid 'elite capture' and the worsening of the living conditions of a significant share of the target group. As several authors Platteau, 2000, 2004;Das Gupta et al., 2004;Tendler, 2000) have argued, there are severe risks in quickly devolving power to communities. Such processes require attentive management and careful social analysis of the target community.
Careful management of participation implies tackling powerful interests which, as mentioned above, requires various resources that may not be available in every context. In Kwa-maji, this was difficult for a number of reasons, amongst which the wider political situation figured prominently, particularly the national post-election violence. However, additional causes included the failure of the previous upgrading project, and a history of resistance and violent reaction from local structure owners, which contributed to the government prioritizing the need to avoid further violence. The government implementers also lacked legitimacy, a key resource needed to tackle powerful interests. There was no political support for a process which might create conflict in the informal settlements in the aftermath of what could have potentially become a disastrous civil war. The project had limited financial resources, and the donor was unwilling to invest in such a risky undertaking. Moreover, while in other contexts the state may have the legitimacy to use its power to manage development processes, implementers in Kwamaji lacked this legitimacy in the community. In Nairobi slums, the state is largely absent as a service provider and it is perceived by residents only in its repressive and brutal dimension, particularly as the main perpetrator of evictions and demolitions. The post-election violence further delegitimized the state. Under these conditions, implementers found the support of the local elite very valuable and aimed at establishing a mutually beneficial alliance.
When discourses of community participation are coupled with limited resources to deal with power imbalances within communities, the likely result is elite capture. In such cases, 'participation' may become, as Rahnema (1992: 119) critically argued, a tool for saving money by passing on some of the costs of development to the poor. Participation can save costs associated with ensuring community compliance and collaboration; however, as we have argued, if 'participation' is to avoid failure with regard to its pro-poor objectives, then it is anything but cheap. An important conclusion of this study is that, in the context of pervasive and entrenched inequalities, it may not be possible to initiate a transformative process of social change without creating conflict. Thus, resources to deal with such conflict are needed -a conclusion consistent with the findings of Mansuri and Rao's (2013) review of participatory projects.
The case study presented here forces us to question the viability of community-driven 'participatory' slum-upgrading projects in the context of conflicts which the implementers do not have the resources to deal with. To implement the project under these conditions means that it may end up, like most of its predecessors in Nairobi, benefitting groups with a higher income instead of the intended beneficiaries (Syagga et al., 2001). This sombre conclusion echoes the predictions of economists that, given the social conditions in Nairobi's slums, standard slum-upgrading approaches based on provision of infrastructure and security of tenure are likely to benefit structure owners rather than the poorest residents (Gulyani and Talukdar, 2008). What this article has shown is how this prediction is allowed to come true, despite the lessons from other, similar cases. Understanding these social processes,