Decentralisation in Kenya: the governance of governors* NIC CHEESEMAN Department of Politics and International Relations and African Studies

Kenya’s March  elections ushered in a popular system of devolved government that represented the country’s biggest political transformation since independence. Yet within months there were public calls for a referendum to * The authors would like to thank Job Bwonye, Alex Dyzenhaus, Amanda Magisu and Betty Okero for writing the county-level reports that this analysis draws upon. We would also like to thank Dominic Burbidge, who collected county level data in support of this analysis and whose work we draw on in the section entitled ‘The design of decentralisation’. Two anonymous reviewers and the editors of JMAS provided comments and suggestions that have significantly improved the quality of the manuscript. J. of Modern African Studies, ,  (), pp. – © Cambridge University Press . This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/./), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited. doi:./SXX terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core significantly revise the new arrangements. This article analyses the campaign that was led by the newly elected governors in order to understand the ongoing disputes over the introduction of decentralisation in Kenya, and what they tell us about the potential for devolution to check the power of central government and to diffuse political and ethnic tensions. Drawing on Putnam’s theory of two-level games, we suggest that Kenya’s new governors have proved willing and capable of acting in concert to protect their own positions because the pressure that governors are placed under at the local level to defend county interests has made it politically dangerous for them to be co-opted by the centre. As a result, the Kenyan experience cannot be read as a case of ‘recentralisation’ by the national government, or as one of the capture of sub-national units by ‘local elites’ or ‘notables’. Rather, decentralisation in Kenya has generated a political system with a more robust set of checks and balances, but at the expense of fostering a new set of local controversies that have the potential to exacerbate corruption and fuel local ethnic tensions in some parts of the country. I N T R O D U C T I O N Kenya’s March  elections brought into being a system of devolved government that represented the country’s biggest political transformation since independence. This reform was undeniably popular: in a  referendum, two-thirds of voters approved a new constitution that included devolution alongside a new Supreme Court and Bill of Rights. The decentralisation measures were extensive, providing for  county governments complete with elected governors and assemblies. In the wake of the  elections, early opinion polls found that % of Kenyans approved of the idea of devolution. Yet just a few months later, there were public calls for a referendum to significantly revise the new arrangements, and by July , two separate campaigns to force a public vote on amendments were under way. The first effort was spearheaded by the recently elected county governors, the second by the political opposition. Both were predicated on a belief that the  constitution and associated legislation were not sufficient to prevent the recentralisation of power by the national government. At the same time, multiple disputes emerged within and between the different levels of the new political system. Focusing on the first of these referendum campaigns – that were led by the county governors – this article seeks to understand the ongoing disputes over the introduction of decentralisation in Kenya, and what they tell us about the potential for devolution to check the power of central government and to diffuse political and ethnic tensions in Africa more widely. We also consider the way in which governors’ choice of strategies has been shaped by competition with other elected members at the county level, and the implications that these  N I C C H E E S E M A N E T A L terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core local struggles have for ethnic relations. In both cases, the focus of these disputes has beenmoney, in the form of salaries and development funds. However, this should not be taken to imply that the new kinds of contestation that devolution has inspired can be reduced to a simple scramble for personal enrichment. Rather, it also reflects the awareness of Kenyan political actors that patronage and development funds are central to sustain a political career, and a system of government (Barkan ; Barkan and Okumu ). Kenya is a useful test case for the impact of decentralisation in diverse and conflict-prone states for three reasons. First, it has suffered recent experiences of extensive electoral violence in ,  and . Second, the political system has historically been over-centralised and dominated by a powerful president. Third, there has been genuine reform: decentralisation was neither killed at birth, as was the case in the nearby Democratic Republic of Congo, nor was it limited to a set of superficial measures with little significance. In addition to the creation of new avenues of local representation in the form of governors and members of county assemblies (MCAs), counties have also been given a voice at the national level through the election of one women’s representative to the National Assembly and one senator to represent each county in a newly created second legislative chamber. When the new constitution was inaugurated in August , most Kenyan commentators, opposition parties and donors focused on the potential benefits of devolution. In line with the most optimistic literature on decentralisation, they extolled both its intrinsic and instrumental virtues. It was hoped that devolution would bring government closer to the people, and provide democratic and development gains, by giving previously marginalised communities an increased stake in the political system and by enabling local solutions to be found for local problems. Most of all, devolution was seen as a means to address Kenya’s chronic ethnic conflicts: ‘the new Constitution establishes national values and principles of governance that seek to diffuse, if not eliminate altogether, the ethnic tensions fuelled by perceptions of marginalisation and exclusion’ (Akech : ). But Kenyan devolution was not without its sceptics. County-level political leaders immediately spoke out about their fear that the government would seek to stymie devolution, retain as much power as possible in its own hands, and manipulate county level politicians and bureaucrats to ensure compliance with central priorities. Knowingly or not, these more cautious voices were echoing academic studies that have identified ‘recentralisation’ as a key obstacle to attempts at decentralisation (Rondinelli  D E C E N T R A L I S A T I O N I N K E N Y A terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core et al. ). Important questions were also raised about the cost of the new system of government, which effectively created  new sets of elected representatives and bureaucracies, and the need to avoid the costly duplication of goods and services (World Bank ). Academics also warned of two additional threats; namely the devolution of patronage-based politics and corruption; and the potential for devolution to create new winners and losers at the local level, which could exacerbate existing social cleavages and, in the worstcase scenario, create new fault lines of conflict (Boone ; Cheeseman et al. ). It is too early to assess the full impact of devolution in Kenya. The new institutional arrangements are still in their infancy: it will take years before they have bedded in, and it would be unfair and hasty to either laud them for their early progress or condemn them for failing to work smoothly thus far. Moreover, this paper does not focus on common criticisms of devolution in Kenya; namely, that it has led to the localisation of corruption (Cornell & D’Arcy Forthcoming), an inefficient duplication of resources, and the exacerbation of inter-communal conflict (Burbidge b). Instead, this article focuses on the continued political debate and competition over the terms of decentralisation, and argues that the Kenyan experience does not fit into the dominant narrative of recent literature, which tends to emphasise the vulnerability of decentralisation reforms to elite manipulation (see Boone ). More specifically, we show that the Kenyan experience cannot be read as a case of ‘recentralisation’ by national government, nor as one of the capture of sub-national units by ‘local elites’ or ‘notables’ (Wunsch ). Rather, we argue that Kenya has established a relatively robust form of decentralisation, in which elected county governors have emerged as the agents and focus of new political struggles, capable of acting in concert to protect their own positions. In making this argument, we draw on field observation, more than  interviews conducted by the three authors spread over two years, three nationally representative surveys, donor reports and six case studies of county level politics in Bungoma, Embu, Kericho, Kiambu, Kisumu and Nairobi. These counties were selected to ensure variation in the degree of political competition (one-party dominant to highly competitive), ethnic composition and regional coverage. Based on this evidence, we suggest that Kenya’s new governors have emerged as influential and in many cases independent political players for two main reasons. First, the process of constitutional reform – which was precipitated by the  post-election violence and the need for a power-sharing  N I C C H E E S E M A N E T A L terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core government to relegitimate the political system – conferred on governors more political and economic authority than has typically been the case in the developing world (Ndegwa ). Indeed, although it is always referred to as a form of ‘decentralisation’, the constitutional protections afforded to governors and senators, combined with the fact that at least % of government revenue must flow to the county level, mean that in practice the Kenyan system is closer to Nigerian federalism than the sort of limited decentralisation practiced in Malawi and Uganda (Crawford and Hartmann ). Second, the strategies selected by governors have been shaped by the fact that they are required to operate in two very different political arenas at the same time. In order to demonstrate this, we draw on Robert Putnam’s theory of two-level games (), which he developed to understand international negotiations in which national governments had to consider two audiences: the domestic (trade unions, NGOs, opposition political parties, voters) and the international (other countries and international bodies). He suggested that in this context, national governments faced two very different ‘win sets’ – one set of outcomes that would be acceptable domestically, and one that would be acceptable internationally. Only outcomes that were acceptable both domestically and internationally were likely to represent a stable equilibrium. On this basis, Putnam argued that we can only fully understand the way in which countries negotiate within this much narrower set of options by using theories that ‘account simultaneously for the interaction of domestic and international factors’ (Putnam : ). We draw on Putnam’s intuition to conceptualise the strategies adopted by county governors, who must also operate in two different political arenas – the county level and the national level – at the same time. In the first, governors must try and retain the support of voters and key opinion makers in the face of constant challenges from local competitors, such as MPs and senators who intend to run for the governorship in future. In the second, governors must decide whether to support or oppose the national government’s policies, especially with regard to decentralisation. Following Putnam, we suggest that recognising the existence of these two games is crucial because the particularities of each game shape the options available to governors in the other. As we will show, the pressure that governors are placed under at the local level to defend county interests has made it politically dangerous for them to be co-opted by the centre, and so has narrowed their options when negotiating with the national government.  D E C E N T R A L I S A T I O N I N K E N Y A terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core However, governors do not all face the same pressures. Rather, the impact of this two-level game is refracted through the country’s party system. Kenya’s governors are drawn roughly equally from parties within the Jubilee Alliance government and the Coalition for Reform and Democracy (CORD) opposition. Being seen to be too close to the national government is far more dangerous for opposition governors given the dispositions of their electorates, especially given the political polarisation and inter-ethnic tensions that have characterised Kenyan politics in recent years (Cheeseman et al. ). Of course, opposition leaders also have other reasons to support devolution, because it enables them to gain access to resources that would otherwise be closed to them (Cornell & D’Arcy Forthcoming). By contrast, Jubilee Alliance-aligned governors are likely to be placed under less pressure to resist central control, and face additional incentives to comply with government demands – such as the promise of safe seats and plentiful finance in future elections. Unsurprisingly, therefore, support for devolution at both the elite and popular level remains highest within opposition strongholds. But as we shall see, despite this qualification, pressure from below has led even those governors who are ardent supporters of the ruling Jubilee Alliance to argue that the funds devolved to the counties should be increased at the expense of the central government, forcing national leaders to go to great lengths to tame their co-partisans. Thus, party politics and ethnic alliances may shape the impact of the twolevel game, but they do not undermine it. Taken together, the existence of this two-level game and the relative strength of county level governments mean that governors – or at least a significant proportion of them – have both the motivation and the capacity to resist capture by central government. One important implication of this argument is that governors’ willingness to challenge the centre has been motivated as much by self-interest and the need to overcome county-level conflicts as it has been about any broader commitment to localism. It is therefore both more reliable, and more sustainable. Another is that decentralisation has generated a polity with a far more robust set of checks and balances, but at the expense of fostering economic inefficiency, corruption and a new set of local controversies that have fuelled ethnic tensions in some parts of the country. The product is a politically significant and popular new tier of government whose overall contribution to democracy, development and national cohesion remains contested and uncertain.  N I C C H E E S E M A N E T A L terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core T H E O R I S I N G D E C E N T R A L I S A T I O N The academic debate on decentralisation has long pitted enthusiasts against sceptics. Enthusiasts have focused on the theoretical benefits of decentralisation, both intrinsic and instrumental; their intellectual lineage stretching back to de Tocqueville’s argument that it was the ability of Americans in the early s to participate in local government that enabled them to experience political freedom and to stave off the threat of the ‘tyranny of the majority’ (Tocqueville  []). More modern iterations of this view have emphasised the benefits to democratic decision-making of having a population that feels that they have a stake in the political system. Others have argued that decentralisation offers an escape valve for regional or religious tensions (Suberu ); or have stressed the consequent benefits in terms of economic flexibility and service delivery (Bodea & LeBas ). Decentralisation has also been posited as one of the most effective ways to protect democratic gains in Africa and beyond. Partly inspired by this literature, decentralisation has long been in vogue within the development community. In the s, such support largely stemmed from a desire to sidestep the centralised state and improve service delivery (World Bank ). In the late s, emphasis shifted to ‘good governance’ and the hope that decentralisation would create new avenues through which notoriously corrupt central states might be circumvented and made more democratic (USAID ). By the s, decentralisation had emerged as something of a cure-all policy prescription. Sceptics, on the other hand, have tended to focus on the practical barriers to effective decentralisation in particular contexts. One line of critique calls into question the impact of decentralisation on intercommunal relations and levels of national cohesion. Thus, in contrast to the hope that decentralisation will decrease spatial inequalities, research has shown that it often ‘accentuates horizontal inequalities between richer and poorer areas as a consequence of differential levels of administrative capacity and ability to raise local resources’ (Robinson : ). Given the kinds of communal tensions that exist in polarised winner-takes-all states such as Kenya (Lynch ), such an entrenchment of regional inequalities would likely facilitate political mobilisation along ethnic lines and thus undermine efforts to foster national unity and cohesion. This combination can be toxic. As Brancati has shown (), decentralisation often fosters ethnic conflict by encouraging the growth of regional parties that reinforce ethnic identities,  D E C E N T R A L I S A T I O N I N K E N Y A terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core produce legislation that favours certain groups over others, and fuel ethnic political mobilisation up to and including secession. A second problem with decentralisation is that its implementation is often manipulated by central political elites to maintain their own power and influence (for a Kenyan example, see Barkan & Chege ). A ‘stocktaking study’ conducted in  by Stephen Ndegwa found that African states are some of the most centralised in the world and that, even when decentralisation is introduced, central governments often find ways of preventing real power and authority being devolved out of their grasp. As a result, the creation of sub-national governments may be a route not to the decentralisation of power, but rather its deconcentration (Rondinelli et al. ). Given this, it should come as no surprise that decentralisation often serves not to reform government, but to localise national-level problems. In this vein, Diane Conyers has posited that ‘administrative performance at the local level is, to a large extent, a mirror of that in the country as a whole’ (: ). This point is echoed by Gordon Crawford and Christoff Hartmann (), whose case studies of Malawi, Uganda and Tanzania demonstrate how devolution of power is far less likely to proceed smoothly where the capacity of central government is weak, political trust is low, and democracy has yet to be consolidated. In this sense, decentralisation is akin to power sharing: it is least likely to work where it is needed most (Cheeseman ). One of the most influential explanations of how such top-down manipulation occurs has been provided by Catherine Boone, who documents how governments can use decentralisation as a strategy to create the appearance of devolving power while in reality defending and entrenching their position. According to Boone, this political sleight of hand is achieved because central elites typically enjoy the capacity to ‘co-opt, demobilise, usurp, bypass or modify’ local notables (: ). Thus, the periphery can be made to do the work of the centre. Boone’s work is an important contribution to the debate, because it emphasises the way in which incumbent governments can instrumentalise processes of decentralisation, and reveals how the type of relationship that emerged between the centre and the periphery can shape the nature of the state. Taking off from Boone’s analysis, we argue that Kenya represents a deviant case in the African context, in which the ability of the central government to recentralise power is comparatively low because the constitution has created both a strong set of county governors and a dynamic set of rival local actors. As a result, governors face considerable incentives to resist central co-option. On the one hand, the intention of  N I C C H E E S E M A N E T A L terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core many senators and MPs to run for governor in the future has encouraged governors to campaign for new powers and sources of revenue with which to secure their re-election. On the other, the likelihood that any evidence of ‘selling out’ to the national government will be used against them by their county level rivals, has deterred some governors from appearing to be too close to the national government – especially in opposition strongholds. The combination of these two developments has given rise to a complex new political landscape in which there is heated competition at and between all levels of government. Under these conditions, there are significant barriers to the recentralisation of power by central elites. We begin our discussion of decentralisation in Kenya by explaining why local level actors in Kenya enjoy relatively strong political and economic powers, despite the country’s recent experience of ethnic tension and political instability (Cheeseman et al. ) before moving on to contests between the governors and other local-level actors, and governors and the centre. D E C E N T R A L I S A T I O N I N K E N Y A To some extent, Kenya fits the pattern of decentralisation in Africa, in which larger and more democratic states are more likely to devolve power. Like Nigeria and South Africa, Kenya is one of the continent’s bigger states by landmass, and has a large and varied society. Decentralisation is naturally more likely in such contexts, where the central government sits at a greater distance from its people (Zanker et al. ). Like Ghana and Senegal, Kenya is also one of Africa’s more open and competitive political systems (although how democratic remains a subject of controversy). The Kenyan political landscape is therefore relatively susceptible to pressure from below, at least when compared with some of the continent’s more closed and repressive states. This is reflected in the fact that while the former Kenyan political system was criticised for featuring an ‘imperial presidency’ (Mutua : ), the country’s ethno-regional politics have always involved some devolution of resources and influence to a local level. In the independence negotiations of the early s, leaders of smaller and more economically marginal ethnic communities (who feared the dominance of a Kikuyu–Luo ethnic elite) successfully pushed for a majimbo or regionalist constitution, which promised to devolve significant powers to eight regional assemblies (Anderson  D E C E N T R A L I S A T I O N I N K E N Y A terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core ). However, majimboism was immediately dismantled by the victorious Kenya African National Union (KANU), who had only agreed to it as a means to hasten independence (Lynch : ). Then in , President Daniel arap Moi granted district administrators new powers to initiate and administer development projects through DFRD. But the early promise of this reform was misleading, and in practice the new political structures were less an attempt to meaningfully decentralise power and more a gambit by Moi to legitimise and strengthen his power through a process of deconcentration (cf. Boone ; see Lynch : ). With the return to multi-party politics in the early s, the idea of majimboismwas revived by the new KANU leadership as ameans tomobilise ethno-nationalist sentiments among self-conceived ‘locals’ against more recent ‘migrants’ in the cosmopolitan areas of the Rift Valley and at the Coast (Lynch ).However, despite the rhetorical resonance ofmajimboism, divisive ethnic politics –which culminated in pre-electoral ethnic clashes in  and  – did not translate into any actual decentralisation of power. Instead, Kenyans had to wait until the introduction of the Constituency Development Fund (CDF) in  by the new National Rainbow Coalition (NaRC) government for the transfer of greater funds to the sub-national level. Under the CDF, ·% of national revenue is allocated to MPs for the purpose of developing their constituencies. However, it is worth noting that in many ways the CDF was not new, but rather codified a set of existing practices known as harambee (self-help), in which political leaders were expected to help organise and fund local development initiatives. Under President Jomo Kenyatta (–), harambee formed a central part of the government’s development strategy, and local communities were promised that if they constructed public goods such as health clinics and schools the government would cover the running costs. The CDF thus cemented existing norms regarding the developmental role of local leaders in the public mindset (Cheeseman b). It is as a result of this less well-documented, if uneven, strand in Kenya’s history that Ndegwa’s rankings put Kenya into the ‘high decentralisation’ category () even before the reforms of . However, neither size nor the country’s relatively democratic politics can explain why a more substantial system of political devolution was introduced in , when minority communities and some opposition parties had been campaigning for it for over  years. The prospect for fundamental political change had looked promising when KANU was finally defeated by NaRC in , but the newly elected President, Mwai Kibaki refused to support a new constitution, despite the fact  N I C C H E E S E M A N E T A L terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core that he had risen to power on the back of a promise to enact a new constitution within  days. As a result, his government disintegrated, and his former running mate, Raila Odinga, successfully led a ‘no’ campaign in a constitutional referendum, by mobilising widespread frustration at the government’s watering down of a more radical and participatory document popularly known as the Bomas Draft (Lynch ; Mutua ). So why was Kibaki prepared to lead the campaign for a ‘yes’ vote in the referendum of  for a constitution that was at least as radical as the Bomas Draft that he had sidelined five years previously? As Nelson Kasfir () has argued, elite support for a reform constitution in  can be explained through an unplanned sequence of responses that followed the unprecedented violence of the / post-election crisis. The election campaign had been one of the closest in Kenyan history, and ended in acrimony when the validity of Kibaki’s narrow victory over Odinga was called into question by both the opposition and European election observers. In the unrest that followed, over , people lost their lives and almost , were displaced (Lynch ), and there appeared to be a real risk that the country would descend into a period of extended civil conflict (Cheeseman a; Chege ). In the wake of the crisis, the general consensus was that, while the violence was triggered by a disputed election, it was fuelled by deep-rooted problems including a top-heavy political system and a tendency towards winner-takes-all politics (Kenya ; Mueller ; Branch &Cheeseman ; Lynch ), an interpretation that revitalised the longstanding campaign for constitutional reform (Mutua ). In this context, the political negotiations that led to an end to the post-election violence and the formation of a coalition government included discussion of the historical causes of the violence, and committed the country’s political elite to constitutional reform, which ushered in the  Constitution of Kenya Review Act. Among other things, this Act established a small Committee of Experts to consolidate prior drafts by resolving inconsistencies between existing draft constitutions. In addition, the public was encouraged to submit their views through memoranda or attendance at public meetings, and a special parliamentary select committee was mandated to consider the draft and offer modifications. High levels of public participation – through the submission of , memoranda and relatively high turnout rates at the Committee of Expert’s public consultative meetings – enhanced the legitimacy of the new draft. Significantly, the proposed constitution drew heavily on the ‘Bomas Draft’, which  D E C E N T R A L I S A T I O N I N K E N Y A terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core contained the strongest provisions for decentralisation of the three previous draft constitutions under consideration (Kasfir ). As a result, county governors were given real economic and political power enshrined under the Constitution. This was important because the Act empowered parliament to offer amendments to, but not reject, the Committee of Expert’s draft. Given that both sides of the political divide had publicly committed to reform, by the time that the new draft reached the legislature, the scope for manoeuvre was also significantly constrained. Nonetheless, the Committee of Expert’s recommendations could still have been undermined. Indeed, the sabotaging of the  draft constitution occurred after it had been sent to parliament. Three other factors appear to have made the new arrangements more palatable to those in power. First, by , Kibaki was in his second and final term as president, and was beginning to think more about his political legacy. Although he was clearly determined to hand power to a sympathetic successor, he followed the pattern of a number of other outgoing African presidents who proved to be more willing to implement reform when their own political fate was no longer on the line (Cheeseman ). Second, by , President Kibaki was also in a much weaker position than in . Not only had he been forced to accept a power-sharing government, but his domestic and international legitimacy had fallen following the post-election crisis of /. Although the Kenyan government remained bullish in its engagement with foreign powers, and continued to benefit from its important role as an American ally in the war-on-terror, Kibaki recognised the need to relegitimise his government. Third, MPs introduced changes to the draft constitution so that it promised to create  countries with boundaries modelled on colonial districts, rather than devolving power to the existing eight provinces, as many long-standing proponents of devolution desired (Willis & Chome : ). The proposed political units were sufficiently small that it seemed unlikely that any individual county would be able to mount a serious challenge to the central government. As a result, Kibaki’s advisors may have underestimated how difficult it would be to re-establish central control once the new system had been established. The combined effect of the strong historical demand for decentralisation, the actions of the Committee of Experts, and the government’s need for political legitimacy, resulted in the adoption of a constitutional draft that conferred real economic and political power on county governments. In turn, the willingness of Kibaki and Odinga to campaign together in favour of the new constitution, recalling the  election  N I C C H E E S E M A N E T A L terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core campaign when the two men worked side-by-side to remove KANU from power, was a remarkable moment of national rapprochement and ensured that the draft passed the referendum with the ‘yes’ camp securing ·% of the vote. T H E D E S I G N O F D E V O L U T I O N The constitution that emerged from the  referendum established a number of new political positions at the county level, which draw heavily on the US model of state government. Each county directly elects a Governor, who is given a free hand to appoint the County Executive Council, and ismandated to take the lead when it comes to budget and development planning. However, the Governor’s legislative agenda must be approved by the County Assembly, which is comprised of members (MCAs) who are directly elected at the ward level. Counties also directly elect senators, who are intended to play a national-level role, representing their county in the Senate. This newly created second chamber of parliament is designed to defend county interests and help craft bills affecting the counties, but also enjoys wider power such as the ability to impeach the President, Deputy President, Governor and Deputy Governors. According to the constitution, no less than % of government revenue –KSh  bn in  –must be devolved to the county level. It is important to note that this only refers to the funds explicitly transferred to county governments – the national government continues to fund county-level services that have remained in its jurisdiction, such as national security and the police force. In this sense, the total government spend at the county level is considerably higher than the % that must run directly through the county governments. In addition to these revenues, the constitution mandates the national government to establish and manage an Equalisation Fund to strengthen the provision of essential services in historically marginalised communities. However, this Fund only receives ·% of national revenue and has received little attention thus far. Another avenue of fiscal transfer has also been overlooked, namely the ability of the national government to make conditional grants to counties, which is perhaps more surprising given the vague constitutional guidelines on how such grants should be awarded, which has created a system that appears vulnerable to manipulation. Although no conditional grants were allocated in –, the amount allocated through this mechanism increased from KSh bn to KSh bn between  and  (Table I).  D E C E N T R A L I S A T I O N I N K E N Y A terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core Despite the decision to write a minimum threshold for the proportion of funds to be devolved to the county level into the constitution, the last three years have witnessed almost continuous debate concerning the basis on which the proportion of revenues owed to the counties should be calculated, and how much they require to be self-sufficient. An early bone of contention was whether counties should pay for the costs of their staff out of their own budget, or whether this bill should be picked up by the national government. Subsequent disagreement has focused on how much more than the minimum threshold of funds should be devolved given the heavy responsibilities that the counties have been given in terms of priority areas such as healthcare. These battles over revenue allocation are covered in detail in the next two sections, but it is worth noting that discussion of this topic has been complicated by confusion over how the proportion of revenue that the counties receive should be calculated. The  constitution stipulates that the % of national revenues must be devolved based on the last audited accounts that have been approved by the National Assembly. Due to the sluggish nature of the auditing and approval process, the last set of relevant accounts dates back to –. Because the Kenyan economy has grown at around % a year since then – while the baseline for the constitutional threshold has remained the same – the national government has been able to make relatively modest increases in the county allocation while using the – figure to claim that the proportion of revenue allocated to the counties increased from % to % between  and  (Daily Nation  March ). By contrast, county governors, who are keen to make the case that they require more resources, are fond of quoting the proportion of revenue they receive based on total government revenue in that year. By invoking the spirit rather than T A B L E I . Revenue distribution under devolution (KSh) County allocation Conditional grants County allocation plus conditional grants Allocation as a share of – government revenue * Allocation as a share of that year’s national revenue –  bn − % % –  bn  bn  bn % % –  bn  bn  bn % % *Last year of fully audited and approved accounts.  N I C C H E E S E M A N E T A L terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022278X1500097X Downloaded from https://www.cambridge.org/core. IP address: 54.191.40.80, on 30 Jul 2017 at 10:19:56, subject to the Cambridge Core the letter of the constitution – which imagined that the auditing and approval of accounts would be done on an annual basis – they are able to argue that the proportion of funds they receive has in fact remained constant (Table I). The confusion caused by two sets of leaders quoting different sets of figures has been magnified by the media, which typically fails to specify the basis on which the numbers that it carries were calculated. Devolved revenues are distributed between the  counties on the basis of a formula designed by the Commission on Revenue Allocation (CRA), a body set up by the constitution for this purpose (Table II). According to the formula, all counties receive an equal share of % of national revenues and a further % based on their fiscal responsibilities. The remaining revenues are distributed on the basis of population, poverty and land area. This has led to significant variation in the amount of funds received by different counties, with counties with large populations, such as Nairobi, and high poverty and large land area, such as Turkana, receiving a considerably larger slice of the pie. In February , for example, the County Allocation of Revenue Bill prepared by the National Treasury allotted KSh · billion to Nairobi, while Lamu received just KSh  billion. Counties are also able to raise their own revenue in certain areas prescribed by the constitution. These include taxes on property, entertainment and any other taxes authorised by an Act of Parliament. In addition, Governors may levy fees and charges in return for directly provided services such as waste management. As with the distribution of national revenues, local revenue generation has been the subject of considerable controversy. Most notably, representatives of the private sector and the National Treasury have accused the counties of infringing on forms of taxation that were intended to be reserved for the national government, resulting in a damaging system of double-taxation that has significantly increased the costs of doing business – especially for those T A B L E I I . CRA Formula for Revenue Allocation