Do Local Power-Sharing Deals Reduce Ethnopolitical Hostility? The Effects of ‘Negotiated Democracy’ in a Devolved Kenya

Abstract How do local power-sharing arrangements affect levels of ethnopolitical hostility? The introduction of decentralisation in contexts previously marked by communal conflict underscores the need to assess local power-sharing mechanisms. However, existing literature on power-sharing has mainly examined national-level arrangements. In this article we contribute to the literature on decentralisation and ethnopolitical conflict by analysing two conflict-affected cases in Kenya. We find that local power sharing in Nakuru made intercommunal relations less hostile than in Uasin Gishu, where no such arrangement was present. The introduction and effects of local power sharing, however, is highly conditioned by national politics.


Introduction
In many societies affected by violent conflict over central government power, decentralisation has been used to mitigate the effects of winner-takes-all politics at the national level (Brancati, 2006;Brown, 2008). However, several analysts have cautioned that the devolution of resources and decision making to subnational entities may instead raise the stakes and increase the risk of communal conflict in local political arenas (Brown, 2008;Diprose, 2009;Erk, 2014, p. 546;Green, 2008;Siegle & O'Mahony, 2010). Communal conflict between groups organised around a shared identity is prevalent around the world, most notably in Africa. These conflicts sometimes become very violent and threaten local livelihoods, human security, and states' ability to provide security for their citizens (Brosché & Elfversson, 2012;Raleigh, 2010;Sundberg, Eck, & Kreutz, 2012). In contexts previously marked by communal conflict the introduction of decentralisation thus underscores the need to assess local mechanisms to regulate such conflict. One potential autonomy and secession have been widely studied (Amoretti & Bermeo, 2004;Bakke & Wibbels, 2006;Brancati, 2006;Horowitz, 1985;Siegle & O'Mahony, 2010;Stepan, 1999;Suberu, 2009). Much less attention has been paid to the consequences of decentralisation on local conflict dynamics in unitary states (for exceptions, see Diprose, 2009;Green, 2008). According to Brown (2008, p. 389), some theoretical expectations about the effects of decentralisation on conflict are similar to those related to federalism and revolve around whether decentralisation accommodates grievances or generates centrifugal fragmentation.
Expectations of conflict resolution through power sharing mainly relate to cases where the lines of conflict run between the national government and relatively homogeneous local units. However, devolution is particularly challenging in contexts characterised not only by conflict over national power, but also by local communal conflicts between identity groups that may or may not be the same as those competing for national power. In cases such as Nigeria and Ethiopia, decentralisation of power has intensified communal conflicts in many locations (Abbink, 2006;Nolte, 2002). In Kenya, analysts have similarly cautioned that decentralisation might intensify communal conflict in some counties (Bosire, 2014;Ghai, 2008). In this context of prior communal conflict and decentralisation, how is ethnopolitical hostility affected by local political power-sharing?

The Possibilities and Limitations of Political Power Sharing
At its core, political power sharing entails two or more social groups ruling jointly and taking decisions within a given political system. Such arrangements can be formal (e.g. as enshrined in the constitution or as a written peace agreement) or informal (Schneckener, 2002). Lijphart's work on consociational democracy illustrated how, in contrast to majoritarian democracy, institutional arrangements to divide power amongst competing social groups can advance political stability in societies characterised by such competition (Lijphart, 1969(Lijphart, , 1977. In line with these arguments, research has shown that power sharing in post-war states, especially when it includes several dimensions (e.g. economic, military, and/or territorial), is associated with a reduced risk of conflict recurrence (Walter, 1997(Walter, , 2002. In the longer term, such institutions may promote a political culture of cooperation and accommodation and foster intergroup trust (Schneckener, 2002).
Others, however, have raised important objections to the supposed beneficial effects of power-sharing. Spears, for instance, finds that 'power-sharing is a surprisingly unstable form of government that, even at the best of times, provides only a short-term reprieve from violent conflict ' (2002, p. 123), arguing that it usually implies that elites are asked to share power with their worst enemies in contexts where power balances are in constant flux. Such agreements are very difficult to reach, and those that are made will be unstable. A peaceful political culture, beyond the strained interactions of elite actors, for as long as it suits their interests, is unlikely to materialise (Spears, 2002, pp. 127-130). The longerterm effects of political power-sharing also depend upon who exactly are sharing power and how such arrangements relate to the broader citizenry. Whereas the theoretical ideas around power-sharing propose that consociationalism is more inclusive than majoritarianism in divided societies, and in that sense more democratic, Mehler argues that powersharing arrangements in post-conflict states mainly serve elite interests and are broadly perceived as 'sacrificing democracy in the name of peace' (Mehler, 2009, p. 453). This point has also been made in regard to power sharing following electoral crises in Kenya and elsewhere, where one of its effects has been to postpone dealing with underlying, but politically sensitive, issues of electoral justice (Cheeseman & Tendi, 2010;LeVan, 2011;Murray & Cheeseman, 2017). More broadly, power-sharing may have several adverse effects on a society's political culture (Rothchild & Roeder, 2005). Not least, the institutionalisation of power-sharing based on societal divisions risks reinforcing these same divisions, preventing the development of crosscutting identities and a joint national identity (Aitken, 2007;Rothchild & Roeder, 2005, p. 37).
This article examines local political power-sharing arrangements, and their effects on ethnopolitical hostility within specific sub-national entities. As noted above, most research on power-sharing has focused on the national level and on arrangements between governments and rebel groups in the context of ending civil wars or settling conflicts after disputed elections. How do the proposed mechanisms translate to communal conflicts at the local level? There is little research to date that specifically investigates local power-sharing as a tool to mitigate local ethnopolitical hostility and conflict. Recent exceptions are Bunte and Vinson (2016) and Vinson (2017), who investigated the role of local power-sharing in preventing inter-religious violence in Nigeria. They found that informal power-sharing arrangements decrease the risk of communal violence through two mechanisms: (1) increasing incentives for elites to seek cooperation and employ conciliatory rhetoric (hostile rhetoric would seem contradictory to power sharing and thus undermine their own positions) and (2) shaping the general public's perceptions and reducing their fear of religious diversity as they no longer feel threatened by total domination. Where power-sharing is absent, by contrast, Bunte and Vinson found that elites are more likely to rely on divisive rhetoric to mobilise support. In line with these mechanisms, we expect local power sharing to decrease ethnopolitical hostility in the Kenyan context.
How and when does local power sharing occur? Vinson suggests that local powersharing is best understood as a strategic decision taken by local leaders (2017, p. 231). However, Simons, Zanker, Mehler, and Tull (2013), who study the extent of local power-sharing within national peace agreements, find that the national level takes precedence. Both the presence and impact of local dimensions of national power-sharing varies with the general importance of local politics and the significance of particular local conflicts and arenas to national politics (Simons et al., 2013, p. 703). While their study focuses on national peace processes following civil war, the cases they analyse also include distinct local conflicts, the implications of which we believe are applicable to local power sharing intended to address communal conflicts. The argument that national politics condition the prospect of local conflict resolution resonates with research on communal conflicts that shows local actors have difficulty making peace if their agreement is not in line with the interests of the central elites (Brosché, 2014;Elfversson, 2019).
In summary, there are compelling and contradictory arguments about the effects of local power sharing. On the one hand power sharing reduces ethnopolitical hostility through increased incentives for elites to cooperate and diminished popular fears (Bunte & Vinson, 2016; see also Lemarchand, 2007;Schneckener, 2002). On the other, there may be important obstacles at both the central/national and local levels that render local power-sharing either unimportant or too complicated (Simons et al., 2013). The latter point supports the recommendation of previous studies that research should look more closely at how local and national power-sharing arrangements interact (Bunte & Vinson, 2016;Mehler, 2013). Combining these strands, we expect that local power-sharing may reduce ethnopolitical hostility, in line with the findings by Bunte and Vinson. However, we also expect that this outcome will be strongly conditioned by local-central relations and, more specifically, by the extent to which local power sharing corresponds to the interests of the national political elite.

Local Power-Sharing in Kenya
To investigate how local power-sharing affects ethnopolitical hostility, we examine two cases in Kenya: relations between the Kikuyu and Kalenjin in the counties of Nakuru and Uasin Gishu. The counties have a similar history of intercommunal violence between the same main groups, implying similar relations to national elites in the context of the national political pact between Kikuyu and Kalenjin leaders for the period covered by this study. Both counties are in the Rift Valley and until 2013 constituted districts in Rift Valley Province. In both counties, the main division has long been between the Kalenjin, who consider themselves indigenous to these counties, and the Kikuyu, many of whom acquired land in the areas during Jomo Kenyatta's rule (Anderson & Lochery, 2008;Klopp, 2001;Lynch, 2011) and who for a long time have constituted the largest ethnic group in Nakuru. In contrast the Kalenjin are the biggest ethnic group in Uasin Gishu. The demographic relations between majorities and minorities are however relatively similar: in both counties the largest group is estimated to amount to slightly more than 50% of the total population (Burbidge, 2015), and the smaller of the two groups between 10% and 15% (the exact figures are difficult to establish). 1 Political elites have repeatedly sought to mobilise support and votes based on these ethnic divisions and on perceived injustices related to land. During the 2007-08 PEV both counties were heavily affected. The case selection thus allows us to control for a number of factors that might otherwise explain variations in our outcome of interest: the intensity and origins of preceding conflict and the ethnic ties between local and national leaders. This means that we can leverage the fact that a local power-sharing agreement was reached before the 2013 elections in Nakuru but not in Uasin Gishu, to analyse the effects of a local power-sharing agreement on local ethnopolitical hostility while ensuring limited variation in several relevant factors.
To situate the case analyses, we first provide an overview of the national context with emphasis on national power-sharing agreements and pacts. Next, we describe our two cases, starting with the history of communal violence and the situation after the 2007-08 PEV. We then examine whether there were agreements between the main ethnic groups on how to distribute elected positions and allocate resources and how such agreements came about. Finally, we assess whether and how local power-sharing has affected the level of ethnopolitical hostility. To this end, we analyse the presence and extent of physical violence, threats of violence, and threatening and polarising political rhetoric. We investigate the time from 2012 to 2017, thus covering two elections (in 2013 and 2017). The analysis is based on extensive field research and a systematic analysis of news sources and statements by the political elite. To analyse the level of hostility between local communities, we consulted existing data sources on political violence and conducted a systematic search for news reports in Factiva, a global news database covering local media and international news wires. To complement the secondary material, we rely on interviews conducted by the authors in 2013, 2014, and 2017. During these research visits, more than 50 semi-structured interviews were conducted with community leaders, civil society activists, analysts, NGO representatives, and local business people. Interviewees were selected from the two communities in focus (Kikuyu and Kalenjin) and other local residents using a snowballing technique with multiple entry points, with the aim to capture all relevant perspectives rather than to achieve a representative sample.

Dispersing Central Government Power: National Political Pacts and Devolution
Local power-sharing in Kenya has national precedents, both formal and informal. Formally, power-sharing was introduced in the peace agreement negotiated by Kofi Annan in 2008 after the PEV and specified in the National Accord, which brought into law the coalition government that ruled the country for the remainder of that presidential term until 2013 (Cheeseman & Tendi, 2010;Murray & Cheeseman, 2017). The National Accord recognised that the 2007-08 crisis was the manifestation of longstanding problems created by the concentration and abuse of power and corresponding group-based exclusion. As a solution to these problems, structural and institutional reforms building on ideas related to power sharing were introduced to disperse and control power and to promote accountability and inclusivity (KNDR, 2008). One such reform brought about by the 2010 Constitution, itself an outcome of the National Accord, was the devolution of power and resources from the national government to 47 counties that were to be operational following the 2013 elections (Cheeseman, Lynch, & Willis, 2016).
In view of historical experiences of politicised ethnicity and claims to exclusive territorial homelands for certain groups (Kanyinga, 2009;Lynch, 2016;Lynch & Anderson, 2014), concerns were raised that devolution would produce similar consequences (Ghai, 2008). The constitution makers, however, sought to avoid giving significant decision-making power and resources to large regions built around the more populous communities; they opted instead for the creation of 47 counties. The counties' boundaries were drawn through spatially concentrated ethnic groups and not around them, dividing all the major ethnic groups into different counties to prevent territorially united sub-national ethnic blocks and to encourage county-based multi-ethnic politics (Barkan, 2012). Caution remained nevertheless about exclusionary identity politics within the new local entities, all of which contained ethnic majorities and minorities (Bosire, 2014). The National Cohesion and Integration Commission (NCIC), a government body established as part of the national-level dialogue following the PEV, was tasked with promoting national unity and a peaceful society. Ahead of the 2013 elections, NCIC endorsed local initiatives to defuse local competition over power and prevent communal violence through 'negotiated democracy' (Forero et al., 2013, p. 37;Lynch & Anderson, 2014), a term used in different African contexts referring to deals between ethnic or customary leaders concerning the distribution of elected posts or other electoral outcomes (Forero et al., 2013, p. 9). In addition to Nakuru, such deals were made in Migori, Busia, Isiolo, Wajir, Mandera, and Marsabit (Lind, 2018;Mitullah, 2017).
Local majority-minority dynamics fuelled the potential for identity politics and conflict in the new counties. Another source of local identity politics was national ethnopolitical alliances, which also constituted the informal dimension of national power-sharing. Despite devolution, mobilisation for national political office remained hugely important and revolved around politicised ethnicity. The national context remained fluid and continued to be shaped by rapidly shifting alliances amongst the five biggest ethnic blocks. From the perspective of power-sharing, the most important such alliance, both nationally and in Rift Valley, was the one forged between political elites of the former main adversaries, the Kikuyu and Kalenjin communities.
The pact between the Kikuyu and Kalenjin political leaders developed in different stages. After a brief period during which actors on both sides remained loyal to their respective parties, some Kalenjin members of parliament instigated a rebellion within the Orange Democratic Movement in early 2009, and the de facto Kalenjin political leader William Ruto developed closer ties to the Kikuyu-dominated Party of National Unity (PNU), particularly with Uhuru Kenyatta (Siele, 2010). The backdrop to this development was the indications that Ruto, just like Kenyatta on the PNU side, was being investigated for having organised the PEV. This was confirmed in December 2010 when Ruto and Kenyatta together with four others were named by then International Criminal Court (ICC) prosecutor Luis Moreno Ocampo as suspects of crimes against humanity. Ruto and Kenyatta turned their court cases, which began in 2011, into the foundation of their political alliance. They portrayed themselves as victims of a conspiracy and claimed to unite the Kalenjin and Kikuyu communities for the sake of peace . The political union was finally announced in December 2012 and registered in January 2013 with the official declaration of the Jubilee Alliance coalition, under whose umbrella Ruto's United Republic Party (URP), Kenyatta's The National Alliance (TNA) and their joint presidential candidate, Kenyatta, were to contest, and controversially to win, the March 2013 elections . The election campaigns gave further shape to the planned sharing of political positions, both nationally and in some counties, and the victory ensured the realisation of ethnic power-sharing.

Land, Politics and Communal Violence and in Nakuru and Uasin Gishu
Nakuru and Uasin Gishu are both in the Rift Valley and are amongst Kenya's most economically and politically significant counties. Both counties contain vast areas of fertile farmland, and considerable sections are owned by influential individuals and families and used for commercial agriculture. Since independence, the great resources invested in Nakuru have made both the town and the region targets of high-level attention and intervention. Uasin Gishu is the most important county in the Kalenjin-dominated North Rift Valley and is the home area of William Ruto; Eldoret, the capital of Uasin Gishu County, is a regional economic and administrative hub. Nakuru County is larger and more populous than Uasin Gishu: a decade ago the area corresponding to Nakuru was home to 1.6 million people, whereas the area that now comprises Uasin Gishu had close to 900,000 inhabitants (KNBS, 2009). Nakuru is often described as cosmopolitan. It includes most of Kenya's ethnic groups, with the Kikuyu the largest at slightly more than half of the county's population and the Kalenjin the second largest (Burbidge, 2015, p. 13;D'Arcy & Cornell, 2016, p. 266). In Uasin Gishu, most of the population consists of different Kalenjin sub-groups: the Nandi (the second biggest Kalenjin sub-group in Kenya after the Kipsigis) are the most populous, followed by the Keiyo. 2 It is estimated that the Kikuyu population in Uasin Gishu amounts to around 10% of the county's total.
The counties have a similar history of land appropriation during colonialism and subsequent politicised land allocation following independence. Under colonialism, the highlands of the Rift Valley were expropriated by European settlers, and the African population was at first transferred to less fertile reserves. After independence, under President Jomo Kenyatta, many Kikuyu moved to the region as part of government-sponsored settlement schemes. In Uasin Gishu, the Kikuyu presence and land ownership in the area has long been contested by the Kalenjin, who have frequently cited this as an historical injustice. Similarly, in Nakuru, the controversial evictions of some communities and settlement of others have shaped Nakuru's political economy in decisive ways; land ownership and access remain sources of political tension (Boone, 2011(Boone, , 2012Wanjohi, 1985).
The combination of communal tensions, high stakes, and influential actors, including those operating at the national level, has promoted hostile politics and contributed to violent intergroup conflicts in both counties (Akiwumi, 1999;Waki, 2008, p. 79). In Nakuru, lingering antagonism between the Kikuyu and the Kalenjin over access to land and rightful belonging was activated with the reintroduction of multi-party politics in the early 1990s. President Moi and his ruling Kenya African National Union (KANU) party rested firmly on a Kalenjin-dominated constituency, whereas most Kikuyu supported various opposition parties. Prior to the 1992 elections, and continuing after it, the government incited violence against opposition-leaning groups all across the Rift Valley, and some parts of Nakuru district were particularly badly affected. Several hundreds of thousands of people were evicted from their homes (HRW, 1993, p. 1). This was repeated on a slightly smaller scale before and after the 1997 elections (HRW, 2002;Kanyongolo & Lunne, 1998). In Uasin Gishu, which formed part of the heartland of support for Moi and KANU, whipped up antagonism against opposition-leaning communities was a strong force in the region during the early days of multi-party politics (interview, religious leader, Eldoret, February 23, 2014). During the violence in 1991 and 1997, these communities, and notably the Kikuyu, were heavily affected in Uasin Gishu (Akiwumi, 1999).
In both counties, the most severe instance of communal violence occurred in 2008 in the wake of the disputed 2007 elections. After the incumbent Mwai Kibaki was sworn in for his second term as President on December 30, 2007, following a chaotic election marred by irregular processes and dubious results, riots broke out in the oppositionleaning parts of the country. In some of these areas, mainly the Kalenjin-dominated North Rift Valley, including Uasin Gishu, but also in parts of what was then Molo District, now located within Nakuru County, this was accompanied by large-scale communal violence targeting communities regarded as having voted for the government, mainly the Kikuyu. A few weeks later, Kikuyu militias, with alleged political backing and financial support from influential actors at the national level, retaliated by evicting, assaulting, and killing civilians from opposition-leaning communities in Nakuru and Naivasha (from 2013 part of Nakuru County). Within the present Nakuru County, it is estimated that more than 200 people were killed and more than 50,000 people were displaced; the exact figures are difficult to establish (HRW, 2008;Waki, 2008). In Uasin Gishu, hundreds of people, mainly Kikuyus, were killed-around 30 of these in one of the most notorious events during the PEV, when a church south of Eldoret in which people had taken refuge was burnt down-and thousands were evicted (HRW, 2008, pp. 39-41;Waki, 2008, p. 49). A smaller number of Kalenjin were also killed in defensive and retaliatory violence.

Negotiating Local Power Sharing
In both Nakuru and Uasin Gishu, local relations between Kikuyu and Kalenjin were unsurprisingly characterised by hostility during the first years of the coalition government which ended the PEV and by lingering suspicion even after the political alliance between Kikuyu and Kalenjin was forged at the national level. The thorny questions of settling and compensating internally displaced people brought up controversies that added to the mutual suspicion, as did investigations into the PEV, for which local politicians were believed to be responsible (HRW, 2013;Mwangi, 2014). In Nakuru, the effects of the budding national alliance between some Kalenjin politicians and the PNU side gradually trickled down into the communities, but were not strong enough to neutralise local tensions. 3 In an attempt to manage these tensions and prevent communal violence ahead of the 2013 elections through community dialogue, a Nakuru-specific peace process was initiated in early 2011 by the National Cohesion and Integration Commission (HD Centre, 2014, pp. 17, 42). The final agreement, the Nakuru County Peace Accord, was signed in August 2012 by (exclusively male) elders from all ethnic groups in Nakuru County. 4 The Kikuyu and the Kalenjin were explicitly the main parties to the agreement, but any other group who so desired could embrace it, and representatives of all locally significant communities appended signatures. The agreement did not contain provisions about sharing political positions, 5 and it was received with disinterest by local political elites (Mureithi, 2012). However, during the making of the Nakuru Peace Accord (and probably necessary for its realisation), the national-level Kikuyu-Kalenjin political alliance was simultaneously being cemented (interview, Kikuyu community leader, Nakuru, March 19, 2013;interview, youth leader, Nakuru, March 19, 2013).
This alliance was finally unveiled in December 2012. Significantly, the rally to announce the news was held in Nakuru. To maximise electoral outcomes and avoid straining relations, the parties in the Jubilee alliance (the Kikuyu-dominated TNA and the Kalenjin-controlled URP) had agreed to zone their strongholds. The North Rift region, including Uasin Gishu, was declared a URP sphere of influence. Nakuru was a special case. While Kalenjin leaders insisted on regarding it as part of their historical homeland, Kikuyu leaders contested this, and their position was backed up by demographic realities. In view of Nakuru's significance, a compromise was struck, anchored at the national level (Cornell & D'Arcy, 2014;. TNA would field the candidate for the governorship, for the senator position, and for the women's representative to parliament, while URP were allocated the position of deputy to the governor (interview, Kalenjin community leader, Nakuru, March 18, 2013). Parliamentary constituencies were also subjected to such 'zoning' (Lynch, 2014, p. 103). TNA committed to abstain from fielding candidates in four of Nakuru's eleven constituencies: Kuresoi North, Kuresoi South, Rongai (even though the Kikuyu were the majority community in both Kuresoi South and Rongai), and Nakuru Town West. In the 2013 elections, TNA won in all constituencies where the party fielded a candidate, while URP lost to KANU (Kalenjin) candidates in Rongai and Kuresoi South (Cornell & D'Arcy, 2014, p. 9;Lynch, 2014, pp. 101-103).
Unlike in Nakuru, no local power-sharing arrangement was adopted in Uasin Gishu ahead of the 2013 elections, despite demands by Kikuyu leaders. Both national and local factors made it difficult for the Kikuyu to push for a strong local political role. Uasin Gishu was considered a key part of the Kalenjin sphere of influence , something further underscored by the relative numerical weakness of the Kikuyu community in Uasin Gishu. Thus, the Kikuyu community experienced political marginalisation in Uasin Gishu under the newly devolved dispensation. In 2013, Kikuyu candidates managed to win only two seats in the county assembly. Furthermore, unlike in Nakuru, the local minority community in the Kikuyu-Kalenjin alliance could not expect to win any parliamentary seat. The perceived gerrymandering of electoral boundaries was an issue of contention for the Kikuyu, who had long been campaigning for making Eldoret Town a parliamentary constituency (interview, lawyer, Eldoret, March 1, 2017).

Ethnopolitical Hostility, 2012-2017
In terms of physical violence, there were no reports of intercommunal violence between Kikuyu and Kalenjin in any of the two counties during the period examined here, which includes the elections of 2013 and 2017 (ACLED, 2018;UCDP, 2018). This absence of violence in both cases has largely been attributed to the existence of a national-level pact (e.g. Malik, 2018), but observers have also credited the local power-sharing arrangement in Nakuru for a role in preventing election-related violence there (e.g. ICG, 2017;Mitullah, 2017). The agreement centres on political positions, and mobilisation therefore becomes pronounced around elections, which present the obvious opportunity to renegotiate the rules and content of the pact. In September 2016, a year prior to the 2017 elections, TNA and URP merged to become the Jubilee Party. Nakuru was a solid Jubilee stronghold, and the most competitive elections were the party primaries (Damary, 2016). As stakes were high, competition between candidates was tense, but by and large tensions did not take on a communal aspect. Instead, the party primary election campaigns activated competition between different camps within the ethnic communities (interview, civil society activist, Nakuru, February 18, 2017; interview, civil society activist, Nakuru, February 19, 2017). One reason for the absence of ethnopolitical conflict was that the powersharing formula was adhered to, with the slight change from 2013 that the women's representative from Nakuru County to parliament was won by a Kalenjin candidate (Damary & Wambugu, 2017), though the position had not been reserved as a group of Kalenjin elders had proposed (Matara, 2017). The power-sharing arrangement also held in the gubernatorial race, with all main candidates making sure to select running-mates from the other group (Wambugu, 2016).
Threats of violence along ethnic lines were present in both cases but to different degrees. In Uasin Gishu, although there were no major instances of intercommunal violence, relations between Kikuyu and Kalenjin remained hostile long after the National Accord and the end of the PEV. The recurrent evictions of mainly Kikuyu traders from the town of Eldoret were regarded by many, not least the targeted, as ethnic victimisation (interview, community worker, Eldoret, February 27, 2017;interview, two Kikuyu politicians, Eldoret, March 1, 2017). When the ICC case against Ruto commenced in 2013, it was reported that intercommunal relations in Uasin Gishu became much tenser, with potential witnesses being branded as traitors and threatened (Stewart, 2013). During the run-up to the 2017 elections leaflets and demonstrations in Eldoret repeatedly targeted non-Nandi communities (Ndanyi, 2017a). The controversy also resulted in some violence between supporters of two Kalenjin gubernatorial candidates, Mandago and Buzeki (Kibor & Ochieng, 2017). Deputy President William Ruto was called upon to intervene (Kiplang'at & Suter, 2017), with Mandago demanding that the government forbid Buzeki to run as an independent, just as it had done in Nakuru (Kibor, 2017b). 6 Immediately around the elections, which Mandago won comfortably, there was not much violence. However, accusations against Mandago of pursuing ethnic victimisation, with special reference to the evictions of traders from Eldoret, returned after the elections and were also voiced by the new governor of Nakuru (Kahenda, 2018).
During the 2017 elections, hate leaflets were distributed in Molo and other areas in Nakuru, warning Kikuyus to prepare to leave if the sitting governor, Kinuthia Mbugua, were re-elected; however, Kalenjin and Kikuyu leaders dismissed this as 'a trick by some politicians' to divide their communities for personal gain (Ensor, 2017;Kerich, 2017). In terms of broader intergroup relations, there were also tensions between the national governing alliance and communities associated with the opposition. The largest of these communities in Nakuru (the Abagusii, the Luhya, and the Luo) are numerically too small to exercise significant electoral influence; together they make up no more than a quarter of the county's population and are concentrated in particular neighbourhoods on the outskirts of the towns of Nakuru and Naivasha. Local ethnopolitical hostilities between these groups and the majority communities waxed and waned with national political events. When the opposition campaigned across the country in 2014 and 2016 to reform the Independent and Electoral Boundaries Commission, opposition-leaning communities were targeted through demonstrations and leaflets threatening them with eviction (Agutu, 2016; Wanja & Mureithi, 2014). Occasionally, and reflecting the national pattern, ethnopolitical expressions of hostility were reinforced by an authoritarian use of executive power to intimidate and silence dissenting voices in ways that victimised both their communal and political identities (Gitonga, 2014;Mureithi & Matara, 2016). Thus, the logic of community-based political exclusion is still prevalent in Nakuru, but its boundary demarcations are drawn differently.
In terms of elite-level relations and political rhetoric, local power-sharing ('a local unwritten constitution', interview, youth leader, Nakuru, February 21, 2017) conditioned politics in Nakuru. The pact was limited to political positions, but also had indirect effects on resource allocation through employment opportunities, contracts, and service delivery (interview, civil society activist, Nakuru, February 17, 2017). It brought about political stability, often credited to the way it safeguarded elite-controlled property and power. 'It locked leaders on both sides into an agreement and established an equilibrium' (interview, political analyst, Nakuru, February 22, 2017). Many on the Kalenjin side were slightly unhappy with the distribution of positions (interview, Kalenjin elders, Nakuru, February 18, 2017), but were aware that they had negotiated from a weaker position. As one informant put it, the Kalenjin lacked both political numbers and economic clout and thus had to engage pragmatically and strategically (interview, political analyst, Nakuru, February 22, 2017). Furthermore, the pact was being carefully monitored by the national political leadership with an eye to stability, which made it difficult to mobilise discontent. 7 In Uasin Gishu, in contrast to Nakuru, intimidating and polarising political rhetoric was common. Governor Mandago, a Nandi, was increasingly accused by leaders of other communities of pursuing an exclusivist line in political positions, government employment, and resource allocation; a recent survey ranks Uasin Gishu amongst counties with poor proportional ethnic representation in local government (Burbidge, 2015). Other groups felt side-lined and some felt harassed and intimidated (interview, religious leader, Eldoret, March 1, 2017). Grievances included the perceived skewed allocation of services such as water and electricity and discrimination in public employment and economic opportunities. The governor created local networks of loyalty around distinct narratives of rightful ethnic ownership for the Nandi. He was deemed to have chosen this strategy because of being a political outsider; in the 2013 gubernatorial race he defeated William Ruto's preferred candidate. This cut both ways: he had poorer connections to the Kalenjin national political leadership, but he was also more difficult for them to control (interview, conflict analyst, Eldoret, March 3, 2017).
The lack of an agreement and the absence of a structured commitment to contain identity politics around ethnic and sub-ethnic groups rendered politics in Uasin Gishu more fluid and hostile than in Nakuru. The absence of power-sharing was an indirect and sometimes direct source and content of tension in Uasin Gishu. Sometimes this played out between the Kalenjin and the Kikuyu, whose demands for negotiations had since 2013 been met by intimidation (interview, lawyer, Eldoret, March 1, 2017;interview, two Kikuyu politicians, Eldoret, March 1, 2017). In June 2016, protests erupted in Eldoret after a Kikuyu governor in Central Province proclaimed that 'his community' would not vote for Ruto in the 2022 Presidential elections (Jebet, 2016). Although Kikuyu and Kalenjin communities both protested against the statement, the incident highlighted the high level of tension and the strong concern that local stability is dependent on the national-level alliance.
Devolution and competition for local power and resources also generated other communal cleavages. As the 2017 elections approached, Uasin Gishu politics became more openly hostile, with the most significant tension developing between the two main Kalenjin groups, the Nandi and the Keiyo. The trigger was the emergence of a strong Keiyo contender, Buzeki, for the governorship (Kibor, 2017a). Because of rumours about an anti-Nandi coalition to support Buzeki, rhetoric and threats connected to this conflict were intentionally linked to the Kalenjin-Kikuyu cleavage. Mandago and his supporters issued both subtle and not-so-subtle threats to Kikuyus (Ndanyi, 2017b), warning them of repercussions should they vote against him (Kipsang & Lubanga, 2017), to the extent that the central government cautioned that he could be charged with incitement (Nation, 2017). 8 Overall, communal relations and political rhetoric in Uasin Gishu point to a broader tendency in Kenya under devolution in which local communities claim rightful ownership to the newly created subnational entities (Bosire, 2014;D'Arcy & Cornell, 2016). In the absence of a power-sharing agreement, electoral politics took on a pronounced communal expression and turned fluid and antagonistic. Various community representatives voiced feelings of being victimised (interview, two Kikuyu politicians, Eldoret, March 1, 2017;interview, Luhya politician, Eldoret, March 2, 2017;interview, member of Luo council of elders, Eldoret, March 2, 2017), although the precise manifestations of community hostility varied with attempts to create alliances in the pursuit of electoral majorities. The main pattern during the period under study, however, was that the county leadership communicated in words and deeds the message of Nandi supremacy over all other groups, Kalenjin and non-Kalenjin (interview, Kalenjin businessman, Eldoret, February 28, 2017;interview, religious leader, Eldoret, March 1, 2017).
To conclude, we find that in Nakuru, the power-sharing agreement stabilised relations between the Kikuyu and the Kalenjin communities and contained, but did not eliminate, ethnopolitical hostility. A relatively stable pact between the political elite has contributed to cohesion and predictability, cautiously monitored by actors at the national level. In Uasin Gishu, however, repeated calls for some sort of negotiated democracy, sometimes with explicit reference to Nakuru, were brushed aside by the political leadership, and other communities experienced neglect and intimidation. The absence of a power-sharing agreement in Uasin Gishu rendered politics there less predictable and more arbitrary and conflict-ridden than in Nakuru. This is true for most of the period under study, but it intensified significantly ahead of the 2017 elections, as the competition for elected positions took on pronounced communal dimensions.
The effects of power-sharing do not seem to run deep. Beneath the surface of political unity, relations between the two communities at the popular level were uniformly regarded as characterised by lingering suspicion and latent hostility, and no meaningful reconciliation mechanisms have been established (interview, civil society activist, Nakuru, February 18, 2017;interview, civil society activist, Nakuru, February 19, 2017). There are widespread concerns that a dissolution of the Kenyatta-Ruto alliance would reignite open conflict in Nakuru and elsewhere in Rift Valley (Ensor, 2017).

Conclusion
Based on previous research, we argued that local power sharing may reduce ethnopolitical hostility, but that this outcome will be strongly conditioned by local-central relations and, more specifically, by the extent to which local power-sharing corresponds to the interests of the national political elite. The findings from Nakuru and Uasin Gishu support our arguments. Neither Nakuru nor Uasin Gishu experienced large-scale violence between the Kikuyu and Kalenjin during the period under study. This absence of violence has largely been attributed to the national-level pact. However, at a more fine-grained level, our analysis indicates that the agreement in Nakuru made relations between the groups less hostile than in Uasin Gishu, mainly due to the greater stability introduced by committed elite cooperation. The absence of such a pact in Uasin Gishu rendered politics there more volatile and antagonistic. A significant result of the power-sharing pact in Nakuru is its apparent dampening effect on open conflict during the gubernatorial race. The opposite can be said for Uasin Gishu, where the combination of devolved power and exclusionist politics rendered the local election an intense and partly violent winner-takes-all contest, especially as very few other political positions were available for minority communities. There is nothing to suggest that specific local conditions for conflict mitigation were more benevolent in Nakuru preceding the pact. If anything, the higher stakes associated with controlling Nakuru would likely have made the risk for conflict even higher there than in Uasin Gishu.
However, our findings also point to the limits of power-sharing. While devolution has changed the conditions for local politics by creating new issues, resources, actors, and arenas in which local power sharing can make a difference, these are strongly shaped by conditions at the national level. In the cases studied, components of the pact between Kenyatta and Ruto affected the bargaining space of local actors in Nakuru and Uasin Gishu in the push for 'negotiated democracy'. In the longer term, the national-level alliance restricts the sustainability of both the pact in Nakuru and the negative peace in Uasin Gishu in terms of time, scale, and scope. Regarding the time dimension, expectations are restricted to the present electoral cycle at most. As actors move towards the 2022 elections, uncertainty is likely to increase and commitment to diminish. The temporary character of the pact was underlined by many interviewees: 'a post-2017 split is factored in ' (interview, political analyst, Nakuru, February 17, 2017). Concerning the scale of sustainable peace, and in line with Spears' (2002) arguments, there are strong indications that the positive effects in Nakuru have not moved beyond the elite pact. There are very few signs of genuine reconciliation between communities at the popular level. This is obviously linked to the restricted time horizon, but also to the absence of serious efforts to anchor the elite pact in institutional or structural reforms. The restriction in scope concerns the actors involved. The pact is between the Kikuyu and Kalenjin communities; the other communities are left out of any agreement. Power-sharing in Nakuru and on the national level thus defused one particular conflict through inclusion but ignited others through deepened exclusion.
Our analysis also indicates some conditions that make local power-sharing more or less likely to materialise, and that may also affect its impact. An important condition is the significance of the unit in the national context. Nakuru is more important to the national elite than Uasin Gishu for both economic and political reasons, which may explain both the emergence of the pact and the better relations between the communities in Nakuru than in Uasin Gishu. Relatedly, the national government also holds more influence over local politics in Nakuru; the most visible expression of this is the autonomy of the governor of Uasin Gishu in relation to the national government. A second possible condition concerns the relative strength in both numerical and electoral terms of different groups (cf. Vinson, 2017, p. 24). This affects their bargaining power and the likelihood that a local powersharing agreement will be made, which resonates with the concept of 'negotiated democracy' used for these types of arrangements in Kenya. However, the similar demographic dynamics between the biggest group and its former adversary in the two counties cannot explain differences in the possibility of local power sharing or the differences in ethnopolitical hostility. The scope for both is more strongly related to the significance of the counties to national political elites and the character and autonomy of local political leaders.
To conclude, our analysis provides support for the argument that local power sharing can decrease polarisation. However, the cases also illustrate that for genuine and sustainable local peace, more profound political change is necessary. While elite Kikuyu-Kalenjin politics in Nakuru and in Kenya are more broadly characterised by temporary accommodation, the underlying logic of community exclusion and victimisation remains, with other groups as outsiders. This in turn points to the inherent instability of the arrangement, which is related to the instability and unpredictability of the national political framework. We expect these dynamics to be similar in other cases characterised by both macro-level conflicts over national power and local ethno-communal conflict. In such cases, local power sharing may reduce hostility, but policymakers and analysts should pay close attention to the national political dynamics that may change the incentives for, and feasibility of, upholding such agreements.
2. According to the 2009 census, the Luhya constituted the second largest ethnic group in the county (Burbidge, 2015). Other communities such as the Abagusii and the Luo also reside around Eldoret town. 3. There was local PNU-ODM collaboration, particularly in the mayor's office, during the 2008-13 term (Zanker, Simons, & Mehler, 2014), although individual politicians switched sides so that actual loyalties did not always correspond to formal political affiliation. 4. While counties would not become operational until after the 2013 elections, their boundaries had been known since the promulgation of the constitution in August 2010. 5. However, an NGO representative involved in the process claimed that power sharing, although not explicitly codified, was implied by the wording 'taking into account other communities' in governance (interview, NGO representative, Nairobi, March 14, 2013). 6. After the incumbent Governor Kinuthia Mbugua lost the Jubilee Party primaries in early 2017 and stated his intention to appeal, national political elites demonstrated both their influence and the importance of Nakuru when they successfully prevailed upon him to concede (Mkawale & Ndung'u, 2017). 7. An exception to this consensus emerged halfway through the political cycle and had its origin in national politics. The outspoken Governor of Bomet, Isaac Rutto, had staged a rebellion within the wider Kalenjin community against William Ruto's leadership, and in 2016 launched his own political party, Chama Cha Mashinani (CCM). The Bomet governor's political allies in Nakuru, including the member of parliament of Kuresoi South, Zachayo Cheruiyot, also decamped to CCM. Cheruiyot and local Kalenjin leaders in Kuresoi articulated and sought to link the socio-economic grievances in Nakuru to the brokers of power sharing, who they claimed showed little concern for this relatively poor part of the County and the Kalenjin population in Nakuru (Mkawale, 2016). In the end, Cheruiyot failed to defend his parliamentary seat and Isaac Rutto withdrew from the opposition. 8. It should be noted that the Kikuyu community was divided into different camps, and some supported Mandago (Lubanga, 2017).