Contested implementation: the unilateral influence of member states on peacebuilding policy in Kosovo

International organizations play an important role in policy implementation. As member states do not necessarily resolve political disagreements before delegating tasks, this article focuses on how individual member states seek to influence policy implementation by international organizations. It argues that the institutional context in which delegation takes place affects the opportunities for such unilateral influence. Particularly when the agent has considerable autonomy, implementation is likely to be a contested process. The article presents evidence on the implementation of peacebuilding policy by three international organizations in Kosovo after independence in 2008. Despite the fact that the member states within the UN, OSCE and EU fundamentally disagree on the legal status of Kosovo, the organizations have deployed substantial peacebuilding missions. The UN, OSCE and EU have, however, different institutional designs: implementing agents in the UN and OSCE have, by default, more autonomy than those in the EU. We analogously observe variation in how and to what extent member states exert unilateral influence during implementation of peacebuilding policy on the ground in Kosovo. Policy Implications • Policy implementation by international organizations can be as politically contested as the decision-making by member states. Rather than avoiding the delegation of policy implementation, states should seek to balance between who gains during decision-making and who gains during implementation. • Strong states have a disproportionate amount of influence over policy implementation. Weaker states should focus on the institutional constrains in which such unilateral influence plays out. This is an effective method of setting boundaries. • The stricter collective control mechanisms in international organizations, the less likely individual member states exert unwanted unilateral influence. States should, however, also recognize that strict collective control often comes at the expense of the quality of policy implementation. • The nitty-gritty institutional rules on staffing and budget allocation significantly affect how individual states can influence policy implementation by international organizations. States should be aware that these politics of bureaucratic resourcing can also undermine the quality of policy implementation. Lobbying international organizations International organizations play a key role in policy implementation. Member states often delegate resources – such as staff or budget – for international organizations to implement their agreements. By delegating implementation functions, member states run two risks. First, the implementing agent may not implement the agreed policy (Hawkins et al., 2006). Second, individual member states may put unilateral pressure on the agent to influence implementation at the expense of the other member states (Sridhar and Woods, 2013; Stone, 2011; Urpelainen, 2012). Implementation by international organizations is thus likely contested. We show in this article that the autonomy agents have in the use of resources for implementation, affects how member states unilaterally influence implementation. We analyse peacebuilding by three international organizations in Kosovo, which presents a unique empirical case to study contested implementation. The member states of the United Nations (UN), Organization for Security and Co-operation in Europe (OSCE) and European Union (EU) fundamentally disagree on the status of Kosovo. As a result, they have launched field missions in Kosovo (UNMIK; OMIK; EULEX) with a ‘status neutral’ mandate. This has moved the political contestation on the status of Kosovo to the implementation phase. Member states which recognize Kosovo – such as France, UK and US – have tried to pressure these field © 2017 The Authors. Global Policy published by Durham University and John Wiley & Sons Ltd. Global Policy (2017) 8:Suppl.5 doi: 10.1111/1758-5899.12455 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Global Policy Volume 8 . Supplement 5 . August 2017 102

• The nitty-gritty institutional rules on staffing and budget allocation significantly affect how individual states can influence policy implementation by international organizations. States should be aware that these politics of bureaucratic resourcing can also undermine the quality of policy implementation.

Lobbying international organizations
International organizations play a key role in policy implementation. Member states often delegate resourcessuch as staff or budgetfor international organizations to implement their agreements. By delegating implementation functions, member states run two risks. First, the implementing agent may not implement the agreed policy (Hawkins et al., 2006). Second, individual member states may put unilateral pressure on the agent to influence implementation at the expense of the other member states (Sridhar and Woods, 2013;Stone, 2011;Urpelainen, 2012). Implementation by international organizations is thus likely contested. We show in this article that the autonomy agents have in the use of resources for implementation, affects how member states unilaterally influence implementation.
We analyse peacebuilding by three international organizations in Kosovo, which presents a unique empirical case to study contested implementation. The member states of the United Nations (UN), Organization for Security and Co-operation in Europe (OSCE) and European Union (EU) fundamentally disagree on the status of Kosovo. As a result, they have launched field missions in Kosovo (UNMIK; OMIK; EULEX) with a 'status neutral' mandate. This has moved the political contestation on the status of Kosovo to the implementation phase. Member states which recognize Kosovosuch as France, UK and UShave tried to pressure these field missions into progressive actions. Status quo-oriented member statessuch as Russia, Serbia and Spainhave tried to ensure that these missions stick to their mandates.
Unilateral influence often takes place through like-minded staff within the international bureaucracy who act on behalf of the lobbying state. But we find that the autonomy of the field missions is an important scope condition. Scholars measure autonomy in international organizations by referring to discretion over the use of staff and budgetary resources (Birchler and Michaelowa 2016, p. 2;Brown, 2010;cf. Heldt and Schmidtke, 2017;Ege and Bauer, 2017). We find that because they have more autonomy during implementation, the UN and OSCE missions are more attractive targets for unilateral influence by member states than the EU missions.
This article is based on document analysis, leaked diplomatic cables and interviews conducted by both authors in Kosovo and mission headquarters in New York, Brussels and Vienna between 2008 and 2015 (Dijkstra, 2011(Dijkstra, , 2013Eckhard, 2014Eckhard, , 2016aEckhard, , 2016b. Given the informal nature of unilaterally lobbying, gathering data on concrete examples presents a challenge. The case studies we present after our theoretical discussion should be seen as illustrations of informal lobbying which inform rather than test the theory.

Theoretical perspectives on contested implementation
The traditional concern with the delegation of implementation is whether the agent faithfully implements the agreed policy. Because agents typically have specialized expertise, they may follow their own interests. Scholars have therefore studied how agents can be given proper ex ante incentives at the moment of delegation and be ex post controlled after the agent has implemented the policy (McCubbins and Schwarz, 1984).
Scholars have paid less attention to the second problem: shirking by an agent often benefits some member states at the expense of others. Individual member states therefore may have an incentive to lobby the agent during policy implementation. While the theoretical literature acknowledges the distributional conflict over the design of policy, it assumes that the member states are not present when the agent implements policy Hawkins et al., 2006;Martin, 2006;Nielson and Tierney, 2003;Pollack, 1997). After all, why would member states delegate if they want to be actively involved in implementation? Yet several scholars have recently shown that member states do often control agents on the spot (Delreux, 2008;Eckhard, 2016a;Kerremans, 2006). Urpelainen (2012) has proposed an advanced model of contested implementation, but has not provided empirical evidence.
In this article, we consider (1) decision-making and (2) policy implementation as two distinct arenas where conflict between the member states plays out. This implies that member states which 'win' in one arena can accept a 'loss' in the other arena (Manulak, 2016;Urpelainen, 2012). It also gives member states a choice where they fight their battles: before or after delegation. Member states may accept an agreement not fully in line with their preferences, if they anticipate that they have influence over the implementing agent. For example, progressive states may accept status quo agreements, if this allows them to move forward toward policy implementation.
Member states vary in their ability to exert influence. While strong states are obviously in a position to substantially affect decision-making and implementation (e.g. Dijkstra, 2015;Panke, 2012;Stone, 2011), they are likely to be even more powerful during implementation (Manulak, 2016;Urpelainen, 2012). During formal decision-making, member states have to collectively agree and often require consensus. This gives weaker states a say. On the contrary, the unilateral lobbying of the implementing agent is an informal (obscure) activity. And informality benefits powerful states with more resources (Dijkstra, 2017;Kleine, 2014;Stone, 2008Stone, , 2011Urpelainen, 2012).
We add to this state of the art the finding that the institutional context in which delegation takes place affects the opportunities for unilateral influence. As Urpelainen (2012, p. 710) writes: '[i]f international bureaucrats are relatively autonomous (. . .) states can expect great benefits if they successfully collude with them'. Autonomy varies along the extent to which member states put in place monitoring mechanisms and restrictions on the agent's discretion in using resources for implementation. Most previous scholarship on autonomy points to the use of budgetary and staff resources as two key dimensions that 'characterize the leeway granted to some public sector organizations' (Maggetti and Verhoest 2014, p. 239; for a more extensive discussion see Ege and Bauer, 2017;Heldt and Schmidtke, 2017). When implementing agents have both the competences to act and the resources to do so, we expect that the member states will unilaterally try to influence how these agents implement policy.
International organizations in Kosovo: variation in autonomy organizations with a mandate to build capacity (Rittberger et al., 2012). The three cases of policy implementation are very similar. The three organizations act in the same environment of post-war Kosovo, implement peacebuilding policy through field missions, and in all three organizations the member states are highly divided concerning the status of Kosovo. Some member states have recognized Kosovar independence; others have not. As such the members have formally agreed that the peacebuilding policies should be 'status neutral'. While the three organizations are similar in most respect, they differ in the autonomy of field missions to use staff and budgetary resources. This variation in institutional design is exogenous to the case of Kosovo. 1 First, autonomy of staff depends on the extent of ex ante managerial discretion (captured by the command and control structure) and the ex post reporting requirements (see Figure 1). There is considerable variation across the three organizations in terms of the discretion of staff. In the UN, command is delegated to the UN Secretary General's Special Representative, who heads the mission on the ground (Karlsrud, 2013). While this civilian may be in regular contact with the UN headquarters in New York, he or she is in charge. UN missions also have limited formal reporting requirements: in the case of Kosovo one quarterly report. The OSCE is slightly less autonomous. Command also lies with the head of mission and the support structure in Vienna is very limited (Eckhard, 2016a). But standard reporting is more extensive, with weekly, monthly and six monthly reports being issued by missions (Eckhard, 2016a). In the EU, on the other hand, operational command is with the Civilian Operations Commander in Brussels, who sits in the European External Action Service (EEAS). The member state ambassadors provide 'political control and strategic direction' on a daily basis (EU Council, 2007, Art. 8). Reporting requirements are as extensive as in the OSCE with additional strategic evaluations being executed every two years (Dijkstra, 2013).
Second, the budget is another indicator of mission autonomy. In the EU, member states, the European Commission, and European Parliament negotiate top-down the sevenyear multiannual financial framework. It is a centralized process, which gives the EEAS and in particular the missions limited authority over budgetary resources . The three organizations thus vary in their autonomy. UN member states have established an independent agent with autonomy on staff discretion (high) and budget allocation (high). In the OSCE, formal command also provides leeway for mission management (medium), but budget decisions can be vetoed by single states (low). The EU constitutes the least autonomous case with Brussels-dominated command (low) and budgeting (low). We therefore expect that UN missions are the most attractive targets for unilateral influence, followed by OSCE missions and finally EU missions.
To provide evidence for this claim we analyse the peacebuilding missions of these three organizations in Kosovo.  Because the membership is split, all three organizations pursue a so-called 'status neutral' mandate. This is close to a status quo arrangement. These organizations formally cannot take progressive actions to bring the recognition of Kosovo closer. Despite disagreement on status, the member states have delegated to all three organizations the mandate to implement peacebuilding policy. This takes the form of the UNMIK, OMIK and EULEX missions. While the scope of these three missions differs and has evolved since they were first deployed (UNMIK/OMIK in 1999 and EULEX in 2008), they all largely deal with assistance and local capacity building in the area of security, rule of law and other key state functions. The purpose of these missions therefore often logically conflicts with the 'status neutral' mandate: they have, in their own ways, the objective of creating and strengthening the parallel Kosovar institutions rather than re-integrating them into the Serbian state. This puts the staff of these missions in a tricky spot. As professionals, most of them want to be successful in strengthening the Kosovar institutions. They therefore hold progressive preferences. Yet this clearly clashes with the interests of the status quo-oriented member states. Because the Kosovo status was not settled prior to delegation, we expect contested implementation: the progressive member states will try to ensure that implementation indeed constitutes a further step towards Kosovar statehood, while status quo member states will try to ensure that the missions stick to their mandates.

The contest over UNMIK's direction
Our empirical analysis starts with UNMIK, which was established after the intervention of NATO in 1999, through the UN Security Council's Resolution 1244. It specified that Kosovo should 'enjoy substantial autonomy within the Federal Republic of Yugoslavia' (S/RES/1244, 1999, p. 3, emphasis added). This was a setback for US diplomacy. Throughout the Balkan crisis, Secretary of State Madeleine Albright had believed that independence was the only solution to the Kosovo conflict, even if this meant force (Daalder and O'Hanlon, 2000). Yet following NATO's intervention, which had not been authorized by the Security Council, Washington sought to ease tensions. In the interest of compromise with Russia, the US accepted an international administration with a status-neutral mandate in Kosovo, which remained part of Serbia as long as there was no political solution. The US therefore agreed to a temporary status quo agreement, with the understanding that independence was 'clearly on the way' (quoted in Ker-Lindsay, 2009, p. 16). As a consequence the disagreement on status 'was passed down to the operational level [in UNMIK] and left to the bodies responsible for implementing the deliberately vague new settlement' (King and Mason, 2006, p. 73). UN missions are decentralized, by default, and have significant autonomy. UNMIK was no different: little operational guidance was provided by the Security Council. In early 2000, UNMIK's mission head, Bernhard Kouchner, travelled to New York to ask member states how to interpret the term 'substantial autonomy' in Resolution 1244. When being left without a concrete answer, he reportedly said that he was planning 'to move ahead as he sees fit, and that he does not intend to wait for a council definition of where he is headed' (New York Times, 2000). Subsequent heads of mission, nonetheless, anticipated a red line that they avoided to cross. Michael Steiner, for instance, declared a politically-charged border resolution by the Kosovo Assembly invalid in 2002, even though he privately heldlike many UNMIK heads of missionprogressive preferences (Ker-Lindsay, 2009;King and Mason, 2006).
On the ground in Pristina, the main battle was between UNMIK and the Kosovar Albanian politicians. The provisional president Ibrahim Rugova and the former political leader of Kosovo's liberation army Hashim Thaci used every opportunity to stress that independence remained their ultimate political goal. In light of its status neutral mandate, UNMIK had to repeatedly put a break on these ambitions. It responded with a piecemeal tactic that foresaw technical state building before deciding on status (Caplan, 2005;King and Mason, 2006). The Provisional Institutions for Self-Government were established in 2001, which the Kosovo Albanians saw as a step towards independence. They were greatly annoyed when UNMIK stated it was not (Ker-Lindsay, 2009). Michael Steiner furthermore announced the 'Standards before Status' policy in 2003: Kosovo's status would not be addressed until the provisional institutions had met several good governance criteria (King and Mason, 2006). UNMIK's stalling tactic of incremental political concessions ultimately failed. In March 2004, organized ethnic riots swept through the country and left 19 civilians dead and hundreds injured. Given the broken relationship between UNMIK and the Kosovar Albanians, which now regarded UNMIK as an obstacle towards independence, it was widely understood that UNMIK 'was no longer the appropriate body to run Kosovo's affairs' (Ker-Lindsay, 2009, p. 22). It became also clear that the negotiations on Kosovo's final status had to start, something that the Norwegian ambassador Kai Eide stressed to the Security Council following his investigation into the March 2004 riots.
In the meantime, the US and EU allies started to unilaterally influence UNMIK on the ground. The key mechanism was through the appointment of likeminded senior staff. Following the riots, the head of UNMIK was replaced by the progressive Søren Jessen-Petersen. Under Jessen-Petersen, 'the whole mission was reorienting towards getting out' (King and Mason, 2006, p. 211 Steiner to slow down on the transfer of powers' (King and Mason, 2006, p. 211), the US now brought in a new UNMIK deputy, Larry Rossin, who 'believed that powers should be transferred to local institutions much faster' (King and Mason, 2006, p. 211). King and Mason (2006, p. 210) conclude: 'Just as NATO has gone around the UN when it undertook the war, now, albeit more subtly, the same great powers were bypassing the UN again'. Despite such massive lobbying, the UNMIK organization itself posed a challenge. Peterson and his team asked all units which tasks they could hand over to the local authorities, but some 'people were trying to hold on to their unit' (UNMIK Interview 3). In particular UNMIK officials with Russian nationality lobbied against transferring government functions from the UN to Kosovar authority (UNMIK Interviews 1 and 2). The 2004 riots and domestic political pressure had nevertheless unleashed irreversible political dynamics. In October 2005, the Security Council launched the final negotiations on status under the leadership of former Finnish President Martti Ahtisaari. This was a significant progressive move, opposed by Serbia, as it put status on the agenda. But the diplomatic effort failed due to Russian discontent (Ker-Lindsay, 2009;Perritt, 2010). Ahtisaari nonetheless tabled his plans for 'supervised independence' which eventually led to Kosovo's declaration of independence in February 2008.
The implementation of UNMIK prior to independence thus shows the relevance of unilateral influence. The Security Council resolution which launched UNMIK was an uncomfortable status quo compromise. It brought international politics to Pristina, where the different parties started to unilaterally lobby UNMIK from the moment it was established. The strongest pressures came from the local Kosovar Albanians (with the tacit support of the US). The international community, on the whole, was long supportive of UNMIK and its attempts to postpone the difficult questions. Yet after the March 2004 riots, the US and EU allies started to push for status negotiations. At the same time they actively worked on getting UNMIK out of Kosovo by increasing the pace of the transfer of tasks to the local authorities. They worked through UNMIK staff. But it is also clear that Russia sought to counter such influences through its own nationals. The UN's decentralized command setup and the absence of high frequency reporting provided autonomy for UNMIK staff to interpret the mandate and therefore opened up contests about the direction of policy implementation.

(Non-)withdrawal of UNMIK and OMIK after independence
The unilateral declaration of independence of Kosovo on 17 February 2008 increased the divisions within the international community. While within days, France, Germany, Italy, the UK and the US recognized Kosovar independence, Russia condemned the declaration as a 'blatant breach of the norms and principles of international law' (S/PV.5839, 2008, p. 6f.). These divisions had a significant effect on UNMIK and OMIK. While the recognizing countries wanted UNMIK and OMIK replaced by EULEX, as proposed in the Ahtisaari plan, for Russia and Serbia the two missions had to stay. Russia and Serbia were in a strong position, as it would require a new resolution by the Security Council. As the Russian representative to the Security Council made clear, 'UNMIK's reconfiguration is inadmissible without the Security Council's authorization' (S/PV.5944, 2008, p. 16). Reconfiguration is essentially a question about resource allocation. Budget plans determine the number of staff deployed in field missions and their financial resources. The UN and OSCE differ with respect to budget autonomy. In the UN, Secretary-General Ban Ki-moon proposed the budget which only needed to be approved by the majority of member states in the Assembly. In the OSCE, by contrast, single member states have a veto on the budget. Our evidence shows that these differences in institutional design explain why UNMIK was significantly downgraded after 2008 while Russian control over the OMIK budget prevented staff reductions.
The original game plan for reconfiguration had already been drawn up by the EU and UN officials prior to independence. When it became clear that the Ahtisaari negotiations had failed, the EU started contingency planning for its EULEX mission during the summer of 2007 (Dijkstra, 2011, p. 201). US diplomatic telegrams, available via Wikileaks, furthermore show extensive consultation between the UN and the progressive member states. A picture emerges of progressive UN officials, such as the head of UN peacekeeping operations who stated that 'the UN wants to leave Kosovo as quickly and completely as possible' (08USUNNE-WYORK407_a, 2008). UN officials expressed also the expectation that the plans for the UNMIK reconfiguration are workable unless Russia 'wildly object[s]' (08USUNNE-WYORK385_a, 2008). The UN strategy was to informally consult with all key players, but importantly not to seek formal Security Council endorsement (08USUNNEWYORK407_a, 2008).
Despite the early progressive messages from the UN officials, the US and EU were particularly worried about Ban Ki-moon and his wish to have the buy-in from Russia and Serbia (08STATE52042_a, 2008). US ambassador Zalmay Khalilzad said that the US need to 'keep the UN's feet to the fire in face of Russian resistance' (08USUNNEWYORK407_a, 2008). Interestingly, even some UN peacekeeping officials themselves asked the US government to lobby Ban Ki-moon for their own plan to 'reconfigure UNMIK . . . even if this must be done in the face of strong Russian and Serbian opposition ' (08USUNNEWYORK465_a, 2008).
When it became clear that Russia would block the UN Security Council from providing guidance, Ban Ki-moon had to take measures himself in light of the changed situation on the ground (08USUNNEWYORK706_a, 2008). The changed situation could be summarized as: (1) the adoption of the Kosovar constitution, which came into effect in June 2008 and rendered the civil administration part of UNMIK irrelevant; (2) the deployment of EULEX which rendered the police and justice component of UNMIK irrelevant; and finally (3) the general unwillingness of the Kosovar authorities to cooperate with UNMIK. He therefore told the Security Council that 'the conditions on the ground require a practical adjustment to the structure and profile of UNMIK' (S/ 2008/354, 2008, para 14, emphasis added).
In the light of his reconfiguration decision, the UN, EU, US, Serbia, and Russia eventually agreed on a so-called 'sixpoint-plan' clarifying the transfer of competences from UNMIK to EULEX. It was presented by the Secretary General in November 2008 and meant that in practice UNMIK would relinquished all operational activities (S/PV.6025, 2008). EULEX would be 'status neutral' and report to the UN on its activities. The Secretary General's proposal for the 2009 budget thus reflected a drawdown in budget from $210 Million in 2008/09 to $47 million in 2009/10. Despite the six-point plan compromise, Serbian and Russian representatives still expressed 'concerns' in the light of these reductions (GA/ AB/3911, 2009). Outvoted by a majority of states in the budget committee, however, they were unable to prevent the reconfiguration which led to a staff reduction by nearly 90 per cent (see Figure 2).
While UNMIK was thus drastically reduced, OMIKformally under the overall UNMIK umbrellawas not significantly affected by the reconfiguration. Naturally, OMIK was considerably smaller than UNMIK proper. And it was less an obstacle to Kosovar independence, as it dealt with democratization and institution-building rather than civil administration or executive police and justice. But it was still part of the ancien r egime and would continue to give Russia and Serbia leverage in Kosovo as OSCE members. Before independence, Russia and Serbia had introduced a motion to henceforth extend the OMIK mandate on a monthly basis (SEC. GAL/27/16, 2016, p. 14). Initially, this was meant to pressure the mission to remain 'status neutral' (ICG, 2008, p. 17). After independence, it provided Moscow and Belgrade with an effective lever to also prevent downgrading.
Indeed, the reason for the OMIK's unchanged field presence was not the continued relevance of its services. The mission was under great pressure to downgrade. Western allies and 'the US tried to shut down the mission which for them presented an obstacle to Kosovo's independence' (OSCE Interview 1). Given Serbia and Russia's control over the mandate renewal, they attempted to instead downgrade OMIK by manipulating the budget preparations. Other than the UN, the OSCE drafts its budget bottom-up: staff in field missions determine their financial requirements which are put together by the secretariat headquarters in Vienna to be decided on by the Permanent Council. In the process leading up to the 2009/10 budget, staff forwarded significantly smaller budget figures. A US code cable demonstrates that this happened due to US unilateral pressure. The cable refers to a situation in which an OSCE official spoke to a US diplomat about the budget cut: [I]f the Quintet and other like-minded states do not now show coordinated and firm support for reductions during upcoming budget negotiations, Serbia and Russia will dominate the discussion and walk the reductions back (. . .). Without the proposed  The episode again confirms the relevance of both unilateral influence and mission autonomy. Similar to the UN, proponents and opponents of independence lobbied the OSCE both in mission and at headquarters. Yet unlike the UN, the OSCE was less autonomous in allocating staff and financial resources. Informal lobbying to downgrade the budget failed due to Serbia and Russia's veto in the Permanent Council. As a result, OMIK staff levels remained almost untouched. Although Russia and Serbia argued that there was still a role for OMIK, the effect of their policy unmasks this as hypocrisy. The renewable monthly mandate introduced by the two states undermined OMIK's performance because long-term planning was made impossible and the mission lost much experienced staff due to the insecure contractual situation (Eckhard, 2016a).

Tight member state oversight and the absence of contest on EULEX
The EU is also highly divided on the Kosovo dossier. This political split in the EU has proved remarkably persistent. Even the progressive ruling by the International Court of Justice on 22 July 2010 (that 'declaring' independence is in accordance with international law) has not helped to overcome EU differences. Furthermore, while all member states agree on the need to have stability and well-functioning institutions in Kosovo, adopting the legal basis for EULEX was a challenge. The member states were, in fact, forced to adopt all the operational documents before the declaration of independence (EU Interview 5, 6). This allowed the EU to launch a mission without any reference to status, as Kosovo was still under the legal framework of UNSC Resolution 1244. It was an agreement clearly favouring the status quo. As could be expected given this background, implementation of the EULEX mandate was a delicate balancing act. In much of the rhetoric surrounding the launch of EULEX, it was made clear that EULEX would present a break from the previous UNMIK mission. While UNMIK and OMIK were part of the pre-independence political structure for Kosovo, EULEX initially was perceived (by the Kosovars and the EU elites alike) as the mission that would help Kosovo towards statehood. It was made clear that EULEX would not be a sort of 'EUMIK' (Solana and Rehn 2005, p. 2). Instead, it would be much lighter in presence and would focus mostly on supporting the Kosovar state. On the other hand, the 'six-point plan' of November 2008, which provided for a 'status-neutral' role of EULEX, was not consulted with the Kosovo government. 'The change in their international supporters' stance ', as D zihi c and Kramer (2009, p. 16) report, 'took them completely by surprise'. This resulted in a situation where Kosovar elites and the local population became quite quickly disillusioned with EULEX.
As in the two other missions, contested implementation was a possible scenario for EULEX. Immediately after the declaration of independence, the Spanish Foreign Minister Angel Moratinos said that the promised contingent of Spanish police officers would not be deployed as a sign of Spain's opposition. 2 But internal EULEX data reveals that eventually all non-recognizers, with the exception of Cyprus, deployed staff to EULEX. In 2011, a total of 260 out of 1,632 international staff came from the non-recognizing states (EULEX, 2011). As such, even the non-recognizing states had sufficient opportunities for informal lobbying. Yet, at the end of the day, the centralization of command and control in the EU peacebuilding system effectively prevented unilateral influence. Most importantly, a high frequency of routine reports EULEX sent to Brussels informed member states at all times of mission activities. Whenever status quo oriented states felt that mission activities implied recognizing Kosovo's independence, they objected in the EU Council. And this was frequently the case, as an internal code cable specifies: Madrid and the other four non-recognizing capitals complain to Brussels each time a routine report even suggests that EULEX is venturing beyond the bounds of status-neutrality and straying from its mandate under UNSCR 1244, and Brussels, in turn, notifies EULEX headquarters of its offense (09PRIS- TINA148, 2009) Confirming this quotation, several interviewees working in EULEX and the Brussels' support structures emphasized that the EU Council committees 'totally micromanage missions' (EU interview 1, 2, 4). One interviewee said that 'CIVCOM formally can only intervene on matters of political and strategic concern, but in Kosovo everything is political' (EU interview 3). Proponents of the status quo therefore used the EU's control capacities to ensure that EULEX would not stray from its mandate. They even put EULEX in irons irrespective of the damage they caused.
Given this background it is hardly surprising that Kosovo's rule of law institutions were rather unwilling to cooperate with a mission that did not recognize their sovereign authority. Instead, they preferred working with the more progressive (and also more autonomous) European Commission that has taken on the 'approach of diversity on recognition, but unity in engagement' (European Commission, 2009, p. 4). Kosovar delegations several times told the EU Council that they preferred to shift from EULEX mentoring to Commission twinning projects (Kosovo Interview 1). But Brussels' oversight regime revoked the head of mission any flexibility in implementing the mission's ambivalent mandate in a way more accommodating to the Kosovars. In 2009, the US embassy warned that '[EULEX's] cautiousness in hewing to status neutrality and defining its mission as purely technical (. . .) is a serious limitation' and that 'it could quickly become immobilized by its inherent political limitations' (09PRIS-TINA148, 2009). Several years later, a number of external evaluations and studies indeed confirm a mismatch between EULEX's significant resources and its disappointing performance (Capussela, 2015;Eckhard, 2016a;European Court of Auditors, 2012). And studies with a focus beyond EULEX confirm the pivotal role played by a divided EU Council: 'EU operations have been most effective when there has been a clear convergence of Member State interests' (Keohane, 2011, p. 202).
After the unilateral declaration of independence in 2008, there was of course less need for the proponents of Kosovar statehood to unilaterally pressure the mission towards recognition. But the need to properly govern Kosovo remained the same to avoid that the country turns into a safe haven for smuggling and organized crime in the middle of the Balkans. There is strong overall political interest in a well-performing mission. This is why member states funded the mission with a budget of over one billion Euro budget since 2008. The finding that only five non-recognizers, of limited power and size, significantly obstructed the execution of the EU's flagship mission underscores the important role formal institutional control mechanisms play in the context of contested implementation.

Conclusion
Member states often delegate the implementation of their agreements to international organizations. While academics have previously focused on how member states attempt to ensure that the agent faithfully implements the actual agreement, we have proposed an alternative model. For us, implementation remains a site of member states' contestation. Member states may unilaterally lobby the agent during implementation to ensure policy outcomes closer to their preferences. We have argued that such unilateral influence is likely to take place if the implementing agent has considerable autonomy and decides on the use of resources for implementation far away from the control of member states. It is easier for individual states to lobby the agent in all obscurity away from the oversight of the other member states. It is therefore important to not only focus on the power of the individual member states (Stone, 2011;Manulak, 2016;Urpelainen, 2012), but also the institutional design of the agent (Eckhard, 2016a). Equally, researchers who study the policy-making influence of international public administrations Eckhard and Ege, 2016;Hawkins et al., 2006;Johnson, 2014;Johnson and Uperlainen, 2014) should consider the possibility of informal lobbying.
Our findings on how the UN, OSCE and EU implemented their peacebuilding mandates in Kosovo demonstrate that single member states in a divided principal indeed make use of unilateral influence. They act through key individuals working within the implementing bureaucracy, with the nationality serving as token of loyalty and common interest. We observed such influence mainly for powerful states with significant resources, such as the US and Russia (but did not systematically test for this). We found that missions vary in terms of the extent of unilateral influence. On the one hand, the autonomous UN mission was subject to significant influence. Both, status quo opponents and proponents lobbied UN staff to implement the mandate along their respective political intentions. On the other hand, the most similar system arrangement of our research design demonstrates the limits of unilateral influence. Similar to UNMIK, autonomy in formal command allowed OSCE mission managers to insert cuts in the budget plan. Control over the budget decision, however, allowed Russia and Serbia to walk back such reductions, other than in the UN. More significant even is the case of EULEX. The centralized formal command structure of EU peace operations from the outset prevented mission managers to deviate from the letter of the mandate.
To be sure, our argument is not one about the consequentiality of unilateral influence. The evidence does not allow discriminating between autonomous bureaucrats' privately held positions and member state influence. For example, although the downgrading of UNMIK in the end appears to be in line with the progressive US position, we cannot unequivocally say that US pressure was the sole cause. Diplomatic telegrams rather imply that Ban Ki-moon acted along functional requirements. Instead, our argument is about the consequentiality of institutional designs in the way they permit or limit unilateral influence. Reductions in the agent's autonomy to use resources for implementation (staff and budget) restrict opportunities for unilateral influence. Future research on unilateral influence should thus consider the institutional design of the agent to better understand member states' influence strategies. Furthermore, those interested in the consequences of bureaucratic autonomy should be aware that there seems to be a discrepancy between formal and informal autonomy when it comes to policy implementation.

Notes
We thank the editors and two anonymous reviewers for their helpful comments and suggestions which have improved the article. We are also grateful to the participants of the 21st century global governance section at the 2016 ECPR conference in Prague for their comments on an earlier version of this paper. This article is based on research carried out in the project "International Public Administration: The Emergence and Development of Administrative Patterns and their Effects on International Policy-Making (IPA)" (http://www.ipa-research.com) funded by the Deutsche Forschungsgemeinschaft under grant number FOR # 1745 KN 891/6-1. It is also based on the results from the EU-CIVCAP project (https://www.eu-civcap.net) which has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 653227.