Regaining control of errant agents? Agency slack at the European Commission and the World Health Organization

Introduction

The principal–agent (PA) literature has extensively discussed why, how and with what consequences states have delegated power to international organizations (IOs) (Dijkstra, 2015; Hawkins et al., 2006b, Nielson and Tierney, 2003; Pollack, 1997; Tallberg, 2000). Using this approach, we have learned the extent to which agents’ actions correspond to principals’ preferences, how delegation design affects agency losses, how this relationship is mediated via control mechanisms, and what strategies agents use to overcome principals’ control. Scholars have studied oversight mechanisms in single IOs and across organizations (Damonte et al., 2014; Graham, 2013; Grigorescu, 2010) as well as how the participation of international bureaucrats results in less stringent state control mechanisms (Johnson, 2013a, 2013b). Even though the literature refers to the possibility of agency slack, that is, independent action undertaken by agents which is contrary to the intentions of their principals (Conceição-Heldt, 2013), it shies away from theorizing and investigating agency slack. Given the concern of PA scholars with runaway agents, it is surprising how little we know about the conditions under which principals are able to regain control of their agents after agency slack has occurred.

Against this background, this contribution argues that once agency slack has happened, the ability of principals to regain control of their errant agents depends on three causal mechanisms: type of monitoring mechanisms; the extent of fragmentation within an IO; and the existence of credible sanctioning mechanisms available to member states. Centralized monitoring mechanisms in the form of police patrol are necessary but not sufficient for principals’ ability to regain control of their errant agents. Equally important are fragmentation within an organization, which is signaled by power dispersion across regional and country offices, and credible sanction mechanisms. The latter refer to principals’ ability to punish agents for undesired actions. Credible sanction can include budgetary contractions, personnel reduction, or principals turning to other agents in the same policy field.

To illustrate this argument, I have selected two case studies – the European Commission (henceforth, the Commission) and the World Health Organization (WHO) Secretariat – which hold fragmentation and credible sanction mechanisms constant and show variance in the monitoring mechanisms. In both cases, recurring to PA literature, I assume that principal and agent preferences diverge (Hawkins et al., 2006b; Pollack, 2003). The selection of these two case studies allows in-depth comparison that satisfies the research design of most similar cases: the cases selected are similar in all independent variables except one and differ in their dependent variable (George and Bennett, 2005). While the Commission offers a case study in which centralized police patrol monitoring mechanisms are in place, the WHO case illustrates a case of decentralized fire alarm monitoring and reporting requirements. This provides the necessary variation to illustrate the main argument, namely, that a combination of conditions, including centralized monitoring mechanisms, fragmentation, and sanction mechanisms will shape the ability of states to regain control after agency slack has occurred. In collecting evidence, I draw on interviews conducted with representatives from these institutions as well as on primary sources, including internal documents from the Commission, the European Parliament, the WHO, and the United Nations Joint Inspection Unit (UNJIU).

The article is organized as follows. The first section discusses conceptualizations of discretion, autonomy, and agency slack as well as how principals use control mechanisms to minimize agency losses. This will allow me to investigate agency slack empirically in the case studies. The second section presents the theoretical framework, highlighting which causal mechanisms are more likely to affect the ability of states to regain control of their agents. The two case studies trace the processes through which agent slack has occurred and the ability or failure of member states to regain control of their errant agents. Thereby, this contribution focuses on the type of monitoring mechanisms which were in place, the level of fragmentation within an IO, and the credibility of sanctions mechanisms at the WHO and the Commission. The final section summarizes the main findings.

Principal–agent literature: discretion, autonomy, and agency slack

One of the central issues for states is how to delegate power without losing control (Johnson, 2013a, Lupia, 2003; Newman, 2010). States (principals) delegate tasks to IOs (agents) as they expect this division of labor to generate some benefits arising from specialization. IOs are, among other things, in charge of representing states, resolving disputes, enhancing credibility, and re-delegating power (Hawkins et al., 2006a). Generally, once states have defined the functions and competences of agents, their degree of interference in agents’ actions is minimal, limited to cases where agents produce outcomes or policies inconsistent with principals’ preferences. There are, however, two challenges that principals face when delegating power. First, agents can hide information from their principals. Second, agents can do things behind the principal’s back, for example, concealing actions that principals would sanction if they were known (Nielson and Tierney, 2003).

Delegation relationships are, therefore, characterized by a trade-off between the benefits of delegation and the losses in control, or agency losses, which can arise in the post-delegation stage. States try to minimize agency losses by using different types of control mechanisms, including screening and selection procedures, a precise negotiating mandate, monitoring and reporting requirements, institutional checks and balances, and sanctions (Hawkins et al., 2006a, Pollack, 2003). The literature has shown that these oversight mechanisms allow principals to extensively monitor their agents, but also that agency losses are endogenous to delegation and can vary in magnitude (Conceição-Heldt, 2010; Hawkins et al., 2006a). Existing scholarship extensively discusses the conditions under which agents are able to act autonomously. For example, several studies indicate that, if members of the collective principal, that is, a principal composed of more than one actor, hold homogenous preferences (Nielson and Tierney, 2003), this will allow them to efficiently use available control mechanisms as long as no fragmentation exists within an organization (see Graham, 2015). Nielson and Tierney (2003) show in the case of the World Bank that when agent behavior diverges from principals’ preferences, the probability of institutional reform will depend on the degree to which preferences converge among multiple principals. Finally, Johnson (2013a) demonstrates that the more proactive international bureaucrats are in designing an institution, the less stringent state control mechanisms are in the resulting IOs.

Although the PA-literature refers extensively to agency slack, scholars have been reluctant either to theorize or to investigate it empirically. This is surprising, as agency slack can only be observed empirically. One explanation for this reluctance is that it is particularly difficult to assess agency slack. What at first glance appears to be agency slack may prove to be simply the discretion that agents need to fulfill their delegation mandates. In order to circumvent this methodological challenge, I assume that agency slack exists when principals have to spend additional resources to control their agents. Before turning to the operationalization of the dependent variable (agency slack), I briefly review the literature with a specific focus on how scholars have conceptualized and operationalized agency slack to date.

The PA-literature employs two close, but different, concepts related to agency slack: discretion; and autonomy. Hawkins et al. (2006a) argue that discretion gives agents a certain room for maneuver in performing delegated tasks and taking decisions without interference from principals. The concept of discretion involves a grant of authority that indicates the principals’ goals, but does not specify the actions agents must take to accomplish their mandates. The opportunity for slack emerges when principals write discretion into the delegation contract. Thatcher and Stone Sweet (2002) conceive the gap between what principals want from their agents and what agents actually do as the ‘zone of discretion.’ This zone represents the sum of delegated authority minus the sum of control mechanisms available to principals to shape or annul policy outcomes after agents have fulfilled their tasks. Other scholars emphasize that discretion includes the scope of issues the agent is authorized to handle, the policy instruments it has at its disposal, and the procedures it should follow to use its policy instruments (McCubbins et al., 1987). Discretion is a necessary condition for autonomy and creates the opportunity for slack. According to Cortell and Peterson (2006), whether IOs that have been granted significant discretion to actually engage in slack depends on staffing rules and on the structure of control or voting rules. They expect that when IOs have the authority to fulfill their mandate without the prior approval of states, it is more likely that they will engage in slack.

Autonomy falls in the middle of the continuum between discretion and agency slack. For Tallberg (2000), autonomy is an agent’s ability to successfully pursue a private agenda. In line with other classical studies in the field, such as Hawkins et al. (2006a) and McCubbins and Schwarz (1984), autonomy refers in this article to the range of independent actions available to agents after principals have established control and sanction mechanisms intended to minimize agency slack in the post-delegation stage. Public administration scholars provide some useful conceptualizations of autonomy. Verhoest et al. (2004), for example, develop six dimensions of the concept in public organizations which can be organized into two broad categories: managerial autonomy; and structural autonomy. Managerial autonomy refers to the autonomy an organization has related to its decision-making competencies, including policy autonomy. In contrast, structural autonomy describes the ability to act without governmental constraints on actual decision-making (Verhoest et al., 2004: 109). In this way, whereas managerial autonomy deals with an organization’s formal competences, structural autonomy focuses on an organization’s room of maneuver and its capacity to act independently from its principals. Caughey et al. (2009) distinguish between autonomy of will and autonomy of action and make a first attempt to map and measure bureaucratic autonomy by developing the following indicators: legal mandate; financial resources; other resources; and case studies of clash between agents and principals. However, these indicators remain at the conceptual level, and the authors’ complex framework has not yet been successfully applied to IOs. Finally, PA scholars have also helped conceptualize autonomy. Brown (2010), for example, developed a metric for delegation based on the services provided by an agent and the resources and autonomy an agent has to deliver those services illustrated with the International Atomic Energy Agency and the International Monetary Fund.

Autonomy, in turn, can lead to agency slack, which falls at the end of the continuum of the PA-relationship. Hawkins et al. (2006a: 8) define agency slack as a situation in which agents are able to act independently from their principals in a way that is undesired by principals. Agency slack can take two different forms: shirking; and slippage. Whereas shirking refers to agents minimizing the efforts they exert on their principals’ behalf, slippage defines a situation in which agents shift policies away from their principals’ preferred outcomes and towards their own preferences (Pollack, 1997). Autonomy and slack differ in two subtle ways from one another. Whereas autonomy refers to the range of independent actions available to agents and can be used either to benefit or to undermine principals’ preferences, slack has a negative connotation and means that agents’ behavior is undesired by their principals. In this article, I define agency slack as agents’ attempts to overreach their delegated authority by deviating from or going beyond their mandates. Consequently, agency slack can only be assessed empirically in concrete policy situations. Agency slack exists when principals have to expend additional resources to control their agents (Hawkins et al., 2006a: 9). Examples include when principals hire a new agent to evaluate the performance of another agent or when they strengthen monitoring or reporting mechanisms, for example, by establishing a new committee to monitor the activities of an IO. After agency slack occurs, principals can punish their agents by sanctioning undesired actions.

Principal–agent studies see agent slack as the outcome of imperfect control. However, the existence of slack does not mean that delegation has failed or that it is disadvantageous to states. The costs of agency losses or slack have to be weighed against the gains from delegation and the best alternatives available to states. As Lake and McCubbins put it, ‘the greater the gains from delegation, the greater the agency slack states will tolerate before reverting to unilateralism or international cooperation. In delegating to IOs, as elsewhere, “no pain, no gain”’ (Lake and McCubbins, 2006: 343). To find the right balance between control by principals and discretion of the agent to fulfill its mandate remains a crucial issue. For example, in dispute resolution, too much control undermines the purpose of delegation. At the same time, in delegation relations mechanisms of control are intentionally designed and used to minimize agency slack (Hawkins et al., 2006a: 31). One of the most frequently used control mechanisms available to principals to reduce slack is centralized monitoring in the form of regular reporting to committees, a mechanism identified in the literature as police patrol (McCubbins and Schwartz, 1984). However, police patrol mechanisms can be both timely and costly, and sometimes principals do not even have the expertise to evaluate their agents’ performance. Further, when selecting a certain degree of delegation and appropriate control mechanisms, principals must also consider the maximization of their overall return, not only just the minimization of agency slack (Hawkins et al., 2006a: 31).

As this brief discussion shows, one of the major dilemmas faced by principals is how not to lose control of their agents. These three dimensions – discretion, autonomy and agency slack – follow a sequential order. Agency slack cannot happen unless principals give their agents discretion and autonomy. The following section introduces the theoretical framework and explains how after agency slack has occurred the type of monitoring mechanisms, fragmentation and credible sanction mechanisms affect the ability of principals to regain control of their errant agents.

Agency slack and regaining control of agents: monitoring, fragmentation, and sanctions

Under what conditions is agency slack more likely? And are principals able to regain control after agency slack has happened? In reference to the first question, existing research focuses on the impact of preference heterogeneity or homogeneity within the collective principal to explain agent behavior (Nielson and Tierney, 2003). In reference to the second, control mechanisms and fragmentation are often used to explain the limits of principal control (Graham, 2013; Grigorescu, 2010). Several studies persuasively demonstrate that when agent behavior diverges from principals’ preferences, credible threats by principals to re-contract with the agent will reduce the gap between principals’ demands and agents’ subsequent behavior (Nielson and Tierney, 2003). However, fewer studies look at the ability of principals to regain control of errant agents. This contribution argues that once slack has occurred, the type of monitoring mechanisms (centralized or decentralized) in place, together with fragmentation within an IO, and the existence of credible sanctioning mechanisms shape principals’ ability to regain control of their agents.

Monitoring mechanisms can be centralized (police patrol) or decentralized (fire alarm) and are central to explaining principals’ ability to regain control of their errant agents. Centralized monitoring mechanisms refer to legal rules and formal texts in which it is stipulated that an IO will be scrutinized actively and directly by the principals with the aim of detecting and discouraging agency slack. Such police patrol oversight might include regular reporting in committees, questioning representatives from an IO, the preparation of reports, or even direct field observations in which principals participate in negotiations. Under a decentralized fire alarm system, principals externalize the costs of monitoring to third parties, including citizens, interest groups, courts, national parliaments, or new agents (McCubbins and Schwartz, 1984). This has the advantage of allowing principals to gather information at a lower cost. In addition, in a well-designed fire alarm system, third parties notify principals when agents violate their mandates (Kiewiet and McCubbins, 1991). Police patrols can be more easily evaded than fire alarm mechanisms, and additionally reflect a more routine monitoring process. As a result, their consequences may not be as dire as those arising from decentralized fire alarm mechanisms, which require less direct intervention by principals. In contrast to ex ante mechanisms of control, such as the delegation mandate, these two types of ex post monitoring and reporting requirements offer the opportunity to detect and can even deter slack. For example, regarding detection, while police patrol allows principals to identify malfeasance directly, fire alarm mechanisms rely upon affected parties outside the PA-relationship to bring evidence of slack to states’ attention (see also Hawkins et al., 2006a).

Even if principals have sanction mechanisms at their disposal, the existence of a fragmented organizational structure may undermine the effectiveness of monitoring requirements in a case of agency slack. While centralized monitoring mechanisms are better suited to help principals regain control over their agents after slack has occurred, decentralized monitoring mechanisms make it more difficult for principals to minimize agency slack, in particular in a situation of IO fragmentation. Fragmentation is the second causal mechanism to explain principals’ ability to regain control of their errant agents. The concept of fragmentation within an IO refers to the existence of the division of power between central and regional or national levels. Institutional fragmentation is more likely to inhibit the efficacy of control mechanisms exercised by collective principals because, when power is dispersed across regional and country offices, principals’ attempts to regain control might lead the collective agent to respond in uncoordinated ways or not even react to principals’ demands (see also Graham, 2013).

Finally, the existence of credible sanctioning mechanisms is, together with the type of monitoring mechanisms and fragmentation within an IO, equally important to explain whether principals are able to regain control of their errant agents. If monitoring and reporting mechanisms are inefficient, principals can also reduce slack by punishing their agents for undesired actions through the use of sanction mechanisms. One of the most effective measures to sanction against errant agents is to cut agencies’ budgets. Sanctions allow principals to punish agents for undesired actions or to reward agents if principals are satisfied with their performance. This carrot and stick approach can take the form of budget expansion or contractions. When principals are satisfied with the performance of their agents they can reward them with larger budgets. By contrast, agents who fail or engage in slack can be punished with a smaller budget (Hawkins et al., 2006a). Other measures include dismissing IO personnel, overruling the IO by enacting new legislation, refusing to comply with IO decisions, or restricting agents’ mandates (Pollack, 2003). Effectiveness of police patrol mechanisms is more likely to be low when states, represented by their agents (e.g. the Director-General of an IO), are not able to exercise direct control within an organization. This could be the case when there is a situation of fragmentation within a collective agent, when the Director-General does not exercise budgetary control over regional and country offices, or when the Director-General does not have the competence to hire and dismiss staff (Graham, 2013). Finally, in cases of dissatisfaction with the performance of an agent, principals can theoretically turn to other agents available in a certain policy field. In practice, however, the pool of agents is limited (Hawkins and Jacoby, 2006), so this is not usually a credible option. ¹ The above argument generates the following proposition:

After agency slack has occurred, the ability of member states to regain control depends on the type of monitoring mechanisms, the extent of fragmentation within an IO, and the existence of credible sanction mechanisms available to principals.

In other words, if control mechanisms are ineffective, there is high level of fragmentation within an IO and sanction mechanisms are not credible, the ability of member states to minimize slack and regain control of their errant agents is likely to be low.

In this contribution, agency slack is a dichotomous variable measured as either high or low. Agency slack is most likely to be high when decentralized monitoring mechanisms are in place, for example via interest groups, courts or parliaments. In the absence of regular written or oral reporting requirements, which presuppose IO officials will report back to the headquarters on the state of activities, agency slack is likely to increase. Et ceteris paribus, agency slack is low when centralized monitoring mechanisms exist, such as regular reporting within a committee. Police patrol is expected to be more effective when members of these monitoring committees are representatives of member states and IOs. Thus, centralized police patrol is less likely to lead to agency slack than decentralized fire alarm monitoring mechanisms. After agency slack has happened, the ability of member states to regain control of their errant agents depends on the type of monitoring mechanisms, the extent of fragmentation within an IO, and the credibility of sanctions available to member states. The following sections analyze two case studies on agency slack: at the Commission during the Eurozone crisis in 2015; and at the WHO Secretariat in the 1990s.

Centralized monitoring mechanisms: agency slack at the commission

During the Eurozone crisis, the European Commission, the Internal Monetary Fund (IMF) and the European Central Bank – the so-called troika – was in charge of supervising the implementation of the conditional bilateral financial loans to Greece, Portugal, Ireland, and Cyprus. This mandate was formalized in a memorandum of understanding (MoU) between the Eurogroup member states and the troika. In this new institutional configuration, the Commission was granted a discretion-based mandate, providing it with the authority to negotiate the conditions attached to financial assistance on member states’ behalf.

The PA literature suggests that discretion-based delegation enhances the policy-making role of agents. Discretion can be useful under two conditions. First, when uncertainty is high and flexibility necessary, it enables agents to fulfill their mandates without too much interference from their principals. Second, when tasks require a high degree of specialized agent knowledge, discretion helps agents in effectively accomplishing their missions (see also Hawkins et al., 2006a: 28). The latter was the case during the Eurozone crisis, in which member states relied on the technical expertise of the troika institutions. Moreover, the IMF and the Commission had a permanent mission with several technical experts in the country concerned (interviews with IMF and European Commission representatives). The Commission’s mission included representatives from the different Directorates-General, such as Economic and Financial Affairs, Social Affairs, Health, Cohesion, Energy, Education, and Environment (interview with European Commission representatives). Discretion can also be helpful when principals have heterogeneous preferences (Hawkins et al., 2006a: 28). Principals can leave it to agents’ discretion to decide where to set policies or how to implement programs to prevent policies and programs from being overturned by a group of disappointed principals (McCubbins and Page, 1987: 418). High discretion can, however, create greater opportunities for agents’ opportunistic behavior, leading them to gain autonomy and to exercise slack (Hawkins et al., 2006a: 28).

In May and June 2015, there was a situation of agency slack between member states and the Commission. Despite the opposition of Eurozone Member States towards bridge financing under the European Financial Stabilization Mechanism, the fund managed by the Commission on behalf of the 28 member states, the Commission’s president Jean-Claude Juncker decided to initiate discussions with the Greek prime minister on bridge financing (Lamer, 2015; interview with officials from the European Commission). This led to virulent reactions by some European Union (EU) member states. In particular, the German finance minister, Wolfgang Schäuble, supported by the Netherlands and Finland, found that the Commission had overstepped its mandate in negotiating Eurozone aid to Greece (Mussler, 2015). The president of the Commission went beyond his mandate because he was not only concerned with the implementation of the MoU. Rather, he also sought to maintain the euro, one of the major milestones of the European integration process. Principals reacted to this betrayal of trust by threatening to disempower the Commission. Germany, the Netherlands, and Great Britain initiated a discussion about how to constrain the powers of the Commission in order to avoid further situations of slack. Schäuble even proposed creating an ‘independent political administration’ with the objective of restricting the Commission’s competences to its role as guardian of the treaties, limiting its mandate to the management of the internal market and competition law. The Dutch finance minister, Jeroen Dijsselbloem, reinforced this position by stating that this issue would be one of the Dutch government’s priorities when it assumed presidency of the Council in 2016 (Lamer, 2015). The existence of these credible sanction mechanisms allowed principals to sanction their errant agent through budgetary and personnel contractions as well as through the restriction of the Commission’s competences.

After slack occurred, member states were able to sanction the agent and to regain control because they had extensive centralized monitoring and reporting mechanisms at their disposal the extent of fragmentation within the Commission was low. Monitoring and reporting mechanisms included the Economic and Financial Affairs Council (ECOFIN), in which finance and economy ministers are represented, and the Eurogroup, ² which is composed of the finance ministers of the euro area member states. ³ The collective principal extensively used centralized monitoring and reporting requirements and had credible sanction mechanisms (budget contraction and disempowerment of the agent) at its disposal.

Extensive centralized monitoring with a strict division of work between the two monitoring committees ECOFIN and Eurogroup was the first mechanism that allowed member states to regain control. In addition, the Commission was required to report to the Council preparatory committees in charge of providing financial assistance to the Economic and Financial Committee (EFC) and to the Eurogroup Working Group (EWG) in the remaining cases. The Commission reported to these two working groups on the design, negotiation and implementation phases of the financial assistance programs. The EFC/EWG president, in turn, informed the members of the EFC/EWG about the Commission’s reports. Following each program review, the Commission provided regular updates on the progress of review missions and macroeconomic developments in the countries concerned. On the basis of quarterly compliance reports prepared by Commission’s staff, the Eurogroup assessed progress with respect to program implementation and then decided on whether to continue the financial assistance program (European Parliament, 2014). These two additional centralized monitoring mechanisms gave principals extensive information about the implementation status of the conditional loans issued to Eurozone countries under the bailout program. At the same time, they allowed principals to effectively monitor the actions of their agent. On the fragmentation side, because Commission officials regularly reported back to the Directorate-General for Economic and Financial Affairs in Brussels and to the EFC and EWG, member states were able to regain control of their agent. The existence of credible sanction mechanisms, in the form of budget reduction and disempowerment of the agent on the competence side, proved to be a further very effective mechanism for principals to punish their agent for undesired actions.

This case illustrates how a combination of centralized monitoring mechanisms, low fragmentation within the Commission, and credible sanction mechanisms, allowed member states to counteract undesired Commission behavior after agency slack had happened. The existence of extensive police patrol mechanisms at the Council – in the form of committees (ECOFIN, the Eurogroup, and the economic and financial working groups) – together with a low fragmentation within the Commission with extensive reporting to the Directorate-General for Economic and Social Affairs in Brussels, and the existence of credible sanction mechanisms in the form of budgetary contractions and restriction of the agent’s competencies allowed Member States to reverse slack and to regain control of their errant agent. Member States also successfully discouraged future slack by threatening to disempower the Commission.

Decentralized monitoring mechanisms: agency slack at the WHO secretariat

The divergence of agent preferences from those of some members of the collective principal has been a constant since the creation of the WHO. During the 1980s, the WHO Secretariat promoted policies aligned with the demands of developing countries, for example health programs that accepted social development as a rationale for funding. This led to dissatisfaction among other member states, in particular the US, a major donor country. Critics blamed the WHO Secretariat for irresponsible budget growth and for managerial incompetence. The US considered that the WHO had overstepped its mandate and decided to turn to other agents, including the World Bank and the IMF, which incorporated health into their programs. For example, the Reagan administration gave the World Bank a budget for health promotion activities that was higher than the WHO’s annual budget. At the same time, member states agreed to cut the regular budget of the WHO, donate more money to discretionary funds and introduce a new category of program-specific extra-budgetary funds (Chorev, 2013). These decisions constituted an attempt to sanction the runaway agent. A third measure used by member states to regain control of the WHO Secretariat consisted of creating a third party, the UNJIU, which is an independent external monitoring body of the United Nations, to oversee the WHO.

The UNJIU has evaluated the WHO’s performance three times since 1991. These reviews assessed the WHO’s ability to achieve its central mandate of promoting and improving global health, with the UNJIU reporting its findings back to WHO member states after each evaluation. In 2001, the UNJIU prepared several reports and recommendations for improving WHO performance and for regaining control of the organization, concluding that the WHO’s impact on the health systems of its Member States was currently ‘very limited’ (UNJIU, 2001). The WHO Secretariat responded by ignoring many of the UNJIU’s recommendations and only agreeing to implement one-third of the suggestions (UNJIU, 2012: 48).

In response to this failure, member states decided to establish two new oversight mechanisms – the Office of Internal Oversight Services (OIOS) and the Independent Expert Oversight Advisory Committee (IEOAC) – and to employ an external auditor to regain control of their agent. Whereas OIOS was placed in charge of conducting internal audits, monitoring the WHO, and evaluating the WHO’s efficiency and effectiveness in implementing the programs and mandates of the United Nations (Office of Internal Oversight Services, 2016), IEOAC is in charge of advising the Program, Budget and Administration Committee and the Executive Board of WHO regarding the adequacy of WHO’s external and internal controls and audits (WHO, 2013). Finally, the external auditor is responsible for monitoring all of the WHO’s operations, including risk management and evaluation of the efficiency of the WHO’s internal control system (WHO, 2016).

The UNJIU inspectors came to the conclusion that the internal audit office, which only employed one full-time evaluator, did not have sufficient resources to effectively achieve its mandate. The UNJIU consequently recommended that WHO member states increase the personnel resources of this oversight committee. The introduction of the OIOS into the WHO illustrates how hiring a third party to monitor the new agent can have the unintended consequence that the monitoring agent recommends even more resources and more decision-making competences be allocated to the new agent, as occurred with the evaluation of the OIOS by the UNJIU.

In 2009, the WHO Executive Board created a second internal monitoring mechanism, the IEOAC, which reports to the Program, Budget and Administrative Committee of the Executive Board (PBAC) (UNJIU, 2012: 47). IEOAC functions include: reviewing the financial statements of WHO; advising on the adequacy of the organization’s internal controls and risk management systems; and controlling the effectiveness of external and internal audit functions (WHO, 2014). The PBAC, in turn, oversees the internal audit office and the external auditor. However, the WHO’s chief internal auditor reports directly to the WHO Director-General. A report by the US Government Accountability Office consequently concluded that, by reporting to the WHO Director-General rather than to the PBAC, the chief internal auditor of the IEOAC lacks the organizational independence necessary to monitor implementation because it is dependent on the entity under audit. The WHO reacted to this report by opposing any alterations to the current rules, arguing changes would create a second external auditor and leave the WHO Director-General without an internal audit function (United States Government Accountability Office, 2007: 49). As a result, the creation of these three new internal monitoring mechanisms, supervised by a third party, has not minimized agency slack.

The fragmented structure of the WHO further inhibits the efficiency of such fire alarm mechanisms. The WHO has a highly decentralized organizational structure, comprising six regional offices and 154 country-level offices located in national health ministries (Graham, 2013). This dispersion of power hinders principals from reining in their multiple agents (e.g. the WHO Secretariat, regional offices and country offices), all of whom have different preferences. It also prevents principals from strengthening monitoring mechanisms at these different levels. For example, member states cannot effectively use threats of dismissal against regional directors, as the latter are not elected at the WHO-level but by representatives in their respective region. In addition, structural fragmentation between the Director-General in Geneva and regional and country offices means in practice that monitoring mechanisms primarily scrutinize the WHO Secretariat in Geneva. In fact, the WHO Director-General does not have a formal system for monitoring country-level offices (Graham, 2013).

When decentralized monitoring mechanisms proved inefficient, member states turned to sanction mechanisms to regain control, cutting the WHO’s regular budget and dismissing personnel. The WHO’s budget is approved by the World Health Assembly on a biennial basis and comprises two elements: a core budget, also called assessed contributions; and additional funding, known as voluntary and extra-budgetary contributions. The core budget includes mandatory annual subscriptions from member states to cover the organization’s operating costs and to fund essential programs. Extra-budgetary funding comes from member states and from institutions that support specific health initiatives or programs. Agency slack at the WHO led member states to create extra-budgetary funds. These funds comprised 40 percent of the WHO budget in 1980, more than 50 percent in 1990 (Vaughan et al., 1996: 254), and over 70 percent by 2011 (Lee, 2009; UNJIU, 2012). As a result, the WHO is strongly dependent on unpredictable voluntary contributions. In other words, by earmarking a majority of the WHO’s resources, member states have created an effective mechanism to regain control over the WHO Secretariat. If a program is not effective in the sense that it improves health standards in a country, member states can simply refuse to continue financing that program in the next biennial budget.

However, several control problems remain. First, in the 1990s, the WHO headquarters had little information about how funds were spent at the regional- and country-levels. Second, UNJIU inspectors could not ascertain what proportion of WHO resources were allocated for regional- and country-level programs and could not find sufficient information to determine whether country health programs corresponded to the stipulated indicators for low-income countries or disease prevention. Third, after the African Regional Office refused to provide the requested information to the WHO External Auditor, the latter resigned, stating she could no longer guarantee the efficient functioning of the auditing mechanisms (Graham, 2013: 15). The lack of transparency regarding several of the regional offices thus inhibited effective monitoring by internal and external audit committees.

Apart from reducing voluntary contributions to the budget, a second effective sanction mechanism consists of dismissing, or refusing to reappoint, staff of runaway bureaucracies. When the WHO was created in 1948, it had a total of 259 staff members. This number increased continuously until 1991, when WHO personnel peaked at 4,693 permanent staff members. Following agency slack in the 1990s, the size of staff was reduced, reaching a low-point of 3,846 in 1998. However, from 1999 to 2014, staff size grew again, and the WHO today has a total permanent staff of 6,233 (WHO Human Resources, 2015). The staff growth starting at the end of the 1990s can be explained by the fact that Member States are no longer the sole source of core voluntary budgetary contributions. Even though, at 300 million dollars, the US still had the highest combined voluntary and assessed contributions to the WHO in 2014, the voluntary contributions of the Bill & Melinda Gates Foundation, at 256 million dollars, position it as WHO’s second largest voluntary donor (WHO, 2015). Thus, the reduction of budget and staff as an efficient control instrument against errant agents only works when member states have a monopoly over the provision of budgetary contributions.

In addition to reducing voluntary contributions to the budget or dismissing personnel, principals can also turn to other agents active in the same issue area to sanction runaway agents. Even though at the time of its creation the WHO was the only global health organization, there are now several other IOs – such as the World Bank, United Nations International Children’s Emergency Fund (UNICEF), United Nations Population Fund (UNFPA), or Joint United Nations Programme on HIV/AIDS (UNAIDS) – that have expanded their activities to include health issues. The mandates of these organizations overlap, leading them to compete with the WHO in acquiring resources to implement health programs. More importantly, IOs are not the only players in global health. In the last decade, several public–private partnerships, such as the Global Fund to Fight Aids, Tuberculosis and Malaria, the Global Alliance for Vaccines and Immunization, as well as private foundations, of which the Bill & Melinda Gates Foundation is the largest, have also assumed overlapping and competitive roles (UNJIU, 2012). These new actors have become important competitors for the WHO, as member states often turn to these new global health actors when pursuing health policy goals (United States Government Accountability Office, 2007: 34). A big pool of organizations in global health consequently impacts the effectiveness of other control mechanisms. Thus, sanctions are more effective when agents know that principals have other options.

This section has discussed a situation where decentralized monitoring mechanisms, sanction mechanisms and high fragmentation within an IO, hindered principals’ ability to minimize agency slack. To regain control over their runaway agent, principals reduced the WHO’s core budget, created extra-budgetary funds for specific programs, and dismissed personnel. However, the WHO was structurally fragmented, and monitoring mechanisms at the regional- and country-levels were practically nonexistent. As a result, agency slack could not be reduced, and the delegation relationship became even more complex as the number of oversight committees and external parties involved in the monitoring process grew. In sum, the case of the WHO Secretariat illustrates how a combination of decentralized monitoring mechanisms, a fragmented structure, and alternative material resources limits principals’ ability to rein in errant agents.

Conclusion

Principals delegate power to IOs in the hope that organizations will fulfill their designated mandates. While delegation can have this effect, in some PA relationships agency slack occurs forcing principals to expend additional resources and time to monitor their agents. In this article, I explore the conditions under which principals are able to regain control of their agents after agency slack has happened. I have argued that a combination of centralized monitoring mechanisms in the form of regular reporting to different committees, low level of fragmentation within an IO, and credible sanctioning mechanisms available to member states allow them to reverse slack and thus to regain control of their errant agents. This was exemplified in the case study of the Commission acting as an agent of the Council of Ministers during the Eurozone crisis. In the case of the European Commission, member states were able to easily reverse Jean-Claude Juncker’s undesired actions because they had extensive police patrol mechanisms available, a low level of fragmentation existed within the Directorate-General for Economic and Financial Affairs, and member states additionally possessed credible sanction mechanisms – such as budgetary contraction, dismissing personnel and reversing competences – that would have allowed them to disempower the Commission. By contrast, the WHO Secretariat case shows that decentralized fire alarm mechanisms, together with a high level of fragmentation, and the inexistence of credible sanction mechanisms did not minimize agency slack. Even though principals reduced the Secretariat’s core budget and dismissed staff this proved inefficient to reverse slack because the agent had alternative sources of financing.

The case of the WHO Secretariat further shows that, when IOs are not dependent exclusively on member state budgetary contributions, sanctions in the form of reduced budget and personnel do not work as efficient means of regaining control over runaway agents. Hawkins and Jacoby (2006) assume that the costs of creating new agents is very high for principals. This contribution brings this research a step further by demonstrating that, instead of creating new agents, principals can minimize costs by turning to existing agents, as the US did in the case of the WHO Secretary by incorporating health issues into the programs of the IMF and the World Bank and by transferring health funds to other organizations working on health issues, such as UNICEF, UNFPA, and UNAIDS. In the long run, if a single principal within the collective principal is dissatisfied with the malfeasance of the agent, as the US was in the WHO case, this can lead to the decline of the organization as a whole. Put differently, agency slack can have a devastating effect when an IO does not have alternative financing resources at its disposal.