Raison d’être. How Political Salience Affects the Independence of Independent Regulatory Agencies in the Netherlands

(1)

August 2020

Master Thesis Charlotte (C.) van Tongeren

MSc. Public Administration: Economics & Governance Leiden University, Turfmarkt 99, The Hague

Supervisor: Maarja Beerkens

Raison d’être. How Political Salience Affects

the Independence of Independent Regulatory

Agencies in the Netherlands.

ABSTRACT: Independence is a necessary requirement of independent regulatory agencies (IRAs) placed at arm’s length of government in order to achieve effective enforcement of and compliance with regulation, and credible commitment of politicians. Accountability and the related control mechanism available to government legitimizes this delegation of regulatory authority to IRAs, but at the same time risks compromising IRA independence. The risk is even greater when IRAs and regulatory matters attract public and political attention directed towards both agents and principals. This thesis researches the effect political salience has on the establishment and use of ex post control instruments and the consequent effect on formal and de facto independence of IRAs. A mixed methods analysis into 74 IRAs in the Netherlands and a subset of eight market regulators concludes that political salience positively affects the scope and variety of ex post control instruments, and has a small positive effect on their use. However, the establishment and use of ex post control instruments does not necessarily affect either formal or de facto independence. The thesis concludes that political salience has a small and statistically insignificant effect on IRA independence.

(2)

I. Introduction

Raison d'être | rey-zohn de-truh - reason or justification for existence

1.1 A Delicate Balance

Production and consumer markets controlled by the central government: it is barely imaginable and highly undesirable in the current liberal global order. But it has only been since a few decades that the terms and conditions of free trade are safeguarded by regulations and institutions that are granted authority by, and are explicitly separate from, central governments. The rise of the regulatory state from the 1970’s onwards heralded the departure from the positive state as planner, producer and employer and introduced the proliferation of independent regulatory agencies, or ‘highly specialized organizations enjoying considerable autonomy in decision making’ (Majone, 1997: 152). Such agencies provide policy makers with essential sectoral expertise, the possibility of making credible commitments to market participants, and the (cost-)efficient implementation of policies (Majone, 1997; Van Thiel, 2019). Crucial in achieving these benefits is the ability of agencies to act at arm’s length of the majoritarian central government, implying a significant degree of autonomy and discretion (Majone, 1997). As a result, most regulatory agencies are formally independent – some to greater extent than others – meaning that their powers and responsibilities are stated by law, that they are organizationally separate from ministries, and that they are not managed by elected officials (Thatcher, 2002).

However, delegating authority to regulatory agencies and placing the enforcement of regulation at arm’s length of those democratically elected impairs the legitimacy of policy making and implementation if left unchecked. For this reason the accountability of independent regulatory agencies to their principal is formalized through statutory control mechanisms that outline what objectives agencies should pursue and audit whether agencies have been successful in doing so (Busuioc, 2009). And so a delicate balancing act emerges: whilst the concept of ‘independence’ implies a necessary institutional distance between central government and regulatory agency, ‘accountability’ implies a necessary nearness between the two.

The result of this continuous balancing act has been that critique on the independence of regulatory agencies is twofold: (i) regulatory agencies are either too independent from government and act according to their own interpretation of what is in the public interest; or (ii) regulatory agencies are de facto not independent from government and political involvement infringes agencies’ raison d’être i.e. their independence. Especially in policy areas where

(4)

delegation has occurred with the specific objective of achieving credible commitment the critique on the balancing act tends to tip towards the side of too little independence. In the field of utilities regulation for example the relationship between politics and independent regulation has been experienced as a return of politics into regulation i.e. an increased role played by politicians in independent regulatory practices over the past years (Finger, 2019).

1.2 A Dutch Balance: Ministerial Responsibility

In the Netherlands especially the debate has been a pressing issue because the balancing act mimics a performance of three acrobats instead of two: agency independence, agency accountability, and ministerial responsibility (ministeriële verantwoordelijkheid). The provision of ministerial responsibility lies at the heart of public governance in the Netherlands ever since the Revision of the Constitution in 1848 stripped the sovereign of its political powers and granted them to a cabinet of elected ministers (Parlement, 2020(a)). The provision states that ‘the Ministers, and not the King, shall be responsible for the acts of government’ (art. 42 lid 2 Grondwet 1815), meaning that ministers can ultimately be held accountable – and can consequently be demanded to resign – by parliament for all acts performed by and in name of their departments. In other words, no matter how far the institutional and organizational distance between the conduct of the agent and the control of the principal, the minister remains accountable for agency results – the good, the bad and the ugly. Whereas in most countries the democratic legitimization of delegation is achieved through direct accountability of agencies to parliaments, in the Netherlands it takes shape as an indirect accountability mechanism that places the minister (or state secretary) between agency and parliament. For a minister this means that, in extremis, holding the agency accountable equals holding oneself accountable. A dangerous position for any executive to be in, especially when the executive is expected not to meddle with agency conduct.

The debate arose in the Netherlands, as everywhere, as a result of the rise of the regulatory state that was marked by large-scale privatization and agencification (verzelfstandiging). Although public bodies operating at arm’s length of central government were not a new phenomenon in the Netherlands – having a long-standing decentralized governance structure e.g. provincial and municipal administrative bodies – their proliferation as part of austerity measures and civil service reform from the early 1970’s onwards was unprecedented. The number of agencies – coined Zelfstandig Bestuursorganen (ZBOs) – increased without an unambiguous or centrally managed policy; sometimes with the purpose of separating policy making from implementation, other times with the purpose of achieving cost

(5)

efficiency. The result was an absence of cross-agency coherence and a diluted relation between agency and the responsible minister (Dijkstal, 1995; Kennis Openbaar Bestuur, 2020). More importantly still, the primacy of politics (primaat van de politiek) and by extension the constitutional provision of ministerial responsibility was at stake. And so, the independence-accountability balancing act in the Netherlands was rather turbulent instead of delicate with the scale tipping over to one side and, consequently, to the other: from unanimous support for large scale delegation and agency independence, to an extensive process of accountability formalization resulting from a perceived and confirmed loss of principal and public control (Yesilkagit, 2004; Scheltema, 1992; Sint, 1994; Algemene Rekenkamer, 1995). To address this loss, the first “purple cabinet” in Dutch parliamentary history – consisting of a coalition of labour (PvdA), conservatives (VVD) and liberal democrats (D66) – decided that ‘in order to restore the primacy of politics, a more accurate control of the executive apparatus by cabinet and parliament is necessary’ (Kok, 1994: 24). The resolution was followed up by a four-part policy proposal to (i) establish new ZBOs only when necessary – not when pragmatic –, (ii) create an unambiguous framework in which ZBOs are required to operate, (iii) scrutinize and potentially undo existing ZBOs, and (iv) provide organizational arrangements to ensure appropriate use of control and as such effectuate the primacy of politics (Dijkstal, 1995). The proposals eventually materialized in the Framework Act ZBOs that came into force in 2007 after years of debate, including a revolt by a small number of ZBOs against “autonomy-limiting measures” following the four-part policy proposal (Yesilkagit, 2004).

1.3 Restoring the Primacy of Politics or Meddling?

It has been thirteen years since the Framework Act has gone into force and the objectives of limiting ZBO-proliferation and increasing harmonization have been achieved, and ministerial responsibility has been safeguarded (Donner, 2011). But little notice has been given as to how the appropriate establishment and use of principal control instruments to effectuate the primacy of politics fits within the broader academic discussion on the independence-accountability balancing act. There are overt similarities between ‘restoring the primacy of politics’ and ‘a return of politics into regulation’, yet the former is defined as an objective and the latter as a risk. Such a fine conceptual line begs the question whether the status quo of IRAs’ independence and accountability in the Netherlands is well-balanced, and what causes potential imbalance.

A cause for tipping the scale may be political salience, or ‘the degree of importance which is attributed to political matters (Koop, 2011: 210). It is argued that the degree of political salience of issues is positively related to the number of accountability measures implemented for

(6)

ZBOs dealing with these salient policy issues, causing differences amongst ZBOs (Koop, 2011). It is, however, not clear whether these differences in agency accountability caused by political salience have an unbalancing effect – i.e. affect the independence of ZBOs – and if so, how they create the unbalance.

This thesis addresses the differences in formal independence amongst ZBOs, attempts to explain whether and how these differences are caused by political salience, and assesses how these differences affect the control behaviour of the principal. To this end the following research question will be answered:

How does political salience affect the establishment and use of ex post control instruments by the political principal and consequently the formal and de facto independence of independent regulatory agencies in the Netherlands?

IRA independence is a highly contested topic that enjoys the continued interest from policy makers – both on a national and international level –, academia and those affected by regulation. Yet the debate remains abstract and the vast variety of IRAs is of little help in deciding on a one-size-fits-all theoretic and policy framework for the assessment of (the functioning of) IRAs. Getting lost in a theoretic and administrative debate on agency independence risks losing sight on whether society actually benefits from IRAs. The research question will address both the academic and societal relevance of the debate on agency independence, because in attempting to answer it the abstract debate is made concrete and IRAs are tested on their raison d’être.

1.4 Making the Abstract Concrete

The research question will be answered using a mixed methods research design. Through the combination of quantitative large-n regression analysis and qualitative small-n comparative analysis the effect of political salience on the design of formal independence and specifically on the establishment and use of ex post control instruments can be uncovered and comprehended.

For the purpose of the large-n analysis a database was constructed of 74 ZBOs in the Netherlands to which the Framework Act applies. These ZBOs are assessed using a number of variables that link political salience to formal independence in order to conclude whether there is a significant relation between these two concepts. This will provide an initial test for the hypothesis that political salience has a negative effect on formal independence. Subsequently, a subset of eight market regulators from the total sample of 74 ZBOs will be assessed in more

(7)

depth. Their general and annual political salience covering a ten-year period from 2009 to 2019 will be measured against their formal independence, the number and variety of ex post control instruments established for each, and the use of these instruments in the period of interest in order to conclude whether the potential relation found through the large-n analysis is the result of the assumed causal mechanism.

In the following the most important concepts and theories leading up to the hypotheses and analysis will be discussed. The benefits and limitations of delegating authority to IRAs will be discussed, as well as the interaction between agency independence, agency accountability and principal control, and the significance of political salience in relation to these concepts.

(8)

II. Theories of Regulatory Independence

2.1 Principal-Agent Theory

The delegation of public authority from political actors to (independent) organizations such as IRAs is an administrative phenomenon that can best be explained through principal-agent theory. Even though relevant questions have been raised about the explanatory value of the theory concerning agency accountability in practice, the theory does provide a suitable framework for better comprehension of the motifs for establishing IRAs and the probable benefits and limitations associated with it (Schillemans and Busuioc, 2014; Thatcher and Stone Sweet, 2003). In public administration, principal-agent theory is concerned with delegation and public accountability (Gailmard, 2012). Who bears responsibility for policy implementation and who in fact implements it? In short, who is accountable to whom? In simple terms, the principal-agent framework describes the relationship between two actors (or groups of actors) of which one, the agent, performs actions on behalf of the other, the principal, and the latter devises an objective and incentive structure for the actions of the former (Gailmard, 2012). The principal-agent relationship is often the product of necessity: the principal simply cannot perform all the tasks for which it is responsible itself due to a lack of resources. As such it is a relationship that occurs naturally and primarily for practical reasons. However, the delegation of authority to an agent is also associated with an increase in flexibility and expertise, which is why the establishment of a principal-agent relationship can actively be pursued by the principal. Such benefits have in no small part been the reason for the proliferation of IRAs across economically liberal countries as can be seen from the vast increase of ZBOs in the Netherlands.

But, risks are attached to the principal-agent framework as well. The principal-agent relationship is contractual in nature and is assumed to be non-cooperative; game theory suggests that actors are rational and aim to maximize their own utility (Gailmard, 2012). The agent can therefore decide to engage in self-interested behaviour and pursue objectives that are not those of the principal (Thatcher, 2011; Busuioc, 2009). Even more problematic, the act of pursuing them may even disadvantage the principal in achieving its own objectives. Such ‘agency loss’ is often the result of information asymmetry, which means that the agent possesses information about the task assigned, its own preferences, abilities and possibilities that the principal does not. This provides the agent with significant discretion about when and how to act, and defers the principal from knowing when and how to apply control (Majone, 1997). The behaviour resulting from the agent’s superior information position often confirms the fear of critics that IRAs have too much independence.

(9)

The problem is that information asymmetry is the natural by-product of utilizing agency expertise, which is often the reason to establish a principal-agent relationship in the first place. In order to address this dilemma, increase its information position and prevent agency loss, the principal can generally act in one of three ways: (i) monitoring the agent’s behaviour, (ii) contractually binding the scope of discretion (ex ante control instruments), and (iii) providing a structure of incentives (Majone, 1997; Laegreid and Verhoest, 2010). Both the risk of agency loss and the actions the principal can take to address the risk apply to IRAs and their political principals.

2.2 Independent Regulatory Agencies

Independent agencies are ‘those governmental entities that (a) possess and exercise some grant of specialized public authority, separate from that of other institutions, but (b) are neither directly elected by the people, nor directly managed by elected officials’ (Thatcher and Stone Sweet, 2003: 2). The most common benefits of establishing IRAs for principals are (i) (cost-) efficiency, (ii) expertise, (iii) credible commitment, and (iv) blame avoidance. Efficiency – and more specifically cost-efficiency – should be understood as the overarching benefit which is the product of the sum of the other benefits discussed. It is assumed that organizations operating at arm’s length of central governments are able to act more competitively i.e. business-like concerning general organizational management, service delivery and meeting demands (Van Thiel, 2019). As such, agencies provide better value for money than does the traditional bureaucracy – an assumption which reflects the application of private sector tools and instruments to the mode of governance in the regulatory state. Closely related to (cost-) efficiency and the subsequent benefit is the notion of specialization. Delegation to agencies that have limited and non-contesting objectives as well as single-purpose tasks should result in improved regulatory performance compared to the traditional bureaucracy (Thatcher, 2002; Laegreid and Verhoest, 2010).

Consecutive to this, delegation most often occurs in regulatory fields that require specific scientific, engineering and economic knowledge which is not sufficiently available within central government administration (Majone 1997). Due in part to the high level of specialization of agencies, IRAs provide and attract the technical expertise to address such complex regulatory fields. In doing so, IRAs are also able to alleviate information asymmetries that traditionally exist between governments and regulatees in favour of regulatees, especially in sectors where competition has been recently introduced (Thatcher, 2002). Aside from being a reason to

(10)

delegate, expertise has also been a source of legitimization for the degree of independence allocated to IRAs (Majone 1997; Koop, 2011).

Even more important than providing expertise, and due to their expertise, IRAs operating at arm’s length of central government provide a solution to the credible commitment problem from which elected politicians, and by extension (top) civil servants, suffer. The credible commitment problem represents the difficulty policy makers who are subject to popular control and short-term politics encounter in achieving policy continuity: ‘under the expectation of alternation, politicians have few incentives to develop policies whose success, if at all, will come after the next election. Hence, it is difficult for political executives to credibly commit themselves to a long-term strategy’ (Majone, 1997: 153; Maggetti, 2007). However, especially in complex and highly internationalized areas, the capacity to create credible long-term policies is vital to secure investments and as a result achieve economic growth and stability. This is even more pressing where central banks and utilities – often natural monopolies – are concerned that were previously owned by governments. In such areas increased political involvement is assumed to harm the functioning of the market as it will discourage investors (Gilardi, 2002; Koop, 2011). Delegating regulatory authority to independent agencies relieves policy makers of their discretion and commits them to the apolitical long-term strategies of experts, increasing the credibility of their policy decisions (Majone 1997; Gilardi, 2002).

If the possibility to credibly commit to long-term policies is regarded as a positive feature of delegation to independent agencies, blame avoidance and reputation management can be regarded as its negative counterpart. Decision making – especially in complex policy areas – poses the risk of being held accountable for unsuccessful and/or unpopular outcomes. This is a risk politicians want to avoid by shifting regulatory authority and as such potential blame to independent agencies (Lodge and Wegrich, 2012).

The widespread introduction of IRAs over the years has also revealed common limitations of delegating regulatory authority to agencies. A general critique is that they have not achieved that for which they have been established, namely (cost-)efficiency (Van Thiel, 2019). More specifically, disadvantages of IRAs are (i) agency losses or regulatory drift, and (ii) the question of accountability.

The disadvantages of delegation already discussed in the principal-agent framework apply to IRAs due to the nature of their relationship with governmental actors: information asymmetries increase the risk of IRAs engaging in self-interested behaviour (Busuioc, 2009). This self-interested behaviour may include achieving the objectives for which the IRA is

(11)

established, but it may also include pursuing outcomes different from those preferred objectives through misusing/abusing the authority given to it (Thatcher and Stone Sweet, 2003; Koop, 2011). More specifically tailored to IRAs the concept is also called ‘regulatory drift’, which includes three means by which an IRA deviates from initial intent. One way is in pursuit of typical bureaucratic self-interest i.e. agency loss. The other two are the result of external pressures either from the regulated industry (capture) or from politicians (Lodge and Wegrich, 2012). The latter is the product of diverging interests that may encourage political actors to exert their influence over the regulatory process in order to achieve specific goals different from, or even contrary to, initial intent (Koop, 2011). Such political involvement is particularly evident in appointment processes and oversight hearings (Majone, 1997). The salience of political matters especially influences the principal’s need to pressure an IRA to behave in a certain manner, thereby achieving regulatory drift. The following sections on independence, control and political salience will further elaborate on this subject.

The final limitation associated with IRAs is the assumed lack of accountability from which agencies as non-majoritarian institutions inherently suffer. IRA-officials are not directly elected and when responsibility to either voters or their political representatives is considered a prerequisite for democratic legitimacy, delegation to IRAs logically implies a democratic deficit (Majone, 1997; Thatcher, 2002). In order to reduce this deficit, the principal – according to principal-agent theory – should establish an institutional framework that enables the principal to influence agency performance i.e. apply formal controls (Thatcher, 2002). And so a dilemma materializes: considering what has been said about agency loss, regulatory drift and the responsive application of control to assure IRA legitimacy, begs the question how a necessary degree of independence that is the raison d’être of the IRA can be achieved without jeopardizing the necessary degree of agency accountability (Busuioc, 2009).

2.3 Agency Independence

Independence is a precondition of allocating regulatory authority to agencies positioned at arm’s length of the central government. It is an essential aspect of guaranteeing high-quality enforcement of regulation and credible commitment (Majone, 1997; Majone, 2001). What is more, without independence an IRA is in fact not an IRA. Independence – interchangeably used with ‘autonomy’ – is generally explained as the synthesis of two components. One describing independence as a technical concept – relating to the institutional and legal basis – and the other describing independence as a behavioural concept – relating to agency discretion and autonomy in decision making (Busuioc, 2009). An often-used conceptualization of this

(12)

distinction – which will be adopted in this thesis as well – is ‘formal’ and ‘de facto’ independence. Formal independence ‘refers to a series of prescriptions, enshrined in the constitutions of agencies, which should guarantee independence from elected politicians’ (Maggetti, 2007: 272), and de facto independence ‘can be seen as a synthesis of two components: the self-determination of agencies’ preferences, and their autonomy throughout the use of regulatory competencies, that is, during the activity of regulation’ (Maggetti, 2007: 272; Majone, 1997; Gilardi, 2002). Why is this distinction relevant to make? Often when there is critique on the position of an agent vis-à-vis the principal it concerns the extent to which the agent is able to exercise its independence in practice; a discrepancy in formal and de facto independence is alleged. Thus, for analytical purposes it is important to distinguish between the degree of independence expected (i.e. granted by law) and the degree of independence observed. The observed or de facto independence sometimes exceeds and sometimes falls short of formal independence as a result of both agency attributes and external factors (Maggetti, 2007; Koop, 2011). Tying everything together, an agent exceeding its formal independence closely resembles agency loss and self-interested regulatory drift, and an agent falling short of its formal independence closely resembles politically pressured regulatory drift. Further stressing the necessity of the conceptual distinction, Maggetti (2007: 273) has found that formal independence is ‘neither a necessary nor a sufficient condition for a high level of de facto independence from politicians’. This suggests that formal independence and de facto independence not only differ from each other on occasion, but are unrelated entirely.

Thatcher and Stone Sweet (2003) provide an explanation of formal independence that better addresses the degrees of variance in formal independence amongst IRAs. Explained as ‘the zone of discretion’, formal independence is the construct of two components: ‘(i) the sum of delegated powers (policy discretion) granted by the principal to the agent, minus (ii) the sum of control instruments, available for use by the principals to shape (constrain) or annul (reverse) policy outcomes that emerge as a result of the agent’s performance of set tasks’ (Thatcher and Stone Sweet, 2003: 4). Within this zone, agency conduct and behaviour is effective, legitimate and accountable, outside the zone it is not. In other words, within the zone of discretion there is no discrepancy between de facto independence and formal independence; outside the zone, there is. To summarize, agency independence can be divided in formal independence and de facto independence, and formal independence in turn is the product of the sum of powers delegated to the agency minus the sum of controls available to the political principal. Figure 1 depicts how these powers delegated and controls established can create a customized formal independence that varies from IRA to IRA.

(13)

Understood as a variable zone of discretion, the degree of formal independence is IRA-specific: it is determined by the motif and the status quo at the time of establishment (Thatcher and Stone Sweet, 2003; Koop, 2011). Furthermore, because formal independence is the product of agency powers minus principal controls – that are necessary elements of accountability – the association with the accountability-independence dilemma discussed in the introduction is obvious. Although independence and accountability should not be considered contradictory concepts, it is apparent that too many powers delegated without controls established – i.e. agency discretion – compromises accountability, and too little agency discretion compromises independence (Busuioc, 2009). This raises the interesting question of what scope of an IRAs zone of discretion can be considered necessary and sufficient, and which actors and factors attempt to increase or decrease its scope. As will become clear from the discussion on principal control instruments, not all controls established limit the agent’s zone of discretion. Namely (the use of) ex post control instruments have this effect. The variable zone of discretion and the relation with the use of ex post control instruments will be returned to in the discussion of the causal mechanism and hypotheses.

2.4 Agency Accountability

Similar to independence, accountability is a precondition of allocating regulatory authority to agencies positioned at arm’s length of the central government. Accountability is ‘the degree to which an actor A is obliged to offer information on, and explanation of, his or her conduct to another actor B, and may be sanctioned or rewarded for this conduct’ (Koop, 2011: 216). Guaranteeing a significant degree of accountability is central to legitimate delegation of regulatory authority because it democratically links agency conduct to principal oversight. The concept consists of transparency on the one hand – from agent to principal – and liability on the other – from principal to agent (Koop, 2011). Majone (1997) distinguishes between ‘procedural

(14)

legitimacy’ and ‘substantive legitimacy’ in achieving agency accountability. Procedural legitimacy implies that the composition and conduct of IRAs follows a formal set of rules about, inter alia, the mode of appointment and decision-making. These rules are often laid down in agency statutes and legal texts of the regulations which the agencies are entrusted to enforce, enabling the principal (and third parties) to monitor and assess whether agencies act within their legislative mandate (Thatcher, 2002; Koop, 2011). Substantive legitimacy builds on procedural legitimacy, but implies more intangible requirements about the behaviour of agencies in line with the purpose of establishment such as professionalism, expertise and consistency (Majone, 1997). As opposed to procedural legitimacy, substantive legitimacy can only be ascertained after the fact. As will be shown, a similar distinction applies to the control exerted by the principal vis-à-vis the agent.

Accountability as discussed in academic literature on IRAs more often than not suggests that democratic legitimacy – either procedural or substantive – is derived from agency accountability to parliament. In many cases, a cabinet member or (top) civil servant will in practice be the go-between in this process of legitimization. However, in the Netherlands legitimacy via a minister or state secretary is a constitutional provision. This indirect link between IRA and democratic representatives further complicates the independence-accountability dilemma in two ways. First – as was discussed in the introduction and the section on regulatory drift – ministerial responsibility encourages the political principal to maintain a certain or significant degree of relatedness to the tasks and conduct of the agent. Second, control exerted by the political principal over the agent cannot unequivocally rely on democratic legitimacy. It is, after all, not the elected parliament that controls the IRA, but the head of the government department or bureaucracy. It may seem a somewhat irrelevant nuance, but when it comes to the credible commitment of governments to regulatory policies and the effects of (use of) principal control instruments on this commitment, there cannot be enough nuance.

2.5 Principal Control

Simply put, control is a means by which the principal can reduce agency loss, achieve the objectives of regulatory delegation and safeguard agency accountability. Control over an agency’s conduct and behaviour is applied before and after the fact – ex ante and ex post –, closely related to procedural and substantive legitimacy as was mentioned. Ex ante control is ‘a form of preliminary control mechanism’ (Busuioc, 2009: 607) and constitutes legal rules, regulations, operating procedures and approval requirements to direct the agency towards the regulatory objectives of delegation. Simply put, ex ante control instruments outline what the

(15)

IRA may do, must do, and may not do, and they establish the IRA’s and principal’s positions towards other.

Ex post control is concerned with objective setting, monitoring, evaluation, audit and sanctions, also based on legal grounds (Laegreid and Verhoest, 2010). Because ex post control assesses in hindsight whether the agency has acted and behaved within its mandate, Busuioc (2009) argues that the concept is synonymous with accountability and therefore both necessary for, and not limiting to, independence. This definition of ex post control is however rather narrow, as it does not include a principal’s ability to ‘shape (constrain) or annul (reverse) policy outcomes that emerge as a result of the agent’s performance of set tasks’ (Thatcher and Stone Sweet, 2003:4) at its own discretion rather than for the necessity of achieving accountability. I will return to this observation shortly. Other than applying to after-the-fact rather than before-the-fact oversight, an important difference of ex post control instruments is that some are applied during an IRAs lifetime. Ex ante control instruments on the other hand are formulated once, potentially mandated once or twice, and affect the agent whether the principal chooses to or not. The fact that the principal has at his or her disposal ex post control instruments does not mean that (i) the principal makes use of these instruments regardless, and (ii) that the use of these instruments in rare cases is illegitimate or limiting to agency independence. In short, the existence of formal control instruments must be explicitly distinguished from the use of formal control instruments. Formal control – similar to agency discretion or autonomy, though operating in the opposite direction – occupies a central role in the accountability-independence dilemma: too much control compromises independence, and too little control compromises accountability.

This does not mean that the zone of discretion in itself cannot be compromised by a principal’s application of control, bringing us to the concept of informal control. Also called ‘ongoing control’, informal control is not limited to application either before or after the fact and is not constituted in the laws governing the IRA. More specifically, it is ‘direct control exercised by a principal vis-à-vis an agent in which the agent’s (future) actions are steered and/or determined by the principal resulting in a decrease in the original mandated discretion of the agent to accomplish the delegated tasks and, implicitly, its decision-making autonomy’ (Busuioc, 2009: 607). From this definition it can be concluded that informal control limits agency independence and even more so voids an IRA’s raison d’être. Although this form of control is highly relevant to research in relation to independence-limiting means and conduct of the principal, it is equally difficult to observe and for this reasons falls outside the scope of this thesis. Focus will therefore be on ex post control instruments that are applied at the discretion of

(16)

the principal. Of all the formalized – and as such traceable – control instruments, these provide the principal with the most freedom to steer and direct, and if used often can be considered a form of meddling.

2.6 Political Salience

Why would the principal – effectiveness and accountability aside – want to exert control over the agent? One of the explanations suggested is that independence is correlated with the degree of political salience of the issues addressed by IRAs (Yesilkagit and Van Thiel, 2008). Political salience is ‘the degree of importance which is attributed to political matters’ (Koop, 2011: 210). Even if regulatory decision making is placed at arm’s length of the political centre, the electorate will still hold politicians accountable for policy and regulatory outcomes, especially when an institutional provision such as ministerial responsibility places ultimate accountability with these politicians. The more importance that is attributed to political matters, the more public and political attention there will be for the decision makers that are (directly and indirectly) involved with these matters. The psychological phenomenon of the empty nest syndrome provides an interesting metaphor to explain the effect such heightened political salience may have on the principal’s conduct vis-à-vis the IRA. It states that (persistent) meddling in IRA-affairs by politicians stems from a (perceived) loss of control. The need for control will be more pressing when the stakes are high i.e. when there is a high degree of political salience, and a loss of control will be perceived when the responsible principal has no direct influence over the salient political matters (Van Thiel, 2019). This is a risk for any principal that places regulatory authority over (potential) salient political matters at arm’s length. Because monitoring resources available to the political principal are scarce and, as a result, holding IRAs accountable is costly, the principal will prefer to direct its efforts and limited resources towards policy issues and IRAs that attract public attention (Koop, 2011).

Political salience can be divided in two types of salience that in this thesis will be conceptualized as general political salience and ad hoc political salience. General political salience refers to importance that is attributed to matters because they inherently and on average require and attract a certain degree of public and political attention. Such topics are often subject to international interdependence, concern formerly state-owned enterprises (often utilities), or are subject to many veto players (Gilardi, 2002; Koop, 2011). The IRAs that oversee the markets relating to these topics enjoy the same or similar degrees of public and political attention by extension. Ad hoc political salience is incident-driven; it can arise at any given moment as a result of any given situation. It can present itself for political matters and IRAs that

(17)

generally enjoy high degrees of salience, but can also manifest for political matters and IRAs that have no precedent for attracting public and political attention.

These two types of salience theoretically affect IRAs differently. General political salience will have an effect at the outset, meaning that with the establishment of an IRA measures will be included in the legal framework of the IRA – i.e. its ex ante and ex post control instruments – that provide the principal with the desired means to address the degree of salience of political matters with which the IRA is concerned. Ad hoc political salience will affect principal conduct vis-à-vis the IRA overtime, meaning that the principal will resort to using the control instruments already available if the sudden and significant public attention given to the political matters with which the IRA is concerned so requires. As such the effect of either form of political salience should be observed in different aspects of the principal control mechanism: general political salience will theoretically affect the establishment of both ex ante and ex post control instruments, and ad hoc political salience will theoretically affect the use of ex post control instruments.

Interestingly, again a dilemma emerges. High degrees of political salience require politicians to credibly commit themselves to policy issues due to the heightened attention from market participants and consumers for these matters. As a result principals prefer to place decision-making at arm’s length. Simultaneously, high degrees of (ad hoc) political salience increase the perceived loss of control and incite politicians to meddle with regulatory decision-making (Van Thiel, 2019). This places political principals in an awkward and difficult position vis-à-vis the IRA and the electorate/parliament.

2.7 Causal Mechanism and Hypotheses

As has become clear from the discussion above, I argue that political salience and perceived loss of control are connected concepts that together influence the principal’s need to exert control over the agent. The influence of political salience has both an a priori effect and an a posteriori effect; general political salience increases the need to establish control instruments at the outset and ad hoc political salience increases the need to use these control instruments during an agency’s lifetime. Such constraint and nullification of agency outcomes at the principal’s will can only be achieved through the establishment and use of ex post control instruments. Both their establishment and their (frequent) use have the potential of limiting the extent of agency discretion and by extension decreasing agency independence. The former (establishment) limits an agency’s formal independence and the latter (use) limits an agency’s de facto independence.

(18)

As a result, overall IRA independence decreases as a result of political salience through ex post control instruments. Figure 2 provides a visual representation of this dynamic.

Figure 2. Causal Mechanism

+ -

The research question is as follows: How does political salience affect the establishment and use of ex post control instruments by the political principal and consequently the formal and de facto independence of independent regulatory agencies in the Netherlands?

At the outset of delegation of regulatory authority to an IRA the principal intends to achieve credible commitment by granting significant decision making and enforcement powers. Simultaneously, the principal will ensure agency accountability by implementing a formal control mechanism consisting of both ex ante and ex post control instruments. Where politically salient policy issues are concerned the need to formalize agency accountability will be greater because politicians remain accountable for policy outcomes i.e. ministerial responsibility. Ex post control instruments provide the principal with the means to correct agency conduct after the fact, which is especially desirable in exerting control over salient regulatory issues. The related hypothesis is the following:

• Hypothesis Ia: General political salience increases the principal’s need for IRA accountability and principal control which increases the sum of ex post control instruments available for use by the principal at the time of establishment.

The fact that the principal has a formal control mechanism at his or her disposal does not entail that he or she will apply it in full regardless. Once the IRA has been established the extent to which the principal intends to make use of the ex post control instruments from the formal mechanism overtime will depend on the extent to which the principal perceives a loss of control. This will be the case when the IRA is engaged with a policy issue that suddenly becomes salient, Political Salience Use of Ex Post Control Instruments Establishment of Ex Post Control Instruments Formal Independence De Facto Independence IRA Independence =

(19)

thereby heightening public interest and emphasizing principal accountability and, in the Netherlands, ministerial responsibility.

• Hypothesis Ib: Ad hoc political salience increases the principal’s perceived loss of control which increases the extent to which use is made of the ex post control instruments available to the principal overtime.

Together, these hypotheses construct the main hypothesis which is also the preliminary answer to the research question:

• Hypothesis I: Political salience increases the establishment and use of ex post control instruments and consequently limits the formal and de facto independence of independent regulatory agencies in the Netherlands.

Its natural counterpart is of course:

• Hypothesis 0: Political salience does not increase the establishment and use of ex post control instruments and therefore does not limit the formal and de facto independence of independent regulatory agencies in the Netherlands.

(20)

III. Research Design

3.1 Operationalization of the Main Concepts

Starting with political salience, the most important concepts used in this research will be defined and their related variables discussed. Political salience is ‘the degree of importance which is attributed to political matters’ (Koop, 2011: 210). In a representative democracy, the activity of importance allocation primarily takes place within parliament, where Members of Parliament (MPs) have at their disposal several powers to respond to issues that are salient in society in order to influence government behaviour. Of the six rights that prescribe how they may influence government behaviour, the right to inquiry – specifically through written questions – is Dutch MPs’ most accessible right and is – as a consequence – most frequently used (Rijksoverheid, 2020). Therefore the main explanatory variable political salience (polsal) measures the degree of political salience as the number of political matter-related written questions submitted by MPs in a given period to the responsible minister. This type of operationalization is also used by Koop (2011) who measures the number of times IRA-names pop up in written questions. As was mentioned in the theoretical framework, issue-salience and the salience of IRAs are related because IRAs are involved with the regulation of the political matters that generate attention. But, relatedness does not mean similarity. The political matters do not always concern their oversight when discussed in written parliamentary questions. Furthermore, IRAs may be mentioned in written parliamentary questions unrelated to oversight of the policy issues – for example when the IRA’s budget, internal affairs or agency leadership is the topic of interest. The focus on only IRA-names in written parliamentary questions therefore holds the risk of not accurately representing the salience of political matters, because it allocates disproportionate weight to miscellaneous issues unrelated to the policy. In order to achieve measurement validity, the variable political salience used in this research measures the averaged number of times IRA-names – being ZBO-names – and the titles of the regulations they implement pop up in written parliamentary questions in a given period. As such, the variable measures a more comprehensive degree of importance attributed to political matters. For the large-n analysis mainly, the general political salience of ZBOs between 2009 and 2019 is used. In the small-n analysis the ad hoc political salience as well is important. This is measured as the annual political salience i.e. the averaged number of times ZBO-names and their related regulations pop up in written parliamentary questions in a year, for each year between 2009 and 2019. It is a continuous variable that ranges from zero to Xn.

(21)

Formal independence is ‘a series of prescriptions, enshrined in the constitutions of agencies, which should guarantee independence from elected politicians’ (Maggetti, 2007: 272). The concept will be operationalized in this research as two outcome variables that are used parallel to each other and provide different levels of precision: a less precise three-category variable that is applicable to the entire sample of 74 ZBOs, and a more precise variable that is case-specific and is applied to the subset of eight ZBOs.

Figure 3. Visual Representation of the Legal Status of ZBOs

(22)

The first outcome variable formal independence simple (findsi) classifies the sample of ZBOs researched in three categories that reflect the legal position of ZBOs vis-à-vis the political principal. The sample of ZBOs to which the Framework Act applies comprises – in ascending order of the degree of independence – (i) ZBOs that are established under public law and are part of the State (a-bodiesI), (ii) ZBOs that are a body of a legal person established under public law (a-bodiesII), and (iii) ZBOs that are a body of a legal person established under private law (b-bodies) (Kaderwet Zelfstandige Bestuursorganen 2006; De Vries, 2000; Kennis Openbaar Bestuur, 2020). Figure 3 provides a visual representation of the distance between each of these three legal bodies and the political principal i.e. central government. The diagram shows how different types of organisations that are placed at arm’s length of central government are related to it legally. The orange vertical oval represents all ZBOs. The part of the oval that overlaps with the inner blue circle represents the a-bodiesI, the part that overlaps with the outer blue ring represents a-bodiesII, and the part that overlaps with the small yellow and pink circles represents full time and part time b-bodies. Findsi is an ordinal variable with scores 1 for the least independent a-bodiesI, 2 for more independent a-bodiesII, and 3 for the most independent b-bodies. This variable will be used in the first part of the research directed at the entire sample of 74 ZBOs to which the Framework Act applies.

The second outcome variable formal independence Gilardi (findgi) is a measure of formal independence using Gilardi’s index of independence (Gilardi, 2002). The measure has been used in several academic papers to measure formal independence and therefore provides valuable validation of the operationalization of the concept as well as an indication of the measurement validity of findsi. Findgi is a more complex measure that comprises indicators which reflect both powers delegated and controls implemented. Gilardi lists 21 indicators about an IRA’s head official, management board, budget allocation and control, accountability provisions and principal control instruments. On each of these indicators an IRA scores a value that represents the degree of independence that that indicator provides – lower scores mean less independence. Appendix I provides an overview of the 21 indicators and the allocation of values for each of the scores. Because this measure for formal independence is more case specific it will only be applied to the sample’s subset of eight market regulators in the small-n comparative analysis. Formal independence Gilardi is a bounded continuous variable that ranges from zero (no independence) to 21 (complete independence) based on the extent to which certain indicators for independence apply to a ZBO.

An additional variable is used in the analysis of the entire sample and the subset of market regulators that measures the number of deviating provisions from the Framework Act.

(23)

The Framework Act provides a one-size-fits-all framework for ZBO powers and controls, but – as has been discussed – political salience affects these powers and controls. The degree of customization of the Framework Act to fit a specific ZBO measured by the variable deviating provisions (devpro) may provide an interesting indication of this effect. Because deviations generally increase the degree of formal independence, the variable is used in part of the analysis as an alternative outcome variable. Devpro is a discrete variable that ranges from zero to Yn.

Formal control instruments are means through which one actor (e.g. the principal) can direct the conduct of another actor (e.g. the agent) and that are formalized in directives, laws and regulations. Formally directing conduct takes the shape of provisions that prescribe what the agent may, may not and must do before the fact – ex ante control instruments – and provisions that prescribe how agency accountability should be achieved and how the principal may adjust agency conduct after the fact – ex post control instruments (Bovens, 2007; Busuioc, 2009). The use of formal control instruments is the extent to which ex post control instruments – specifically those that increase the zone of discretion of the principal – are applied overtime. The comparative case analysis will focus on the number of ex post control instruments established, and the application of those increasing principal discretion in the most politically salient year in a given ten-year period (2009 to 2019) for all eight market regulating ZBOs. The moderator variable ex post control instruments (expcoi) is a discrete variable that expresses the number of ex post control instruments established and ranges from zero to Mn. The variable has a double purpose: on the one hand, the actual number provides information about the degree of formal dependence. On the other hand, the types of ex post control instruments (see Appendix II) and the frequency of their application provide information about the de facto independence of the market regulators – ‘the self-determination of agencies’ preferences, and their autonomy throughout the use of regulatory competencies, that is, during the activity of regulation’ (Maggetti, 2007: 272).

In the regression analysis of the entire sample of 74 ZBOs to which the Framework Act applies two variables are added to control for the effect of political salience on formal independence. These are ZBO age (zboage) and ZBO size (zbosize). ZBO age is theoretically relevant because older ZBOs may generate more political and public attention – either because of increased familiarity or because older ZBOs are concerned with political matters for which credible commitment was most urgent. The age of a ZBO is determined based on the number of years between its establishment and the year 2020 and is therefore a discrete variable that ranges from zero to Cn. ZBO size is theoretically relevant because of the political and public attention it generates as well. This often has to do more with scope than with actual size, but

(24)

since these concepts are arguably closely related and scope is difficult to operationalize, zbosize is a variable that measures the total of full time equivalents in the last year of interest (2018). It is a continuous variable that ranges from zero to Cn.

3.2 Case Selection

The sample of interest comprises all ZBOs to which the Framework Act applies. Although this excludes some (semi-)independent governing bodies – some of which are a ZBO – the remaining 74 ZBOs are relevant to research, because (i) they include the majority of regulatory agencies in the Netherlands and (ii) they are the political principal’s mould on which existing and future ZBOs will be (re)modelled (De Vries, 2000). The first part of the research analyses the relation between the variables political salience and formal independence simple for the entire population. In order to provide a more in depth analysis of the relation between political salience, the establishment and use of ex post control instruments, and formal and de facto independence, a subset of ZBOs from the sample will be compared. It is important that these selected ZBOs differ in their degrees of political salience and are similar on other important accounts (ceteris paribus), being: (i) the motif for their establishment and (ii) their relation vis-à-vis regulatees. Market regulating ZBOs provide the most suitable subset of the population because they are equal on these accounts, agency independence is especially relevant for these types of regulators in achieving credible commitment, and there are only eight such ZBOs in the sample which makes comparisons relevant and feasible (Gilardi, 2002; Koop, 2011). Furthermore, market regulators are – according to theory – generally more politically salient. Aside from differences in degrees of political salience, the presence of political salience as such is of vital importance to test hypotheses Ia and Ib. The eight market regulators are entrusted with monitoring regulatee compliance and enforcement of regulations that aim to improve consumer protection and the competitiveness of markets (AFM, 2020). Table 4 provides an overview of these market regulators including their scores on the relevant variables polsal, findsi, findgi, devpro, zbosize and zboage.

The time period chosen to analyse the cases in both the regression analysis and the comparative analysis is a ten-year period that comprises the years 2009 (starting January first) to 2019 (ending January first). This decade provides the most recent and complete information and spans the interesting post financial crisis period.

(25)

3.3 Method of Data Collection and Analysis

All the required information to test the hypotheses and answer the research question is available through public records published on the website of parliament (TweedeKamer.nl), the websites of the central government (Rijksoverheid.nl and Almanak.Overheid.nl), the online archives of legislation and regulations (Wetten.Overheid.nl and OfficieleBekendmakingen.nl), and the websites of the market regulating ZBOs (ACM.nl, AFM.nl, AutoriteitPersoonsgegevens.nl, BureauFT.nl, CvdM.nl, DNB.nl, KansspelAutoriteit.nl and NZa.nl).

From these sources a database is constructed that includes the entire sample of 74 ZBOs to which the Framework Act applies including – inter alia – their scores on the variables formal independence simple, political salience and deviating provisions. This dataset will be used for the large-n quantitative analysis.

For the qualitative analysis the legal texts that include the provisions that attribute powers and limitations to the selected ZBOs will be scrutinized to collect all relevant provisions to construct a score on the variable findgi. Appendix I provides a schematic overview of the score for each market regulator on each indicator from Gilardi’s index and the references to the legal provisions from which these scores are derived. The legal texts will be further scrutinized to collect all ex post control instruments – i.e. all provisions that prescribe how agency accountability should be achieved and how the principal may alter agency decisions.

Subsequently, the method used to derive the score for polsal will be applied for each market regulator to each of the years of interest (2009 to 2019) to retrieve the most politically salient year. Based on the ZBO-specific ex post control instruments the annual reports submitted by the ZBOs for these years and publications related to these instruments in the Government Gazette (Staatscourant) will be analysed in order to assess whether the control instruments applied at discretion of the principal have been used. The cabinet is obligated to publish activities it performs vis-à-vis ZBOs in the Government Gazette. This obligation applies to the vast majority of, though not all, ex post control instruments e.g. appointments, dismissals, and ministerial regulations. Appendix II provides an overview of all ex post control instruments that apply to the market regulators – split in accountability provisions and principal discretion provisions.

The research uses a mixed methods strategy, combining a large-n regression analysis with a small-n comparative analysis, and using both quantitative and qualitative analytical instruments. The regression analysis is applied to roughly test the relation between the main explanatory variable and the outcome variable, i.e. hypothesis I, and the comparative analysis is used to test the causal mechanism described by hypotheses Ia and Ib. The mixed methods

(26)

strategy builds on the strengths of large-n and small-n analyses whilst compensating – to some extent – for the weaknesses of each.

3.4 Reflection on Validity and Reliability

The mixed methods design of this thesis makes it possible to utilize the benefits of small-n and large-n research designs, whilst significantly correcting for the limitations of each. The sample of 74 cases in the first part of the analysis makes it more plausible that a potential relation between political salience and formal independence will be revealed - even when the correlation is relatively weak. However, large-n designs are not suited best for inferring causation from observations. There are several large-n strategies to test for a causal mechanism, but a small-n comparative analysis in addition to results from a large-n quantitative analysis as well enable the researcher to infer causation. The second part of the analysis will therefore compare eight most similar – as well as most likely to manifest an effect – cases from the sample that allow for the analysis of the potential effects of political salience on the use of ex post control instruments. The other side of the coin is that when the causal mechanism has been established through the small-n analysis in the second part of the research, it is difficult to generalize the results to the entire sample and consequently the entire population of ZBO’s c.q. organizations at arm’s length in the Netherlands.

This last point is further complicated by the vast heterogeneity within the sample. There is a significant number of differences between ZBOs to which the Framework Act applies that may to some degree affect the establishment and use of ex post control instruments (e.g. organization size, embeddedness in international networks, number of veto players, et cetera). The research attempts to address this problem in two ways. First, the regression analysis includes two possible confounders: a variable for agency age and a variable for agency size. Second, cases selected from the sample for the comparative case analysis are market regulators from the population that – due to their responsibilities and activities – should affect the establishment and use of ex post control instruments similarly even when political salience is not the main explanatory variable.

The heterogeneity is not limited to the characteristics of the ZBOs. The data used for the comparative analysis varies significantly as well because the annual reports of the market regulators differ (significantly) from each other: some are elaborate whilst others are concise, and some have a schematic layout whilst others are more narrative. This makes it difficult to filter out whether and which ex post control instruments are used. The additional information from the Government Gazette is therefore highly relevant.

(27)

A final note is on the operationalization of the main variables. In order to compare the entire sample of 74 ZBOs a simple and straightforward measure of formal independence is required. The legal status of ZBOs opposite the political principal is the low hanging fruit because it represents the legal arm’s distance between principal and agent. However, formal independence – as has been explained – is the product of powers delegated and controls implanted, which does not necessarily correspond with a ZBO’s legal status. To address this problem, the research uses a more comprehensive measurement tool for formal independence where possible – in the comparative case analysis. But the quality of this tool as well requires some discussion. Gilardi’s index is an elaborate instrument to measure an IRA’s formal independence that has been tried and tested. The 21 indicators and their multiple scores to some extent correct for potential problems with measurement validity. Still, it can be argued that – for example - the potential of the principal to annul agency decisions without cause is far more limiting to an agency’s independence then the ability of a principal to reappoint an agency head multiple times. The analysis will show how both measures of formal independence compare and what conclusions can therefore be drawn about their measurement validity. In systemically applying the instruments equally to the cases, the relative scores at least will provide valid information.

(28)

IV. Analysis

4.1 Part I: Large-n regression analysis

The first part of the analysis focuses on all 74 ZBOs to which the Framework Act applies. Analysis on this scale will provide information on the general relation between the main concepts under consideration: political salience and formal independence. The regression analysis will enable the isolation of the effect of interest from effects potential confounders may have on formal independence. In order to assess the relation between political salience and formal independence, the regression therefore controls for ZBO age and ZBO size. Before turning to the regression analysis, however, some descriptive statistics of the dataset and pairwise correlations will be discussed.

Table 1. Descriptive Statistics ZBOs to which the Framework Act applies

Central Tendency Spread

Variables Mean Median Mode Minimum Maximum Stand. Dev.

Findsi - - a-bodyII a-bodyI b-body -

Polsal 31.75 11 - 0.00 232.50 53.11

Devpro 1.24 1 0 0 10 1.76

Zboage 30.73 20.50 16 1 222 38.52

Zbosize* 548.24 60 17 3 15304 1864.71

N = 74

*For the variable Zbosize there are three missing cases. N is thus 71.

Table 1 shows the descriptive statistics for the entire sample of 74 ZBOs to which the Framework Act applies. For the most relevant variables – being the outcome variables findsi and devpro, the main explanatory variable polsal, and the control variables zboage and zbosize – the degree of similarity amongst the ZBOs is given through indicators measuring the central tendency and spread of the observations. Because findsi is an ordinal variable consisting of three categories, not all statistics are relevant to report.

A first noteworthy remark is that most ZBOs appear to be a-bodyII – a public IRA that is legally not part of the State. Most sections in the Framework Act are directed at this kind of ZBO, which is the most balanced form of independence and accountability of the three.