115 Abstract The paper explores the patterns of Romanian ministerial control over executive agencies, based on the data collected in 2010 using the COBRA survey. Seven research questions were constructed starting from the major theories existing in the field aimed at describing and understanding the control patterns. The paper’s main conclusion is that ministries still prefer ex ante types of control in exchange to those after-the-fact, although some rather incremental shifts toward ex post control can be noticed. Keywords: Romania, autonomy, ex post and ex ante control, executive agencies, trust, rewards, sanctions, performance measurement, accountability. PATTERNS OF CONTROL IN THE ROMANIAN EXECUTIVE AGENCIES: AN EXPLORATORY RESEARCH*1 Călin HINŢEA Adrian HUDREA Dan BALICA Călin HINŢEA Associate Professor, Public Administration Department, Faculty of Political, Administrative and Communication Sciences, Babeş-Bolyai University, Cluj-Napoca, Romania Tel.: 0040-264-431.361 E-mail: hintea@apubb.ro Adrian HUDREA Researcher, Public Administration Department, Faculty of Political, Administrative and Communication Sciences, Babeş-Bolyai University, Cluj-Napoca, Romania Tel.: 0040-264-431.361 E-mail: hudrea@apubb.ro Dan BALICA Researcher, Public Administration Department, Faculty of Political, Administrative and Communication Sciences, Babeş-Bolyai University, Cluj-Napoca, Romania Tel.: 0040-264-431.361 E-mail: dan.balica@apubb.ro * The research was also supported by UEFISCU-CNCSIS, IDEI, grant no. 2595/2008, within the project entitled ‘Decision-making models in the Romanian public institutions at local and county levels’. Transylvanian Review of Administrative Sciences, Special Issue, pp. 115-139 116 1. Introduction The creation of governmental agencies is probably one of the most studied manifestations of New Public Management (NPM). Seen as entities structurally disaggregated from the core governmental apex that operate under more business- like conditions than their bureaucratic counterparts (van Thiel, 2009), agencies were meant at increasing the general quality of public services by specialization (Bouckaert and Peters, 2004, p. 29) and simplification (World Bank1, not dated), facilitating performance management (through performance contracting for example), deregulation (Verhoest et al., 2004a), depoliticization (OECD, 2005) and by facilitating competition (Peters apud Bouckaert and Peters, 2004, p. 29). Moreover, it is considered that structural disaggregation will encourage innovation, adaptation and responsiveness (Schick, 2002) and also will increase bureaucratic commitment to task accomplishment (Christensen, 2009). Still, as probably any decision concerning public issues, agencification is not a Pareto optimal decision. The literature evokes some downsides to agencification, which vary from learning to be autonomous (Schick, 2002), fragmentation of public sector (Bouckaert, Peters and Verhoest, 2010; Yesilkagit and van Thiel, 2011) or deficit of democracy (Christensen and Lægreid, 2007, p. 20; Papadopoulos, 2007). Therefore, agencification, also known as autonomization, requires a special focus on the control-related aspects meant at maximizing the pros and minimizing the possible cons. Following, the paper examines the theoretical framework of agency control and then approaches the issue related to Romanian agencies. 2. Why should agencies be controlled? When it comes to public organizations, control is fundamental for two main reasons: political and managerial. On the political side, control is meant at guaranteeing that democracy is safeguarded. In this understanding it is included the principal-agent paradigm: citizens elect politicians to represent their interests and politicians appoint bureaucrats to make sure that citizens’ needs are being addressed. Without the control of the politicians on the bureaucrats, the democratic chain is broken and we are confronted with what is called a democratic deficit. The problem is that elected officials depend on the information their subordinates provide them and unless they have a strong corpus of advisors and inspectors, we are confronted with some sort of information asymmetry between the two parts (Moynihan, 2006, p. 1033; Christensen and Lægreid, 2005; van Thiel and Yesilkagit, 2008; Talbot et al., 2005). This deficit has the premises to be more significant when it comes to agencies where task specialization offers an information asymmetry hard to overcome. This ministerial incompetence is closely connected to the idea of failure to monitor and evaluate agency performance. 1 The label does not appear as such in the WB document. It was defined as such in order to facilitate the understanding of the logic behind the argument. 117 The political control deficit is an important critique of the agency model and of the NPM movement at large (although the agencification process, as part of NPM, generated most of the issues on this level – Christensen and Lægreid (2007), and an argument for measures aimed at limiting the autonomization process (New Zeeland or Australia are relevant examples in this sense). On the managerial side, control is needed to make sure that the objectives are attained (the classical managerial functions elaborated by Fayol – Planning, Organizing, Command, Coordinate, Control – or Gulick’s POSDCORB are of extreme relevance in this sense) and improvements are brought where they are needed. Regarding this last statement, Christensen (1999) makes a difference between the desired consequences of granting autonomy and the perverse effects they might generate. These perverse effects act both on the democratic dimension (bureaucratic exploitation of information asymmetry and bureaucratic politics, unresponsiveness and low accountability) and on the managerial one (budget maximization and slack maximization). Because of these perverse effects, the control of agencies is justified. 3. What are the patterns of control faced by Romanian agencies? One way of analyzing the alternatives that the principals have at their disposal for controlling the agents is by classifying them depending on the moment in which they take place. In this perspective, control can happen before the agency takes any action (before-the-fact) or after such situations. In the literature this dichotomy is known as ex ante and ex post control (Thompson, 1993; Epstein and O’Halloran, 1994; Verhoest et al., 2010; McCubbins, Noll and Weingast, 1987; de Kruijf, 2010; Eichenberger and Schelker, 2006; Huber, Shipan and Pfahler, 1998; Aberbach, 1990). In the first scenario, known as ex ante control, the parent ministry controls the resources (inputs) the agency uses for accomplishing its objectives, the process for fulfilling its mission and the discretional decision making power the management has, by demanding prior approval for some type of decisions (or by possessing nullification rights), while the grounds for which the agency is held accountable are rule compliance, fairness and equity, and integrity. In some views (Berry and Gerse, 2010), structural aspects can also be included in ex ante control since, by means of structure, like assignment of the CEO or of board members, the principal ensure a before-the-fact form of control. In the second scenario, known as ex post control, the parent ministry liberates the agency from control over inputs or pre-established objectives or rules and regulations, but monitors the result the agent has and evaluates its performance, while the grounds of accountability are represented by the efficiency and effectiveness of its activity, its responsiveness and its overall performance. In theory, agencification demands a shift from the former to the latter (Verhoest et al., 2004; Christensen and Lægreid, 2007) in line with the idea of freeing managers, as was evidenced by Norman (2001, p. 68): Liberation management concepts can be seen at work in a delegation of authority to managers, letting them make decisions about choices of inputs and staff instead of holding them subject to complex centrally driven procedure manuals. 118 Following, these two large groups of control alternatives are described in more depth. 3.1. Ex ante agency control As aforementioned, one alternative of controlling agencies is by providing the resources they need to fulfill their objectives. Resources can vary from financial ones, to HRM and even material ones, if, for example, the agency cannot make acquisitions on its own. Regarding financial resources, one way of using them as control mechanisms is by making agencies dependent, as much as possible, on the resources coming from the principal. In this respect, what matters is how much of the agency’s income comes from state budget and how much from alternative sources or it is self-generated. If there is a great level of dependency on the state budget, it can be argued that parent ministries can use financial resources to steer executive agencies by reducing or increasing the level of budget granted to the agency (Verhoest et al., 2004; Vershuere, 2006; Verhoest et al., 2010; Christensen, 1999). Therefore, the more an agency depends on the state budget, the stronger the financial control it has to face. But financial control can be expanded by taking also into consideration how much discretion agencies have in deciding over aspects like setting tariffs and raising resources through taxes, shift between budgets or contracting loans (Verhoest et al., 2004; Barbieri et al., 2010; McGauran, Verhoest and Humphreys, 2005). The latter are usually included in the larger framework of financial autonomy. At this stage we approach only the former understanding of financial autonomy, while the latter is approached later when analyzing the decision-making restrictions agencies face. But financial resources are just one type of resources an organization needs to fairly conduct its activity. Staff represents another type, an even more important in anything but few opinions. Because of this, controlling such an important resource would suggest that the ministries practice the so called hands on agency management (Hood, 1991, 1995), in line with ex ante control. From this perspective, the parent ministry can control an agency’s activity by controlling the number of employees the organization has (Eisner and Meier, 1990). It is understandable that if, in order to better accomplish its objectives, an agency needs to increase the number of its employees, but cannot do this or can do it only with the ministry’s consent, the control it faces is fairly strong. Control over personnel can be further operationalized in control over the employee’s salary, the career or evaluation policy, but at this stage we will focus only on the degree to which agencies can control the total number of their employees Starting from these two input dimensions – financial and personnel – a first research question for the current study can be formulated. Q1: How strong is the level of input control agencies have to face in terms of main sources of income and setting the staff policy? Besides controlling agency inputs, the parent ministry can control the process that agencies need to follow to fulfill their objectives (McCubbins, Noll and Weingast, 1987; Verhoest et al., 2004). At this level, the involvement of the parent ministry in aspects like choosing implementation instrument, target group and implementation 119 process can vary from none, case in which agencies solely decide on the mater, to full, when the parent ministry makes all the decisions, regardless of the organization. In the latter case in fact the level of process control the agency faces is total since it plays no role in this regard. But the extent of control the ministry exerts over the manner of task fulfillment is not the only criterion for judging principal control. The ministry’s involvement can be more or less generous in terms of scope. In other words, the ministry can set the manner of task fulfillment, but it can also decide (or not) which policy instrument should the agency use in fulfilling its objectives (Lægreid, Roness and Rubecksen, 2005) or which part of the society should benefit from its activity (setting the target group) (Epstein and O’Halloran, 1994). Starting from this, the second research question can be developed. Q2: How strong is the level of process control agencies have to face in terms of choosing the target group, choosing the policy instrument and in deciding the manner in which the task is fulfilled? By merging the two, it can be observed that ministries can control agencies by restricting their access to the inputs they need to conduct their activity and by telling them how to fulfill their mission. But ministries can also demand agencies to consult them when certain decisions are made (Epstein and O’Halloran, 1994; Langbein, 2004, p. 55). In this case the subject of control is neither the input, nor the process, but rather the discretion agencies have in making some decisions2. This also qualifies for ex ante control because it takes place before-the-fact, or before any decision is actually made and before any action is truly implemented. We will call this discretional control. Moreover, in some interpretations (Vershuere, 2007; Verhoest et al., 2010, p. 25; Berry and Gerson, 2010) in the same category can also be included veto rights (or nullification rights) and the possibility to influence the decision-making process through agents, like board members. We subscribe to this interpretation because, as is the case for control over inputs or process, the control takes place before the result, in all cases the subject of control being the decision, itself, and not its results. Starting from the above, a third research question is constructed. Q3: How strong is the level of discretional control agencies face in terms of prior approval, veto right and agents of influence? Last but not least, as was aforementioned, one way of ex ante controlling agencies is by structural means. This is not commonly included in ex ante form of control, 2 There is a thin line in this respect between the control faced by agencies regarding the policy process and that faced in reference to the scope of its decision-making competences. The latter case regards those situations in which agencies have decision-making abilities, whereas the former describes whether or not the organizations have such decision-making possibility. In other words, in order to face discretional control, you need to possess some decision-making powers. In any case, the two categories are not mutually exclusive, but rather complementary, the distinction being introduced mostly for the sake of analysis, to have a better insight of the profile of ministerial control agencies face. 120 but we adhere from this perspective to Berry’s and Gerse’s (2010) and Epstein and O’Halloran (1994) view, which do describe this as a form of ex ante control (Berryand and Gerse, 2010, p. 5): Given the challenges of ex post monitoring (McCubbins and Schwartz, 1984), the structure and process literature tends to emphasize ex ante restrictions that mitigate the informational advantage enjoyed by agencies and stack the deck in favor of certain interests to ensure the durability of the original bargain. One way of defining structural control is as the inverse of structural autonomy3. The latter, as Christensen (1999)4 highlights, is determined by the presence of an intermediary board between the agency and the parent ministry (see also Yesilkagit and Christensen, 2010, p. 63; Verhoest et al., 2004; Painter and Yee, 2011). This board will act as a buffer between the ministry and the agency head, increasing his autonomy. Still, the mere presence of a board does not mean total structural autonomy. This board’s independence (defined as who appoints its members) and its relationship with the CEO (if it appoints the CEO and if the CEO is accountable before the board for its activity) are other variables in building a more comprehensive view on what is the level of structural autonomy (in the same vein, Gilliardi and Maggetti, 2010; Berry and Gerson, 2010). In other words, an agency faces a strong level of structural control if it has no governing board and the CEO is assigned and accountable before the parent ministry. On the contrary, an agency faces a low level of structural control if it has a governing board, whose members are assigned by other actors than the executive government and the board appoints the CEO, the latter being accountable for its actions only before the governing board. Q4: What is the level of structural control that Romanian agencies have to face? 3.2. Ex post agency control The above questions describe the so called ex ante control agencies have to face, but as we previously mentioned, ministries can control agencies not only before-the- fact, but also after decisions are made and actions taken. This is called ex post control and, as Sulle (2010, p. 247) puts it, (…) can be regarded as a management process that consists of interrelated sub- systems: a planning system (setting goals for agencies), a monitoring system 3 Defining control as the inverse of autonomy is promoted also by Verhoest et al. (2004, p. 106): ‘Control is defined here as the mechanisms and instruments used by the government to influence the decisions and the behavior of the agency intentionally in order to achieve government objectives (White, 1991, p. 189) (…) In this view, control is clearly the inverse of autonomy, taking into account that autonomy and control both are multidimensional concepts’. 4 ‘Structural autonomy, i.e., the insertion of either an alternative or a competing level of political supervision’ (Christensen, 1999, p. 9). 121 (measuring agency performance results) and an evaluation system (where sanctions and rewards are applied). Regarding the planning system, ex post control arrangements demand the existence of pre-established organizational objectives, against which agency performance can be evaluated. Several aspects are of interest at this level. Firstly, the actor that establishes the objectives; if these are set by the parent ministry than it is reasonable to assume that the agency faces a stronger control than in the case in which it formulated its own objectives. Second, whether these objectives are measurable or not. Effective performance management is dependent on measurable objectives (Kaplan and Norton, 1992). Unless these objectives fulfill this characteristic, the de facto level of control the agency faces is weaker. Planning is followed by monitoring, which takes the form of reporting5 and auditing arrangements. At this level what matters is not only if such arrangements are at work or not, but also how strong these are, one indicator of their strength being the frequency with which they take place. Therefore, the more agencies have to report, the stronger the ministerial control. The more agencies are audited, the stronger the ministerial control. Last but not least, the planning and monitoring systems, in order to be truly effective, need to be associated with reward and sanction systems (Verhoest et al., 2004). Therefore, in case the agency did fulfill or even overachieved its objectives, it receives some sort of benefits, while, on the contrary, in case it failed to do so, it faces some sort of sanctions. This ‘stick and carrot’ logic is translated into tangible consequences the conclusions of the assessment process described above. Moreover, assuming that individuals tend to maximize the utility, they will be more sensitive to rewards and sanctions aimed at them than at those aimed at the organization, at large. In the latter case the possible benefits are exported to those that may or may not have something to do with the success, while the sanctions are being externalized on others, so the costs of failure are not fully born by those responsible. In both cases, responsiveness is lowered and so is the effectiveness of rewards and sanctions as control instruments. Because of this we argue that the level of ministerial control is stronger for those agencies that face rewards and sanctions aimed at individuals rather than at the agency, since in this case the benefit/sanction is not diffused. Moreover, since according to prospective theory (Kahnema and Tversky, 1979, p. 279) losses are weighted stronger than gains, we argue that the control an agency faces is stronger if sanctions are present. Starting from the above, several questions need to be answered when it comes to Romanian agencies. Q5: How strong is the level of ex post control agencies have to face in terms of measurability of objectives and reporting and auditing arrangements? 5 Epstein and O’Halloran (1994, p. 698) include this in ex post arrangements. 122 Q6: In what degree do agencies face reward and sanctions systems and are these aimed at the individual or at the organization? All the instruments previously described can, with relative certainty, be included in the so called dichotomy of ex ante and ex post control alternatives. But there is another form of control that, on its one, is somewhere in between, and that is the grounds on which the agency and the agency management are accountable before the parent ministry. It could be argued that this is an ex post instrument since it is based on after-fact assessments, but if the grounds of accountability are base more on legality and rule compliance and less on results, it clearly represents an ex ante control instrument. On the contrary, if the CEO or the agencies are held accountable for the efficiency, effectiveness or responsiveness, the accountability offers a clear sign of ex post preferences (Verhoest et al., 2010). Therefore, a seventh research question can be formulated. Q7: Are the grounds on which the agencies are held accountable more charac- teristic to ex ante or ex post control? These would be the main facets of agency control. Following, by answering the questions that were formulated, we intend to find out what is the pattern of control Romanian agencies face. Is it more ex ante or ex post focused? Did a shift from before-the-fact to after-the-fact control take place, as the classic agencification model suggests? In this respect, these are the two central questions our present study strives to answer. Furthermore, we are interested to see how much the different types of control agencies face vary in terms of agency size, defined as full time employees (FTE), agency task (policy formulation, development and advice; regulation, scrutiny, control and inspection; other kind of exercising public authority; providing general public services – COBRA, 2010) and agency structure (defined by the presence or absence of a governing boards). The latter therefore represents both an independent and a dependent variable (Berry and Gerse, 2010; Verhoest et al. 2010). The conclusions are aimed at representing the grounds of future analysis of patterns of autonomy and control of Romanian executive agencies. 4. Data and methodology The findings highlighted in the present study are based on the COBRA survey (COBRA, 2010) that was carried in Romania in 2010. The survey targeted 125 Romanian executive agencies. The response rate was 36.85 % (46 agencies answered the COBRA questionnaire). In order to differentiate between executive agencies and other forms of public organizations the following criteria were used to select the public entities that would participate in the study as executive agencies (adapted after Pollitt and Talbot, 2004): 1. structurally separated from the main hierarchy of the ministry (defined as separate legal personality); 2. carries out tasks at the national level; 3. not commercial activity; 123 4. not totally independent from its ministry from a legal standpoint (subordinated to the Executive, not the Parliament); 5. vested in public law (either law or other governmental act); and 6. staffed with public employees. 5. Data analysis Q1: How strong is the level of input control agencies have to face in terms of main sources of income and setting the staff policy? 5.1. Financial control This dimension, as was previously stated, describes the level to which agencies depend on transfers or allocations from other government bodies to accomplish their tasks. The more an agency depends on such resources, the stronger the financial control the organization faces. This dependency can be used, theoretically, by the parent ministry to steer the agency in the direction it pleases, therefore reducing its de facto autonomy. Based on the COBRA survey methodology the following alternatives were constructed: Category 1: The agency is solely funded by some kind of transfer/budget allocation/ subsidies from governmental bodies Category 2: The agency has a mixed funding, but mainly represented by some kind of transfer/budget allocation/subsidies from governmental bodies Category 3: The agency has a mixed funding, but most of its income comes from other sources than transfer/budget allocation/subsidies from governmental bodies Category 4: The agency is solely funded by other sources of income Therefore, the further an agency is from Category 1, the lower the financial control it faces. On the contrary, the closer an agency is to Category 1, the higher the financial control it faces. How do Romanian agencies present themselves from this perspective? Figure 1 offers a first insight on the issue. Source: Authors’ calculations Figure 1: Agency sources of income 124 Almost half of the responding agencies are solely funded by some kind of transfer/budget allocation/subsidies from governmental bodies, suggesting therefore that they face a high level of financial control. Moreover, for over 80% of agencies, governmental income sources represent the main funding. In other words, from a financial perspective, most Romanian agencies face a high level of control. Regarding the level of financial control and agency characteristics as size, task or structure (defined as presence or absence of a governing board), several aspects were noticed: it appears that there is no difference between the level of financial control agencies face based on task or size. There seems to be though a difference between agencies that have and agencies that do not have a governing board in the sense that those that do have a board tend to be less dependent on government sources then agencies that do not have a governing board, although most of their income still comes from government sources. One possible explanation for this situation could be the presence of representatives of other actors in the agencies’ board. These board members could bring extra sources of income for the agency. We do not have specific data to test this but we did noticed that, for example, out of the six agencies that are solely financed from other sources than the state budget, two cases include in their boards representatives with voting rights coming also from other government levels. Still, further investigations are needed to see if there is a real and significant connection between the level of financial control agencies face and their board membership. 5.2. Staff control As was said before, financial resources represent only one alternative of controlling the agency by inputs. The principal can also control agency staffing and by this directly influencing its activity6. At this level we were interested to see in what degree the parent ministry keeps control over the agencies’ employment policy. The following were noticed: Less than a fifth of responding agencies can set the number of staff it needs while the majority can do this, but only with prior approval from the parent ministry. From this perspective agencies face an average towards strong level of ministerial control since most of them cannot do this or can do this only with prior ministerial consent. There is no difference from this perspective between agencies depending on structure, size or task. 6 The COBRA survey included a question aimed at assessing the level of autonomy agencies have in setting the staff number (under budgetary limits). The question had three possible answers: yes, the agency can set the policy regarding the agency staff, yes, but only with prior approval, and no, it cannot decide on the matter. The question was designed to assess the level of autonomy, not necessarily the level of ministerial control. Because of this, not being able to decide on setting the policy regarding the agency staff does not directly imply that the ministry can. It only suggests that de decision-making power lays somewhere else. But the second answer choice (yes, but with prior consultation of the parent ministry) sug- gests ministerial control on this type of organizational input. 125 Q2: How strong is the level of process control agencies have to face in terms of choosing the target group, choosing the policy instrument and in deciding the manner in which the task is fulfilled? The COBRA questionnaire asked agencies to answer who has the leading role in setting the target group, policy instrument and in deciding the manner in which the task is fulfilled. Regarding the setting of target group, in the large majority of cases (73.9%) parent ministries either do not get involved whatsoever (10.9%) or they intervene in a small degree, by demanding prior consultations from the agencies’ behalf (63%). Therefore, agencies face a rather low average level of ministerial control when it comes to setting the target group. The same trend was noticed also in reference to the level of ministerial involvement in the setting of the implementation instrument. In 67.4% of the cases the ministry either does not get involved (10.9%) or, as the previous case, demands only prior consolations (56.5% of cases). Therefore, again, agencies face a rather low level of ministerial control. Things are even more radical when it comes to deciding the manner in which the task is fulfilled. In this case, in 91.3% of the cases agencies either take this decision themselves (23.9%) or the ministry demands only prior consultations, but lets the agency to decide on the matter (67.4%). In no case does the ministry, solely, decide the manner in which the task is fulfilled. There is no statistical difference between agencies that have and those that do not have a governing board in terms of process control on all three variables (instrument, manner of fulfillment or target group). The same were observed also in reference to agency size. There was a difference noticed, though, between agencies with different primary task when it comes to choosing the policy instrument. It appears that agencies that have as primary task benefit from the lowest level of control, while those responsible for other kind of exerting public authority face a tougher control. Overall, primary task explains about 32.6% (Eta2) of the difference between groups of agencies with different tasks. No differences were noticed though between agencies with different tasks in selecting the target group or choosing the manner in which the task is fulfilled. Source: Authors’ calculations Figure 2: Agency employment policy 126 The results suggest a rather hands off ministerial attitude regarding procedural control over agencies. Even though in the large majority of cases prior consultations are needed, the decision is taken by the agency not the ministry. This is especially valid for choosing the manner of task fulfillment, the dimension which is consider as being the most sensitive to ministerial control (Verhoest et al., 2004, p. 109) Q3: How strong is the level of discretional control agencies face in terms of prior approval, veto right and agents of influence? Discretional control is seen as that specific form of control in which agencies do possess the capacity to decide on certain matters, but their discretion on the matter (or scope of their decision) is restricted by specific means, like the need to ask the ministry for its point of view, the possibility to influence the decision itself through influence actors (like board members or other kind of agents) or by threatening to nullify the decision. In all these cases, the agencies do possess the right to decide but this right can be more or less restricted by the parent ministry. In order to see if such restrictions are at work, agencies were asked to tell if, in their specific case, parent ministries have veto-rights, demand to be consulted before certain decisions are made, or influence the decision-making process by internal agents. First, regarding the degree to which the parent ministry has veto right on organizational decisions, in 50% of cases the parent ministry has such powers. This suggests an average level of discretional control. Second, when it comes to influencing the decision-making process through internal agents (the COBRA discriminated between board members with voting rights and other ministerial agents) in 54.3% of agencies, the parent ministry does use agents to steer the decisions as desired. Moreover, in 10 cases (22.2%) even though the principal does not influence the agency using inside agents, it does have nullification rights. Overall, this means that in 76.5% of cases the parent ministry does exert some sort of influence through agents or nullification rights. Discretional control also refers to those cases in which the ministry lets the agency to decide on certain aspects, but requests the organization to consult it before making any decisions. Requiring before-the-act consultation represents, from where we stand, a clear sign of control the ministry exerts on the agency and a limitation of its decision- making competences (or autonomy). The COBRA survey included several questions regarding human resource management and financial management autonomy and these questions ask the responding agencies to say if they can make those decision or not, and if they can, with or without prior consultations with the ministry. On most aspects related to setting HR policies, like deciding on salary, promotion, evaluation, appointment or downsizing policy, the majority of agencies have decision-making powers (except setting the salary policy), and, except deciding on matters regarding the downsizing policy, the ministry, as a general rule, prefers not to demand prior consultations. The table below shows on which dimensions parent ministry tends to demand more often prior consultations. 127 Table 1: Agency discretional control Policy No. of agencies that candecide on the matter (T1) No. of agencies that can decide on the matter, but after consulting the ministry (T2) % T2/T1 Salary 22 14 63.6% Promotion 33 16 48.4% Evaluation 36 6 16.6% Appointment 37 11 29.7% Downsizing 31 26 83.87% Source: Authors’ calculations It seems that the lowest level of ministerial control regards the evaluation and appointment policy, while the strongest control is kept on the salary and downsizing policy. This makes sense considering the budgetary impact these decisions usually have and the rather political salient character of downsizing. If we corroborate the results with what was observed at Q2 (that most agencies can decide on those matters but after consulting the ministry), it can be concluded that, overall, agencies face a rather strong ministerial control. Q4: What is the level of structural control that Romanian agencies have to face? At this level several variables are of interest, as was specified above: Whether the agency has a governing board or not? Who appoints the board members? Who appoints and evaluates the CEO? Regarding the first dimension, presence of a board, 37% (17) of agencies have a governing board while 13% have an advisory board; all in all, half of agencies have some sort of structural buffer between the parent ministry and the organization. Still, when it comes to the strength of this buffer, the data suggests that structural control is rather strong. First, only 37% of agencies have a governing board, which means that in only 37% of cases the board has decision-making powers related to the agency. Second, out of these 37%, in 64.7% of cases the board members are assigned by the parent ministry, suggesting therefore that even if the board has decision-making powers, it is still very exposed to ministry interventions. Moreover, as the COBRA data shows, 76% of agencies that declared as having a governing board also declared that the parent ministry uses its board members to exert influence on the organization. In none of the cases the board assigns the CEO (see below). Table 2: CEO appointment Who appoints the CEO? Percentage 1. The government or minister 82.6% 2. The government or minister, after nomination by or consultation of the board or of the organization itself 6.5% 3. Another actor 10.9% Source: Authors’ calculations 128 As it can be noticed from Table 2, in 82.6% of cases the government or the ministry assign the agency CEO. This suggests a high level of structural control that the executive agencies have to face. If we add the fact that in 78.3% of cases the agencies are evaluated by the government or parent ministries, and only in 10.9% by the board then it can be concluded that agencies face a strong level of structural control. Table 3: CEO evaluation Who evaluates the CEO? Percentage 1. The government or minister 78.3% 2. The board 10.9% 3. The Parliament 2.2% 4. Another actor 8.7% Source: Authors’ calculations Q5: How strong is the level of ex post control agencies have to face in terms of measurability of objectives and reporting and auditing arrangements? All three dimensions describe the core of ex post control and are very much dependent one on the other. Regarding the planning system or the degree in which agencies face pre-established organizational objectives and their level of measurability, the COBRA data shows that only three agencies have no objectives, suggesting that in Romanian agencies at least some steps were made towards the implementation of performance management principles. However, when analyzing the measurability of these objectives, a different picture appears before us: from the total number of responding agencies, 26 (56%) have measurable objectives7 mentioned in some sort of document, percentage that puts the agency population under an average level of control8. There is no relationship, though, between the level of measurable objectives and the agency structure, task or size (tested by compare means and correlations, the test not being statistically significant). Taking into consideration the fact that 78.2% of agencies formulate these objectives themselves, with or without prior consultations with the parent ministry, it can be concluded that the control by the oversight authority over the agencies is rather limited. If regarding planning ministries exert a rather low level of control, a different picture is observed with respect to the frequency with which the agency has to report on objectives and overall results to the oversight authority. The average reporting frequency, on a scale from 1 (no reporting) to 6 (monthly) is 4.2, which means around twice per year. This could suggest that despite the relative freedoms regarding the setting of organizational objective, there is still a strong degree of control by the parent ministry. Findings also indicate that the most frequent reporting arrangements are 7 Another study found a contrary trend (Stan et al., 2005). 8 The average agency scores 0.75 (scale from 0 to 1), so somewhere between having objectives and having measurable objectives. 129 with agencies that have no objectives, followed by agencies with objectives which are difficult to measure. This would suggest that the parent ministry uses reporting requirements to compensate for a lack of measurable objectives. Reporting is not related to agency task or structure, but it is related to agency size, meaning that there is a positive but rather weak relationship between the frequency of reporting and the number of FTE. Apparently, bigger agencies tend to report slightly more often than the smaller ones. Related to reporting arrangements are the auditing arrangements. Only one agency said that no audit takes place. Most agencies run regular internal audits (73.9%) while only 17.4% hire an external entity to conduct it. 45% of agencies declare that they are subjected to regular audits conducted by other executive audit entities, while only 26.1% are subjected to regular audits of a non-executive court of audit. 37% of agencies are subjected to audits conducted by other entities assigned by the superior political or administrative entity. Regarding the strength of this audit, 29 agencies out of 46 are subjected to some form of executive auditing system, conducted either by another unit of the executive government or by an entity assigned by the parent ministry. This means that 63% of agencies are placed under some sort of external regular auditing system, which makes the executive control rather present. Table 4: Type of regular audit arrangements agencies have to face Type of regular audit Percentage 1. Regular internal audits 73.9% 2. Others, contracted by the organization 17.4% 3. Another unit of the executive government 45% 4. Court of Audit linked to the Parliament 26.1% 5. Another actor commissioned by the oversight authorities 37% 6. No regular audits 2.17% Source: Authors’ calculations Regarding the possible relationships existing between organizational characteristics like structure, size or agency tasks, on the one hand, and presence of external auditing system, several mentions must be made. Firstly, no such relationship was noticed regarding agency structure or size. Therefore, it can be assumed that it makes no difference if an agency has a governing board, or how big it is; it has the same chances of being subjected to regular external auditing systems. Second, there appears to be a difference between agency task and the presence/absence of an external executive regular auditing system (see Figure 3 below). Further inquiries are needed though to better understand how agency tasks influence the presence of external regular auditing systems, conducted by executive government entities. 130 Source: Authors’ calculations Figure 3: Frequency of external auditing arrangements and agency task Q6: In what degree agencies face reward and sanctions systems and are these aimed at individuals or at the organization? As the COBRA data show, in 84.4% of cases no reward system for the achievement of the objectives is found. This suggests that here is a low level of ‘carrot’ interventional control in Romanian agencies. Three agencies report as having wage increase or bonus for manager or other personnel member as rewards, while other three report as benefiting from increased resource allocation for the organization, if they appropriately or even overcome their objectives. No agency on the other hand benefits from increased autonomy (for the organization or individual manager) as reward for the accomplishment of its objectives. A different picture emerges when it comes to the presence of sanctions for failure to accomplish the pre-established results. 25 agencies (54.3%) declared they would face sanctions in case of failure of achieving their objectives. Table 5 below shows what types of sanctions are most common. Table 5: Sanctions agencies have to face in case of failure of reaching their goals What sanctions does the organization face in case of failure to reach its goals? Percentage 1. Wage decrease or bonus reduction for manager or other personnel members 15.2% 2. Decreased resource allocation 19.6% 3. Less autonomy for the organization or individual manager in relation to senior politico-administrative units 4.3% 4. Restructuring or abolishment of the organization 19.6% 5. Other sanctions 10.9% Source: Authors’ calculations 131 As it can be noticed, in those cases where a sanction system is implemented, there is a tendency to punish the organization and less the management; more precisely, 34.8% of agencies face sanctions aimed at the organization (such as the reduction of financial resources, reduction of autonomy or restructuring/abolishment) while only 15.2% face sanctions like wage decrease or bonus reduction for the manager or other staff members. This could give us an insight into the level of de facto accountability of agency’s management, since they rarely have to face sanctions for their failure (that is if we interpret organizational failure as being the management’s responsibility). But, as before, a question arises: how much agencies differ on rewards and sanction systems in terms of size, structure and task? The possible relationship was tested using correlations and it appears that the answer is: not that much. Except the correlation between size and rewards, all other correlations are not statistically significant, meaning that structure, task and size (at least for sanctions) make no difference when it comes to ‘stick and carrot’ arrangements. Regarding the rewards, it appears that there is a relationship between agency size and the presence of a reward system, meaning that bigger agencies have higher chances of being rewarded if they accomplish or overcome their goals. The relationship is rather weak, though, Pearson Correlation being 0.364. But even though a weak one, a relationship between size and goals still exists. Could this be explained by the fact that such organizations, with more employees, have stronger negotiation power and obtained more concessions from the parent ministry? We do not have the data yet to test this hypothesis, but it is worth debating. An interesting hypothesis can be developed starting from the aforementioned observations: agencies face an average level of sanctions, but at the same time face an average level of objective measurability. Do the agencies that have a low level of measurability of objectives overlap with those that face sanction systems?9 And if yes, how does the interaction between the agency and the parent ministry, in terms of mutual trust (in the agency’s eyes), presents itself? From where we stand, three major possibilities can be met in this scenario: 1. Agencies with low objective measurability that do face sanction systems tend to have a more trustworthy relationship with the parent ministry. In this case it could be argued that trust compensates (or even justifies) the lack of measurable objectives, or that ‘trust and a trust-based organizational culture could be an option which limits the necessity of explicit performance indicator-based steering’ (Greiling, 2007, p. 12). 9 If yes, this could suggest that in those cases parent ministries could have higher level of discretion on steering agencies; since there is no term of reference which can be used by agencies to protect themselves, ministries could sanction agencies for not accomplishing their objectives. How can the agency prove otherwise if it has no terms of reference – mea- surement of objectives? In such a scenario it can be argued that the parent ministry has the premises for exerting stronger informal steering power over the agency. 132 2. Agencies with low objective measurability that face sanction systems tend to have a less trustworthy relationship with the parent ministry, one possible explanation being that the ambiguity in terms of objectives makes a more uncomfortable relationship between the two parts. This alternative is based on the idea that (at least) relational trust depends on the level of positive expectations, vulnerability and risks (Choudhury, 2008, p. 595): ‘risk herein in the very grant of discretion where determining success or failure is difficult to ascertain due to goal ambiguity, measurement problems, and the long duration of policy effects’10. 3. There is no difference between agencies with measurable objectives and those without measurable objectives in terms of trust in relation with the parent ministry. In order to test the new hypotheses, a cross tabulation between the presence of measurable objectives in some sort of document and the presence of a sanction system was made. The results are the following: Table 6: Cross tabulation between the presence of measurable objectives in some sort of document and the presence of a sanction system Measurable objectives Total Yes No Sanctions Yes 18 7 25 No 8 12 20 Total 26 19 45 Source: Authors’ calculations Apparently, most agencies that face a sanction system also have measurable objectives present in some sort of written document – 18 out of 25. This is explainable since a fair system of sanctions and rewards depends on fair assessments and one criterion for a fair assessment is transparency, hard to imagine in the absence of a measurable system of objectives attainment. So, in order to answer our research questions: agencies that have a low level of measurability of objectives coincide only in a small degree with those that face sanction systems11. Still, it remains the question of patterns of trust between agencies that do not have measurable objectives and the parent ministries. Even though their number is fairly low (7) it is worth testing which of the three assumptions are more probable. The figure bellow offers a first insight on the issue: 10 This interpretation of trust can also be related to the idea of procedural justice, defined as ‘fairness of the procedures used to determine organizational outcomes’ (Connell, Ferres and Travaglione, 2003, p. 572). 11 This invalidates the construct presented in note 4. 133 Source: Authors’ calculations Figure 4: Level of trust of agencies which face sanction systems that have measurable objectives (1) and that do not have measurable objectives (0) At a first glance, it appears that agencies with measurable objectives (the set of agencies situated on the right side of the figure) tend to describe in more trustworthy terms their relationship with the parent ministry than those without measurable objectives (the left category). But when comparing the average assessment of trust12 on both categories, there appears to be no difference: agencies without measurable objectives describe their level of trust with the parent ministry around 0.840 (on a scale from 0 – no trust – to 1 – high level of trust), while agencies with measurable objectives describe their level of trust around 0.847. Basically, there is no difference between the two categories of agencies in terms of the reported level of trust with the parent ministry. This means that the presence of a sanction system influences in no way the level of trust between the agency and the parent ministry, in terms of how measurable the organizational objectives are. Q7: Are the grounds on which the agencies are held accountable more charac- teristic to ex ante or ex post control? Regarding what matters when auditing or evaluating the CEO’s results, the follo- wing were noticed: 12 For agencies without measurable objectives, the average assessment of trust, per group of agencies, was calculated as follows: 0.75*(4/7) + 1*(3/7). For agencies with measurable objectives, the average assessment of trust, per group of agen- cies, was calculated as follows: 0.5*(2/18) + 0.75*(7/18) + 1*(9/18) 2 4 7 3 9 0 1 2 3 4 5 6 7 8 9 10 .00 1.00 .50 .75 1.00 Level of trust Goal measurability: 1 - measurable; 0 - unmeasurable 134 Table 7: Criteria used in evaluating the CEO Rules/legality Results/objectives accomplishment Audit 19.6% 10.9% CEO accountability Average of 0.95013 Average of 0.93014 Source: Authors’ calculations13 14 As Table 7 shows, there is a tendency towards focusing more on rules and legality than on results in assessing agencies, 19.6% of agencies declaring that regular audits are predominantly focused on legality aspects, while a little over 10% declared that audits focus on results. The data suggests that even though result related aspects are stressed by the superior ministry, legality still remains the rule. If we add the fact that the main criteria used by the parent ministry for steering and controlling agencies is legality (see Table 8 below), then we can conclude that the shift from a priori to a posteriori control is rather low, legality and rule compliance still being the main assessment criteria. Table 8: Criteria used by the parent ministry for steering and controlling agencies15 First score Second score Third score 1st most important criterion for control Legality/compliance to rules Economy criteria Effectiveness/goal attainment 89.1% 2.2% 2.2% 2nd most important criterion for control Effectiveness/goal attainment Effi ciency Economy criteria 37% 23.9% 15.2% 3rd most important criterion for control Effectiveness/goal attainment Effi ciency Economy criteria/equal access to service 30.4% 23.9% 4.3% Source: Authors’ calculations 6. Conclusions and future steps One of the aims of the present paper was to have a better understanding on the patterns of control Romanian agencies face. This is welcomed since few inquiries in understanding different consequences of more or less coherent administrative reforms were made and even more when it comes to agencification, in particular. Therefore, our endeavor was focused on finding whether the expected shift from ex 13 On a scale from 0 to 1, 0 meaning it is not accountable, 0.5 meaning it is accountable to some extent and 1 that it is accountable to a large extent. 14 The same as for rules/legality. 15 Agencies were asked to rank which criteria were used by parent ministries in the steering controlling process. They had to prioritize between legality, equity, economy, efficiency and others. As table 4 shows, 89.1% of responding agencies declared that legality is the most important criteria for control, while only 2.2% of agencies declared that the main criteria of control are economy and effectiveness. 135 ante to ex post control was met, as it is evidenced in theory. As was observed, there is a certain shift in this direction, but apparently ministries still keep a rather firm ex ante control over agencies. Input control seems to be the preferred alternative of ex ante control, ministries keeping agencies rather dependent both in terms of budget and staff. This could be owed to the fact that in 2009 the Government enacted an ordinance that significantly reduced agencies’ abilities to find alternative sources of income, like fees, in order to have a better control over governmental spendings. Process control seems to be also present, although it is less stressed by parent ministries. In the large majority of cases agencies can decide on aspects like selecting the target group, the policy instrument or the manner of fulfilling, but only after consulting the parent ministry. In other words, agencies face what is called discretional control, or that type of control where agencies have decision-making capacity, but this capacity is bordered by certain ministerial tools, like demand prior consultations, keeping veto rights or using internal agents as influencers. As we found, ministries exert a fairly strong discretional control over agencies, in the large majority of cases having either veto right or using internal agents as tools to influence the organization’s decisions. Last but not least, ministries keep a firm hand over agencies even from a structural perspective, very few of them having governing boards, which should put the agency and the management at a safer distance from the principal. Moreover, even in those cases the large majority of board members are assigned by the ministry, as is the case also for the CEO (appointed solely by the ministry in over 80% of cases). But as we specified, there are some ex post alternatives of control valued by the ministries. Reporting is one of them, most agencies have to report around twice per year. A similar trend was noticed also regarding audit arrangements, over 60% of agencies facing some sort of external executive audit. It is hard to tell how effective and truly useful are these instruments for parent ministries, considering that only about half of these agencies have measurable objectives and these, in the large majority, are set by the organization not by the principal. A lower level of control was noticed though regarding rewards and sanctions instruments. Few agencies ‘benefit’ from these tools, and even when they do, as is the case for sanction (met in over half of agencies), are rather focused on the agency and not on the individuals. From where we stand this strongly reduces the factual level of accountability in the principal agent relationship and, we believe, deserves to be paid more attention in future studies on the matter. Overall, as we said before, it seems that ministries still prefer ex ante types of control in exchange to those after-the-fact. It is hard to explain this trend based only in the COBRA data we have, but some alternative explanations can be advanced. The first has to do with the rather exceptional character of these times. Financial crises and recession reduced the resources available to the government. Moreover, the media is more sensitive than ever to public spending and decisions. Since the 136 ministries are one of the first entities that have to answer in case of mismanagement, this could represent one incentive for hands on approach. An alternative explanation has to do with the lack of capacity to shift control from ex ante to ex post arrangements. Reporting, auditing, and assessing, all demand resources, both in terms of HR and finance. This is even harder to do when agencies’ tasks are more technical and specific. It can be argued that this could be an explanatory reason, but we do not have any specific data to support this perspective. The aforementioned conclusions, even though they bring more light on the relationship between executive agencies and their parent ministries, explain only one side of the agencification coin. These observations need to be supplemented by similar tests regarding the autonomy side of agencification. Moreover, patterns of trust, briefly approached in the present study, can also provide alternative explanations for the patterns of control and autonomy (the latter briefly triggered elsewhere – Hinţea, Hudrea and Balica, forthcoming) and can also help better understand the informal side of the relationship between the ‘principals’ and their ‘agents’. References: 1. Aberbach, J., Keeping A Watchful Eye, Washington, DC: The Brookings Institution, 1990. 2. Barbieri, D., Bellè, N., Fedele, P. and Ongaro, E., ‘Determinants of Autonomy of Euro- pean Union Agencies’, 2010, Paper for ECPR Porto, Panel Impact and Consequences of the EU Agencification Process, June 25, 2010, [Online] available at http://www.jhubc. it/ecpr-porto/virtualpaperroom/018.pdf, accessed on August 10, 2011. 3. Berry, C.R. and Gersen, J.E., ‘Agency Spending and Political Control of the Bureau- cracy’, Paper presented at Conference on The Politics of Federal Spending, May 26-29, 2010, [Online] available at http://polisci.ucmerced.edu/sites/polisci/files/public/docu- ments/Berry%20Gersen%20March2010.pdf, accessed on August 19, 2011. 4. Bouckaert, G., Peters, B.G. and Verhoest, K., The Coordination of Public Sector Organi- zations: Shifting Patterns of Public Management, Basingstoke, UK: Palgrave MacMillan, 2010. 5. Choudhury, E., ‘Trust in Administration. An Integrative Approach to Optimal Trust’, 2008, Administration & Society, vol. 40, no. 6, pp. 586-620. 6. Christensen, J.G, ‘Bureaucratic Autonomy as a Political Asset’, Paper presented at the workshop Politicians, Bureaucrats, and Institutional Reform, ECPR Joint Sessions of Workshops, Mannheim, March 26-31, 1999, [Online] available at http://www.essex. ac.uk/ECPR/events/jointsessions/paperarchive/mannheim/w1/Christensen.PDF, acces- sed on August 19, 2011. 7. Christensen, T. and Lægreid, P., Transcending New Public Management. The Transfor- mation of Public Sector Reform, Surrey: Ashgate, 2007. 8. COBRA Survey 2010, [Online] available at http://soc.kuleuven.be/io/cost/survey/surv_ core.pdf, accessed on August 19, 2011. 9. Connell, J., Ferres, N. and Travaglione, T., ‘Engendering Trust in Manager-Subordinate Relationships. Predictors and Outcomes’, 2003, Personnel Review, vol. 32, no. 5, pp. 569-587. 137 10. De Kruijf, J.A.M., ‘Lessons Learned: Ex-Ante Control of Externally Autonomised Enti- ties in the Netherlands at the National Level’, 2010, Public Administration Quarterly, vol. 34, no. 4, pp. 479-512. 11. Eichenberger, R. and Schelker, M., ‘Independent and Competing Agencies: An Effective Way to Control Government’, 2006, Public Choice, vol. 130, no. 1, pp. 79-98. 12. Eisner, M.A. and Meier, K.J., ‘Presidential Control versus Bureaucratic Power: Explai- ning the Regan Revolution in Antitrust’, 1990, American Journal of Political Science, vol. 34, no. 1, pp. 269-287. 13. Epstein, D. and O’Halloran, S., ‘Administrative Procedure as Instruments of Political Control’, 1994, American Journal of Political Science, vol. 38, no. 3, pp. 697-722. 14. Giliardi, F. and Maggetti, M., ‘The Independence of Regulatory Authorities’, 2010, [Online] available at http://www.fabriziogilardi.org/resources/papers/gilardi_maggetti_ handbook.pdf, accessed on July 18, 2011. 15. Greiling, D., ‘Trust and Performance Management in Non-Profit Organizations’, 2007, The Innovation Journal: Public Sector Innovation Journal, vol. 12, no. 3, [Online] avai- lable at http://www.innovation.cc/scholarly-style/greiling9final1draft.pdf, accessed on August 10, 2011. 16. Hinţea, C.E., Hudrea, A.V. and Balica, D.O., ‘Evolution of the Agency Population in Ro- mania’, in Verhoest, K., van Thiel, S., Bouckaert, G. and Lægreid, P. (eds.), Government Agencies in Europe and Beyond: Practices and Lessons from 30 Countries, Hampshire: Palgrave Macmillan, forthcoming. 17. Hood, C., ‘A Public Management for All Seasons?’, 1991, Public Administration, vol. 69, no. 1, pp. 3-19. 18. Hood, C., ‘The “New Public Management” in the 1980s: Variations on a Theme’, 1995, Accounting, Organizations and Society, vol. 20, no. 2/3, pp. 93-109. 19. Huber, J.D., Shipan, C.R. and Pfahler, M., ‘The Influence of Divided Government on Policy-Making and Bureaucratic Autonomy’, 1998, Paper prepared for presentation at the Annual Meeting of the American Political Science Association, Boston, MA, September 3-6, 1998. 20. Kahneman, D. and Tversky, A., ‘Prospect Theory: An Analysis of Decision under Risk’, 1979, Econometrica, vol. 47, no. 2, pp. 263-292. 21. Kapla, R.S. and Norton, D.P., ‘The Balanced Scorecard – Measures that Drive Perfor- mance’, 1992, Harvard Business Review, vol. 70, no. 1, pp. 71-79. 22. Lægreid, P., Roness, P.G. and Rubecksen, K., ‘Autonomy and Control in the Norwegian Civil Service: Does Agency Form Matter?’, 2005, Paper presented at the Scancor/SOG workshop on ‘Autonomization of the State: From integrated administrative models to single purpose organizations’, Stanford University, April 1-2, 2005. 23. Langbein, L., ‘Bureaucratic Discretion’, in Rowley, C.K. and Schneider, F. (eds.), The Encyclopedia of Public Choice, vol. 2, Dordrecht: Kluwer Academic Publishers, 2004, pp. 53-58. 24. McCubbins, M.D., Noll, R.G. and Weingast, B.R., ‘Administrative Procedures as In- struments of Political Control’, 1987, Journal of Law, Economics, and Organization, vol. 3, no. 2, pp. 243-277. 138 25. McGauran, A.M., Verhoest, K. and Humphreys, P.C., ‘The Corporate Governance of Agencies in Ireland. Non-commercial National Agencies’, Institute of Public Admi- nistration, Dublin, 2005, [Online] available at http://www.cpmr.gov.ie/Documents/ The%20Corporate%20Governance%20of%20Agencies%20in%20Ireland.pdf, accessed on May 5, 2011. 26. Norman, R., ‘Letting and Making Managers Manage: The Effect of Control Systems on Management Action in New Zealand’s Central Government’, 2001, International Public Management Journal, vol. 4, no. 1, pp. 65-89. 27. OECD, ‘Agencies: Their Benefits and Risks’, 2005, OECD Journal on Budgeting, vol. 4, no. 4, [Online] available at http://www.oecd.org/dataoecd/3/7/43487848.pdf, accessed on August 10, 2011. 28. Painter, M. and Yee, W.H., ‘Task Matters: A Structural-Instrumental Analysis of the Autonomy of Hong Kong Government Bodies’, 2011, American Review of Public Ad- ministration, vol. 41, no. 1, pp. 395-410. 29. Papadopoulos, Y., ‘Problems of Democratic Accountability in Network and Multilevel Governance’, 2007, European Law Journal, vol. 13, no. 4, pp. 469-486. 30. Pollitt, C. and Talbot, C. (eds.), Unbundled Government. A Critical Analysis of the Global Trend to Agencies, Quangos and Contractualisation, London: Routledge, 2004. 31. Ruffner, M. and Sevilla, J., ‘Public Sector Modernisation: Modernising Accountability and Control’, 2004, OECD Journal on Budgeting, vol. 4, no. 2, pp. 123-141. 32. Schick, A., ‘Agencies in Search of Principles?’, 2002, OECD Journal of Budgeting, vol. 2, pp. 7-26. 33. Sulle, A., ‘The Application of New Public Management Doctrines in the Developing World: An Exploratory Study of the Autonomy and Control of Executive Agencies in Tanzania’, 2010, Public Administration and Development, vol. 30, no. 5, pp. 345-354. 34. Thompson, F., ‘Matching Responsibilities with Tactics: Administrative Controls and Modern Government’, 1993, Public Administration Review, vol. 53, no. 4, pp. 373-318. 35. van Thiel, S., ‘The Rise of Executive Agencies: Comparing the Agentification of 25 tasks in 21 Countries’, Paper presented at EGPA Conference 2009, Malta, September 2-5, 2009, [Online] available at http://soc.kuleuven.be/io/egpa/org/2009Malta/papers/ EGPA%202009%20thiel%20The%20rise%20of%20executive%20agencies.pdf, acces- sed on August 10, 2011. 36. van Thiel, S. and Yesilkagit, K., ‘Good Neighbors or Distant Friends? Trust between Du- tch Ministries and Their Executive Agencies, [Online] available at http://soc.kuleuven. be/io/cost/act/pdf/20081014_PhD%20Training%20School/Dag%202/7%20Thiel%20 and%20Yesilkagit%20autonomy%20and%20trust.pdf, accessed on August 10, 2011. 37. Verhoest, K., Peters, G., Bouckaert, G. and Verschuere, B., ‘The Study of Organizational Autonomy: A Conceptual Review’, 2004, Public Administration and Development, vol. 24, no. 2, pp. 101-118. 38. Verschuere, B., ‘Autonomy and Control in Arm’s Length Public Agencies: Exploring the Determinants of Policy Autonomy”, 2006, PhD thesis, [Online] available at http:// soc.kuleuven.be/sbov/rapport/s0105015_verschuere.pdf, accessed on August 20, 2011. 39. Verschuere, B., ‘The Autonomy – Control Balance in Flemish Arm’s Length Public Agencies’, 2007, Public Management Review, vol. 9, no. 1, pp. 107-133. 139 40. World Bank, ‘Policy and Implementation – Together or Separate?’, not dated, [Online] available at http://www1.worldbank.org/publicsector/civilservice/pol-imp.doc, acces- sed on May 7, 2010. 41. Yesilkagit, K. and Christensen, J.G., ‘Institutional Design and Formal Autonomy: Politi- cal versus Historical and Cultural Explanations’, 2010, Journal of Public Administration Research and Theory, vol. 20, no. 1, pp. 53-74. 42. Yesilkagit, K. and van Thiel, S., ‘Autonomous Agencies and Perceptions of Stakeholder Influence in Parliamentary Democracies’, 2011, Journal of Public Administration Resear- ch and Theory, Available at: http://jpart.oxfordjournals.org/content/early/2011/06/02/ jopart.mur001.abstract, accessed on June 24, 2011.