ANNEX 1: LITERATURE REVIEW 2 Autonomy as formal-legal design 3 Policy autonomy 10 Managerial autonomy and control 19 Autonomy-control balances 26 ANNEX 2: ORGANIZATIONS SURVEY 2002-2003 33 TYPE 1 34 TYPE 2 34 TYPE 3 35 TYPE 4 35 ANNEX 3: QUESTIONNAIRE SURVEY 2002-2003 36 Autonomy 37 Control 38 Variables explanatory research 39 ANNEX 4: MEASUREMENT VARIABLES VIA DOCUMENT ANALYSIS 42 Financial control 43 Structural control 45 Size and agency type (geographical structuring) 46 ANNEX 5: TRANSFORMATION SURVEY DATA 47 ANNEX 6: ORGANIZATIONAL SCORES ON THE INDEXES OF AUTONOMY AND CONTROL 49 Autonomy 50 Control 51 ANNEX 7: ORGANIZATIONAL SCORES ON THE EXPLANATORY VARIABLES 54 External variables 55 Internal variables 56 ANNEX 8: DESCRIPTION CASE ORGANIZATION VVM 60 Structure of the VVM 61 Personnel of the VVM 63 Managerial autonomy & control 65 Task and Mission 65 Budget (income) 67 Performance & Results 68 Culture 77 Environment 78 ANNEX 9: DESCRIPTION CASE POLICY-PROGRAMS 87 Basic mobility 88 Pegasus 98 ANNEX 10: DOCUMENT ANALYSIS CASE STUDY (09/2005 02/2006) 103 Step 1: Analytical tables 104 Step 2: Assessing policy autonomy via document analysis 105 ANNEX 11: QUESTIONNAIRE CASE STUDY (03/2006 04/2006) 117 Step 1: The questionnaire 118 Step 2: Descriptive statistical analysis 132 Step 3: Bivariate tests for differences (basic mobility & Pegasus) 143 ANNEX 12: INTERVIEWS CASE STUDY (04/2006 05/2006) 147 Step 1: The interviews 148 Step 2: Analysis of the interviews per respondent 153 Step 3: Bringing the 32 individual sheets together 156 1
ANNEX 1: LITERATURE REVIEW Annexes 2
Autonomy as formal-legal design The point of departure is Terry Moe s Politics of Structural Choice (Moe 1990). The argument is that the structure and the design of public agencies (hence also their autonomy/discretion) is the product of processes that evolve in a context of political and technical uncertainty and political compromise in which three actors are present; politicians, bureaucrats and interest groups. The political uncertainty stems from the fact that political property rights are uncertain in a democracy. Actors can be very powerful today, but can loose this power tomorrow. Political decision-makers know that decisions made or structures put in place today can be overturned or become subject to the authoritative direction of political opponents tomorrow. Therefore those that are in power today (the interest groups that are in the position to create a public agency e.g.) will tend to demand structures that insulate or shield off the public agency from strict political interference, for the reason that the structure will also survive in an eventual uncertain future situation (anticipation). Politicians must thus deal directly with strategic interest driven actors. In structural politics, choices made within these politicians-interest groups relationships have implications for the kind of bureaucracy they will create. In an ideal-typical situation where there is one dominant powerful interest group, one would expect that this power is used to tell everyone what to do, hence to bind the public agency very tightly (detailed legislation, close inspections and monitoring, ). The problem is however that when policy problems are of any complexity, the knowledge to impose the will upon the public agency, and to control and bind the agency tightly, will lack. This is an expertise problem that causes technical uncertainty. Some policy problems require technical expertise which is sometimes not present for those that have an interest in the design of agencies. In the case of expertise problems due to the complexity of the policy problems, powerful interest groups will find it worthwhile to let the experts in the agency some degrees of freedom, and only steer on general lines. However, granting to experts discretion to solve complex problems raises control problems on the side of the interest groups. These problems are argued to be common in principal-agent relations (adverse selection, moral hazard), and make control imperfect. According to Moe, these problems can (partly) be solved by relying on reputation, professionalism and predictability when selecting experts to do the job for them. Thirdly, real life politics will also face problems of political compromise (political opponents have also a say in structural design, and by doing so they can enhance their own control in the agency). Via the presence of veto players, negotiations will lead to compromises about the structural design and the autonomy of agencies. As a result of political compromise, political opponents will try to gain control by fostering mechanisms of political intervention in the agency, or by favoring agency decision-making procedures in which they can participate for example. In sum, the structure and design (and the level of autonomy that stems from this design) will be the result of processes of political and technical uncertainty, and of political compromise. Political environment: -uncertainty -need for compromise Technical (task) environment: -complexity of problem Principals (those that are in power) rationally assess: -the possibilities for agency design -the consequences of certain agency design in order to advance favoured policies agency is designed/structured with a certain level of autonomy Annexes 3
The claim about structural politics as advanced by Moe is deeply grounded in the particular US situation. This has consequences for assessing the outcomes of structural choice in parliamentary systems. The nature of political compromise will be different (e.g. number and status of veto players). Next to that, taking interest groups as a starting point seems to be more appropriate in pluralist than in neo-corporatist and consociational polities. One can also question the explanatory power of Moe s theory in policy domains without a powerful interest representation. On the other hand, several key features of the theory of structural choice such as political and technical uncertainties can be held upright, as these are intrinsic to democratic governance in general (Yesilkagit 2004a). Kutsal Yesilkagit elaborates Moe s claim in a theoretical article (Yesilkagit 2004a) by translating the structural politics into contexts of parliamentary systems. He starts from two types control problems the principal (oversight authorities) faces in designing agencies: the agency problem and the commitment problem (Horn 1995). The agency problem is about being sure that the agent acts in line with the principal s demands and thus retains from shirking. This agency problem has been identified in economic principal-agent theory, but in a political setting it goes right to the heart of the relationship between politicians (principals) and bureaucrats (agents). The commitment problem is about being sure that the institutional arrangement that was created for implementing certain policies will remain unchanged also in the future (after a regime change e.g.). In terms of Moe, the commitment problem for politicians stems from political uncertainty. Overcoming both kinds of problems involve certain costs (Van Leerdam 1999). Agency costs can (partly) be paid by imposing control on the agency (which can take several forms, e.g. structural checks such as ex ante control or ex post control). Commitment costs can be paid by trying to reduce political uncertainty (in Moe s terms this means distancing the agency from future political opponent s interference). According to Yesilkagit, agencies that have low degrees of autonomy reflect the agency-cost logic, while agencies with high levels of autonomy reflect the commitment-cost logic. The logic of the agency problem will induce principals to create agencies in which the information-asymmetry is easily bridged (low autonomy), while the logic of the commitment problem will induce principals to design agencies that are insulated from political interference (high autonomy). The agency logic is expected to prevail when politicians want to benefit from a responsive bureaucracy, for being able to give benefits to their supporters in the short run. The commitment logic is expected to prevail when the beneficiaries of the program in case are able to lock in the long term distributional consequences in the policy. Hence from this line of reasoning it can be expected that politicians will want to let the agency logic prevail, while certain interest groups will want to let the commitment logic prevail in the structural design of agencies. The model for explaining structural design that is then developed by Yesilkagit is based on two dimensions: (1) the structure of interest representation and (2) the macro-political institutional structure of a political system. Making abstraction of the hostility between agency and dominant interest group over policy goals, a fourfold typology is constructed: (1) a majority political system with dispersed interest representation, (2) a majority political system with concentrated interest representation, (3) a consensus political system with dispersed interest representation, and (4) a consensus political system with concentrated interest representation. Depending on the situation (1-4) and the prevalence of agency or commitment logic, the agency will be designed in one or another Annexes 4
way. It is hypothesized that the pure agency logic will prevail in policy domains with dispersed interests in majority systems (1), where the main problem of the party in power is technical uncertainty and bureaucratic responsiveness. The commitment logic will prevail in domains with concentrated interests in consensus (multiple veto players) systems (4), as the strong interest representation will want to lock in policies in a high autonomy agencies to which it has strong links. This commitment is possible due to the many veto-points in a consensus system, and especially when the interest representation is necessary for the implementation of the policy or for the expertise. There are, however, also asymmetric cases which are more complex. For example the consensusdispersed case (3) yields three hypotheses based on the terms of the coalition between the parties (assuming that there are two parties in the coalition): (a) both parties want to control the agency, (b) only one party wants, (c) no party wants to. In the case both parties want to control the agency, the party not obtaining the agency may want to have influence by drafting detailed legislation upon the agency or by imposing structures and procedures upon the agency. In the case where only one party wants the control over the agency, a pure agency problem occurs for the party willing to control the agency. In the case no one wants control over the agency, the controlling party might apply the strategy of blame-shifting (avoiding responsibility, as the responsibility for the agency might be considered harmful by the controlling party). In the case were there is a majority system and concentrated interest representation, it will depend on intermediate factors such as e.g. technical uncertainty (is the input of the interest representation necessary for having the policy implemented?), the strength or the power of the interest representation (can government afford to act unilaterally?) or the need to legitimize (will the input of interest representation be considered as more legitimized?), what logic will prevail in determining the structural design of agencies. Technical environment: -structure of interest representation Institutional environment: -macro-political situation (majority/consensus) Prevalence of a certain logic (commitment or agency) will induce principals (those that are in power) to design agency in order to advance favoured policies agency is designed/structured with a certain level of autonomy In an empirical paper on regulatory agencies in Denmark, Sweden and the Netherlands, Yesilkagit and Christensen (2005) assess the impact of political conflict on the choice political decision-makers make about the structural forms for implementing policies (autonomous arm s length agencies versus hierarchically subordinated and strictly controlled organizational forms). The key dilemma is whether rational politicians choose for strictly controlling the bureaucracy, or whether the rationality lies in the choice for designing policies (putting policies at arm s length) such that political opponents (once they come to power) find it hard to reverse. The theoretical framework applied is the politics of institutional choice. Agency costs (stemming from goal-conflict between principal and agent e.g.) can induce politicians to closely overview the agency, or commitment costs can induce rational politicians to design agencies that are far removed from political supervision. In the latter case the politicians want to be sure that desired policies persist in the long-run, even after political coalitions have been reversed (political uncertainty), or they might consider that putting policies at arm s length is beneficial for the legitimacy and credibility of the enacting coalition. Hence as Yesilkagit and Christensen argue, delegation Annexes 5
involves dilemmas and the way out of the dilemma is function of political decision-making. The level of political conflict is the prime determining factor and itself a product of the institutional rules and procedures of administrative decision-making, which differ between political systems. The explanatory variable (policy conflict) is made operational by looking at minority and majority government and by looking at single-party and coalition governments. The assumption is that the level of policy conflict increases from political system with single-party majority, over single party minority and coalition majority to coalition minority. The dependent variable (agency design) is made operational by looking at formal structural bureaucratic autonomy (the formal exemption of an agency head from full political supervision by the departmental minister through the insertion of an extra layer of hierarchy between the agency head and the minister, a governing board), and by looking at the legal basis of agencies (primary legislation such as legal rule through which agency owes its establishment e.g.). Although these are formal variables, the authors believe that these formal features of agency design do affect the real autonomy of agencies, as formal structures can affect information-gathering and reporting procedures within an organization e.g., and also tell something about politicians intentions about attacking agency and commitment problems. As the level of conflict between the main political actors increases, the hypothesis is that (1) the number of agencies whose head reports to a board decreases and (2) the number of agencies established through a formal law decreases. The empirical analyses (correlation analyses) showed that the more the level of policy conflict decreases, the more enacting politicians can design agencies through legal instruments that are difficult to reverse. In addition to this finding, the authors also recognize that other key variables need to be added to the analysis, such as other measures of policy conflict (e.g. range between veto players in a political system), the legislative capacity, or the level of corporatism (like influence of interest groups) in a sector. Macro-political environment as a proxy for "political conflict": -type of government *majority/minority *coalition/single-party Rational choice by agency designers depends on prevalence of agency or commitment logic Formal design of agencies (legal status, direct or indirect structural control by minister) Thatcher and Stone Sweet (2002) make a claim for combining the pure functional approaches in explaining the delegation of authority to non-majoritarian institutions with other less functional approaches. Non-majoritarian institutions are those governmental entities that possess some degree of specialized public authority but are neither directly elected nor directly managed by elected officials. Delegation is the authoritative decision, formalized as a matter of public law, that transfers policy making authority away from the established representative organs to public or private non-majoritarion institutions. The authors claim that functional explanations for these transfers, mainly taken from principal-agent theories, are incomplete. These functional logics of delegation principals find it rational to delegate because of commitment problems, informationasymmetries, blame-shifting etc. need to be completed with other, less functional explanations for several reasons. Firstly, the desire to delegate might be socially constructed instead of functionally induced. Symbols and social constructs, more than rational assessment of the principals functionality, induce the perception that Annexes 6
delegation is the best way for dealing with problems. Secondly, the functionalist explanations alone do not account for the variation observed (across countries and across policy sectors) about why, when and how to delegate ( European politics face similar functional pressure to delegate partly because they are similarly developing and integrating into international and regional economies, yet outcomes [of delegation] vary ). Thirdly, functionalist explanations do not account for the timing of delegation. Thatcher and Stone Sweet see two sets of responses to these problems. One response is to adapt functional explanations to local conditions and contexts. A second response is to replace the functionalist principal-agent framework with richer accounts for explanations. For example, the definition of interests of the key actors involved in the decision to delegate, or institutional isomorphism leading to spread of appropriate delegation practice over Europe ( orgy of borrowing, Wilks and Bartle), and historical-institutional legacies that play at the domestic level such as existing state structures that influence the possibility, ease and attractiveness to delegate. Being more a research agenda rather than an explanation for delegating autonomy to agencies, the contribution of Thatcher and Stone Sweet pleas for extending purely functional explanations. They suggest to investigate decisions to delegate in their context (country, time, policy domain) acknowledge constraints on the choices of principals (cognitive, normative, political, constitutional) and consider the explanatory weight of inherited structures, norms and experience the politics of delegation calls for analyses going well beyond simple functional logics, involving how actors interests are defined, policy is made and enforced and the legitimacy of government is conceived. Hence they propose two possible research agenda s, one completing the functionalist approach (1), one replacing the functionalist approach (2): 1 Functionality for the principal Context Decision to delegate to agencies 2 -Actors interest -Isomorphism -Historical legacy Decision to delegate to agencies In a recent article (2006) Elgie gives an overview of the literature that has addressed the explanation of why governments do delegate authority to quasi-autonomous agencies. He sees at least eight hypotheses and (sets of) factors that might explain this delegation: policy complexity, credible commitment, blame shifting, political uncertainty, institutional isomorphism, state tradition, political leadership and the spread of NPM norms. These hypotheses are derived from several theoretical and research traditions. He argues that initially the literature was dominated by rational theories such as principal-agent and the work by Moe and McCubbins, Noll and Weingast e.g., but that later also other less rational explanations were investigated, for example Thatcher and Stone Sweet, Gains, Pollitt et al e.g. The eight hypotheses are tested by Elgie by examining the reasons for the creation of independent administrative authorities (IAA s) in France. He combines the large-n method and the case study method. First, the reasons for creating IAA s are studied in a general way applying the large N quantitative Annexes 7
method for 34 IAA s. Secondly, he studies two cases (IAA s) in detail 1. Elgie defends this combined methodological approach by stating that we firmly believe that there are distinct advantages to an approach that combines large-n and case studies our findings are likely to be more robust if they are based on information from different types of investigation rather than only one it is the accumulation of evidence across the different types of study that matters as no single piece of information proves or disproves any of the hypotheses advanced. The findings of the combined research show strong explanatory power for the crediblecommitment and policy complexity variables. In highly complex policy fields, governments can benefit from delegating authority to technical experts. Governments can also gain from appearing credible. They commit credibly to various policy decisions by delegating these decisions (to agencies). The need to delegate to experts, and the desire to make a credible commitment, seem to explain a large part of the decision to create arm s length agencies. This evidence adds to the existing literature showing the importance of these variables, and as the hypotheses are proven now both quantitatively (e.g. Huber and Shipan) and qualitatively (this study by Elgie) their salience is strengthened. But as Elgie states, also other hypotheses have proven to have considerable influence. Institutional isomorphism seems to play a role because a cross-national policy transfer seems occurring. There is, for example, pressure from the EU to harmonize regulatory structures. Next to that, foreign agencies can act as models for the creation of domestic agencies. Thirdly, also political uncertainty and political leadership seem to play a role but only a minor one in the decision to create arm s length agencies. The author concludes so, due to the observation that most IAA s were created under two governments, and by the observation that one of the cases (CSA) was closely related to the president (Mitterand). Other theses, such as the influence of the NPM-doctrine, the domestic state tradition, the influence of blame-shifting seem to be less influential in the decision to creation IAA s. -Institutional isomorphism (imitation and external pressure) -Complexity of the task -Credible commitment Assessment of rationality and appropriateness by the government... Decision to delegate to agencies Sandra Van Thiel (2000, 2004) tries to offer an explanation for the proliferation of quangos 2 in Dutch government. She observed an increased use of agencies, ZBO s (zelfstandige bestuursorganen) and RWT s (rechtspersonen met een wettelijke taak) despite the lack of evidence about the performance of such arm s length organizations (see e.g. the discussions in the Netherlands about the accountability of quangos following critical reports by the Dutch Court of Audit). Hence according to Van Thiel this raises the question whether the observed quango proliferation is a well-informed choice, or rather an autonomous trend. In trying to find an answer, Van Thiel departs from a rational model explanation. This model explains the establishment of quangos as being beneficial for politicians that strive to be re-elected. Firstly, in quangos the members of interest groups can be 1 Cases: the Conseil Supérieur de l Audiovisuel and the Commission de Regulation de l Electricité. 2 Quangos are quasi-autonomous non-governmental organizations», a heterogenous group to which belong organizations ranging from contract agencies, public bodies, voluntary organizations and governmental owned enterprises (Greve, Van Thiel and Flinders 1999). Somewhat comparable to the range from type 2-type5 organizations in my formal-legal typology (cf. supra). Annexes 8
engaged, for example by having them seated in the boards of quangos, which ensures the support of these groups as voters to the politicians ( patronage ). It is assumed that arm s length agencies offer better possibilities to engage diverse groups in the organizational structure than governmental bureaucracies that are supposed to impartial. Secondly, the responsibility of politicians for organizations operating at arm s length diminishes. In cases of ill performance the chances that politicians will be held accountable (and loose electoral support) decrease. The backlash is that in case of ill performance, the distance created also might lead to fewer intervention possibilities for the principal (politicians), or that intervention will involve high monitoring costs. Hence, the fact whether an agency will be created will depend on the assessment of politicians about the costs and benefits of distancing service delivery in an arm s length agency. Which choice they make Van Thiel argues, depends on the situation at hand. A number of conditions that constrain the choice were evaluated (Van Thiel 2000) 3. Political ideology (benefits are similar for all politicians), and economic reasons (assumed benefits of agencies such as the reducing of fiscal pressure through efficiency gains outweigh the expansion of public tasks, hence the establishment of more agencies, in times of prosperity: in both cases the N of quangos is assumed to increase) do not have effects on the establishment of quangos, as was hypothesized. Neither does the degree of corporatism and the fact whether elections were being held, contrary to the hypotheses. Corporatism and elections are assumed to increase competition for votes, hence politicians wishing to reduce responsibility (especially in times of elections), and politicians wishing to include interest groups (especially in sectors with high levels of corporatism) are inclined to establish more agencies in order to gain support. The only factor found having significant effect on the politician s choice to establish a quango was the task. Tasks requiring specific investments (involving low re-deployability and high sunk costs) will force politicians to keep implementation at close range. Van Thiel concludes that the rational model (politicians making deliberate and rational choices, mediated to some conditions) has weak explanatory power in determining why quangos are established. Therefore, she suggests an expansion of this model. The starting assumption for this expansion is the belief that politicians have no full information about the (dis)advantages and consequences of their choices: their rationality is bounded, which implies that decision-makers face a certain level of uncertainty about the affects of their choices (to delegate or not). Additional analyses by Van Thiel showed that politicians use imitation in order to reduce this level of uncertainty. They imitate what has done before by simply repeating others choices or own previously made choices. According to Van Thiel imitation proved to be an important cause of the increase in the number of quangos. This raises the question about how and why this imitation occurs. Therefore she relies on DiMaggio and Powell s idea of institutional isomorphism. Organizations in the same organizational field will adopt the same structures and ways of working in order to survive. Institutional isomorphism can occur through three mechanisms: coercion, imitation or normation. Coercive isomorphism is related to the concept of resource dependence. Powerful actors such as the EU, other international institutions (e.g. OECD, World Bank) or domestic actors demand the use of quangos. As the lower governmental tier is dependent of the higher, it has no choice but to adhere to these demands. Mimetic isomorphism is a response to uncertainty. In situations where a clear course of action is unavailable, organizational leaders may decide to imitate a peer that they perceive to be successful. Normative isomorphism is the result of professionalization. The training or socialization of public 3 Regression analysis of data of decisions of ZBO-establishments (392 decisions in the period 1950-1993. Annexes 9
managers could include a favorable opinion towards quangos as a modern form of public management e.g. Or governments may employ a new type of managers that is very good informed about NPM and believes in this doctrine. A second source for normative isomorphism is the interaction between members of the profession, for example between managers and consultants. Consultants proposing similar ideas in different establishments of quangos could be called opinion-leaders. Van Thiel claims that institutional isomorphism can help to explain why quangos have proliferated in the Netherlands. Powerful actors may have imposed this model, sometimes implicitly and indirectly (coercion), the spread of NPM made the concept of the quango available, and perceived as useful to managers (imitation). Finally, as more and more adopted the concept, it became a kind of a normative standard in the profession (normation). To conclude, Van Thiel claims that the concept of institutional isomorphism can explain the variety observed among quangos ( quango proliferation can lead to a large variety ). The type of isomorphism at work, hence the source of information available, will influence the choice to create a quango ( the network in which governments operate, the type of manager, the type of consultant, all can influence the choice about the type of quango to adopt ). The variation in such variables will thus affect not only the proliferation in the use of quangos, but also the variety of quango s observed in practice. These new predictions, according to Van Thiel, offer a promising direction for ongoing research. Functional rationality (benefits vs. cost) for principal Isomorphism (coercive, mimetic, normative) Task features Bounded rationality / incomplete info -Choice to create a quango -Variety amongst quangos in practice Policy autonomy In a volume of 2001 Daniel Carpenter puts emphasis on the pro-activity of agencies as an explanation for the level of autonomy agencies have. He asserts that bureaucratic autonomy is politically forged. Bureaucratic autonomy occurs when bureaucrats take actions that are consistent with their own wishes, actions to which politicians and organized interests defer, even though they would prefer that other actions or no actions at all be taken (the discretion within the agency to take the first decisive step towards a new policy, establishing an agenda or to establish the most popular alternative in that it becomes too costly for politicians to alter). Empirical evidence (a historical account of 3 American agencies 4 in the period 1862-1928) shows that this bureaucratic autonomy was manifested by strong bureaucratic leaders which controlled their programs and organizations, using networks and internal socialization to transform the agency into politically distinct entities. Bureaucratic autonomy then occurs when bureaucrats successfully practice a policy of legitimacy. Legitimacy is assumed to be present when reputations (for efficacy, uniqueness of service, moral protection, expertise) are built by the agency leaders. This reputation should be grounded in a diverse coalition wrought from the multiple networks in which they are engaged. Organizational reputations are the result of organizational capacity (talent, cohesion and 4 Cases: the US Department of Agriculture, the Post Office Department and the Interior Department. Annexes 10
efficiency) and are socially rooted (grounded in diverse political affiliations maintained by career bureaucratic officials). Capacity was built by assembling talented offices where turnover was minimized. The process was completed by participating in the construction of lasting legitimacy in the form of organizational reputations. According to Carpenter, bureaucratic autonomy is thus premised upon stable political legitimacy of the agency, and not upon the popularity of a policy, administrative fiat, or a lobbying. The argument made by Carpenter is also a critique on two important theoretical schools; the Structural Politics School (Terry Moe) and the Procedural Politics School (McCubbins, Noll and Weingast). The main critique on the procedural politics is that procedural politics is unlikely to control agencies with stable political legitimacy agencies with esteemed officials who have publicly recognized capacity and expertise and who have access to organized citizens exercise power over the procedures of their agency. The point is that legitimated agencies can make it extremely costly for politicians to constrain their agency by using various control-mechanisms. Hence according to Carpenter, bureaucracies can be autonomous even when it appears that they are in perfect political alignment with politicians and the public mood: political harmony and the appearance of control can mask autonomy. The school of structural politics claims that the characteristics of an agency are the product of deliberate and strategic design by politicians and organized interests. The result is a chaotic politics of design that leads to the design of agencies capacities and procedures. The problems with this argument are that (1) much policy innovation occurs within agencies that were created before, (2) design cannot foresee or determinate the roles the agencies see for themselves and (3) organizational capacities cannot easily be designed, capacities grow evolutionary. Environmental networks Organizational capacity AGENT build reputations, resulting in political legitimacy, which enables AGENT to "forge" bureaucratic autonomy In an article by Hammond and Knott (1999) a theoretical model is developed for explaining the influence of toplevel public managers in making significant policy choices. According to Hammond and Knott, this debate is clustered around two extreme positions. Firstly those authors that claim that public managers are heavily constrained by political and institutional forces so that they have little room for significant choice. On the other hand there are the authors that claim that public managers are able to forge changes in policies, hence being less constrained by political and institutional forces. Two sets of variables are crucial in the model proposed by Hammond and Knott: (1) the constraints by the political environment and (2) the leadership strategy of public managers. A first factor in the environment is the extent to which the policy preferences of the powerful actors in the environment diverge. When the external actors in the agency s environment have diverging preferences, the agency has potential large influence over the policy choice. On the other hand, when all external actors agree the agency will not be able to alter policies to its own preference. This configuration of preferences in the agency s environment (the core ) sets the boundaries of the space within which the agency can move around. A second environmental factor concerns the distance between the status quo policy point (the policy that is issued in time being) and the agency s ideal policy point (the policy fostered/favored by the agency). Considering both environmental factors, Hammond and Knott claim two situations in which the agency is able to exert a large Annexes 11
influence on the policy: (1) when the agency s ideal point and the status quo policy point diverge, but both lie within the zone of points that are acceptable for all external actors. (2) When the agency s ideal point lies outside the zone of points that are acceptable for all external actors (core) and the status quo point is inside the core but at a considerable distance from the agency s point. To put it otherwise, an agency will potentially be able to exert some influence when there is potential to bring the status quo policy closer to the agency s ideal policy, as long as the policy fostered is acceptable for the major actors in the environment of the agency. This is an explicitly strategic view on agency discretion, as the agency is considered to be able to develop a strategy that takes advantage of the opportunities offered by the external environment. This environment is taken as a given, and its features (size of the core, distance between status quo point and agency s preferred point) will determine the boundaries for action of the agency (fixed constraints). However, an important question remains: how can agency heads try to loosen these fixed environmental constraints? The solution lays in a second important set of factors. Agency leadership-skills will be necessary to effectively develop a strategy to modify environmental constraints. According to Hammond and Knott, the successful manager is concerned with understanding and then developing a strategy for taking advantage of the opportunities offered by the environment this requires understanding of other actor s preferences, the relationship of those preferences to the status quo policy and the agency s own preferences. For developing such a strategy that is successful some leadership-skills are needed. In particular four skills will be important: (1) identification-skills, or the capacity by the agency leadership to accurately estimating what is politically possible, (2) anticipation-skills, or the capacity to be able to move first and so influencing other actor s moves, (3) persuasion-skills, or trying to persuade other actors to change policy and (4) issue framing-skills, or the capacity for defining reality in a for the agency advantageous way. Hence, a successful leader can identify a favorable policy dimension, but also has the rhetorical skill to impose this dimension on how other actors (president, Congress the theory was developed for and in the US context, but also general public) think about the problem. Political environment: -other actors preferences -agency's preference -current policy and the relation of these features define... Constraints to agency to "forge" own preferred policy Skillful agency leaders develop strategies to successfully "loosen" these environmental constraints Extent to which agency is able to impose its preferred policy This theory by Hammond and Knott was originally developed for studying the situation in the United States, but in later work by Hammond (2003) the applicability for other democratic systems has been assessed. Hammond starts with the observation that bureaucrats do vary in the extent to which they are controlled by elected officials, hence also in policy-making capacity. He identifies two major reasons why bureaucrats might develop an independent policy-making capacity. One reason is that bureaucrats may know more than the elected officials about what needs to be done, and how to do it. The second reason the focus of the argument by Hammond here is related to the divide et impera strategy. If the politicians are divided among themselves then the agency may be able to conquer. On the other hand if the politicians are unified, the agency may have to do what they are told. Two major kinds of divisions among politicians might occur. One involves veto-points that can reject proposals for policy (change). The more veto points, the more problematic policy change as more politicians are Annexes 12
able to block change proposals. The concept of veto-points is operationalized by looking at political or governmental systems (e.g. presidential systems, parliamentary systems, uni-cameral or bi-cameral). The second potential division line between politicians is the extent of policy disagreement amongst politicians or amongst veto-points. The more disagreement, the smaller the chances for policy change. The argument by Hammond is that differences in policy stability will be function of the interaction between the number of veto-points and the distribution of preferences. By applying these two variables, Hammond tries to answer the question whether different kinds of political systems will produce different kinds of bureaucratic autonomy (as the result of the room for autonomy the kind of equilibrium policies produce). Bureaucratic autonomy is the result of the size of the set of equilibrium policies (the more equilibrium policies, the greater the set of choices of policies the agency can adopt without fear that the choice will be overruled by the decisive coalition). This set of equilibrium policies is the core (the set of policy that no decisive coalition can overturn with some alternative policy, cf. supra). Hammond shows that in any political system (from presidential to parliamentary, each with its own institutional characteristics such as veto points and decision rules) various sets of equilibrium policies are theoretically possible. From large sets various equilibrium policies thus a large core, to small sets only one equilibrium policy thus a small core, depending of preferences of decisive actors. Thus, resulting from this theoretical observation Hammond claims that for explaining bureaucratic autonomy just relying on institutional features such as variety of veto-points and other decision-making rules (which differ per political system) is not enough. This is due to the fact that for every system the core can theoretically vary from a single policy to a range of policies. Thus there is no logical necessary relationship between a political system s institutional features (decision-rules etc.) and the size of the system s core. Hence also the preference variable in interaction with the institutional variables will have to be taken into account. Hammond claims that in studying bureaucratic autonomy it is necessary that both the institutional and preference variables are taken into account. The implication is that since policy preferences of politicians in a system may vary per issue area, also the size of the core may vary per issue area, following the reasoning deployed above. This suggests that the bureaucratic autonomy in a system can be expected to vary from agency to agency. Next to that, since the preferences of politicians may vary over time, also the size of the core may vary over time. Hence bureaucratic autonomy can also be expected to vary over time as well. Interplay between: Institutional features of a system -veto points -decision making rules Preferences of decision-making actors that vary per issue and over time Core of equilibrium policies (large or small) Level of bureaucratic discretion left for the agency Suzanne Dove (2002) has done research for explanations of the sustainability of agencies independence in Bolivia. Two primary case studies (agencies) at two moments in time were performed 5. Independence was made operational by looking at three concepts of agencies autonomy: personnel autonomy (the freedom to make 5 Cases: the Comptrollers General Office and the Superintendency of Banks. Annexes 13
HR selection and promotion decisions), financial autonomy (the degree of insulation of the agency s budget from pressures by the department of Finance) and technical autonomy (the freedom to take mission related operational decisions. The research by Dove has shown that sustained autonomy partly results from the formal structures contained in the establishment legislation of the agency. Elected officials can exert considerable controls over the agencies hence the autonomy of an agency depends on the favorable disposition of elected officials. But Dove also showed that despite many control constraints agencies sometimes are able to maintain their autonomy even in potentially hostile political environments. She identified 2 strategies employed by the agencies leadership to sustain the agencies independence. These strategies can be undertaken at three levels: (1) building clarity and consensus about the agencies mandate, (2) showing the value of the agency expertise and (3) developing or having clear measures (PI s) to show how well one is working. If these three conditions are fulfilled, agencies are likely to have more autonomy (or less political oversight). The insulation-strategy (agency shielding itself off of the environment) seems less successful in fulfilling the conditions above, resulting in lower autonomy of the organization (the Comptrollers General Office). The limited contact with the task environment reduces the ability to (1) develop a clear mandate for the agency on which there is consensus, (2) carve a role for the agency as a valuable source of expertise, and (3) the PI s developed by the agency will look meaningless for the environmental constituencies. On the other hand, the cautious engagement strategy is a better strategy as communication with environment leads to a better fulfillment of the conditions shown above (case of the Superintendency of Banks). Formal structural checks -Relation with environment -Task & program Control capacity for the principal Agencies' strategy (reputation-building) Autonomy is the result of -control capacity and will of principal -success of agent's strategy In a research paper, Huber, Shipan and Pfahler (1998) investigated the influence of (1) the political context, (2) the professional capacity of the legislature and (3) alternative institutional arrangements for influencing agencies behavior, on the propensity of legislators to limit agency s discretion in the legislation they adopt ex ante. The authors argue that these factors should not be treated as having merely independent effects, but rather that they are important in conjunction with each other. Empirically they rely on data on state-level bills enacted by US states related to health care and Medicaid in the period 1995-1996. By looking at a single policy area over several states, they account for variation in the three sets of explanatory variables which allows for testing the extent to which these factors influence the use of ex ante control. The claim they develop is simply one of costs and benefits. Limiting agency discretion by specifying precise detail in legislation is costly to legislators, hence the amount of detail placed in legislation depends on the magnitude of these costs and the benefits of paying Annexes 14
them. The key is to specify how the political and institutional environments (and the resulting actor s preferences) affect the expected cost and benefits of limiting agency discretion. The variable of political environment is deeply grounded in the US situation, distinguishing between divided (legislature and executive dominated by different parties) and unified government. The claim is that in a divided government the incentive to detail legislation will be bigger than in unified government. The need to limit discretion (by writing detailed legislation) will be smaller if legislators and agencies want to achieve the same objective (which is more likely the case in unified government where executives that are able to steer agents share the same goals as the legislative as belonging to the same party). Hence the political context induces the level of need/motivation to constrain the agency (see also Moe, Yesilkagit). Secondly, the professional capacity of the legislature affects the capacities to limit agencies discretion. Even if the environment induces the need for writing detailed legislation, the legislators must be able to do so. This ability or capacity will depend on the costs involved in writing detailed legislation. Huber et al. discern two types of costs. The first type of cost is informational. Legislators need information and expertise about what policies will yield what output and about what groups will need to be consulted. The cost of obtaining this information will be a function of the legislator s knowledge and expertise, which is amongst other a function of the staff legislators have at their disposal and of the extent to which legislative committees are specialized. The second kind of cost is about opportunities. For example the time available for the legislature. If there is a lot of time available then the cost for studying an issue will be lower (or the assessment of the time they cannot spend to other issues). A second example of an opportunity cost is the extent to which the individual legislator depends on other professional activities for his/her income. The more the legislator is dependent on other activities, the less inclined he is or she is for micro-managing agencies. As a third variable, there is the presence of alternative options for controlling agency behavior. The question is here how the institutional context shapes the costs of alternative control strategies (and thus the benefits of limiting discretion). Three alternatives are discussed by Huber et al.: the role of the courts, the legislative review of rule-making activities of the agency (cf. ex ante control), and the legislative influence in appointments of agencies (cf. structural control). The propensity with which these alternatives will substitute writing detailed legislation will depend on the extent to which the legislature is involved in appointing judges (that are favorable to them), by the extent to which the legislature can overrule agency decisions, and by the extent to which the legislature is involved in the appointment of agency personnel. Hence the claim by Huber et al. is that the political context affects the need to limit agency discretion and that the professional capacity will affect the ability to limit discretion. The interplay between both factors creates incentives and possibilities for control, and the extent to which the legislature seeks to limit discretion will depend on the existence of and the cost and benefits of alternative devices for control. This claim was empirically investigated. The dependent variable discretion was measured by looking at the total number of words enacted in new state legislation in the period 1995-1996 on issues related to medical care (Medicaid). More words in the act is implying less discretion for the agency. The dependent variable is regressed with the explanatory variables of political environment (level of divided government, capacity of the legislature in terms of compensation for the legislators, and the existence of alternative control devices such as appointment of judges, legislative veto for agency decisions, appointing personnel). The findings generally support the claims made. Firstly, the political environment cannot by itself account for the variance in the dependent variable. Secondly, when controlling for professional capacity and for Annexes 15