The role of agencies in policy-making: explaining variation in. ministry-agency relations in Germany

Transcription

1 The role of agencies in policy-making: explaining variation in ministry-agency relations in Germany Paper prepared for the EGPA Annual Conference, September 2010, Toulouse Study Group on Governance of Public Sector Organizations Tobias Bach Leibniz University Hannover Department of Political Science

2 Introduction This paper addresses how federal agencies in Germany perceive the relationship to their parent ministry when it comes to agency involvement in ministry decision-making. 1 The dominant assumption regarding ministry-agency relations is that agencies implement policies, whereas ministries are responsible for policy-making and direct support of the political leadership. The separation of policy and operations is a major claim of the New Public Management (NPM) rhetoric for improving public sector performance (Verschuere 2009). In Germany, which is not among the usual suspects of NPM reforms in the public sector, the separation of policy and operations is at the core of the ideal-type model of public administration (Döhler 2007). However, both anecdotal evidence and a growing body of empirical literature suggest that agencies are frequently involved in policy-making (Bach 2010; Döhler 2007; Egeberg 1995; Verschuere 2009). In addition, the literature suggests substantial variation both within and across countries (Gains 2003; Elder/Page 1998, 2000; Pehle 1998; Pierre 2004; Yesilkagit/van Thiel 2008), but few authors provide systematic explanations of the variation in agencies involvement in policy-making. The main task of an agency is considered as highly relevant for its relationship with the parent ministry, including the implementation of performance management techniques and various dimensions of agency autonomy and ministerial control (Döhler 2007; Hood and Schuppert 1988; Pollitt et al. 2004; Pollitt 2006; Verhoest et al. 2010; Wilson 1989). The literature on the involvement of agencies in policy-making puts forward similar claims about the effect of agency task characteristics (Elder/Page 1998, 2000; Gains 2003). However, there clearly is a need for a more nuanced analysis of the effects of agency tasks on policy involvement. The task-perspective is supplemented by structural and cultural explanations, 1 I would like to thank Per Lægreid and Sandra van Thiel for their very helpful comments on an earlier version of this paper which was presented at the 6 th meeting of COST-Action CRIPO in Cluj (Romania) in May

3 which offer both contrasting as well as complementary perspectives on agencies role perceptions (Roness 2009). This paper analyzes the effect of task characteristics, structural and cultural features on the agencies role perceptions. I distinguish two role perceptions which are associated with different types of ministry-agency relations, namely the agency as policy-maker and the agency as expert. These distinct role perceptions are outlined in the following paragraph. The paper is structured as follows: The next paragraph summarizes the literature on the role of agencies in policy-making, especially focussing on federal agencies in Germany. After that, I present the theoretical framework of the paper and develop several hypotheses on how task, structural and cultural characteristics will affect agencies role perceptions. Then I outline the method and data for the empirical analysis, which is a comprehensive survey of federal agencies in Germany conducted by the author in Next the results from the statistical analysis (descriptive statistics, ANOVA and multivariate regressions) are presented and discussed, followed by some concluding remarks. The role of agencies in policy-making: normative models and empirical findings The doctrine of separating policy and operations is a powerful and appealing principle which has influenced administrative reforms for many years (Verschuere 2009: 24-27). For instance, this doctrine is at the core of the New Public Management (NPM) reform ideas and manifests itself in structural devolution (e.g. by creating more autonomous agencies) and the use of contracts: NPM has through devolution and contracting sought to separate policy making more clearly from policy administration and implementation. Policy makers make policy, 2

4 and then delegate implementation to managers, and hold them accountable by contract. (Christensen/Lægreid 2001: 79). Although NPM reform ideas have not had a major impact on ministry-agency relations in Germany (Bach/Jann 2010), the separation of policy and operations is strongly embedded in the country s Rechtsstaat administrative tradition. Among other things, the normative model of public administration in Germany suggests that policy-making only takes places in parliament, whereas the agencies function is to impartially execute those policies which are laid down in statutory law (Döhler 2007). Moreover, agencies are not considered as political actors, but rather as neutral expert bodies, e.g. in the context of parliamentary hearings. Finally, agencies have a subordinate status in relation to their parent ministry, even if explicit hierarchical interventions are rare (Bach 2010: ; Döhler 2007). In spite of widely held normative doctrines of separating policy and operations, both anecdotal evidence and a growing body of literature suggest that agencies are frequently involved in policy-making. For instance, drawing on empirical studies from different countries on the de facto involvement of agencies in policy-making, Pollitt et al. (2004: 41) argue that the separation of policy and operations is not a core element of the agency idea and agency practice. Hence, these authors consider the policy-operations split as a subvariant within the agency trend (Pollitt et al. 2004: 42) and do not include it in their idealtype model of a modern agency. However, if agency involvement in policy-making is the rule rather than the exception, there is surprisingly little systematic research on this topic. In the remainder of this paragraph I summarize the main findings of the empirical literature and flesh out the article s research question in more detail. A number of studies focus on the role of agencies in the different phases of specific decision-making processes. Verschuere (2006, 2009) studies the involvement of a large service delivery agency in Flanders in the formation of two policy programs and attributes differences in the dependent variable to the agency s capacity for policy development, the 3

5 attitude and expertise of the agency regarding the policy program in question, and the willingness of the political oversight authority to include the agency in the process of policy development. Maggetti (2009) analyzes the role of regulatory agencies in two policy areas (finance and general competition) and three countries (Netherlands, Sweden, Switzerland) through assessing the respective agencies centrality in six legislative processes. He finds that the regulatory agencies generally had central positions in the decision-making processes under scrutiny and explains these differences as joint outcome of high de facto agency independence from politicians, a non-professionalized legislature and low independence of agencies from regulatees (Maggetti 2009: 465). Although not explicitly acknowledged by the author, this analysis clearly reveals country-specific patterns of agency involvement in policy-making. Egeberg (1995) compares the role of central agencies in two comprehensive planning processes in the transportation sector in Norway and finds that a lot of policy-making takes place at the agency level, albeit with some differences between the agencies under scrutiny. Also, several scholars study the role of agencies in policy-making without referring to specific decision-making processes. Yesilkagit and van Thiel (2008) find that public sector organizations in the Netherlands on average report frequent involvement in policy-making. Also, they find significant differences between legal types. Those organizations with the least degree of formal autonomy (i.e. those which are closest to the parent ministry) report the highest degree of involvement in policy-making (Yesilkagit/van Thiel 2008: 146). In the case of the UK, there is substantial variance as to the separation or integration of policy and operations between ministries and agencies: some agencies only do operations, others do both policy and operations, and a third type shares both functions with the parent department (Talbot 2004: 10). This variety is also reflected in executive agencies framework documents and their actual involvement in policy-making (Gains 2003: 70-71). Elder and Page (1998, 2000) compare the influence of agencies in policy-making in Sweden and Germany (with the purpose of drawing lessons for the UK). Among other things, 4

6 they find that agencies in Sweden supposedly have a somewhat broader role in policy making than in Germany and Britain (Elder/Page 2000: 234) which they explain by the consensual nature of the Swedish system (i.e. many issues which in other jurisdictions would be considered policy are here understood as operational ) and the consultation procedure during which draft reports of commissions of inquiry intended to lead to legislation are circulated to interested administrative agencies inter alia for their comments (Elder/Page 1998: 37, italics in the original). Swedish agencies may independently comment on those reports (unless there is a government directive on the issue at hand) and all comments are public domain. Moreover, they describe several episodes of policies that originated in the agencies under scrutiny, but they also find clear variation according to the agency and the political circumstances (Elder/Page 2000: 234). Pierre (2004: ) also reports that several Swedish agencies have become sources of policy in the 1990s and have either put forward extensive policy proposals or by virtue of their institutional autonomy formulated programs to guide their actions. However, it should also be noted that Swedish agencies enjoy an internationally unmatched degree of autonomy due to the lack of the principle of ministerial responsibility. In the German context, the literature on federal agencies suggests that some agencies are frequently involved in policy-making, whereas others operate at a rather large distance from policy-making. Döhler (2007) studies the steering of four large federal agencies by ministries and reports several episodes of policy-making by federal agencies (e.g. drafting laws and regulations, negotiating EU regulations in Brussels together with the parent ministry). Also, he argues that agency involvement in policy-making (which may involve diverging preferences between ministry and agency) is generally considered as part of the internal decision process within the executive and does therefore not take place in public. Pehle (1998: ) finds that the contribution of agencies in drafting laws are of high or fairly high importance for almost 60% of the ministerial bureaucrats in the Federal Ministry 5

7 of the Environment. However, he also finds substantial variation among ministry divisions, which suggests that the agencies belonging to this ministry (which are supervised by different divisions) are involved in policy development to various degrees. Also, while some authors report a frequent involvement of agencies in policy-making by drafting laws and regulations (Frach 2008; Welz 1988), other studies suggest rather low levels of policy involvement. Boswell (2009) shows how a large agency was largely excluded from formulating policy (172) until the agency was given new tasks. As a corollary, a research unit was established in the agency, which subsequently was increasingly used by ministry officials for providing evaluation reports and various kinds of support activities such as requests for briefings, speech-writing and responses to parliamentary questions (184). Moreover, survey data show that a substantial number of federal agencies report various policy-related activities, such as giving feed-back on policy effectiveness, the formulation of regulations and guidelines and the support of political initiatives, whereas only few agencies comment on draft laws from other ministries or prepare draft laws and decrees themselves (Bach 2010: 97-98). Elder and Page (1998, 2000) also find that federal agencies are involved in policymaking by offering policy advice to the ministries, which usually takes the form of consultation on policy proposals of the ministries, while some agencies may also initiate policy proposals. However, while they find little evidence of policy-making in a small agency with research functions, they emphasize that generally the involvement by the agency in policy making was predicated upon the primacy of the ministry in policy making; agencies were invited to participate in policy making forums within the ministry. (Elder/Page 1998: 35; see Boswell 2009: 164 for a similar argument). Finally, Elder and Page (1998, 2000) emphasize that agency officials usually will avoid challenging the (political) authority of their parent ministry and try to keep their hands out of politicized issues: Sensitive areas of policy are, to many bureaucrats, trouble. (Elder/Page 1998: 40). Nevertheless, these efforts to distance themselves from policy-making seem to be more prevalent in Germany than in 6

8 Sweden. This observation is in line with the normative model of public administration in Germany (Döhler 2007), according to which agencies are considered as expert bodies, rather than policy-makers. However, the self-perception of an agency regarding its own role in policy-making may change over time. Verschuere and Bach (2010) discuss the case of a large agency in Flanders which has changed over time from a predominantly technical, operative agency towards an agency responsible for broader policy issues (see also Verschuere 2006, 2009). The review of the empirical literature can be summarized as follows: The involvement of agencies in policy-making is quite common, even in contexts where norms of separating policy and operations prevail. Also, the policy-making by agencies may vary across agencies (even if they have the same parent ministry), policy-decisions, or countries. Also, the involvement of agencies in policy-making may increase or decrease over time. Moreover, agency involvement in policy-making comes in many guises, such as developing policy proposals, drafting laws and regulations, commenting on draft laws (especially in order to assess the practicability of policy proposals), negotiating EU regulations, evaluating existing policies, or preparing answers to parliamentary questions. Furthermore, especially in the German context, parent ministries have a critical function with regard to agencies policymaking role as they decide whether any policy proposal (or any kind of input provided by the agency) will be accepted. Finally, several studies point out the importance of the agencies own perceptions regarding their role in policy-making, which supposedly vary across agencies, over time, and across countries. This article focuses on the latter aspect and studies how federal agencies perceive the relation with their parent ministry with regard to agency involvement ministry decisionmaking. It distinguishes between two role perceptions, namely policy-making ( policymaking role ) and the provision of expertise ( expert role ) and aims at explaining the variation in these distinct role perceptions. This distinction thus takes into account embedded 7

9 norms of the role of agencies in policy-making (i.e. expert bodies), but it also takes into account a different model which emphasizes a more prominent role of agencies in policymaking (or at least a different perception of the kind of input agencies deliver to their parent ministries). The following paragraph outlines how different analytical perspectives potentially contribute to explaining variation in agencies role perceptions. Explaining variation in agencies perceptions of ministry-agency relations This article uses different analytical perspectives in order to develop expectations about the effects of different variables on agencies role-perceptions. It distinguishes between task characteristics, structural characteristics, and cultural features of ministry-agency relations (see Verhoest et al for a broadly similar approach). This approach takes into account theoretical diversity regarding the study of agencies (Pollitt 2004; Roness 2009). In essence, it is based on the assumption that using different theoretical perspectives is a useful strategy to grasp a complex phenomenon like ministry-agency relations. The approach which is chosen fits neatly in what Pollitt (2004) describes as a traditional social science approach which aims at the testing and development of middle-range theories. Also, the kind of data on which this research is based allows for testing the relative explanatory power of different analytical perspectives by contrasting them with each other, as well as assessing how these perspectives complement each other in explaining agencies role perceptions (Roness 2009). Task The task perspective highlights the importance of task characteristics for how agencies are managed and how they relate with external actors (Gains 2003; Pollitt 2006; Verhoest et al. 2010; Wilson 1989). Many observers find that the political salience of an agency has a strong effect on the relationship with its parent ministry. Dudley (1994: ) argues that 8

10 ministers are likely to intervene in highly politically salient agencies and thereby undermine formal lines of authority and create conflicts with the agency management. In contrast, ministers will pay much less attention to agencies with low levels of political salience, which can operate as an enclosed world with its own procedures and values (Dudley 1994: 234). However, his analysis exclusively focuses on ministers propensity for controlling agencies, leaving aside the function of ministry officials in managing ministry-agency-relations (Gains 2003: 68). In a study of the use of performance management in ministry-agency relations, Pollitt (2006: 39) finds that political salience is perhaps the most powerful and immediate influence and the most unpredictable in its effects in comparison to other task characteristics. Also, his findings suggest that politicians show little interest in performance management, which is conducted ( ) by managers and for managers ( ) unless and until disasters, scandals, or breakdowns come along. (Pollitt 2006: 41). In the German context, Döhler (2007) also finds that ministers seldom deal with agencies; the oversight function is by and large performed by ministry officials. In sum, the literature suggests that ministers only deal with agencies under conditions of high political salience. Yet the question remains whether political salience also affects ministry-agency relations in terms of policy-making? Also, how can political salience be measured? Gains (2003: 74-75) argues that the more a minister, the ministry and an agency are dependent on each other for different types of resources (e.g. political authority, financial resources, information), the more the agency will potentially be involved in policy-making. The degree of interdependence will depend on various characteristics of the agency s tasks. Agencies in high dependency relations are politically salient, have big budgets, provide monopoly services, or have multiple goals. The operational considerations of those agencies are more likely to be taken into account in policy development and cannot easily be ignored by departmental policy makers. Also, these agencies tend to have high-rank contacts with the parent ministries and possibly the minister, which is not the case for small agencies with 9

11 mainly executive tasks (Gains 2003: 68). However, these less politically salient agencies may play an important role in determining the overall policy agenda (Gains 2003: 75) through close bureaucratic networks between ministries and agencies, especially with regard to administrative policies aiming at the networks themselves (such as privatization initiatives), but also in terms of substantive policy-making (which is the focus of this article). As a result of the ambiguity of the concept, no single indicator for measuring the political salience of agencies exists. Among other things, the literature suggests that agencies will display relatively high political salience if they are large in terms of budget and/or staff, have direct contact with a large proportion of the public (i.e. provide services to the public rather than government itself), are frequently addressed in parliamentary questions and debates, and are frequently covered by news media (Dudley 1994: 235; Hood/Dunsire 1981: 76-94, Judge/Hogwood/McVicar 1997). It is important to note that some of these variables can be expected to be stable over time (budget and staff, contact with the public, policy area), whereas others are more sensitive towards contingent events such as crises, scandals, lawmaking etc (attention by parliament and news media). Moreover, as will be argued in more detail below, in the present article agency size is considered as a structural variable rather than a task variable. In this article, the level of parliamentary control and the degree of contact with the public are used as indicators for political salience. The literature suggests that the level of control by parliament is a good indicator the political salience. Agencies which are subject to high levels of parliamentary control supposedly operate in controversial policy areas. It can be assumed that high levels of political conflict about the agency or the policy area in which it operates (or both) make it hard for the ministry to ignore the agency in policy-making. Although there is empirical evidence that agencies which have traditionally been strongly involved in policy-making may be bypassed by their parent ministry in extreme crisis situations (Handke 2010), the following hypothesis can be formulated: 10

12 H1: Agencies which are subject to high levels of parliamentary control will perceive themselves as policy-makers rather than experts. The policy area in which an agency operates may serve as another proxy for political salience. Verhoest et al. (2010: ) argue that agencies operating in welfare and social policy are politically salient because they have direct contact with larger proportions of the population in comparison to agencies operating in other policy areas. Their empirical findings confirm the assumption that agencies in welfare and social policy have less policy autonomy (i.e. discretion in policy implementation) than other agencies (Verhoest et al. 2010: 251). In the case of Germany, there are only few federal agencies in those policy areas, as most social policies are implemented at the state and local level. However, one can assume that agencies having the federal administration itself as main target group (as an indicator of low political salience) will display traditional role perceptions, whereas agencies having citizens as main target group (as an indicator of high political salience) are more likely to perceive themselves as policy-makers: H2: Agencies with the federal administration as main target group will perceive themselves as experts rather than policy-makers. H3: Agencies with citizens as main target group will perceive themselves as policymakers rather than experts. Finally, a considerable number of federal agencies have policy advice as their (legally defined) main function, among them many of the so-called government research agencies (Ressortforschungseinrichtungen), such as the Federal Agency for Nature Conservation (Bundesamt für Naturschutz) or the Institute for Employment Research (Institut für Arbeitsmarkt- und Berufsforschung) which is part of the Federal Employment Services. Moreover, many agencies performing functions such as regulation or service delivery have policy advice as a secondary function. For instance, the main task of the Federal Office for Migration and Refugees (Bundesamt für Migration und Flüchtlinge) is to handle asylum 11

13 procedures, yet it also has a formal advisory function in the area of the integration (Boswell 2009: ). Finally, there are examples of agencies which do not have policy advice as either main or additional function, but which nevertheless are involved in policy-making, either by providing expertise to their parent departments or as experts in parliamentary hearings. The agency which is mainly responsible for overseeing social insurance bodies (Bundesversicherungsamt) is a case in point. According to its latest annual report, the agency has been involved in several recent law-making processes by submitting policy proposals, and the report actually indicates which policy proposals were eventually accepted and which were rejected (Bundesversicherungsamt 2009). It seems particularly interesting to investigate the role perceptions of government research agencies, as this particular type of policy advice based on scientific findings has a long tradition in Germany (Weingart/Lentsch 2008). The work of government research agencies is characterized by an inbuilt tension between safeguarding and upholding their scientific credibility by not getting too close to the ministries and not drifting away too far from their parent ministries by which they risk being excluded from internal governmental and parliamentary negotiations (Lersner 1996: 92). In sum, taking into account the specific kind of expertise these agencies provide to the ministries, the following hypothesis can be formulated: H4: Government research agencies will perceive themselves as experts rather than policy-makers. Structure The underlying assumption of the second analytical perspective is that formal organizational structure matters for decision-making in agencies and ministry-agency relations (Verhoest et a. 2010: ). Among other things, organizational size, type of specialization (vertical, horizontal) and mechanisms of coordination (collegial, hierarchical) 12

14 are considered as relevant dimensions of organizational structure (Egeberg 2003). In this article, four structural variables will be considered for explaining perceptions of ministryagency relations: ministerial oversight, legal type, agency size, and the number of policy units. The formal relationship between ministries and agencies is essentially defined by the ministries instruments to oversee the agencies. In terms of ministerial oversight, the administrative law in Germany distinguishes between functional oversight which may include hierarchical, top-down instructions on both administrative and substantial policy issues, and legal oversight, which formally leaves much more room of manoeuvre to the agencies (Bach/Jann 2010; Döhler 2007). In Germany, legal type and oversight relationship with the parent ministry are the main parameters of formal-legal agency autonomy. In other jurisdictions, there is rather weak empirical evidence of a straightforward relationship between formal autonomy and the levels of autonomy as they are perceived by the agencies themselves (Lægreid et al. 2006; Yesilkagit/van Thiel 2008; Verhoest et al. 2004). In the case of Germany, there is some evidence of a positive relationship between formal-legal autonomy and perceived autonomy, however depending on which dimension of formal autonomy is analyzed (Bach 2010: 106). It can be assumed that agencies which are subject to legal oversight by the parent ministry (rather than the more comprehensive functional oversight) operate at a greater distance from their parent ministries, which leads to the following hypothesis: H5: Agencies which are subject to legal oversight will perceive themselves as policymakers or experts to a lesser extent than agencies which are subject to functional oversight. In addition, the legal type of an agency may have an effect on the (perceived) degree of agency involvement in policy-making (Yesilkagit/van Thiel 2008: ), which leads to the following hypothesis: 13

15 H6: The role perception of agencies will vary according to legal type. The relevance of agency size is frequently discussed in the literature on agency autonomy, while the empirical findings show mixed results (i.e. both positive and negative relationships between agency size and different dimensions of agency autonomy) (Bach 2010: 103; Verhoest et al. 2010: ; Lægreid et al. 2006: 259). Also, agency size (i.e. staff numbers) is sometimes understood as structural feature (Egeberg 2003; Bach 2010: 94), as indicator of cultural heterogeneity (Lægreid et al. 2006: 242) or as both structural and cultural feature at the same time (Verhoest et al. 2010: ). 2 Moreover, agency size may be considered as an indicator of political salience (Dudley 1994; Gains 2003), hence belonging to the task perspective outlined above. There is obviously no single-best-way of grasping this variable in analytical terms. Hood and Dunsire (1981: 131) find that organizational size is highly interrelated with various organizational features: environment/type of work, political salience and structure are all to some degree related to size in their study of UK government departments. However, they argue that understanding agency size as a function of task reflects a rational-instrumental model of how organizations develop, thereby neglecting other explanations such as structural inertia (Hood/Dunsire 1981: ). Agency size may be a function of its task, yet in this article it is considered as a structural variable. With regard to ministry-agency relations, it seems plausible that large agencies (in terms of staff numbers) are less easily overlooked by ministries in policy-making than small ones. Large agencies imply an unequal relationship and a greater degree of dependency between the ministry and the agency (Gains 2003: 68). Verhoest et al. (2010: 207) follow the claim of Egeberg (2003: 120) that organizational size may be considered as reflecting the structural capacity of an agency. In that sense, large agencies will be able to develop their 2 With regard to size as an indicator of organizational culture, the argument is that small agencies will have a more homogeneous organizational culture than large agencies, which allows them to become more autonomous (Verhoest et al. 2010: 210). 14

16 own expertise, which may on the one hand increase information asymmetries between ministries and agencies (if this relationship is conceptualized in terms of principal-agenttheory) and hence lead to higher agency autonomy (Verhoest et al. 2010: 207). Their expertise and structural capacity potentially give large agencies a more prominent role in policymaking. In sum, the following hypothesis is suggested: H7: Large agencies will perceive themselves as policy-makers or experts to a greater extent than small agencies. The capacity of agencies which is explicitly devoted to policy issues or fundamental issues (Grundsatzangelegenheiten) may also affect the degree to which agencies act as policymakers or experts. The empirical literature suggests that ministries delegate tasks such as drafting laws and regulations to agencies because of problems of capacity (staff) and expertise in the parent ministry (Frach 2008: 114). Hence, whether an agency has one or several organizational unit(s) dealing with policy issues is likely to affect ministry-agency relations: H8: Agencies with large numbers of policy units will perceive themselves as policymakers or experts to a greater extent than agencies with small numbers of policy units. Culture In contrast to the structural approach, the cultural approach focuses on informal norms, values, and practices as key explanatory factors for organizational behavior (Christensen et al. 2007). The decision-making behavior of agencies and ministry-agency relations will reflect the norms and values of the broader politico-administrative context (Pollitt 2006), but agencies will also develop specific patterns of decision-making and norms of appropriateness over time (Döhler 2007; Gains 2003). In particular, the literature suggests that quite substantial differences among ministries exist as to how they deal with their agencies (Döhler 2007; Gains 2003; Weingart/Lentsch 2008; BRH 2005). However, these distinct patterns cannot be analyzed without referring to the (constitutionally sanctioned) principle of 15

17 departmental sovereignty, which allows each ministry a great deal of freedom as to how it organizes its agencies, how much decision-making autonomy it grants to the agencies, and how it exercises its oversight duties (Bach/Jann 2010). Here, it is assumed that agencies will be allowed more influence in ministry decision-making if there is a high degree of mutual trust between these organizations (see e.g. Rommel/Christiaens 2009). H9: Agencies perceiving high levels of mutual trust will perceive themselves as policy-makers or experts to a greater extent than agencies perceiving low levels of mutual trust. Moreover, it can be hypothesized that frequent contacts enhance mutual trust (Rommel/Christiaens 2009). Yet the kind of agency involvement in decision-making will depend on the hierarchical level at which this contact takes place. In a study of a large federal agency, Welz (1988) finds that the involvement of the agency in policy-making first and foremost takes place via the agency president. In their comparative case study, Elder and Page (1998, 2000) show that the agency which has the most frequent high-level contacts with the ministry also was most intensely involved in policy-making. Hence, it can be assumed that frequent (formal) contacts at the top level of ministry and agency will strengthen the role perception as policy-maker, whereas contacts among expert staff can be associated with expert advice in policy-making. 3 H10: Agencies having frequent contacts with the expert staff in the parent ministry will perceive themselves as experts rather than policy-makers. H11: Agencies having frequent (formal) contacts with the ministerial leadership will perceive themselves as policy-makers rather than experts. 3 The dataset also contains a variable measuring informal contacts among the agency and ministry leadership, yet this variable s distribution is highly skewed (informal contacts are very frequent for most agencies) and therefore the more normally distributed variable measuring formal contacts is used in the analysis. 16

18 The age of an agency may also be considered as a cultural variable. Verhoest et al. (2010: 210) argue that the older an agency, the stronger its institutional identity, and hence the more it is likely to resist (fundamental) changes. With regard to the dependent variables in this research, it could be hypothesized that old agencies have accumulated a large amount of expertise regarding their policy field and policy instruments. Hence, the following hypothesis may be formulated: H12: The older the agency, the more it will perceive its role as policy-maker or expert. Finally, it is assumed that the position of an agency s contact person in the formal internal hierarchy of the ministry will affect the degree to which the agency contributes to ministry decision-making. Although it could be assumed that this is a structural variable as it is related to formal positions and lines of authority, I argue that the position of an agency s contact person is a largely informal practice which tends to develop over time, which reflects the relative importance assigned to the agency by the ministry and ministry-specific patterns of managing ministry-agency relations. H13: The higher the level of access in the parent ministry, the more an agency will perceive itself as policy-maker or expert. Data and methods The analysis is based on a standardized survey of all federal agencies which was conducted between April and August The questionnaire was sent directly to the agency heads who were asked to respond to the questions themselves or to delegate this task to a senior manager who is well-informed on the organisation and its relation to the parent ministry(s). The survey was distributed via mail (together with a short note on the research project), and the agencies could respond either by sending back the paper questionnaire (which almost 40% did) or via an online platform (about 60%). About 45% of the questionnaires were filled in by the agency heads themselves, more than 27% were filled in 17

19 by division heads, and another 18% were completed by heads of section. The fact that the questionnaire was completed by different types of respondents may cause bias (Yesilkagit/van Thiel 2008: 142). In order to deal with this problem, respondents were asked to answer on behalf of the organization and not according to their own subjective perspective. Also, several respondents indicated (via telephone or written comments) that the survey was completed in a group effort in which the most knowledgeable persons concerning specific items were consulted. In sum, there is no reason to believe that the data are systematically biased through the respondents position in the agencies. The survey was an adaptation of similar research conducted in several countries in the context of an internationally comparative research project. 4 It included almost 70 questions on agency functions and target groups, various dimensions of autonomy, the involvement of agencies in ministry decision-making, the formal structure and accountability of the agency leadership and the use of management instruments. The majority of the questions were asked in all countries that are involved in the project. In addition, the German questionnaire included several questions on how the agencies assess their own role in relation to the parent ministry, including their involvement in policy-making and the use of their expertise by the parent ministry. Also, the agencies were asked whether they perceive changes over time of their in ministry-agency relations. In this paper, the answers to these questions are analyzed as dependent variables, whereas other items from the survey are used as independent variables. The questionnaire was distributed to 122 agencies, and the response rate is almost 60% (N=73). The total population includes all agencies operating under the direct supervision of a federal ministry (although hierarchical instructions may be limited) and having a public law legal status. In this paper, public law foundations are excluded from the analysis. These are 4 Verhoest et al. (2010: 45-53) provide an overview of the project and its research design. 18

20 usually very small organizations managing museums and memorial sites, which results in a slightly reduced sample size (N=66). In terms of legal categories, the sample includes both direct and indirect agencies (Bach/Jann 2010). The agencies belonging to the first category could also be labelled core administration; they are included in their parent ministries budget; and they are subject to various types of generic laws and regulations (financial management, civil service) and collective agreements for federal employees. Within the broader category of direct agencies, three legal types of agencies can be distinguished: Higher Federal Authorities (Bundesoberbehörden) are subordinated to a federal ministry (in some cases, they report to several ministries); they generally operate at the national level without regional or local units and are created by statutory law. Federal Institutions (nichtrechtsfähige Bundesanstalten) are created by ministerial decree and usually perform service functions and non-coercive tasks (e.g. research, information gathering), whereas higher federal authorities may take legally binding decisions towards external parties. Intermediate Federal Authorities (Bundesmittelbehörden) are regional administrations which are directly accountable to a federal ministry, including the administration of inland waterways, customs and tax administration, the administration of the military, and the border police. In contrast, indirect agencies have a higher degree of formal autonomy with regard to management decisions (among other things, they have a separate budget and may be exempted from applying financial management regulations). They also have at least in theory higher degrees of implementation autonomy due to the legal principle of selfgovernance. For example, indirect agencies usually have oversight boards in which stakeholders are represented (Bach/Jann 2010). Institutions of Public Law (Anstalten des öffentlichen Rechts) are mainly state financed and operate in highly diverse areas such as regulation, culture, and business development. In contrast, Statutory Bodies (Körperschaften des öffentlichen Rechts) mainly operate in the field of social security (e.g. unemployment 19

21 insurance, health insurance) and are financed via member contributions (e.g. employees paying unemployment insurance). 5 In the analysis, these two categories are analyzed together, as the sample only comprises a small number of statutory bodies. Results This section is structured as follows: First, I report descriptive results on how the agencies perceive the relation with their parent ministries. In addition to perceptions on policy-making and expertise, I also report whether agencies perceive changes in ministryagency relations over time. Second, using ANOVA statistics, I look for differences in role perceptions by legal type. Finally, the results from the multivariate regression analyses are reported and discussed. Descriptive statistics The agencies self-perceptions regarding different statements on their role vis-à-vis their parent ministry can be found in the Annex (Table A1). A majority of the agencies perceives that their expertise is used by the ministry to a (very) great extent. In contrast, a substantially smaller number of agencies reports that they can develop policy proposals in their area of responsibility to a (very) great extent, and most agencies perceive themselves as performing very little or no genuine policy work at all. For the ANOVA and regression analyses the items development of policy proposals and policy work are combined into an additive index of policy-making. The index was constructed by adding the item values and dividing them by two (Cronbach s Alpha = 0.715). 6 The use of agency expertise by the parent ministry in its decision-making is the second dependent variable of this research. 5 The indirect administration also comprises public law foundations (Stiftungen des öffentlichen Rechts), which are excluded from the analysis. 6 N= 64, minimum = 1, maximum = 4, mean = , standard deviation =

22 In order to track possible changes of these role perceptions over time, the agencies were asked whether these role perceptions had changed during the past five years (Table A2). A general finding here is that the majority of agencies reports no changes, and only few agencies report negative trends. However, almost 40% of the agencies report that their expertise is used to a (much) greater extent, and 28% of the agencies perceive an increase in their policy-making role. In sum, the descriptive findings show that federal ministries make use of their agencies in their decision-making, and this interaction is generally perceived in terms of using the agencies expertise, rather than agency involvement in policy-making. That being said, the data also show that some agencies clearly do consider themselves as policy-makers. However, in general these findings are in line with the dominating norms of appropriate behaviour, emphasizing that agencies operate at a certain distance from the ministries and usually do not actively seek policy conflicts with their parent ministries (Döhler 2007; Elder/Page 1998, 2000). The results on changes over time suggest an increasing importance of agencies in ministry decision-making, both as experts and policy-makers. These findings could be interpreted as part of a supposed trend towards an increasing use of external expertise in the policy making process by ministerial bureaucrats (Goetz 2007). In this context, earlier research points out that agency involvement in policy development depends on the consent of the parent ministry which formally has the task of drafting laws and other types of policy work (Elder/Page 1998). Hence, assuming that the normative model is still valid today, the observed increase in policy-making and use of agency expertise by the ministries supposedly results from a growing demand by the ministries for agency input in decision-making, rather than the other way around (agencies actively seeking to influence ministry decision-making). 21

23 ANOVA The results of the analysis of ministry-agency relations per legal type (ANOVA) are reported in Table 1. This analysis reveals no statistically significant differences regarding role perceptions between agencies having different legal types. It is worth noting that the differences between mean values are particularly small with regard to agencies role perceptions as policy-makers. In contrast, differences between mean values are larger in terms of agencies perceptions as expert bodies (although not statistically significant). Higher Federal Authorities on average perceive themselves as providing more expertise to ministry decision-making than other legal types, most notably Intermediate Authorities, but also Institutions of Public Law. However, as indicated by the standard deviations, there is a lot of heterogeneity within each group, which explains the lack of significant differences in the ANOVA statistics. In sum, H6 gets no support. Table 1: Perceptions of ministry-agency relations by legal types (ANOVA, p<0.05, mean values, standard deviations in brackets) Policy-making Expertise Higher Federal Authorities 2.34 (0.847) 4.00 (0.939) Federal Institutions 2.41 (0.769) 3.82 (0.603) Intermediate Authorities 2.40 (0.652) 3.20 (0.447) Institutions of Public Law 2.46 (0.749) 3.57 (0.852) F=0.075 F=1.817 p=0.973 (n.s.) p=0.154 (n.s.) Total (N)

24 Table 2: Results of ordinary least square regressions (beta coefficients) Policy-making Expertise (1) (2) (3) (4) (5) (6) (7) (8) Task parliamentary control.349* ** target group: federal administration target group: citizens government research agency.274* * Structure legal oversight staff (log) ** ** number of policy units * Culture trust **.431** contacts with ministry: expert staff (formal/informal) contacts with ministry: leadership (formal) agency age level of access in ministry *.264* Model statistics Model F-value Significance N Multiple R Adjusted R Notes: ** significant at 0.01-level; * significant at 0.05-level; - not included in the model 23

25 Multivariate analyses The remaining hypotheses were tested using multivariate regression analysis (OLS). The regression models are reported in Table 2. For each dependent variable, the table reports four different models. In a first step, I tested the explanatory power of each analytical perspective (task, structure, culture). Second, for each dependent variable I developed a model by combining those independent variables which have a significant effect in one of the singleperspective models. Also, I tested whether variables with beta coefficients larger than 0.20 in the single-perspective models (but without making a significant contribution) had a significant effect if included in the multi-perspective model. In most cases, these variables turned out insignificant and were not kept in the final model, with the exception of the number of policy units in the model explaining expertise (Model 8). 7 The regression models show that agencies which are politically salient (as measured by the level of parliamentary control), which are classified as government research agencies, and which have a relatively high level of access to their parent ministry (in terms of the internal ministry hierarchy) consider themselves as being more involved in policy-making than other agencies. These findings give support to H1 (political salience) and H11 (level of access in the parent ministry), but they are the opposite of H4 (government research agencies). As suggested by Gains (2003), high degrees of political salience imply strong dependency relations between ministry and agency, suggesting a relatively close ministryagency relationship, but potentially also less autonomy in comparison to other agencies (Elder/Page 1998, 2000). In that sense, parliamentary attention could be considered as 7 The models were tested for multicollinearity (VIF) and autocorrelation (Durbin-Watson). Also, scatterplots of regression residuals were used to detect heteroscedasticity. In addition, I checked the data for outliers (standardized residuals less than -2 or greater than +2), but the number of cases exceeding these limits is very small in all models, and all outliers are well below three standard deviations. In sum, the OLS regression method seems appropriate for the analysis of the survey data. 24

26 indicator of political salience as a rather stable phenomenon (the data on parliamentary control reflect the perceptions of the agencies concerning the past five years). However, parliamentary attention towards an agency may also be an indicator of variation in political salience, which would lead to a different interpretation of the findings. For instance, assuming that this variable (at least partially) reflects attention due to legislative activity concerning an agency (and the policies it administers), the findings suggest that agencies are typically involved in ministry decision-making which is linked to the legislative process. The findings give support to H11 (level of access in the parent ministry). The higher the levels of access, the more agencies get involved in policy-making. The majority of agencies (above 50%) have a head of section (Referatsleiter) as their contact person in the ministry, which suggests that policy-making agencies tend to have a contact person in the ministry which is either a division head, secretary of state or possibly the minister (these are the remaining answer categories). In contrast, the frequency of contacts at the leadership level does not have any significant effect in the regression model when controlling for other (cultural) variables. However, the findings are not in line with the hypothesis on how government research agencies perceive their input into ministry decision-making (H4). (It should be noted that the hypothesis suggested that government research agencies perceive their own role as experts rather than policy-makers. The regression models on the use of agency expertise in policymaking do not support this hypothesis either.) The regression models suggest that the role of government research agencies is best described as policy-makers rather than (scientific) experts, at least when it comes to the relation with their parent ministry. In general, these agencies are apparently well aware of the (supposedly increasing) role of scientific findings in policy-making. In sum, the level of policy-making by agencies is best explained through task and cultural characteristics, whereas structural characteristics do not have any significant effect. 25