Competing principles of agency organization – the reorganization of a reform

Introduction

Administrative reforms are normally characterized by intrinsic constraints, dilemmas, limitations, trade-offs and paradoxes (Pollitt and Bouckaert, 2011). The principles of organizational design tend to come in contradictory pairs (Hood and Jackson, 1991; Peters, 1998; Simon, 1946) and administrative reforms often happen in cycles and waves (Talbot and Johnson, 2007). Modernization of the state implies both integration and specialization, and to achieve its intended effects a reform needs to balance these partly competing reform measures (Margetts et al., 2012). Bouckaert et al. (2010) argue that there seems to be a stimulus–response pattern between specialization and coordination.

Thus public sector organizations are complex multi-functional entities trying to balance partly contradictory goals and considerations (Perrow, 1972). This seems to be a systemic feature of public sector organizations that needs to be taken into consideration when reorganizing the administrative apparatus, rather than regarding it as a disease that must be eliminated. Trying to find one best way of organizing based on a ‘one size fits all’ approach is normally not a successful reform strategy, partly because administrative reforms are often a political exercise with rather ambiguous roots in organizational or management theory (Peters, 1998). We need to understand the competing principles, institutional complexity and the coexistence and mixes of different organizational forms (Brunsson and Olsen, 1998).

Modern reforms in the welfare administration often have rather general goals such as increasing efficiency, increasing user attention, improving quality of services, etc. Such goals seldom produce unambiguous guidelines for how to organize or reorganize public services, and therefore corresponding reforms tend to result in a rather varied design of public sector agencies. One reason for this is that there is a layering process going on in public organizations (Streck and Thelen, 2005), where traditional Weberian features are combined with elements from the NPM and post-NPM reform waves, resulting in complexity and hybrids of the most important principles of structural design. This may give the executive leadership more flexibility in dealing with diverse interests and considerations, but may also potentially further conflicts and ambiguity.

The focus in this article will be on the following research questions:

What are the principles of structural organization of governmental agencies evident in the welfare administrative reform adopted by Norway in 2005?

The main argument in the article is that reorganization of the welfare administrative reform shows how through implementation bureaucratic leaders change the balance of organization principles originally decided on by the politicians, making the new system more complex and hybrid, something that creates challenges of steering and fulfilling public goals. The broad theoretical point of departure in the article will be structurally or instrumentally oriented organization theory (Christensen et al., 2007b), specifically the work of Gulick (1937) and Simon (1957). Different forms of specialization such as specialization according to geography, clientele, process and purpose as well as vertical and horizontal coordination will be addressed. The empirical basis for the article will be public documents and internal reports as well as interviews with key actors in the reform process, taken from a large evaluation project.

The instrumental point of departure

According to a structural-instrumental perspective, based on Simon’s (1957) theory of ‘bounded rationality’, reform processes in public organizations will be characterized by political and administrative executives controlling those processes (Christensen et al., 2007b). This may happen either through strong hierarchical control by top executive leaders or through negotiation processes that lead to compromises (March and Olsen, 1983). The perspective also presupposes that central actors will score rather highly on rational calculation or unambiguous means–ends thinking (Dahl and Lindblom, 1953). This implies that they have relatively clear intentions or goals, know which organizational solutions will fulfil those goals and have insights into the possible consequences of different choices for formal structures or organizational forms. An instrumental ideal is that leaders both have control and think clearly about reforms. This may happen through hierarchical steering, but because such steering may have legitimacy problems, negotiations furthering participation and legitimacy can be an alternative, but these often result in less clear solutions (Mosher, 1967).

Luther Gulick (1937) stated some main principles of formal organization. He stressed that overall the selection of goals could lead to certain formal structures being chosen that later on would change the content of a policy and eventually fulfil the stated goals. He did not think that this chain of relationships was a simple or unambiguous 1:1 relationship, but instead emphasized that every formal structure offered both advantages and disadvantages, so it was not a question of ‘one size fits all’. To strengthen the basis for structural design, he formulated several principles for formal organization of public organizations. A core organizational mechanism is specialization, which determines which tasks and relations can be grouped together and coordinated and which can be separated (Christensen and Lægreid, 2006). Specialization implies the creation of new organizational units with limited objectives and specific tasks out of core administrations that have many tasks and a wide range of objectives (Bouckaert et al., 2010).

Gulick (1937) stressed that the basic dimensions of specialization were vertical and horizontal. Vertical specialization relates to the hierarchical command structure and says something about how and to what degree authority is divided between hierarchical levels. It entails differentiation of responsibility between hierarchical levels, describing how tasks and authority are allocated between forms of affiliation (Lægreid et al., 2010). The horizontal dimension says something about how tasks and authority are divided on the same level. Horizontal specialization implies splitting organizations at the same administrative level and assigning tasks and authority to the different units. Often reorganization processes involve both vertical and horizontal specialization. In NPM, vertical specialization took the form of structural specialization while horizontal specialization resulted in the establishment of single-purpose organizations; both were considered key mechanisms for enhancing performance (Bouckaert et al., 2010).

The four basic types of horizontal specialization are related to purpose, process, clientele and geography (Gulick, 1937). In public organizations specialization according to purpose is often the primary principle and produces, for example, ministries or agencies, responsible for sectors such as defence, education, social services, health, etc. It is a principle that is easy to use, understand and recognize and encompasses several other sub-principles of organization. It is broad and flexible, but also rather unspecified as a design principle and may make it harder to deliver integrated services across sectors in ‘wicked issues’. Specialization based on process groups all specialists concerned with a specific sub-process or in one educational field – such as jurists, economists, personnel specialists or those responsible for budgeting or planning – together in the same organizational unit. The advantage of such an organizational principle is that it utilizes expert knowledge; the disadvantage is that a more holistic view of cases or regulations is often lacking. Specialization based on clientele, often used in health or social administrations, produces a special organizational focus on certain client groups, giving them special attention and treatment, but implies an organizational bias vis-à-vis other potential groups. Specialization based on geography increases the likelihood of standardization within a geographical area, but at the same time increases the probability of discrepancies between geographical units. Gulick’s principles are illustrated in rather complex and varied combinations in public organizations (Meier, 2010).

Specialization increases the need for coordination (Bouckaert et al., 2010). The emergence of specialization and fragmentation tends to trigger increased efforts to coordinate in order to enhance the alignment of tasks and efforts in organizations. These mechanisms are used to create greater coherence and to reduce redundancy, lacunas and contradictions in policies, implementation or management (Bouckaert et al., 2010). Gulick (1937) also formulated some principles of coordination. One principle was coordination according to ideas. The other and main coordination principle is coordination by formal organization, which has both a hierarchical and a collegial version. The hierarchical one alludes to the principle that the higher the leadership level the more coordination authority it has, while the collegial one focuses more on actors on the same level working together in different ways. Hierarchical coordination may be furthered by collegial coordination, but can also potentially be undermined by it because such coordination challenges hierarchical authority and control. Bouckaert et al. (2010) add networks and markets as two other coordination mechanisms. In our setting network types are especially relevant when they are linked to the partnership arrangements in the welfare administration reform.

We will argue that the awareness of the above-mentioned principles by central actors was there but was weakly developed, something that participated in driving the hybridity and eventually challenges to control and obtaining the intended goals.

Summing up, the structural-instrumental perspective, supplemented by the main ideas of Gulick (1937) about principles of specialization and coordination, is used in the empirical analysis in several ways. First, it helps us analyse the changing patterns of actors and influence, from domination of political actors to administrative leaders heading the implementation. Second, it is useful in analysing the organizational thinking or rational calculation of the main actors reflected in public documents and interviews. Third, taken together, the perspective contributes to explaining the persistent and expanded complexity and hybridity of organizational principles revealed in the reform and reorganization processes, through the rebalancing which takes place.

A methodological note

The analysis is based on empirical data gathered in a large evaluation project. In two rounds of elite interviews, covering the reform process leading up to the reform decision in 2005 and the reorganization process in 2008, between 50–60 and 20–30 respondents were interviewed, respectively. They came from the political executive leadership, administrative leadership in the superior ministry and from the leadership in the welfare agency. Adding to this, we base our analysis on internal reports from the processes and public documents, for example those sent to the parliament for decision.

Choosing Norway as the case study is, as shown, partly a pragmatic choice combining evaluation and research. But the welfare administrations that exist in Europe have been continuously changing during the past decade, and the Norwegian case illustrates some of the complex trends of reforming through combining NPM and post-NPM related reform elements, as also seen in the UK and Germany (Christensen et al., 2009).

Historical and institutional background

The principle of local self-government in Norway goes back to 1837. Since then the local level has been partly an extension of the central government, implementing its policies, and partly an independent, directly elected political level that can take the initiative on its own policies and services. Local self-rule, local knowledge and participation are therefore balanced towards central concerns.

The first agencies were introduced in Norway in the 1850s (Christensen, 2003). In 1955 the government formulated a policy of ‘hiving-off’, meaning that technical and professionally related tasks should be left to agencies and more of them should be established. However, by assuming responsibilities for policy development the agencies soon became more important and many of them became even larger than the ministries. The central agencies might also have regional and local organizational units.

Historically, the three types of administrative organization involved in the welfare administrative reform of 2005 – national insurance or pension administration, national employment service and social service administration – were separate entities on different levels, all of them with a history stretching back a hundred years or more and with different ministerial affiliations (Christensen et al., 2007a). The national pensions administration was traditionally a typically Weberian rule-oriented organization with a lot of complex case-work. Based in a large central agency, it had regional and local units, the latter in every municipality. The national employment administration, also centrally based, but in a somewhat smaller agency than the pensions administration, focuses on programmes to get more people into the workforce. During the last two decades the agency has changed quite a lot to become a modern agency forced to accept competition from private providers. The social security benefits were historically locally based and were traditionally quite different from the other two services. Their employees had less higher education and more local affiliation, including a local political committee, and overall more discretionary powers. Over the years, however, the service has become more professionalized and bureaucratized and hence similar to the other services.

It is also worth mentioning as a context that the unemployment rate in Norway has been among the lowest in Europe, that most economic indicators are scoring among the highest in Europe, indicating that Norway has been little affected by an economic recession. Accordingly, external factors have had minor influence on the welfare administrative reform and reorganization processes studied.

The process leading up to the administrative reform of 2005

One of the major challenges for the welfare services as they existed in the 1980s and 1990s was institutional fragmentation, i.e. they belonged to different ministries, agencies and levels. This created problems for clients with complex problems, in particular the so-called multi-service clients (Christensen et al., 2007a). In 2001 a majority in the Storting asked the government to start a process that would eventually lead to the merging of these services. This became a rather ambiguous and symbol-oriented initiative – labelled ‘one welfare administration’.

The conservative-centre minority government responded to the Storting’s request by basically arguing that the existing fragmented structure was the best one. The Storting sent the report back, saying that the government had not reacted in an appropriate way to the wish for a unified service. The government then established a public committee consisting of academic experts on welfare policy and administration (Christensen et al., 2007a). The committee basically supported the government in its resistance to a full merger between the services. The incoming minister in 2004, now the head of a labour and social ministry embracing all three services, realized that it was impossible to go back to the Storting with a proposal that still insisted on a fragmented structure. He proposed a compromise that was accepted by an overall majority in 2005. The main goals of the new welfare administration were to get more people off welfare and into work and to be more efficient and user-friendly.

The new welfare administrative reform had two major elements. First, it merged the pension and employment agencies, from top to bottom, into a new welfare agency (Askim et al., 2009). In terms of its relationship with the ministry this was established as a traditional agency meaning a combination of independence and political control, which reflected the political salience of the policy area. Second, a mandatory ‘one-stop shop’ was established as a physically co-located local partnership between the three services, which produced the rather unusual combination of a central and local hierarchy. The local partnership agreements had both mandatory (co-location, minimum number of services) and discretionary elements (choice of leadership structure, adding extra local sub-services) (Fimreite and Lægreid, 2009).

The reform implied both stability and changes with respect to formal structural principles. The ministerial responsibility principle did not change as such, but the minister now had both control of all three services and the agency responsibility for a newly merged agency, i.e. potentially making the minister both more powerful and more vulnerable to criticism from the Storting. The most critical part of the reform, however, was the local partnership that potentially was both complex and hybrid. What this amounted to was a mandatory partnership between central and local government, meaning that employees at local welfare offices worked both for the central government, i.e. the ministry and agency, and for the municipalities with their local political leadership (Fimreite and Lægreid, 2009).

Merging the two former agencies into a new central welfare agency, with units at the regional and local levels, increased both structural and cultural complexity, i.e. units and employees had to be merged and moved around, and professional milieus with different norms and values were pressured to collaborate and develop a new and more holistic culture. The new agency was also rather complex in its internal structure, with several central staff units and a major organizational division between the ordinary line organization on all levels and an all-encompassing ‘specialist units division’ (Askim et al., 2009).

Most local welfare offices chose a united leadership structure (93 percent), meaning either a leader from one of the central government services (most often) or a leader from the municipalities. But how many and what type of local services were moved to the local welfare offices varied a lot, making national standardization very difficult. The local offices that changed the least were those in the large cities where the common model had been a divided leadership whereby former service units were retained inside the new local offices, potentially making it more difficult to develop a real unified service.

When the new organization was established, the unions managed to get an agreement saying that no employee should lose their job, making it more difficult to fulfil the goal of more efficiency and aggravating the complexity of the organization (Askim et al., 2010). It was also decided that there should be one welfare office in every municipality, which, given the number of services and sub-services, presented a challenge for local competence. Politically, however, this was understandable given local policy concerns. The regional level in the new welfare organization thus lost out both to the central and to the local level, but this was later to change.

The reorganization of the reform in 2008

The reorganization of the reform in 2008 had two central components. First, it established six regional pension units, which meant moving employees from the local offices up to the regional level and implied a vertical de-specialization or integrative movement. The units were subordinated to a central agency department for special units, entailing a kind of horizontal differentiation. This reorganization to a large extent removed pension services from the local level and coincided with a large pensions reform and the introduction of a new ICT system for pensions (Førde, 2011).

Second, at least one administrative welfare unit was established at the regional level in each of Norway’s 19 counties to handle rights-based services and benefits. Altogether there are now 37 regional administrative welfare units. They were placed in the main line organization, under the leadership of the NAV county director, representing a vertical de-specialization. The reorganization also moved more local employees in the NAV administration up to the regional level. Altogether this implied a clear weakening of resources in the local NAV offices.

The more general principles of casework also changed with the reorganization from a generalist to a specialist one. The reorganization in 2008 formally left this principle intact but subscribed to the principle of economies-of-scale and the importance of having specialists on the regional level handling most questions, as a way of promoting efficiency, professional quality and the equal treatment of cases. Since moving a lot of employees up to the regional level left the local offices more vulnerable, even though they had also lost a lot of tasks, it became more common to urge the local offices to collaborate or specialize on certain task portfolios within a county.

The 2008 reorganization was based on a principle of level differentiation in the decision-making process related to five phases or sub-processes. The local offices were to be given the task of informing clients about the various welfare policies and opportunities for support and services and of receiving all the different types of applications. However, most of these were now to be handled in the new regional/county administrative units. Once decisions had been taken about payments these were to be made by the regional and national level, while it was the local offices’ job to support clients, in getting work, for example. So the local offices were assigned tasks 1 (informing), 2 (receiving applications) and 5 (follow-up) in the decision-making and handling chain, while the regional and/or central level were given tasks 3 (deciding) and 4 (paying).

The reorganization also changed the system in which clients approached the welfare administration. The original reform was based on a ‘one-door’ principle, meaning that the clients, particularly multi-service users, only had to come to one physical location. In the reorganization of 2008 there was more talk of a modern ‘three-channel strategy’, meaning that the number of clients who had to actually show up at the local welfare office was reduced and services were also to be provided via the internet or by telephone, with the latter entailing the establishment of large call centres.

Why were some of the central principles of such a huge administrative reform reorganized again after only a few years? First, at the time when the reform was decided on it was important politically to have some kind of merger of services combined with using the local partnerships as an incentive for the municipalities and social services to go along with it. However, the reform was mainly decided against the will of the administrative leadership in the ministry and the former employment administration; hence the post-reform repositioning (Patashnik, 2008). The reorganizations of 2008 brought back elements from the pre-2005 process in the sense that the pension services became more of ‘an organization within an organization’, along with the old core element of letting the employment service and social services collaborate more locally. Added to this, following the reorganization of 2008, the employment services, which had managed to keep more central positions in the central NAV agency than the pensions service, ended up with an organization, including regional units, that looked more like their old employment organization, in other words the structural reforms exhibited path-dependency (see Krasner, 1988).

A second explanation is that it was natural to develop the NAV organization further after 2005. Seen from a capacity point of view it was difficult to cope with more than merging two national services, establishing a local partnership and gradually establishing all the new local welfare offices. The public documents that prepared the way for the decision on the reform in 2005 mentioned that some of the services later on needed to develop further and to establish larger units for handling cases. Even though it may sound somewhat paradoxical and controversial, particularly from the local level, to say that moving resources and tasks away from the local level and up to the regional level was a win-win situation for both levels, the respondents’ arguments were that larger units on the regional level all represented improvements in efficiency, quality of competence and equal treatment of cases, not to mention a strengthening of the local offices.

A third reason for the reorganization of 2008 was that a huge pension reform was imminent that would anyhow result in some kind of reorganization of the pensions administration. Whether that would also have led to the reorganization of the other two services is maybe more debatable, but that was the respondents’ argument. The chances of the reorganization of the pensions administration succeeding were greatly improved by major investment in a new ICT system, while the other two services and their new administrative units on the county level still had to struggle to cope with four or five different old ICT systems for their services, even though they too were in the process of digitalizing and scanning documents.

After the reorganization of the reform in 2008 there was a lot of public debate about the NAV reform, including several periods of crisis and a public hearing in the Storting based on a very critical report by the General Auditor’s Office. Our elite respondents point to their experience of the opposition in the Storting and employees’ unions and client organizations becoming increasingly critical. The media also became involved using a number of distressing individual cases in their critique. All this is in rather stark contrast to the main sentiments of the elite respondents we interviewed who were rather supportive of the effects of the reform and the urgency of the reorganization of the reform in 2008. One explanation for this is that most of the external actors lost influence in the implementation process, more than the executive politicians, and tried to score political points in the media through some kind of ‘worst case scenario’ argument.

As a result of all the criticism the Ministry of Labour in 2010 established an expert group to analyse the organization of and activities in the local welfare offices and the county administrative units, and in particular their interaction and division of tasks – in other words to look primarily at one central part of the reorganization of the reform. The group’s assessment was rather critical, pointing to problems of productivity and quality in service provision, but also to increases in the number of clients included in employment-related activities. The problems outlined were primarily related to the implementation of the reform, to an increase in the demand for pension services, primarily brought about by an increase in unemployment and to lack of ICT support. The argument was also made that some of the problems experienced by NAV were, however, also evident in the former services, showing some path-dependency.

Even though the group was critical, like our elite respondents, it mainly supported the reorganization of the reform in 2008 and also thought that NAV had been able to counteract some of its negative effects with compensatory measures. Specialization through administrative units on the county level was an important precondition for local offices having their foci on individual guidance and following up clients. It was also argued that services with a high volume gave economy-of-scale benefits and also enhanced equal treatment of similar cases. All these arguments were in line with the basis for the reorganization of the reform in 2008.

The main features of the actors and organizational principles in the processes are summed up in Table 1.

OPEN IN VIEWER

Table 1.

Actors and organizational principles in reforming and reorganizing the Norwegian welfare administration 2005–08

	Reform in 2005	Reorganization in 2008
Most influential actors	Political leaders, supported by external experts	Administrative leadership in welfare agency
Main organizational principles used	Combining central hierarchical steering with local self-government, in local partnerships	Rebalancing of hybrid model More centralization and regionalization
	Much emphasis on geographical specialization and coordination	More use of sector specialization and less emphasis on geography

Analysing the competing principles of rebalancing

The theoretical departure for this article is an organizational perspective based on the concept of bounded rationality (March and Simon, 1958). Formal structures and procedures organize some actors, cleavages, problems and solutions into decision-making processes in the welfare administration while others are excluded (see Schattschneider, 1960). Thus, organization is politics by other means, and structures are important because they influence outcomes (Meier, 2010). Therefore the formal organizational structure of public organizations represents an important selection mechanism that constrains as well as enabling the service delivery process. Their quality depends on their success in balancing unity, integration and system coordination on the one hand and diversity, flexibility and local (government) autonomy on the other (Olsen, 2004).

The Norwegian welfare administration reform focused on introducing more coordination in service delivery. But it did not use a ‘pure’ set of post-NPM principles (Christensen and Lægreid, 2007). Formal vertical and horizontal reintegration or de-specialization were blended with NPM instruments such as performance-management systems, purchaser-provider elements and devolutionary elements such as strong local welfare offices (see Christensen and Lægreid, 2001). The reform was also reorganized and therefore modified after it had been decided on, combining a complex mix of horizontal specialization and vertical de-specialization with a focus on the regional level.

In the first phase of the process the politicians were the main reform agents, dominating both the experts and the central bureaucracy with respect to the organizational model chosen. In the second phase of the reform the actor constellation was the other way around, proving Patashnik’s (2008) assertion that it is important to understand the regrouping of the actors’ field. The politicians were now less active participants and the reorganization of the reform was mainly seen as an internal managerial process, i.e. the main reform process and the reorganization process had different actor structures and therefore different results concerning organizational principles. In both phases, however, the process typically scored rather low on clear organizational thinking, which instead was changing, ambiguous and not well founded, despite the inclusion of experts. Actors shared common goals but proposed widely differing routes for arriving at them. This seems to be typical of reform processes in many countries, because the societal and public structures, cultures and interests that must be catered to are becoming more complex (Pollitt and Bouckaert, 2011).

One way of understanding the NAV reform process is to look at it as a two-phase model of reform implying a sequencing path, whereby the local one-stop shops should be completed before embarking on developing regional units. Capacity problems, reform complexity and goal conflicts imply a sequential attention to goals (Cyert and March, 1963). First, the merger and the establishment of local welfare offices had to be implemented, then this model had to be recalibrated, because the whole new organization lacked the capacity to do everything at the same time.

Another way of understanding the reform dynamic is to look at it as a learning process. The trajectory of the NAV reform follows a sort of stimulus–response pattern regarding specialization and coordination (Bouckaert et al., 2010). The reform itself had a clear holistic integrating ambition focusing on coordination issues. But in the reorganization of the reform in 2008 the organizational model was somewhat rebalanced towards re-specialization. This implied increased internal horizontal specialization between pension and employment/social services and between different phases in the decision-making and service-providing process, which taken together represented a geographical integration process strengthening the regional level vis-à-vis the local one. This represents a mixed order of different organizational principles. All types of specialization have advantages and disadvantages and if one principle becomes too dominant it may trigger a counter-reaction whereby the opposite principle is reactivated and the mixture of complementary principles rebalanced.

Generally, increased specialization results in an increased need for coordination, but which specialization principle is selected will be of considerable significance for the choice of coordinating mechanisms. The first question is, therefore, whether the same specialization principle shall apply at both the central and local levels or whether these principles can be at variance. The next question concerns the implications this may have for multi-level coordination as well as internal coordination at the different levels (Fimreite and Lægreid, 2005). For example, if the central level is organized by sector and local government by clients (or process or area), will this imply weak vertical coordination between the central and local levels while horizontal coordination within local government is well-established? Will the result of this be increased autonomy and holistic thinking locally? And will this, in turn, present a challenge to integration between the two levels of government, resulting in the need for new coordination measures designed to counteract the consequences of autonomy?

Coordination and coordination mechanisms are challenged when principles of organizational specialization undergo change (Verhoest and Bouckaert, 2005). The principles of specialization are concerned with tasks and relationships which should be regarded in conjunction and coordinated, and which could be kept detached. And different specialization principles will enhance different networks, identities and conflict patterns. Redesigning sectorally specialized organizations into geographically structured ones would thus tend to transform functional conflicts into territorial conflicts (Egeberg, 2001, 2004). This was a major concern with the reform in 2005 with the focus on the local level – a focus that shifted to the regional level in 2008. When the administrative units were reorganized in 2008 there was some discussion about whether larger regional units were more feasible. However, the geographical-political interests of the counties led to a decision against this. Thus, structures and specialization principles are value-laden and they institutionalize biases in favour of one set of clients or users over others (Meier, 2010).

The NAV reform is an administrative reform aimed also at strengthening the steering capacity of the welfare administration. A complex multi-level system including a mixed order of hierarchy and network has been set up. The network is represented by the partnership model between the central and local government while the hierarchy obviously extends from the central government – i.e. the ministry, via the central agency to the regional units and below. The network represents coordination through mutual adjustment rather than hierarchical steering (Bouckaert et al., 2010). An organizational model that implies use of these two forms at the same time is challenging.

To make it even more complicated the organizational specialization in NAV has an inbuilt tension between specialization by geography, by purpose, by client and by process. NAV encompasses both tasks that are independent of place based on national standardization and equality, like pension issues, and place-related tasks based on local geographical discretion and leeway, such as employment-related issues and in particular social services (Fimreite, 2011). The challenge is to combine specialization principles in such a way that both considerations are addressed.

The starting point in this reform process was a welfare administration scoring high on proliferation and fragmentation – owing to strong specialization by purpose in central government combined with coordination based on geography in local government – and low on coordination and consolidation owing to weak horizontal coordination within central government as well as between central and local government. The main problem with this organizational model was that some of the problems and tasks did not follow the organizational borders but were trans-boundary. The wicked issues that the welfare administration was supposed to handle blurred the borders between organizations/services and administrative levels and were thus difficult to handle properly. There was a mismatch between the problem structure and the organizational structure.

The welfare administration reform of 2005 tried to solve this problem by implementing two main organizational changes – a merger of the employment and pensions services in central government both at the ministerial and at the central agency level, and the establishment of a social services partnership model at the local government level. This was done without altering the two partly contesting doctrines in the Norwegian political-administrative system – the principle of ministerial responsibility and the principle of local self-government. The focus was on the local partnership models, which were supposed to combine these two principles by upgrading specialization by geography and coordination by networks.

In 2008 the reform went into a second stage. Some of the organizational measures introduced in 2005 were modified or partly reversed. The bureaucracy bounced back and restored specialization by purpose. Tasks and resources were moved from the local partnership agencies in the municipalities to governmental bodies at the regional level. In contrast to local government with political decentralization to politically elected bodies, the regional units were branches of the central agency and thus represented administrative decentralization or delegation. Thus the main specialization principle was by purpose or task and not by geography. It was supposed to bring about standardization across regions and within the same tasks, but variation between different tasks. To some extent this last reorganization represents one step back towards the original organizational model, but the pendulum has by no means swung back to the starting point. It is, however, a paradox that integration and improved inter-organizational coordination as well as increased coordination between central and local government, which was one of the main goals behind the NAV reform, is still a big challenge.

Our analysis of the Norwegian case reveals a reform process that has produced complex and unstable solutions that in different ways attend to a balance of different principles of specialization and coordination, and their combination. The complexity that emerges reflects the fact that hierarchical efforts to control the reform process are constrained by problems of rational calculation. The case also shows that expected effects are problematic to fulfil for political and administrative leaders. The overall performance of the new system has not lived up to expectations, so even though central control has been achieved, the local partnerships and offices are struggling to deliver on the main reform goals (Askim et al., 2010).

Conclusion

One general lesson is that administrative reform does not end when the formal decision is made but tends to change during the implementation process, when the pattern of actors changes (Patashnik, 2008). It seems to be difficult to find a stable equilibrium between different considerations. Second, administrative reform is not only about internal administration but is also a political process, where political, administrative and professional logics clash and are balanced or rebalanced. Third, organizational structures are not only about efficiency but also tend to favour some processes, ideas, clients, users and actors over others (Meier, 2010).

We would tend to subscribe to the argument that reform movements are characterized by combining, complexity, layering and hybridization, rather than by dominance, substitution and pendulum swings (Christensen et al., 2007b; Streck and Thelen, 2005). Public administration faces increasingly complex environmental and internal conditions, reflected in multi-functional organizational forms, and the administrative reforms in the public sector can be understood as compound reforms that combine different organizational principles based on multiple factors working together in a complex mix (Egeberg and Trondal, 2009). Compound administrative reforms are multi-dimensional and represent ‘mixed’ orders and combinations of competing, inconsistent and contradictory organizational principles and structures that coexist and balance interests, values and claims to power (Olsen, 2007, 2010). Our argument is that this multi-functional order is a systemic feature of public sector organizations rather than a disease that should be cured.

Multi-dimensional orders are considered to be more resilient to external shocks and therefore preferable to uni-dimensional orders (March and Olsen, 1989). Compound reforms thus depart from ‘either/or’ theorizing by assuming that executive governance rests on the mobilization of multiple and complementary sets of institutions, actors, interests, decision-making arenas, values, norms, and cleavages, reflected in what we call a transformative approach to reforms (Christensen and Lægreid, 2001). In a pluralistic society, where there are many criteria for success and different causal understandings, we have to go beyond the idea of a single organizational principle to understand how public organizations work and are reformed and look at them as composite organizations (Olsen, 2007). Our argument is that we face a dialectical development in which the old public administration has been combined with NPM reform elements as well as post-NPM features, producing a complex and hybrid organizational model, as shown in this study.

Compared to welfare state reforms in other countries there are both similarities and differences. As in Denmark and the UK, the one-stop shop arrangement at local level is also introduced in Norway, but with somewhat different content (Askim et al., 2011). In contrast to other countries such as Germany and the UK, the Norwegian reform is more a pure administrative or structural reform not involving changes in substantive policy and service delivery content (Christensen et al., 2009). In contrast to Denmark, but in line with Germany, the welfare administration is not moved to the local government but remains a multi-level government responsibility involving both central and local government. The Norwegian case also supports the findings from the UK regarding the difficulty of effecting change in public services when it is based on political attention that is unlikely to be sustained (Neuman, 2011).

What may be learnt from this in general theoretical terms concerning coordination and specialization in a multi-level system? We would argue that a main lesson is that different specialization principles at three organizational levels that are so mutually dependent upon each other as central, regional and local government within the Norwegian welfare state system present major challenges to coordination. The coordination mechanisms that develop are largely directed towards re-establishing the managerial capabilities that are lost when the specialization principles diverge. One lesson is that to balance integration and diversity/autonomy in a multi-functional political-administrative system, combining different organizational forms will probably be more successful than adopting only one principle of specialization or one single coordination mechanism (Egeberg, 2005; Fimreite and Lægreid, 2005). Such organizational forms combine sector specialization and geographic specialization and are able to handle the fact that decision-makers and actors often have multiple identities and have to attend to both primary and secondary loyalties. The challenge is to weigh carefully the dynamic relationship between the different forms of specialization and coordination. How to analyse the different combinations, however, is a matter that has yet to be settled both theoretically and empirically. In practice different forms of specialization and coordination occur simultaneously, in parallel or in mutual interaction and thus produce big challenges both for researchers and for practitioners.