This research analyses steering practices of national executives where important tasks are delegated to central government agencies. Delegation to agencies became a popular solution to overload and legitimacy problems in the New Public Management-era, with manifest representation in public management models like Management by Objectives and Results (MBOR) and so called agencification. Such solutions are associated with the need to control the delegated mandates top-down, if only ex post and with a focus on performance. How ex post- control is applied to control semi-autonomous agencies has been observed to vary in cyclical patterns of escalation and de-escalation (cf. Kristiansen 2015; Kristiansen 2017; Kristiansen, Dahler-Larsen and Moll Ghin 2017; Wockelberg and Ahlbäck Öberg 2021). Regardless of the form and intensity of control, the possibility of establishing a link between the democratic leadership (the elected government) and the operational part of the executive (the agencies) is crucial. While the legitimacy of contemporary government is dependent upon extensive delegation to others, it is also dependent upon the possibility to keep track of performances and enable accountability. At stake is nothing less than the legitimacy of the government, or democracy as such. How governments manage delegation to and control of agencies are thus important issues for political science and public administration scholars alike, and the cyclical nature of government steering mentioned above is spurring renewed interest in these issues.

In terms of thinking about delegation, contemporary public management models are inspired by economic theory, more precisely Principal-Agent theory (PA). Post-NPM, public management ideas proposing that delegation could be defined in equal rather than hierarchical terms and that agencies could operate on trust have gained renewed attention. This has prompted a development in ministry-agency research where analytical concepts from Principal Steward- theory (PS) is used alongside the more established Principal Agent-perspectives on delegation and control. This study contributes to the novel line of research where Principal-Agent- and Principal-Steward-theories respectively are probed for their capacity to capture fundamental traits in ministry-central government agency relations (cf. Bjurstrøm 2020; Schillemans 2013; Schillemans and Bjurstrøm 2020; Van Thiel and Smullen 2021).

PA theory assumes that actions are explained by a logic of consequence and actors’ efforts to achieve rationally defined goals. Since all actors are considered equally rational, political principals must find ways to balance necessary delegation to agents with control. PS- perspectives offer a different view of situations where power is delegated, stating that agents can focus on collective goals and thus principals, under some circumstances, enjoy agents that are working with them towards shared goals. If subordinates act as stewards, the principal can build a relationship based on trust rather than control. The two perspectives are used here as analytical tools with fundamentally different core assumptions of the costs and benefits associated with delegation, and different ideas about what to control and when. In actual public management, governments are however expected to apply and mix elements from both perspectives. They can also switch from a control-oriented strategy towards agencies to one that is trust-based, perhaps after learning to trust the delegate or simply because they find the costs of ex post control too high.

This research contributes to our understanding of multi-dimensional ministry-agency relations by analyzing developments in the Swedish national executive 2000–2020. The research questions to be answered here are, first, what this mix of instruments contains and whether it is predominantly best described as being PA- or PS-oriented. Since the application of public management models such as MBOR is known to change over time (Kristiansen 2015; 2017; for the Swedish case see also Wockelberg and Ahlbäck Öberg 2021), a longitudinal analysis is potentially fruitful. Thus, and second, we ask what, if any, major regime level shifts can be observed during the period 2000–2020?

While some previous studies have aimed at systematically disaggregating data to capture potential variation across policy areas or agency specific traits (cf Ahlbäck Öberg and Wockelberg 2020), this research aims to describe regime-level ideas and practices over time, a task that implies the identification of general and fundamental traits of public management policies. The management regime in focus here consists of three main dimensions: its constitutional preconditions and application of formal and informal steering tools respectively. These perspectives are inter-related, as the constitutional doctrine historically frames government practices both when it comes to formal and informal steering (Ahlbäck Öberg and Wockelberg 2015; Wockelberg 2003).

Combining constitutional and management perspectives in the analysis is crucial, as these respective dimensions points our expectations in two directions. On the first hand, constitutional requirements relate governments’ relationships with agencies to the meta-goal of democratic accountability. In the parliamentary chain of delegation, the principal also has a principal and governments thus apply their management instruments in the shadow of the constitution. Constitutions, including the Swedish one, are more about control and accountability than about trust – governments’ referring to their trust in agencies are ultimately not likely to appease a critical Parliament. With delegation thus comes a set of pressing issues, such as determining who is accountable for delegated mandates; how these mandates are to be controlled, and whether the power of appointing agents is linked to the accountability of the appointing principal (cf. Barberis 1998; Denton 2006; Denton and Flinders 2006; Flinders 2011; Molander, Nilsson and Schick 2002). On the other hand, and as is discussed above, contemporary public management ideals such as trust-based management has been rhetorically embraced by policy- makers, supporting the hypothesis that Swedish governments have increasingly cultivated stewardship among central government agencies.

The Swedish case was selected for the following reasons. First, it is a typical case of an “at arm’s length relationship” between ministries and semi-autonomous central government agencies and as such provides findings of general interest. The need to delegate and the problems involved are issues that permeate the constitutional debate (Ahlbäck Öberg and Wockelberg 2015; Wockelberg 2003), are at the core of public management reforms (Ahlbäck Öberg & Wockelberg 2016; Wockelberg & Ahlbäck Öberg 2016; 2021 and 2022) and, as will be shown here, affect the everyday operations of the national executive. This said, the Swedish dual executive also has some features that makes it less typical, such as the small Government Offices (Regeringskansliet) and the many (about 360) resourceful central government agencies and a constitution that does not allow ministerial rule (more on this below). Swedish governments and agencies have dealt with delegation and control issues for a very long time (centuries). These institutions have long historical roots and managerial reforms in any direction must thus relate to a robust institutional frame. This feature makes it relevant to discuss Sweden as a critical case when it comes to the possibilities of introducing managerial reforms. Second, this case has undergone, at least rhetorically, changes toward a more trust-oriented application of ministerial steering of agencies (cf. Ehn & Sundström 2020). The idea of a trust-based model was introduced in 2014, in a low-key manner typical for Swedish public management reforms (Government Budget Bill 2014/15:1; see also Ministry of Finance 2016). Analyzing the Swedish dual executive during the period 2000–2020 using concepts from PA- and PS-theory respectively adds a new case to this sub-field of public management research. Third, the Swedish case is of general interest as a typical context also in a broader sense. According to Jacobsson, Pierre and Sundström (2015), contemporary governments are challenged by globalization, Europeanization, medialization and managerialization, all factors present in the Swedish case (Jacobsson, Pierre and Sundström 2015, p. 24). Fourth, while just mentioned work by Jacobsson, Pierre and Sundström (2015) is an important point of reference, it builds its theoretical claims on data that is quite different from the data used in this research. Moreover, said publication rests primarily on studies that predate the above-mentioned trust-oriented ideas and other changes of potential importance for ministry-agency relations, such as an increased use of temporary assignments for agencies. In addition, Swedish governments have had the ambition to shift short term formal steering from annual appropriation directives to more stable and long- term agency instructions. Together these factors motivate us to revisit the case.

This research offers new insights to a field of research predominantly focused on the application of formal steering tools, also regarding the Swedish case (cf. Askim, Bjurstrøm and Kjærvik 2019; Dahlström and Holmgren 2017; 2023; Holmgren 2018). This comprehensive analysis builds upon a mixed-methods design and contributes with a unique and updated analysis of the case. It offers new and systematic analyses of the constitutional doctrine as it has developed over the period 2000–2020; a re-interpretation of the most comprehensive prior studies of formal steering tools (Ahlbäck Öberg and Wockelberg 2020; Wockelberg and Ahlbäck Öberg 2021) and novel interview data that provide a unique insight into informal steering across parts of the dual executive. The focus on informal relationships in government is a contribution in itself since studies in this field are rare (but see Rommel and Christiaens 2009) and somewhat dated. Prior research describes informal contacts between political and administrative actors in Sweden as common and as including interaction between ministry level and agency level actors. Contacts have been used to reach a common understanding of political goals, in the drafting of appropriation directives and in agency assignments. According to prior research, informal contacts are mainly (but not exclusively) described in terms of exchanges on rather equal terms, rather than as control instruments (de Fine Licht and Pierre 2016; Hwang 2000; Jacobsson, Pierre and Sundström 2015). Relying on up-to-date interview data, this study offers insights into the role informal steering plays in national executives and the problems associated with informality. Finally, as recent applications of PA and PS theory feature different views on the role of informal contacts (Schillemans and Bjurstrøm 2020; Van Thiel and Smullen 2021) this study proposes a theoretical clarification on the role of informal contacts in PA- and PS-regimes respectively.

The following sections of this article defines the central concepts and problems to be studied, and presents the methodology applied. The empirical analysis that follows is organized in three parts, dedicated to constitutional prerequisites, application of formal instruments, and use of informal contacts respectively. The concluding section discusses the main findings.

It is hard to think about political power in operational terms without including elements of delegation (Bendor, Glazer and Hammond 2001). Delegation can affect the costs of decision-making by adding higher quality or credibility to analytical and operative performances, or by lowering risks. It also, as was noted above, raises questions about principal control over decisions. Assumed benefits and costs following from delegation are core features of public management policies such as Management by Objectives and Results (MBOR), which prescribe a dual structure that separates “politics” from “administration”. The risk of losing control is acknowledged and consequently an additional main feature is that delegation must be followed- up with reporting on results and ex post evaluation (Moynihan 2006, with reference to Schick 1999).

The concept of delegated governance places extensive operative mandates in organizational forms outside ministries (cf. Denton 2006; Denton and Flinders 2006). This at an arm’s length- strategy requires someone to delegate to, such as central government agencies, i.e., permanent and often publicly financed organizations, structurally separated from ministries (Pollitt et al. 2004). The New Public Management-era spurred agencification of national executives and brought managerial autonomy and ex post results control into the design and steering of agencies (Ahlbäck Öberg and Wockelberg 2020; Askim, Bjurstrøm and Kjærvik 2019; Hood 1991; Osborne and Gaebler 1992; Smullen 2010).

The fact that elected leaders are responsible for central government agencies’ activities (to some extent) delimits agency autonomy (Christensen and Lægreid 2006). Jacobsson, Pierre and Sundström (2015) describe the balancing act that is inherent in delegated governance as follows:

Even if a sector is organized with great care with the aim of keeping issues at a distance, events occur and unpredicted things happen that force politicians to become more active. […] [I]t is not possible for politicians to remain at a distance and claim that they cannot interfere in ongoing events. They must act, in one way or another. (Jacobsson, Pierre and Sundström 2015, p. 78)

This means that also in a context that relies heavily on delegation to agencies, politicians are still expected to get involved.

The management instruments available to governments can be formal and/or informal. Formal instruments include e.g., agency ordinances, assignments, budget allocation decisions and ex post scrutiny. This type of steering is easy to verify for the parties involved and for outsiders. Both the instruments and the mandates needed to use them are law-based. In contrast, informal instruments are applied outside formal processes, making it harder to establish who the decision- maker is or whether a decision has been made. Implementing a system or code for informal interaction can make it more predictable for the organization, but it will not make it formal in the way this concept is understood here. Only lawmakers can formalize steering tools and mandates. To be formal, a steering signal must stem from a legitimate mandate and be decided on in a formal process. To otherwise “manage” informal contacts will not make them formal.

Government choices regarding management instruments is best understood from within a multi-dimensional perspective on steering and autonomy. In the terms of Verhoest et al. (2004), agencies’ operational autonomy to make managerial decisions and influence policy making is influenced by constraints in the form of structural, financial, legal and interventional measures applied by the government. How agencies are affected by informal steering is not a well- researched issue, but if informal contact is intense and frequent, it will require agency attention and resources (Rommel and Christiaens 2009), perhaps in ways that undermine agency operations. At the same time, informal steering is opaque and thus easier for agencies to ignore than formal steering signals.

Finally, it is likely that government use of formal and informal steering are interdependent, but exactly how is an empirical question. In theory, a government could choose between formal and informal steering. Moreover, if formal steering is vague, a need for clarification may spur informal messaging or signaling. In such a scenario, informal steering complements but does not replace formal steering. In this latter scenario, the more formal steering, the more informal steering.

Before taking a closer look at the PA- and PS-perspectives on management tools, it is of relevance to clarify their common ground and thus the relevance of using them to capture ministry-agency relationships. Both perspectives make assumptions about a relationship between two parties that for political scientists commonly are collective actors, not individuals. This is also a common feature of public administration research and a way of thinking that informs studies of ministry-agency relations. In addition, both perspectives theorize about how the principal can benefit from delegation and make assumptions about the gains and risks involved. Finally, both perspectives are about trust and distrust, but make different assumptions in this regard. They also, consequently, advocate different ideas about suitable follow-up mechanisms in relation to delegation. PA assumes that delegation carries risks but also that top- down control can reduce these risks. It is not, however, a core assumption in PA that control leads to trust, or, in our understanding, that the possibilities of learning to trust is a factor in the principal’s cost-benefit analysis. By comparison, PS is different in both respects: trust is assumed to be possible, and top-down control mechanisms are defined as secondary to collaborative interaction, that is, the preference is for a relationship between equal parties where the delegate is allowed to play an independent role also in ex post reporting and evaluation.

PA and PS perspectives on ministry-agency relations identify both formal and informal instruments as important. While the perspectives offered in previous research on these types of regimes are to a high degree similar, there are also some interesting differences. Prior definitions of typical PA and PS instruments echo Verhoest et al.’s (2004) multidimensional conceptualization in the ways that important actions are identified in terms of ex ante or ex post attempts to control agency autonomy. According to Schillemans and Bjurstrøm (2020)

[w]hen public tasks are delegated to agencies, government departments perform (at least) six different tasks, both ex ante and ex post […]. Government departments need to (1) select an agency (who is going to do the job?); they need to (2) transmit their preferences and (3) stipulate additional due procedures in their contracts. Ex post measures include (4) monitoring requirements, (5) incentives (how will the agency be rewarded?), and also (6) relationship management. (Schillemans & Bjurstrøm 2020, p. 655)

Governments in a PA-oriented management regime apply instruments designed to minimize the risk of a deviate or slack agent. The relationship is hierarchical. Ex ante steering instructs the agency top-down about political goals. Expected results are expressed in detailed contracts and management involves detailed monitoring. In terms of forms of interaction, PA-oriented management is defined as predominantly formal, informal contacts between principal and agents are rare or minimized. In a PS-regime, governments and agencies interact in a way that allows both parties to participate in policy development. Ex ante and ex post steering is relaxed and allows the agency to decide on means, methods and performance assessment (performance reporting). According to Schillemans and Bjurstrøm, the interaction between governments and agencies in a PS-regime is equal and ongoing. While the authors do not make the formal/informal dimension explicit, the type of interaction described is arguably one that involves informality.

Van Thiel and Smullen (2021) also define PA-regimes in terms of hierarchical relationships. Typically, the instruments used are those that limit the autonomy of agency management and agency influence over financial matters and policy. Moreover, PA is “an intense regime of reporting, for example using multiple performance indicators, frequent contacts and detailed legislation” (Van Thiel and Smullen 2021, p. 762). In contrast, PS-instruments aim to grant agencies a high degree of autonomy, for example by appointing a board to lead the agency and by using framework legislation instead of detailed formal steering. Ex post control is less frequent than in the PA-regime and focused on results rather than input-factors. Importantly, agencies are able to influence how performance is reported. Regarding contacts between ministries and agencies, Van Thiel and Smullen (2021) define frequent contacts as a feature typical for PA-oriented management and as motivated by a lack of trust between the parties. According to the same logic, PS-regimes feature moderate contacts “in [a] trusting atmosphere” (p. 763, Table 1). The theoretical frame for the up-coming analysis is a synthesis of prior theorizing on management instruments that shape ministry-agency relationships and how. Its definition of management instruments builds upon Verhoest et al’s (2004; for an empirical application on the Swedish case see Ahlbäck Öberg and Wockelberg 2020) conceptualization of agency autonomy, which is here supplemented with informal contacts. It thus theorizes about instruments in a way that is similar, but not identical, to Van Thiel and Smullen’s (2021) framework. The ambition of the conceptualizations in Table 1 (below) is to chisel out the basic logic in the proposals made in previous research. Table 1

Ministry-agency relationships in terms of management instruments and regimes

The choice of agency management model is in essence the choice between delegating to one or many, or to a sole Director General (DG) or an agency board. According to Verhoest et al.s’ (2004) theory, to choose a sole DG allows the government more control over the agency than choosing a management board does, thus a sole DG here represents the PA-model. In the Swedish case, this line of thought is also official policy (Vägledning för statliga myndighetsstyrelser, Regeringskansliet 2016). Even if it is theoretically possible for a Swedish government to change the management model to increase or enhance control over a specific agency, perhaps due to underperformance or scandal, such an invasive measure is not expected. Rather, the government would replace individual managers, e.g., the DG.

Ex post monitoring in performance management models is a core feature of contemporary ministry-agency relationships. Governments differ in terms of how much performance information they request and what type of indicators they prefer (Kristiansen, Dahler-Larsen and Moll Ghin 2017). It is here assumed that all else equal, the more performance information that is demanded from agencies, the more controlling this type of steering is. Reporting is one task among others and consumes agency resources. At the same time, reporting on performance can be more or less controlled top-down – there is a qualitative element to reporting. If an agency is invited to present its activities in ways of its own choice, it is less controlled than an agency that is being told what to report and how. If the government seeks to impose hard top-down control, it asks for many precise indicators. In a PA-regime, performance indicators are used to control operations from the top down while a PS-regime is focused on agencies’ own understanding of outcomes.

Previous theorizing about steering in ministry-agency relations also includes government decisions about two types of delegation. One is the possibility of delegating financial decisions to agencies, for example in terms of possibilities to collect fees and thus extract revenue outside the state budget. An agency that is allowed this type of financing is considered, all else equal, to be less controlled by the government than an agency that is not allowed to extract its own revenues. This variation here translates to the PS and PA-perspectives respectively. Another delegation decision concerns whether agencies are allowed to issue regulations. This choice is assumed to be between imposing regulations on to the agency – a top-down control mechanism – and allowing the agency to issue regulations. It is assumed that even if it is demanding to issue regulations, it is more demanding to be regulated top-down.

For each agency, there are theoretically a multitude of possible combinations of PA- and PS-oriented use of the formal steering tools discussed above. However, at the aggregated regime-level, which is in focus here, it is hard to define a priori e.g., how many performance indicators that makes a regime PA-or PS-oriented. This means that when interpreting measures and results, only clear findings should be considered as valid indicators of respective type. According to the same principle, while the longitudinal perspective introduces a relative element to the analysis, only major shifts from one type of steering to another over time will be defined as the regime changing from one perspective to the other.

Turning to informal contacts, these equal informal interaction, and not formal exchanges of steering signals, reports etc. Previous research offers different views on the type and frequency of contacts that are expected in respective management regime. Schillemans and Bjurstrøm (2020) associate PA-regimes with few or no contacts and PS-regimes with “ongoing” contacts. Van Thiel and Smullen (2021) offer opposite expectations, associating frequent contact with the control-oriented PA-regime. In this research, which develops the theory further, we propose a definition in which the frequency is considered less important than the content or purpose of informal contacts in distinguishing PA and PS characteristics. Contacts that represent top-down attempts to control agencies are seen as a PA feature – they are made in order to underline or complement other types of steering. In contrast, contacts that are made to exchange information and/or establish relationships on equal terms are here seen as a PS instrument. This latter type of contact can be initiated by either party.

Finally, given that the relationship between ministries and agencies is here defined as multi-dimensional, the issue of how to interpret the observations comes to the surface. The regime is probably not best described by the PA- or PS-model exclusively, but rather by a mix of both perspectives. Variation is expected, both over time and between different types of steering at the same point in time.

In general terms, and confirming prior research, informal contacts are described as necessary to make government steering efficient (IP1), and as essential (IP6; IP9). At the same time, the risk of doing something wrong is a factor holding civil servants back when it comes to making informal contacts across the dual executive.

Agency-specific characteristics such as political salience (the importance of specific policy issues, the size or weight of the agency) affect the use of informal contacts (IP1; IP2; IP3; IP4; IP5 and IP6). Not unexpectedly, this is a positive relationship, with high salience leading to a high number of contacts. In addition, the interest of individual ministers in specific issues is described as affecting the interest in informal contacts (IP2), as are individual level traits (personality) (IP3; IP8), indicating that the informal relationship between ministries and agencies may change over time. Yet other agency specific features inhibit informality. If an agency’s operation is mainly controlled by formal laws and regulations, informal contacts are relatively uncommon (IP3). Furthermore, and possibly related to the previous observation, one respondent points out that when an agency’s main task is to process individual cases of the type protected by the Instrument of Government (IG ch. 12, art. 2), informal contacts will be more carefully designed (IP5).

A main use of informal contacts is to communicate or exchange information, indicating a PS-oriented regime. This conclusion is supported by a stable narrative stating that there is no such thing as informal steering (IP1 and IP5), which rules out the possibility of using informal instruments to control agencies. Another PS-oriented feature is that contacts are not exclusively top-down initiatives. Central government agencies use these channels in a bottom-up manner to communicate ideas for assignments, deliver input to upcoming appropriation directives, discuss priorities, or confirm their interpretation of steering signals (IP6; IP7 and IP9).

A closer look at the data reveals that ministries do use informal contacts in a controlling manner, to supplement and clarify steering signals and to ensure that agencies are well informed about upcoming steering, including the political intention behind specific steering signals. The drafting of appropriation directives and decisions to give an agency a special assignment is mentioned as regular processes in which informal contacts are used to clarify, and follow up on, steering (IP1; IP2; IP3; IP5; IP6; IP7 and IP8b). According to the theoretical framework of this study, clarifying and informing about steering signals are control-related activities of the kind that is associated with PA-regimes. This said, the narrative described here seems to foster another perspective in which clarifying steering is not seen as steering per se. However, some interviewees express that the difference between discussing things with agencies on the one hand, and overstepping the constitutional limits on ministerial rule on the other is rather small (IP5; IP6 and IP7).

The previous section established that informal contacts are used both in a top-down manner to control agencies as well as in a cooperative, equal way. This subsection focuses on problems that interviewees associate with informality. A main problem identified by the interviewees is that even though informal steering is unconstitutional, steering is sometimes the intended or unintended effect of informal contacts.

In constitutional terms, steering must be collective and formal and cannot target agencies’ decision-making in individual cases. One problem with informality, as one interviewee points out, is that it neglects the principle of collective decision-making in government (IP4). Political officials sometimes pursue their own or their ministry’s pet projects or policy goals outside the formal processes, without the support of the (collective) government. In the examples given, this is a habit of some high-ranking political officials, who are not well educated about the constitutional principles, or who simply disrespect them (IP2 and IP4). Staying clear of the “forbidden area” that protects agencies from outside interference in individual cases is not the main problem; rather, the challenge is to meet the demand that the government operates based on the basis of collective decision-making. In the words of one of the respondents: “That is what makes them trip […]. Civil servants working with a silo mentality […] ‘I work for the Ministry of …,’ and the agencies ‘I listen to the Minister of …’” (IP4, authors’ translation). The risk of informal contacts being used to advance individual agendas that are not anchored in the government as a whole is not limited to the level of political officials. It is also identified as a problem at the level of the merit-based public service (IP3; IP4; IP6; IP7 and IP9). The fact that the careers of the merit-based staff often involve moving between agency and ministry level employment creates personal networks that are used for informal interaction (IP4 and IP6).

Another risk of using informal contacts is that agencies may be overly interested in any initiative or signal coming from the ministry level and end up being too receptive or misunderstanding cooperative communication as being a steering signal (IP7). If agencies perceive, or choose to perceive, the regime as control-oriented, ministries will fail to establish a PS-oriented informal relationship.

Finally, a general problem with informal contacts is their potential to disrupt or blur the information flow (IP2). Knowing where informal messages come from and what they mean is thus important. This is also true when the government and the Government Offices receive information from agencies: One of the interviewees describes that it is not always clear whether ideas communicated informally represent the formal position of the agency or that of individual officials at the agency level (IP6). Moreover, informal contacts leave those involved with the challenge of remembering what was discussed or decided. Documentation would formalize informal contacts but at the same time make them less relevant (IP4; IP5; IP7 and IP8). Adding to this problem is the fact that the Government Offices is an organization with little continuity, in terms of staff and political leadership. The use of informal contacts requires knowledge of the Swedish constitution, the skill to identify illegitimate steering and the courage to ignore it (IP2; IP3, IP4 and IP8). The practice of informal contacts is described as something that civil servants need to learn along the way, sometimes guided by senior colleagues or supported by policies that specify whom to contact about what and when. Such strategies are the topic of the next subsection.

Informal contacts are seen as crucial to the effective functioning of the national executive, but in order to be constitutionally legitimate they must not constitute top-down control (or steering). While PA-oriented use of informal contacts is thus to be avoided, suggestions on how best to manage informal contacts are mainly of a kind inspired by PA features such as clear mandates and hierarchical relationships.

Ministries deal with informal contacts in different ways, including how they train staff and whether they issue actual policies to guide this part of the work. Knowledge of constitutional and practical aspects seems to be developed through learning by doing at the individual level (IP2; IP7 and IP9). Efforts to increase knowledge about the legitimate use of informal contacts within the Government Office have been sporadic and local.¹

Some ministries impose order in the form of explicit policies on how informal contacts should be managed. Some of these agreements describe contacts in terms of specific individuals, while others aim to increase management skills and develop general ideas about how to practice informal contacts. Among the measures taken we also find the appointment of a gate keeper to manage all contacts between a ministry and a particular agency, with the instructions to alert higher level managers if a tricky situation arises (IP1; IP2; IP3; IP6; IP7 and IP9).

In the debate on informal contacts in the 1980s, the level at which informal contacts take place was identified as a key issue for the legitimacy of this type of interaction. As noted above, informal contacts should be located to the level of civil servants, and at an as low level as possible (Ahlbäck Öberg and Wockelberg 2015; Wockelberg 2003). While the issue of who it is that are in contact is still considered important, there is no evidence in our data that the strategy to keep informal contacts between low echelons of civil servants is recognized or considered important today. When asked who is involved or in contact, interviewees collectively describe contacts at most levels of the respective ministry and agency hierarchies. High-level contacts are not common and typically involve state secretaries or sometimes ministers on the one hand and agency heads on the other. Contacts between merit-based civil servants in respective organizations are very frequent in some parts of the executive. In large organizations there are many potential areas of interaction, and more people may be involved (IP3; IP5; IP7; IP8 and IP9). This is described as potentially problematic because the government runs the risk of losing control over the signals sent and received in these exchanges (IP5). This type of situation can also be problematic from the agency’s perspective, leading agency heads to protest that the agreed channels of contact are not honored by ministry level civil servants who break down the hierarchy, making phone calls directly to the agency head (IP2).

That informality as such challenges memory has been described above. Informality also undermines transparency, which in turn challenges democratic claims to accountability. This problem was acknowledged by the Committee on the Constitution, which concluded that informal contacts should be documented. While the interview data show that the Committee’s demands for documentation are known and discussed in the Government Offices, there is little evidence of any systematic efforts dedicated to meet these demands. Notes are taken, but not in a formal way. In the words of one interviewee: “There are napkins to write on.” (IP8) The strategy seems to be to avoid documentation that can be traced by outsiders, relying instead on personal notes and phone calls to validate one’s recollection of informal exchanges. The reasons given for this strategy are that documenting informal contacts would formalize them, making them less effective and less relevant as a tool; that documentation would be cryptic at best, and that documentation is very time-consuming (IP2; IP6 and IP9).