Changing public service delivery: learning in co-creation

Co-creation – where citizens and public organizations work together to deal with societal issues – is increasingly considered as a fertile solution for various public service delivery problems. During co-creation, citizens are not mere consumers, but are actively engaged in building resilient societies. In this study, we analyze if and how state and governance traditions influence learning and policy change within a context of co-creation. We combine insights from the co-creation and learning literature. The empirical strategy is a comparative case study of co-creation examples within the welfare domain in childcare (Estonia), education (Germany) and community work (the Netherlands). We show that state and governance traditions may form an explanation for whether co-creation, learning and policy change occurs. Our paper suggests that this seems to be related to whether there is a tradition of working together with citizens and a focus on rule following or not.


Introduction
Co-creation can be described as the involvement of citizens in the initiation and/or the design of public services to develop beneficial outcomes (Voorberg, Bekkers, & Tummers, 2015, p. 1347. In co-creation initiatives, citizens are regarded as relevant partners, who have specific resources and competences which are valuable for (re)designing public service delivery (e.g. Alford, 2009;Bason, 2010;Bovaird, 2007;Parrado, Van Ryzin, Bovaird, & Löffler, 2013). Public officials and politicians are increasingly taking up co-creation with citizens as a way to address many of the public sector's problems. This seems to mark a paradigm shift, in which the dominant consideration of citizens as passive consumers of public services has changed toward a consideration of citizens as co-creators. Citizens are given the opportunity to participate in the joint framing of what relevant services and service outcomes are and how they should be organized. The rise of co-creation can also be considered a learning process, in which actors learn how to use each other's competences to develop new ways to confront public sector challenges (e.g. aging, unemployment, decline KEYWORDS co-creation; learning; policy change; state and governance traditions of legitimacy of public institutions). Despite its growing relevance, the role of learning in co-creation processes and the influence it might have on policy change has not received much academic attention yet (for an exception, see: Cassia & Magno, 2011). There is much that we still do not know about 'who learns' , 'what is being learned' and 'why it is being learned' in co-creation (Bennett & Howlett, 1992). These are important issues, since the co-creation framework forces contemporary public organizations and public officials to consider alternative sources of knowledge, information and experiences, which are bypassing established venues (Brandsen & Honingh, 2015).
This study's aim is to understand whether co-creation and the learning process behind it has led to a change in the frames of the involved stakeholders (primarily citizens and public officials) and consequently, to policy change. We examine how the co-creation project's institutional and policy context may affect this process of frame adaptation and policy change. We expect that this is the case, since how and whether frame adaptation will occur depends on the institutional setting (Schön & Rein, 1995). So far, the literature on co-creation and co-production has left the influence of macro-level elements relatively unexplored, partly due to a lack of international comparative research (Brandsen & Honingh, 2015;Voorberg et al. 2015; see for an exception Pestoff, 2006). We address this gap in the literature by exploring two interrelated questions regarding co-creation: Does co-creation lead to frame adaptation and policy change? And how can this be explained by the state and governance traditions in which co-creation is embedded? The empirical analysis contributes to the literature by focusing on specific macro-level institutional elements, i.e. state and governance traditions, in three different countries characterized by different state and governance traditions (Estonia, Germany and the Netherlands).
The remainder of this paper is structured as follows: in Section 2, we develop a theoretical framework. We elaborate on the concept of co-creation and why this implies changes in public service delivery. We also discuss how learning can be understood as a process of frame reflection and frame adaptation. Additionally, we analyze how the institutional context affects policy change and frame adaptation. In Section 3, we explain our research strategy. This is followed by a discussion of the results in Section 4 and a brief conclusion as well as suggestions for further research.

Co-creation with citizens and the importance of learning
The definition of co-creation as the involvement of citizens in the initiation and/or the design process of public services in order to (co)create beneficial outcomes (Voorberg et al., 2015(Voorberg et al., , p. 1347) emphasizes the 'initiation and/or design process' and implies a more fundamental role for citizens in public services. This role implies more than asking citizens just to participate in the production of public services. This idea has been illustrated in the literature by Ugo Rossi (2004), who described how the historic city center of Naples (Italy) was restored due to a citizen initiative. The municipality recognized this initiative as a token of urgency given by the local community and decided to participate in the initiative. As such, it shows how co-creation changed conventional ideas about who is responsible for public service delivery and how decisions are made about the allocation of public resources. As Rossi noted, the success of this initiative was influenced by citizens' and public officials' willingness to learn from each other. The municipality of Naples showed this willingness by deciding to take the citizen initiative seriously and gave citizens ownership over the reconstruction process. Other examples where citizens have taken up the role of initiators in public service delivery involve participation in childcare services (Pestoff, 2006) and participation into budgeting procedures of municipalities (e.g. Ackerman, 2004;Maiello, Viegas, Frey, & Ribeiro, 2013).
The importance of learning as an inherent part of co-creation has been recognized for almost two decades in the marketing literature. Prahalad and Ramaswamy (2000), for example, have argued that involving consumers in the production process implies that companies should strive for an active dialog with them and mobilize consumer communities to create tailor-made products. Also in the public sector, the importance of learning in co-creation is gaining more attention. Public officials are urged to consider citizen 'lay knowledge' as a valuable source of insight into how to (re-)design public services (e.g. Cornwell & Campbell, 2012;Guston, 1999;Hardey, 1999;Peters, Stanley, Rose, & Salmon, 1998). Nonetheless, a broader notion of learning has not received much academic attention yet and learning in (public) co-creation is a relative unexplored concept. This paper aims to open the 'black box' of co-creation, by focusing on whether and how co-creation facilitates this learning process. In the next section we analyze what learning in co-creation may imply.

Learning in co-creation: a process of frame adaptation
During co-creation, multiple actors participate in mutually dependent relationships (e.g. Bovaird, 2007;Lelieveldt, Dekker, Voelker, & Torenvlied, 2009). We can therefore argue that learning in co-creation is a social process, in which shared convictions about problems and solutions are the result of a dialog between actors. In accordance with Dunlop and Radaelli (2013), we define learning as the updating of beliefs. Approaching learning as a social process implies that updating of beliefs is the result of sense making about the meaning and interpretation of facts and events (Weick, 1995). To determine whether learning has taken place, we analyze whether stakeholders' frames about what the problem is and how it should be solved are indeed updated to new, shared convictions about problems and solutions. We define frames as images, influencing the convictions and actions around a policy matter by offering a problem definition, causal explanation, target group and a solution (Benford & Snow, 2000;Fischer, 2003;Schön & Rein, 1995). According to Schön and Rein (1995), learning involves a process of critical reflection and of changing actors' individual frames. Benford and Snow (2000), p. 616, 617 distinguish three kinds of frames. These are (1) diagnostic framing, referring to the causes of a specific policy problem, the effects and the target group the problem involves, (2) prognostic framing, referring to the identification of possible and relevant solutions and approaches and (3) motivational framing, which provides a rationale for engaging in collective action. Co-creation creates a new division of roles in public service provision, where citizens become partners to public officials. This implies that co-creation can be considered a prognostic frame: co-creation as an option to deal with a specific policy challenge involves a change in the traditional division of labor between citizens and public official. The question is whether this frame is being adopted and shared by the various stakeholders. If this is the case, we can predict that a policy change will probably occur.
Downloaded by [207.241.229.243] at 02:27 28 October 2017 Dunlop and Radaelli (2013) stressed that the combination of different individuals in learning processes and the control of these individuals ('learners') over the objectives of learning leads to different types of knowledge production. In other words, different sets of actors lead to different kinds of learning results and therefore, also to different forms of policy change. To Hall (1993) policy change involves a change in 'the overarching goals that guide policy in a particular field, the techniques or policy instruments used to attain those goals, and the precise settings of these instruments' (p. 278). As such he considers policy change as a reliable indicator for learning. He distinguishes three levels of policy change. First order change refers to an incremental modification of the existing objectives and instruments of policy. This modification involves an adaptation of current policy to current times, without changing overarching policy goals. Second order change refers to a major change of instruments of policy 'without radically altering the hierarchy of goals behind policy' (Hall, 1993 p. 282). As such, second order change involves a change in ideas of how policy objectives should be accomplished, but not which policy objectives should be accomplished. Finally, third order change involves a policy paradigm shift. A paradigm shifts occurs when there are radical changes in the hierarchy of policy objectives, the instruments used to achieve those changes and the discourses used to legitimize these changes. Third order change is more radical than first and second order change. In this study, we examine whether co-creation causes policy learning (in terms of prognostic frame adaptation) and whether this may lead to different levels of change. We treat learning and policy change as separate concepts, which enables us to see whether indeed policy change is an indicator for learning.

State and governance traditions
Since frame adaptation is always 'situated' in a certain institutional context, it can be the case that dominant state and governance traditions influence learning processes. Co-creation and co-production scholars have predominantly researched how and which institutional elements influence co-creation, such as having a risk-averse administrative culture (Maiello et al., 2013). Institutional elements may influence the prognostic frames of involved actors in terms of how suitable co-creation is as an approach to deal with specific policy challenges. As mentioned before, due to the lack of knowledge about how contextual elements on the macro-level influence co-creation, we focus our comparison on state and governance traditions. In line with Painter and Peters (2010), we treat state and governance traditions as a set of grown ideas and established practices that often act as structures that influence the policy style and substance of public administration in a country. The influence of state and governance traditions may explain why governments respond very differently to conceptually identical challenges (Pollitt & Bouckaert, 2011). Parting from the literature, the factors that define state and governance traditions can be categorized along two dimensions (Lijphart, 2012): sharing of authority and the culture of governance.
The first dimension refers to sharing of authority with non-governmental parties, ranging from a consultative to an authoritative style. A government that can be characterized as consultative or consensual on this dimension is one where multiple collaboration structures have grown between government and social partners, civil society and private actors. One that can be characterized as authoritative is one where decisions are made predominantly by governmental bodies. In this regard, Pollitt and Bouckaert (2011) speak about majoritarian countries, whereby the majority of the electorate, as represented, for instance, by parliament (50 + 1) can determine policy directions in an authoritative manner, based on this majority. This may lead to sweeping changes in policy and reforms. In contrast, in consensual countries, decision-making takes place in consultation with partners who do not necessarily belong to this majority.
The second dimension refers to the dominant normative convictions about how government should act. We label this dimension as the culture of governance. A state can be characterized as either 'Rechtsstaat' oriented or as 'Public Interest' oriented (Pierre, 1995). In Rechtsstaat-oriented states, stately efforts are focused on the preparation, promulgation and enforcement of laws. These states are characterized by a culture of governance aimed at maintaining legal correctness and legal control (e.g. Germany). The 'public interest' model (e.g. Anglo-Saxon countries, like the UK) accords a less dominant role to rules and regulations in society. Government acts rather as a referee, safeguarding equal distribution of resources and deciding which party (for instance competing interests groups) serves the public interest best. In such Public Interest cultures, pragmatism and flexibility is favored over the technical and legal expertise that is dominant in 'Rechtsstaat' countries (Pollitt & Bouckaert, 2011 p. 62).
We will analyze how these different state and governance traditions influence learning and policy change in co-creation. Figure 1 shows a heuristic overview of the main concepts. In the next section, we present the research methods to study this.

Research strategy
Given the limited empirical knowledge about the relations between state and governance traditions, co-creation, policy learning and policy change, a case study is an appropriate method for the study. Case studies allow us to analyze whether and how state and governance traditions influence learning and policy change in-depth. This study is as a co-variational international comparative case study. The study is co-variational, since we selected the cases based on the independent variable, i.e. state and governance traditions. As such, this study is aimed at exploring how a specific cause (X) may affect a certain outcome (Y) (Blatter & Haverland, 2012;Yin, 2009 We selected two cases with similar state and governance traditions (Germany and the Netherlands), and one case with a most different state and governance tradition (Seawright & Gerring, 2008) (Estonia). Estonia (with its communist background) is characterized by an authoritative state tradition. The culture of governance is formally a Rechtsstaat-based legal and governing system but just as in Finland (Pollitt & Bouckaert, 2011), the culture of governance is changing from a Rechtstaat to a more plural one (Sarapuu, 2012). In most social policy areas, the role of the state is described by familial responsibility rather than by a legal one (Tõnurist & De Tavernier, 2017). We argue, therefore, that its governance culture is closer to a 'public interest' culture, in comparison to the culture in Germany and the Netherlands. An overview of the state and governance traditions is shown in Table 1.
A second set of selection criteria was used to ensure that our cases were as similar as possible on other important elements. First, all selected cases involved co-creation, in which citizens took the initiative. Second, all selected cases involved co-creation within the welfare domain. Third, all selected cases represent co-creation projects that had been running for at least one year.
We acknowledge that since we examined only one case per country, the external validity of the findings is limited. However, our study aims at exploring how and whether there is a plausible relation between learning and policy change due to the institutional context. Hence, our aim is analytical generalization, focusing on an enhanced theoretical understanding of co-creation by exploring what makes sense in a reasoned way. To enhance the internal validity, we operationalized the concepts in the model into an interview protocol. This protocol was used as a template to conduct our interviews among the involved actors in all our co-creation cases across the three selected countries (see Table 2).
We interviewed 10 key actors in each case. Following Brandsen and Honingh (2015) we made a distinction between citizens (people who are voluntarily involved in co-creation) and public officials (people who are involved in co-creation on a professional basis and represent a public organization). All interviews were recorded and transcribed. Table 2 summarizes the concepts that were used as a coding scheme to analyze the data. The interviews were analyzed and concepts were coded and classified into categories related to plausible relations between co-creation, policy change and state and governance traditions. These were then compared to the concepts in Table 2.
To increase the internal validity further, we also analyzed relevant policy documents published by both the co-creation projects themselves and the involved municipalities. These documents are listed in Appendix 1. Third, we organized focus groups of twenty academic experts in the field of co-creation/co-production to analyze our findings (held on 11 September 2014).

Results
For each country, we include (a) a short introduction to the case, (b) whether and which kind of policy change occurred, (c) whether new prognostic frames can be detected with the key stakeholders, and (d) whether these observations can be explained by the surrounding state and governance traditions.

Case introduction
The Estonian example, Maarja Küla (Maarja Village) was initiated in 2001 by a group of parents of children with learning disabilities and school staff of Tartu Maarja School. The parents were concerned that after finishing Tartu Maarja School, the youths with learning disabilities had very few opportunities and little support outside of their families. Maarja Küla was established to provide an inclusive living environment and development track (with educational and working opportunities) for young adults with learning disabilities. The value of the service was acknowledged after some time by the state with support to Maarja Küla and similar services being developed across Estonia with support from the government. The German case, Dialog macht Schule (DmS), was founded as a non-profit organization in 2008. It offers a civic educational program targeted at public schools in German neighborhoods where the majority of school-aged children have a migration background and also live in socially disadvantaged communities. Their model consists of introducing open dialog sessions focused on civic education into the school curriculum. The sessions are moderated by university students or recent graduates, who also have a migration background, and who have been trained by DmS to moderate the dialog sessions. During these sessions, themes such as identity, culture, nationalism, racism and other political and civic ideas are discussed. The aim is to create a space where students can learn to formulate their own opinions on political issues, and through that process, develop their political awareness and civic participation skills. The Dutch case, Starters4Communities (S4C), was initiated in 2013 by a social entrepreneur whose goal was to combine the talents of young, highly educated students and citizen initiatives. She noticed a lot of citizen initiatives in the east of Amsterdam aimed at increasing the livability within the city. However, these initiatives often failed because they lacked staff with administrative skills and/or did not have sustainable business plans. She also noticed that a lot of graduated youngsters were having trouble finding access to the labor market because they had no or little relevant work experience. S4C brings the young students and the initiatives together. By combining the skills of these youngsters and the enthusiasm of the civil initiators, the civil initiatives are being uplifted and the students receive valuable work experience. The municipality of Amsterdam recognized the potential of S4C for their own policy objectives and is now actively connected to this initiative.

Policy change
In Estonia, the shift toward co-creating services for disabled children can be considered a third order change, involving a paradigm change regarding who deserves care, how children's disabilities and future development prospects are defined and how children with disabilities should be taken care of (Ministry of Social Affairs, 2010, 2011). Before the 1990s, disabled children usually placed in residential institutional care since birth. If families chose to diverge from this system, they would not receive any state support . In time, it became less common for people to think of disabled children as unintelligent or less deserving of equal treatment. As a result instead of putting children in institutions, children were placed in group-or family-based homes . As one of the parents involved with Maarja Küla put it: 'the overall special child care situation in Estonia at the end of 1990s was rather depressing and dark' . This created pressure for the families involved, who sought out alternatives for childcare provision. By supporting this movement to alternative youth care provision, the government changed its policy on youth care provision and started to actively educate public officials in these alternative services.
In Germany, the co-creation initiative demanded a second level change in how the goals of integration policy are achieved. The co-creation initiative of Dialogue macht Schule aims at strengthening the civic education of all school-aged students in schools where a large proportion of the students have a migration background (i.e. who have immigrated to Germany or whose parents or grandparents immigrated to Germany). In German cities, about 40% of students with a migration background are concentrated in underperforming schools, whereas only 6% of non-migrant students are in such schools (Morris-Lange et al., 2013). Consequently, migrant students are underperforming compared with native German students and have more trouble finding a job. Civic education has been proposed as a new strategy for achieving successful 'migrant integration' . The purpose not to assimilate migrant cultures into the host one but to foster individual commitments to the underlying social and citizen values of a democracy. Students learn 'how to express themselves eloquently' (German citizen). Such skills are expected to help these students succeed in the labor market and in society. However, the required policy change was only partly realized. Although enjoying the support of many school principals and teachers, as the Berlin Department for Education, the program has faced resistance among the teachers' union and some schools. The ultimate goal -introducing Dialogue macht Schule as a permanent part of the curriculum -has not been realized yet.
In the Netherlands, the shift toward co-creation can also be considered a second level change in public service provision. The Dutch King stated that the Netherlands has turned into a participation society, implying that every citizen is asked to contribute to the wellbeing of themselves and others within their direct environment (State of the Throne, 2013). The underlying idea behind this shift is that the current welfare state is unaffordable and needs to be reformed: public organizations should take a step back and allow citizen initiatives to come up with smart solutions. Starters4Communities offers one such smart solution. The basic idea is that well-educated young people help existing civil initiatives become more sustainable by making business plans and organize external communication. Instead of providing these initiatives, government policy changed to financially supporting these initiatives and connecting people, organizations and other resources. This changed role has become part of the official policy of the municipality of Amsterdam (Gemeente Amsterdam Stadsdeel Oost, 2014). As a result Starters4Communities represents a fundamental shift in how public services are delivered and who is responsible for what part of service provision.

Change of prognostic frames
The cases indicate a clear relation between the change in the prognostic frames of involved actors and whether policy change indeed occurred.
In Estonia, the prognostic frames of both citizens and public officials changed. Citizens became increasingly convinced that the conventional way of providing youth care services was not suitable to address the needs of disabled children. One of the initiators explained: By the time we started Maarja Küla, there was a common belief shared in the community of the disabled people and their family members that social services and especially teaching methods of children with disabilities (shunning them away from society) were unacceptable.
Parents became increasingly aware that these children have a need for personalized care and more substantive daytime activities (Praxis, 2011). Involved public officials (and also politicians) also changed their prognostic frames by deciding to (financially) support alternative ways of providing youth care services. As one public official mentioned: 'I think Maarja Küla was an important breaking point as an institution that was motivated to involve different stakeholders and to fill the gap that was missing in Estonia'.
In Germany, the involved public officials' prognostic frames were only partially adapted in favor of the co-creation initiative. In recent years, the discourse on the integration of migrants has intensified in Germany (SVR, 2012). Integration has been mainly understood as the extent to which incomers adopt the mainstream socioeconomic, legal and cultural norms of the host community . German public opinion has been concerned about whether migrants and their children adopt the host country's cultural norms, or maintain allegiance to those of their homeland. Diagnostic frames about the issue at hand have changed somewhat: there are teachers and headmasters who really see a need for this kind of civic and personal education -many of them simply cannot reach their students on any level in the classroom because they are so separated from them, so they are desperate for solutions and have really welcomed Dialogue Macht Schule into their classrooms.
(German public official). Schools are believed to provide an important arena to address integration problems. The respondents indicated that Dialogue macht Schule has changed Downloaded by [207.241.229.243] at 02:27 28 October 2017 ideas about how this problem should be dealt with. For instance, regarding how integration should be approached, respondents indicate that it is not assimilation of cultural traits by the migrant community, but an adoption of the underlying values of a liberal democracy (i.e. a partial change in diagnostic frames). However, public officials differ in opinion whether Dialogue macht Schule should be fully incorporated within the school system, noting that ' [We] oppose to letting non-professional staff into the classroom' (German public official). This shows that prognostic frames about how to address integration in schools are only partially changed.
In the Netherlands, public officials indicated that prognostic frames have changed, and in particular, those of high-level public officials. From a policy perspective, prognostic frames were changed in favor of co-creation (Municipality of Amsterdam, 2013 [1,2], 2014). This is also confirmed by our interviews. As one public official mentioned: 'We don't organize anything, that's the big change with the past. We change from 'taking care for' to 'make sure that' . Instead of taking care for people we only facilitate [them]' . However, public officials also indicated that not everyone shows this change of prognostic frames. One mentioned: 'Responses [to co-creation] are different. Especially people in the 'line' of the organization are like, first, show me what you got' (Dutch public official). Furthermore, some public officials are not convinced that co-creation is a better way to provide welfare services than the professional institutions: 'What [co-creation projects] I see is nice and useful [……….] but it's not going to be better, when you chase away the professionals and replace them with volunteers'. Interestingly, the involved citizens did not show a clear change in their prognostic frames either. They thought of co-creation as a collaborative relationship with existing organizations but one in which public officials remain responsible for the quality and safety protocols. As one participating citizen mentioned: 'The supervisors [of the project] are professionals. They know the [safety] criteria. Of course you need to comply with the conditions. That's their job'. So in this case, high-level public officials (and public policy documents) show clear changes in prognostic frames, but 'street' public officials and citizens did not. Now that we have explored the relationship between frame adaptation and policy change, we examine in the next section how this can be explained by surrounding state and governance traditions.

Dimension 1: authority sharing
Estonia has been independent from the Soviet Union Since 1991. As a result, the parliamentary structure and the tradition sharing of authority with non-governmental parties is relatively young and still carries traces of the authoritative state tradition that was dominant under the Soviet regime (Praxis, 2011;. One of the interviewed public officials described: I think that our democratic system is still quite young: on the one hand, the government is not accustomed to being in dialog with citizens, reacting to them. On the other hand, citizens are not used to talk about tutelage in social policy, fight for their rights, get their message across -basically influence the government, reach an agreement and then defend it. This authoritative tradition allowed a third level policy change to be implemented in quite a short time period; governmental action was not 'hampered' by institutional barriers of consultation and negotiation with other stakeholders. As one of the initiators of Maarja Küla said: 'Siiri Oviir, the then minister of social affairs, removed the obstacles in about 20 min that stood away in the development of the village in 2003' (Estonian citizen). So once prognostic frames were changed about how youth care should be provided, adapting the corresponding policy was relatively 'simple' .
In contrast, the German administration acts according to rule-bound and legalistic procedures. Moreover, there are strong interdependencies between the many government levels. The implementation of federal legislation in most policy areas is delegated to the state (Länder) level, where in turn, it is often delegated further to local authorities. Regarding the Dialogue Macht Schule program, there is 'a clear separation between formal education that takes place in school and non-formal education, like what Dialogue Macht Schule is teaching, which should remain outside school' (German public official). This can explain why the initiators of co-creation have encountered difficulties in convincing stakeholders to adopt the initiative as an integral part of school curriculum (education policy is made at the federal level but implemented locally). The program mainly co-operated with the city district level but they only have limited policy-making power. Consequently Dialogue Macht Schule needed to lobby on many different administrative levels in order to convince administrators of the usefulness of the program. This is a gradual and time-consuming process.
The Netherlands have a tradition of consensus gaining with social partners. This implies that Dutch governmental bodies are used to collaborating with non-governmental bodies, resulting in special budgets and funds being available for innovation and collaboration (Gemeente Amsterdam Stadsdeel Oost, 2013 [2]). As a result, the structures for collaboration needed to make co-creation work were already present. As one public official mentioned: ' Alliances [with social partners] were already there […] they are becoming real good neighborhoods if citizens wants to invest in it' . This could explain why some public officials adapted their prognostic frames toward co-creation. That said, while collaborating with other parties is not new in Dutch policy, working so closely with citizens in a relatively informal manner is new (see dimension 2).

Dimension 2: culture of governance
The Estonian culture of governance can be characterized as Public Interest (although it works differently than in Anglo-Saxon countries). Law -at least in the field of social policyis more in the background compared to the Rechtsstaat states. As a result, government acts as the referee deciding which party serves the public interest at best and other parties make efforts to win the support of government. One of the initiators of Maarja Küla described: ' At the end of the 1990s a lot of things were still in flux and we wanted to prove to the government that we could do things in a different way' . This relatively informal character means that projects depend on the continued favor of government to maintain political and financial support. Where in Rechtsstaat states policy is relatively stable, in Public Interest states, interest groups need to keep fighting for attention and financial support from government. As one citizen mentioned: 'Maarja Küla is in effect in 'competition' with similar initiatives and other social projects for the same funds' .
Germany can be characterized by a Rechtsstaat culture of governance. This implies that the division of responsibilities among actors for public service delivery is governed by formalized rules and procedures. As a result, public officials' flexibility in decision-making is restricted. In the case studied here, this implies that changes in a policy area as fundamental as education can only occur slowly and are constrained by the regulatory framework. For example, several school headmasters said they would like to hire teachers who are more willing to work on programs like Dialogue Macht Schule, but this is not possible because schools have little control over hiring staff: 'these decisions [hiring staff] are made at the state level according to a strict hierarchy' . Consequently, changing the curriculum and introducing an alternative way of education provision is slow and requires changes of prognostic frames of both administrative and political actors, which then need to be formalized into regulation. As one German official mentioned: 'Only now are we seeing a gradual expansion of extracurricular activities and changes in this attitude' (German public official).
The Netherlands is also characterized by a Rechtsstaat culture of governance. Combined with a tradition of consensus gaining, it means that collaboration structures in the Netherlands are institutionalized and formalized in laws and regulation. Consequently, policy is relatively difficult to change. As one citizen argued: 'I think that the administrative context is rather sloth. Decision-making is just too slow' . As a result, professional organizations have gained an authoritative position when it comes to knowledge and expertise, which is formalized in rules and regulation and they do not want to 'surrender' this historically gained position to citizens. They argue that public service quality will decrease if co-creation is embraced as the new way of public service delivery: '[public service provision] not going to be better, of that I am convinced. If you just replace professionals, it's not going to be more effective'. Because of this formalized way of collaborating, the involved public officials reacted in two different ways. Some of them argued that public officials feel restricted by the rules and the bureaucratic procedures to seek alternative ways of service provision that are not based on a regulative framework. As one of the citizens mentioned: 'What I heard from civil servants is: I really want to, but I can't, due to all that bureaucracy. And if I want it I have to consult with others, and they all want something else' . On the other hand, this perceived excess of rules and regulations is considered by some public officials as a reason to consider co-creation as a viable new way of public service delivery, as these projects are not yet 'locked-in' administrative procedures. As one of the officials argued: 'those new initiatives don't have those evaluative frameworks yet' . So the Rechtsstaat culture offers an explanation for why some actors changed their prognostic frames in favor of co-creation, while others held on to the more conventional ones, although in a more nuanced way than in the German case.
The cases indicate that state and governance traditions may affect how and whether co-creation changes prognostic frames and policy change. The Estonian case shows that an authoritative state tradition may help actors implement co-creation and change policy in favor of co-creation in a top-down manner. The German and the Dutch case show that having a consensus gaining tradition means many actors are responsible for different parts of public service delivery and that decision-making and implementation processes are slower. Combined with the Rechtstaat culture of governance, which sees collaboration structures as institutionalized and formalized, policy change is slow and locked in administrative rules and regulations. This was different in Estonia where regulative frameworks are less rigid that in Germany and the Netherlands. However, this also shows that welfare policy in Estonia is less stable than in the two other countries, which creates competition between co-creation initiatives.

Conclusion
Co-creation is increasingly seen as a viable way to address contemporary challenges in public service delivery (Voorberg et al., 2015). As such, frames about what the fundamental problems are and how these should be dealt with are changing. This frame adaptation is a learning process between public officials and citizens. Given the lack of academic attention in the co-creation and co-production literature to this learning process and its relation to policy change, we have addressed this issue. The study took into account the state and governance traditions in which co-creation and learning processes take place to address the following research questions: Does co-creation lead to frame adaptation and policy change? And how can this be explained by the state and governance traditions in which co-creation is embedded? To answer the research questions we have conducted an international case comparison in Estonia, Germany and the Netherlands.
The study showed that co-creation does lead to frame adaptation and policy change. The Estonian and Dutch cases indicated, that once prognostic frames (i.e. the identification of possible and relevant solutions and approaches for a problem) were changed, policy was also changed in favor of co-creation. In Germany, policy change occurred to a lesser extent, and the prognostic frames of German public officials were also changed to a lesser extent. However, to conclude that where we observe learning, there has been a policy change is too simple a conclusion. Our cases show that how policy changes is affected by the macro context of state and governance traditions in which actors and policy are embedded. The Estonian case showed that due to an authoritarian state tradition, policy change in favor of co-creation was relatively easy. This was strengthened by the fact that in Estonian welfare policy, the rule of law has a less prominent role than in the Rechtstaat cultures (Netherlands and Germany). Policy change is not 'obstructed' by regulative frameworks, which was the case in the other two countries. Here, we recognize how the consultative state tradition and the shared responsibility over many actors meant that the prognostic frames of many more actors needed to change in order to create policy change. In sum, to create a more comprehensive understanding of how and whether policy is changing accordingly, we need to take the context of state and governance traditions into consideration. Doing so offers a plausible explanation for contrasting relations between learning and policy change.
These conclusions have theoretical implications for both the literature on co-creation and the literature on learning and policy change. To start with the latter, this research shows that due to the involvement of citizens as co-creators in public service design, learning is required to cause policy change. However, contrary to the assumptions of Hall (1993), we did not find that policy change is a reliable indicator to whether learning has occurred. Our analysis showed that in co-creation, the relationship between learning and policy change is more complicated. For instance, policy change does not by definition occur when actors show adapted frames. Furthermore, not every policy change is a result of frame adaptation of actors. Institutional arrangements -such as regulatory frameworks -should be considered as well.
We add to the co-creation literature by taking a learning perspective. This allows us to open the black box of co-creation a little bit and offers an explanation why co-creation was successfully implemented in some cases. We've shown how actors from different backgrounds develop comparable ideas about problems and solutions and learn based on those contexts. In particular, in our Estonian case, an alignment of frames became visible.
Related to this, the co-creation project in Estonia was the most smoothly implemented one of all are three cases. This way, a learning perspective provides us the possibility to identify a potentially key determinant of successful co-creation.
Furthermore, we add to both bodies of literature by arguing, in line with Schön and Rein (1995), that in order to go one step further and to understand whether learning (i.e. frame adaption) and policy change have occurred or not, one needs to consider the institutional context surrounding the co-creation initiative. Our research shows that this macro-level context could potentially influence whether actors are willing and able to align their frames in co-creation processes. Analyzing learning as a process of frame adaptation (Schön & Rein, 1995) and making a distinction between diagnostic and prognostic frames (Benford & Snow, 2000) is a useful approach to empirically examine to what extent learning occurs.
We must, of course, place our conclusions into perspective. In the first place, this paper was focused on whether change in frames occurred in processes of co-creation. In our analysis, we were therefore focused on only one aspect of co-creation. Given that co-creation is a processual concept (i.e. consisting of different stages), it may be important to examine how learning can be related to different aspects of this process, i.e. decision-making, implementation and evaluation. Future research is needed to conclude whether learning is such an inherent part of co-creation in the public domain, as suggested by Prahalad & Ramaswamy (2000). Secondly, we were only able to examine one case in each country. Variation in learning and policy change between the cases is possibly also explained by other characteristics of the case than just state and governance traditions. Including multiple cases from one country may nuance the findings. However, our aim was not to make statistically generalizable claims about the influence of state and governance traditions, but merely to explore whether they could offer a plausible explanation for learning and policy change. Future research must show to what extent there is a significant correlation between the kind of state and governance traditions and whether learning and policy change does occur. By focusing research on this topic, we can elaborate to what extent there is a significant relation between learning and policy change (Hall, 1993). Furthermore, in doing so, we can conclude what kind of institutional context can be considered as a fertile breeding ground for co-creation and which one is not.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by FP 7 Learning from Innovation in Public Sector Environment (LIPSE) 320090.

Notes on contributors
William Voorberg is a doctoral candidate at the department of Public Administration and Sociology at Erasmus University Rotterdam.