The limitations of policy learning: a constructivist perspective on expertise and policy dynamics in Dutch migrant integration policies

Abstract This article zooms in on the role of policy learning in non-incremental policy change. Can policy learning contribute to ‘policy punctuations’ or ‘paradigmatic change’? This question is addressed from a constructivist angle. Within the constructivist approach debate rages on whether, and if so under what conditions, there could be a relationship between policy learning and policy change. The discourse coalition framework renounces the cognitivist concept of policy learning, whereas the critical frame analysis framework claims that critical reflection at the level of policy frames can lead to fundamental ‘frame shifts’. This article reviews these two constructivist frameworks for policy analysis in terms of how they conceptualize and theorize the relation between policy learning and policy change. Besides offering a discussion of the theoretical assumptions of the two constructivist approaches that have been selected, this article offers an empirical congruence analysis of learning and change. This congruence analysis will be applied to one specific case: migrant integration policy-making in the Netherlands (2000–2015). This involves a clear case study where various non-incremental policy changes have taken place, from an integrationist to an assimilationist approach, which makes it a revelatory case for an in-depth study of policy learning and policy change.


Introduction
Policy learning and policy change are two traditional topics within policy sciences. However, the relation between the two has remained under-theorized (see also Moyson, Scholten and Weible, this volume). This concerns in particular the question what type of relation between learning and change can emerge under specific conditions. One such relation, which has been pertinent within policy sciences, is whether policy learning can lead to non-incremental policy change (Dunlop & Radaelli, 2013;Hall, 1993;Sabatier & Jenkins-Smith, 1999). Some scholars argue that non-incremental change always requires exogenous forces, or as Sabatier defines it, external perturbations (Sabatier & Jenkins-Smith, 1999). In this view, endogenous processes such as learning could only lead to incremental change. However, others argue that non-incremental change could be triggered by actors and factors endogenous to a policy subsystem. This would include processes of policy learning or as Hall (1993) describes it 'social learning' or what Schön and Rein define as 'critical frame reflection ' . This article zooms in on the role of policy learning in non-incremental policy change. Can policy learning contribute to 'policy punctuations' or 'paradigmatic change'? Nonincremental policy change is understood as change at the level of problem definition and deeper premises about what causes a problem, what groups are involved and why something is a problem in the first place (Baumgartner & Jones, 1993, 2015. Policy learning is understood as the updating of policy beliefs based on knowledge and information on the policy problem at hand (Hall, 1993;Sabatier & Jenkins-Smith, 1999).
Whereas rational approaches argue that learning can contribute to fundamental policy change (William, 1994), political as well as institutional approaches see fundamental policy change as result of learning as much more unlikely (Howlett and Koppenjan this volume;Ostrom, Cox, & Schlager, 2014;Pierson 2004;Sabatier & Jenkins-Smith, 1999). Yet, much less research has approached the role of policy learning from a constructivist angle. Rather than focusing on the iteration between 'powering' and 'puzzling' or on how institutions structure the policy process, constructivism focuses on how policy problems are defined and interpreted through language, narratives and symbolism. The central puzzle for this article is whether such problem definitions and interpretations can be changed on a profound level by processes of policy learning.
However, also within the constructivist approach debate rages on whether, and if so under what conditions, there could be a relationship between policy learning and policy change. The discourse coalition framework (Hajer, 1995) renounces the cognitivist concept of policy learning. Fundamental policy change does take place, in the form of shifts in policy discourses, but this is explained primarily on the level of language, symbolism and social interaction rather than on a cognitive level in terms of 'learning' . In contrast, the critical frame analysis framework (Schön & Rein, 1995) claims that critical reflection at the level of policy frames can lead to fundamental 'frame shifts. ' Although not using the concept of learning per se, this framework does recognize that if specific institutional conditions are met, cognitive reflection on the level of deeper policy frames is possible.
This article reviews these two constructivist frameworks for policy analysis in terms of how they conceptualize and theorize the relation between policy learning and policy change. In their contribution on how to systematize thinking about policy learning, Howlett and Cashore (2009) argued for more empirical analyses to verify the vivid and sometimes almost ideological debates on policy learning. Therefore, besides offering a discussion of the theoretical assumptions of the two constructivist approaches that have been selected, this article offers an empirical congruence analysis (Blatter & Haverland, 2012) of learning and change. This congruence analysis will be structured by a set of assumptions deduced from both perspectives regarding specific aspects of the relationship between learning and change: what actors are involved, what knowledge is utilized, how is the learning process organized and what policy outcomes are anticipated in terms of policy change.
This congruence analysis will be applied to one specific case: migrant integration policy-making in the Netherlands (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). This involves a clear case study where various non-incremental policy changes have taken place, from an integrationist to an assimilationist approach, which makes it a revelatory case for an in-depth study of policy learning and policy change. The case study is based on material collected through extensive document analyses (policy documents as well as research and advisory reports) and interviews with key actors in learning-change relations.

A constructivist perspective on policy learning?
Constructivist approaches to policy learning and change involve an amalgam of approaches. They share an interest in the social definition and interpretation of problems and policies, but can differ significantly in terms of how this is conceptualized and theorized. This also concerns the conceptualization and theorization of non-incremental policy change. Schön and Rein (1995) define non-incremental change as policy frame shifts, Baumgartner and Jones speak of policy punctuations that divide periods of relative stability, Haas speaks of changes in or between epistemic communities, Hajer speaks of shifting discourse coalitions. However, most constructivist perspective do share a perspective in which changes on the deeper level of how we define and understand policy problems do occur, however rare such 'reality shifts' (Fischer & Mandell, 2012) may be.
Policy learning as a concept is not used much explicitly in constructivist literature. This can be explained by the strong cognitivist connation of the concept learning in the policy literature. As such, learning is often portrayed by constructivist as an overly rationalist notion of how actors would behave in policy processes and an overly objectivist model of the role of knowledge in policy processes. From a constructivist perspective, the distinction between 'powering' and 'puzzling' is problematic, as both elements are inherently entwined in the ideational sphere of policy-making; knowledge is power, and power is knowledge.
Yet, despite the concept being contested in the paradigm dialogue, constructivist approaches do speak to the same aspects of policy processes that are addressed by other approaches in terms of 'policy learning. ' Constructivist studies have contributed significantly to our understanding of how knowledge can play a role in policy far beyond the rationalist model of 'speaking truth to power. ' Also, constructivists have contributed significantly to the debate on the role of knowledge in changing actors' beliefs and problem perceptions. For instance, work on 'knowledge utilization' has shown how knowledge can be utilized in different ways and also contribute to an updating of policy beliefs in other ways that only the cognitivist way (Boswell 2009). Hence, in the scope of this article, it is important to conceptualize 'policy learning' as an updating of policy beliefs in response to knowledge, information and experiences.
This article discusses and applies two specific constructivist perspectives, with different assumptions on the role of policy learning in policy change. On the one hand, Schön and Rein's perspective of frame critical policy analysis speaks directly to concerns about policy learning in terms of their notion of 'frame reflection ' (1995). If specific conditions are met, frame reflection can contribute to policy frame-shifts or non-incremental policy change. On the other hand, Hajer's discourse coalition framework provides a very different perspective in which knowledge and expertise are much less likely to contribute to fundamental shifts in policy discourse, but rather help to sustain specific discourses (structuration, institutionalization). However, both approaches share a focus on, what is described in this article as, learning from an ideational perspective.
To apply these perspectives in an empirical congruence analysis, I will first explicate and juxtapose the core theoretical assumptions from both perspectives. I will deduce a set of assumptions from the perspectives that we subsequently be confronted with the empirical case of migrant integration policy-making in the Netherlands. First, what types of actors are involved in learning and change, including the influence of scientific experts on policy discourses and framing; secondly, the utilization of different types of knowledge in the policy process and differentiating between instrumental and more symbolic modes of knowledge utilization. Thirdly, how the learning process is structured, such as how discourses interact and have impact (discourse institutionalization, discourse structuration) as well as the institutional conditions that would lead to critical frame reflection. Finally, the outcomes of learning in terms of policy change, involving incremental as well as non-incremental changes.

I. Frame reflection and policy learning
The framework of frame critical policy analysis, developed by Schön and Rein (1995), has become widely used in contemporary policy sciences for its focus on policy framing. Frames are defined as 'underlying structures of belief, perception and appreciation' (ibid, p. 23) that are generally 'tacit' or unknown to actors themselves (ibid, p. 34). However, they do play an important role in actual social practices: to provide a 'way of selecting, organising, interpreting, and making sense of a complex reality to provide guide-posts for knowing, analysing, persuading and acting' (ibid, p. 32).
Situations where multiple frames are used by actors within a policy setting are referred to as 'intractable controversies' , or situations characterized by 'multiple social realities' . Such situations could generate frame conflicts, or struggles over the naming and framing of a policy situation [...] symbolic contests over the social meaning of an issue domain, where meaning implies not only what is at issue but what is to be done.
The type of actors involved in frame reflection or learning should therefore, in this perspective, not be restricted to a limited set of certified actors. Precisely because frame controversies involve actors with very different frames, it is important that actors with very different frame can participate equally in the policy process, and that no frames are excluded or ignored.
Importantly in the context of learning and change, frame conflicts are clearly differentiated from disagreements about more structured problems, or problems where the problem definition as well as the proposed policy solution are little contested (Hisschemöller & Hoppe, 1995). Intractable policy controversies defy resolution by merely studying 'the facts' , because actors with their own unique frames tend to select different sorts of factual evidence and, even if they agree on a selection, tend to interpret it differently (see also Scholten, 2011). When frames are contested, facts do not speak for themselves. Therefore, there is no way of 'objectively' distinguishing between strong or weak knowledge claims, and therefore frame reflection requires the involvement of any type of knowledge claim that is seen as relevant by one of the participating frames. In terms of knowledge utilization this means that instrumental knowledge utilization is not considered likely in settings of intractable policy controversies.
Whether actors are actually able to reflect on their frames in real-life settings, depends, according to Schön and Rein, significantly on the context or 'situational setting' in a policy area. They argue that 'design rationality' is required within a policy situation to enable critical frame reflection. Design rationality would involve a specific set of conditions (see also Scholten, 2009). Firstly, actors should communicate (ibid, p. 182) in order to engage in reflection. Frame reflection must be an open process in which actors communicate about their frames and no important frames are ignored. Secondly, actors must be able to stand in the shoes of other frames; there must be a certain empathy in order to understand how actors with specific frames make sense of a problem situation. (ibid, p. 176). Thirdly, introspection: actors must become aware of their own frames, possibly in interaction with the identification of alternative frames (ibid, p. 174). This creates the possibility to reflect critically on one's own frame and to search for possible design flaws. Fourthly, there must be a willingness to change one's frame if critical reflection would point in that direction, for instance when internal inconsistencies are traced This requires a certain pragmatism -not in terms of the framing itself but of an unbiased willingness to adapt one's frame if necessary. Finally, trust as an important condition for critical frame reflection (ibid, p. 179). Without trust, necessary capacities such as communication, empathy, being able to correct one's own frames and pragmatism are unlikely to thrive.
In terms of policy outcomes, the frame reflective perspective theorizes that nonincremental policy changes can be the result of critical reflection. Such non-incremental changes are conceptualized as 'frame shifts' , or fundamental changes in how actors define and understand problem situations. In relation to the role of 'learning' in such frameshifts, believe that actors can become aware of their own frames, critically reflect on them and subsequently change frames. Thus, frame reflection could contribute to the 'situated resolution of frame controversies' (ibid, p. 176). They argue that frame reflection, as they frame 'learning' , calls for critical reflection on a frame's internal consistence and coherence, as well as on its relationship to developments in a problem situation and to society at large. Actors therefore would be able to reflect critically on the extent to which their frame offers a convincing story about a problem situation, whether it fits the evidence and whether it gels with the broader normative perspective.
Thus, according to Schön and Rein, learning or 'critical frame reflection' can lead to nonincremental policy change. In terms of the dimensions of the learning-change relationship that are examined in this article, this involves specific assumptions. It should be open to participation of actors with different frames, involve those knowledge claims that are considered relevant by the actors involved (multiplicity of frames means multiplicity of knowledge claims), and involves what they describe as 'design rationality' as a mode of organization (involving openness, empathy to other frames, introspection, willingness to change and trust). And if these assumptions are met, then, Schön and Rein argue, nonincremental policy change or 'frame shifts' can be the result.

II. Discourse coalitions and policy learning
The perspective of the discourse coalition framework on learning, non-incremental change and the role of experts therein is however essentially different from the frame reflection perspective. The discourse coalition framework was developed by Maarten Hajer (1995). Hajer defines discourse as 'an ensemble of ideas, concepts and categories through which meaning is given to phenomena' . This emphasizes the discursive production of reality, particularly the role of language therein. Learning is a contested notion when applied to the discourse coalition framework. The framework does not speak of learning, nor of closely associated concepts like reflection, explicitly. Rather than learning, the framework speaks of persuasion and the affinity of storylines that convincingly construct a comprehensible discourse out of a complex and intractable reality.
The discourse coalition framework recognizes, just as the frame reflective approach, the plural nature of the policy process, in which various discourses compete for attention and influence. However, it differs in terms of the key actors involved in this process. Hajer speaks of 'discourse coalitions' , or 'an ensemble of a set of storylines, the actors that utter these story lines and the practices that confirm to these story lines, all organized around a discourse' (ibid, p. 45). Such coalitions do not necessarily mean that interests are known nor necessarily aligned, nor that a discourse necessarily matches the cognitive belief systems of all members of a discourse coalition. Rather, it means that they have been persuaded by a specific discourse and (have come to) share a discourse's construction of a problem situation. Rather than cognition or beliefs, Hajer speaks of 'discursive affinity' with the storylines that are created by discourses, which glue various elements of a problem situation in a more or less coherent whole.
When it comes to the type of actors involved in 'learning' in this perspective, it is important to observe that learning would take place primarily within discourse coalitions rather than between discourse coalitions. Experts, scientists and intellectuals are considered parts of discourse coalitions rather than operating above or beyond discourse coalitions. This is an essential difference with the frame reflective approach that positions the frame reflective expert beyond or between frames.
Regarding the outcomes of learning in terms of policy changes, the discourse coalition framework is much more sceptic regarding the opportunities for non-incremental policy changes. Because discourses just as frames form ways of constructing problem reality, changes in discourses could also be seen as reality shifts. However, such discourse shifts would be difficult to achieve, precisely because of the deeply entrenched nature of discourse not just because of power relations or cognitive beliefs but rather because these discourses constitute reality as we (the policy actors) know it. A process of 'discourse structuration' takes place when 'a discourse starts to dominate the way a society conceptualizes the world. ' This happens in particular when key actors in a field are persuaded by a discourse. Such discourse structuration will be accompanied by a process of 'discourse institutionalization' , when discourses 'solidify into an institution, sometimes as organizational practices, sometimes as traditional ways of reasoning' (ibid, p. 46). Think of how institutions were created around the 'integration industry' , which also defined their interests in terms of promoting integration (Scholten, 2009).
However, two different types of roles can be distinguished for expertise and knowledge, of which only one approximates our conceptualization 'learning' as the updating of policy beliefs based on knowledge and information. First of all, it is recognized that knowledge and expertise could be key factors in discourse structuration. In particular, they can contribute significantly to the persuasiveness and credibility of discourses, contributing to discourse structuration. However, this positions expertise and knowledge distinctly within discourse coalitions, rather than constituting a more distanced party that operates beyond multiple discourses. Secondly, knowledge and expertise can contribute to discourse institutionalization amongst others by operationalizing discourses into organizational practices or by introducing specific discourses into common language. Whereas structuration can be seen as a form of updating of policy beliefs, institutionalization should rather be seen as the establishment of these updated policy beliefs.
Much more than the frame reflective approach, the discourse coalition framework emphasizes the importance of the social construction of credibility and authority of knowledge claims as a central stake in these processes of discourse institutionalization and structuration. Hajer makes a strong connection himself to the sociology of science here (1995), showing how the mobilization of discourses of objectivity and the construction of scientific authority can help establish specific discourses or discredit other discourses.
Thus, the discourse coalition defines learning, or the adaptation of policy beliefs based on knowledge and information, as a process within rather than between discourse coalitions. Researchers, scientists and experts can play a key role in this, especially when their knowledge claims are socially constructed as authoritative and credible. Then, knowledge and information can contribute to discourse structuration and subsequently the institutionalization of a discourse. Especially this discourse structuration process is seen as a form of policy learning. However, the discourse coalition framework is much less optimistic rather non-incremental change; usually the establishment and shift of discourses is expected to be incremental of nature.

Methods
This analysis focuses on migrant integration policy-making in the Netherlands over the period 2000-2015. This case is selected as a revelatory case for studying the role of learning in change for several reasons. First, migrant integration policies have in the Netherlands, as in many other European countries over the last decade, been prone to various nonincremental policy changes. This includes the dramatic turn in Dutch policies from an integrationist approach in the 1990s to a more assimilationist approach in the 2000s, in which the Netherlands preceded many other countries that would also experience a multiculturalism backlash and assimilationist turn in the early 2000s (Joppke & Morawska, 2003;Vertovec & Wessendorf, 2010).
Secondly, migrant integration policies have been strongly contested, even politicized, while at the same time remaining knowledge intensive. This involves various research institutes and advisory bodies, as well as a strong role of public intellectuals (see also Scholten, Table 1. comparison of conceptualization of the role of policy learning in policy change in the frame reflective approach and the discourse coalition framework.

Frame reflective approach
Discourse coalition framework conceptualization of policy learning Frame reflection learning interpreted as persuasion or 'discourse structuration' types of actors involved in learning Reflection requires openness to a broad range of actors with different frames actors operate in the context of specific discourse coalitions Knowledge utilization in learning process (symbolic/instrumental) controversies cannot be resolved by reviewing 'the facts' , therefore symbolic rather than instrumental knowledge utilization instrumental knowledge utilization only within discourse coalitions; but also symbolic knowledge utilization to lend credibility and authority to discourses Structure of the learning process Situated resolution of controversies requires design rationality (communication, empathy, introspection, willingness to change, trust) Process of discourse institutionalization and discourse structuration outcomes of learning (incremental/ non-incremental change) Frame reflection can lead to non-incremental policy change within settings characterized by design rationality Primarily incremental shifts between discourse coalitions 2009). This makes this a well-suited case for studying the role of learning in fundamental changes. This analysis is designed as a 'congruence analysis' . This methodological design is fit in particular for situations where a researcher is interested in exploring the explanatory power of different theories in a specific case (Blatter & Haverland, 2012). A congruence analysis involves first a deduction of relations between variables from different theories (this was done in Table 1 for the relation between learning and change). Subsequently, these relations are operationalized in specific observations for the case under study, mostly in the form of predictions in term of what should be observed if one theory holds (see Table 2 in the next session). The empirical analysis then involves the identification of 'congruence' between the findings and theory-based expectations. This allows for a deeper understanding of to what extent the theoretical assumptions behind the selected theories indeed provide explanatory leverage in the selected case.
In order to be able to fully grasp the role of learning in the development as well as the aftermath of non-incremental policy change, this article covers a relatively long period (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). In terms of research methods, the analysis is based on a review of secondary literature (in particular analyses of integration policies), analysis of primary documents (including parliamentary hearings, policy memoranda and scientific or advisory reports) as well as interviews with key policy-makers and experts (from the DIAMINT and UPSTREAM projects, a total of 21 interviews has been used specifically for this analysis).

Migrant integration policies in the Netherlands
In the following a reconstruction will be made of changes in Dutch migrant integration policies and the role of learning therein (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). This reconstruction will focus on the key aspects of the learning-change relationship that have been identified above and elaborated for the two constructivist perspectives. These elements are operationalized for the case of migrant integration policies in Table 2 below. This will provide the basis for the congruence analysis.

From integrationism to assimilationism in the early 2000s
There has been strong discontinuity in the development of Dutch migrant integration policies over the past decades. In the literature on migration policies, the Dutch case has for a long time been described in terms of its so-called multiculturalist paradigm of integration. This refers in particular to the Ethnic Minorities Policy of the 1980s that was based on a formal recognition of specific immigrant minority groups in Dutch society and an intensive government approach aimed at sociocultural emancipation and socio-economic participation. However, a first non-incremental change took place in the early 1990s, turning away from this internationally renowned multicultural paradigm. It then adopted an 'Integration Policy. ' The focus on integration put immigrant integration in the perspective of participation in central societal institutions (education, labour, housing). Rather than recognizing groups, individual immigrants now became the focal point for integration into Dutch society. However, this integration policy would change dramatically after a decade or so. In 2000, a second national minorities debate emerged, the so-called Scheffer debate, which focused attention on an alleged 'multicultural tragedy' . In addition, a series of events widely discussed in Dutch media and politics drew further attention to an alleged 'clash of civilisations' . This included violent events that involved immigrants, as well as moral events that focused attention on the dilemmas of cultural and religious diversity, such as imams making radical statements about homosexuals or refusing to cooperate with the female Minister for Integration. Especially path-breaking was 'the long year of 2002' when the populist politician Pim Fortuyn made immigrant integration the centre of public and political attention. Fortuyn called for 'zero-immigration' as the Netherlands was 'full' , calling for a 'cold war against Islam' and dismissing Islam as 'an idiotic culture' . When running for the 2002 parliamentary elections, Fortuyn was assassinated by an animal-rights activist on the same day that polls indicated that his party would come out first of the elections.
This 'long year of 2002' set the stage for another non-incremental policy change in Dutch immigrant integration policies. The reformulation of the integration policy in response to the Dutch 'multiculturalism backlash' became a clear political priority. Various centre-right government coalitions established after 2002 developed strong political leadership in the domain of immigrant integration. In particular the Minister of Immigration and Integration from 2002 to 2007, Ms. Verdonk, was a key policy entrepreneur for a more assimilationist policy approach. In one of her first policy memoranda, Minister Verdonk described the contours of a so-called 'Integration Policy New Style' (Treaties of Parliament, TK 2003. Whereas the Integration Policy had focused primarily on social-economic participation, the focus now shifted toward social and cultural distance between migrants and Dutch society. In order to support 'the continuity of society' , the focus would have to be put on the bridging of differences rather than on 'the cultivation of the own cultural identities' . Cultural differences were now framed as problematic cultural distances (ibid). It was argued that 'a too large proportion of minority groups live at too great a distance from Dutch society' . In this context, the goal was to 'diminish the distance between minorities and the native population in social, cultural as well as economic respect' (ibid).

Attempts of policy learning at the paradigmatic level
The reconstruction above clearly shows that there was non-incremental policy change in the early 2000s. It also shows that political and public developments played a key role, such as the debate on the multicultural tragedy and the rise of populism and anti-immigrant parties in parliament. Furthermore, there was a key role of focus events such as the 9/11 attacks and the murder of Fortuyn. Have there also been moments or opportunities for policy learning? What role did knowledge and expertise play in this key turning point in Dutch migrant integration policies?
A key opportunity for policy learning evolved around a 'Temporary Parliamentary Investigative Commission on Integration Policy' that was established in 2003. This commission was named after its chairman Stef Blok from the Liberal Party (VVD). A parliamentary motion that led to the establishment of this commission already included a conclusion that the integration policy had been a failure. It stated: 'Concluding that the integration policy has thus far been insufficiently successful, (...) it would be desirable to evaluate what the cause of this have been' (Parliamentary Treaties, TK 2002-2003. The commission involved representatives from various political parties with different views on immigrant integration, including mainstream parties as well as a new populist party that had been established by Fortuyn. The commission proceedings involved an extensive literature study, a series of closed and open interviews with various persons that had been involved in the national and local integration policy over the past decades, and a series of 'hearings' in four cities that were open to the public (Scholten & Van Nispen, 2008). Also, an extensive literature review provided a provisional answer to all the commission's research questions, provide factual information on policy developments, and address issues such as goal attainment and effectiveness.
Very much in contradiction with the political message within the parliamentary motion that had led to the establishment of this committee, the committee concluded that 'the integration of many immigrants has been a total or partial success, and (...) this is quite an achievement for the immigrant citizens concerned as well as for the host society' (Blok, 2004, p. 105). This conclusion was grounded in a very specific idea about migrant integration, attributing a key role to education. It was argued that progress in the area of education would be a predictor of progress in other areas as well at later stages. In this way the commission dissociated itself from the dominant cultural tone of political debate at that time, but also implicitly adopted a specific frame or discourse regarding migrant integration that put more emphasis on socio-economic aspects of integration, in particular education (interviews with members of the Blok commission).
Furthermore, the commission became contested already before it could present its final conclusions. Internal disagreement within the commission emerged when one of the commission members decided to leave the commission (interview with members of the Blok Commission). This triggered controversy over the commission's research approach, in particular about the type of knowledge and expertise it had selected, its focus on education and its tentative conclusion that policy had been relatively successful (which had already leaked to the press). In the evolving debate on the commission's findings, the credibility of the researchers involved was clearly on the line. A key focus of the debate was whether the commission had indeed evaluated what it was supposed to evaluate. Here, the definition (or framing/discourse) of migrant integration was clearly a central stake in the debate.
A member of parliament argued that 'there has never been a consensus on a definition of the word integration, parliament has given the research commission an assignment that was too vague' (Parliamentary Treaties, TK 2003. Within parliament, parties with very different perspectives on migrant integration all pointed at issues that the commission would have ignored (such as the institutionalization of Islam or criminality amongst migrants) or had interpreted differently than they would (such as educational performance). Also, a parliamentarian questioned why the commission had ignored the mission that it had been given, which was to examine why policy had failed, not to determine that it had been 'relatively successful' (Parliamentary Treaties, TK 2003. Two other sources of knowledge and expertise played a role in this period: the Scientific Council for Government Policy (WRR) and the Social and Cultural Planning Office (SCP). Of these two institute, the WRR had developed a tradition of involvement in the field of immigrant integration (Scholten, 2009). Previous reports of the WRR (1979WRR ( , 1989 had laid the foundation of Dutch integration policies in the past. In 2001 the WRR published a new report, which once again asked for a dramatic change in policy framing/discourse regarding migrant integration. It asked for a systematic connection between central institutions of Dutch society, such as the social state, and policies of immigration and immigrant integration, to adapt the structure of Dutch society to the reality of being an immigration society. Already during the preparation of the report a national debate on the Multicultural Tragedy had erupted (in 2000), in which the Council chose not to engage (interview with two of the authors of this WRR report). Once published, the report met with negative responses from government, and did not receive the amount of public attention that previous reports had. The events of 9 September 2001 played a key role in this, and made the report outdated already at the time of its publication.
A similar faith was due for the 2007 report 'Identification with the Netherlands. ' This report developed different perspectives on integration and identification with Dutch society, claiming that besides culturalist forms of identification, more functional forms of identification take precedence in highly diverse societies as the Netherlands. It argued that identification with Dutch society could be best enhanced in functional domains as education and labour, rather than more normative and emotional ways of identification (interview with author of this WRR report). Again, this contrasted sharply with the dominant public and political discourse that stressed the need for more normative and emotional identification with Dutch society. This once again led to public criticism towards the WRR, especially due to a statement of a Dutch princess (who would later become queen) Maxima during the press presentation of the report. She stated, being born in Argentina herself, that she had not discovered 'the Dutch identity' yet. This drew the report immediately into the strongly politicized climate of that moment, and led to open political criticism towards the WRR from various political parties.
What is clear is that the reports from the WRR, as well as the report from the Temporary Parliamentary Investigative Committee on Integration Policy, all took odds with the dominant assimilationist policy discourse. But rather than triggering policy learning, they seem to have triggered controversy that only fed assimilationist discourse rather than questioned it. In the sharply politicized climate of the 2000s, the questioning of assimilationism by experts led to a questioning of the expert authority of themselves. In this case, their (scientific) authority clearly did not contribute to policy learning of any kind. Furthermore, these experts created a sort of counter-discourse of academic support of multiculturalism that only put more fuel on an emerging storyline that academics had been too strongly involved in past policies, too biased, and too little involved with 'the voice from the street' .
A clear political primacy emerged in this period, with a special minister for Migration and Immigration (Ms. Verdonk) exhibiting a clear political leadership over the development of a more assimilationist approach. Actors like the WRR and the Blok Commission played only a marginal role in this regard; these acts were primarily a result of successful political advocacy (interview with policy-makers). In fact, political primacy was actively maintained also by putting the credibility of scholars (such as from the WRR and the Blok Commission) on the line. It was this political pressure that was responsible for the policy outcome, in terms of a non-incremental policy change towards assimilationism.
However, this does not mean that policy learning did not play a role at all. Rather, we see that a different kind of knowledge and expertise did play a key role. This involved in particular the provision of detailed statistics on the process of integration of various migrant groups in Dutch society. Although contested in various countries, the Netherlands had developed an elaborate system for the collection of 'ethnic statistics. ' This was done by government-associated institutes such as the Central Bureau for Statistics and the SCP. This type of mandated and instrumental information became very prominent in the 2000s (see Scholten & Van Nispen, 2008).
The SCP had attributed attention to immigrant integration from the early 1990s. In annual and later biannual Minorities Reports, it provided data on the position of immigrants in various domains. This included primarily socio-economic domains as education and labour. These Minorities Reports were a product of a contract between the SCP and the Home Affairs Department. In the context of the emerging assimilationist discourse on migrant integration in the early 2000s, the SCP and the coordinating Department of Justice together decided to involve sociocultural integration in their Minorities Report for the first time in 2003. This decision reflected how the SCP saw its role as being to respond to shifts in public and political mood. Also, it shows how the SCP coordinated its growing focus on sociocultural integration with actors in the field of policy-making. These reports would play an important role in legitimizing the new policies in the early 2000s; for instance, the 2003 government memorandum that announced the 'Integration Policy New Style' was formally written as a governmental response to a 2003 report from the SCP.

The mainstreaming of integration policies (late 2000s)
Following this assimilationist turn in the first half of the 2000s, the end of the 2000s and the early '10s witnessed yet another non-incremental policy shift in Dutch policies; the mainstreaming of integration policies. Mainstreaming means that policies are no longer directed at specific migrant groups and any form of labelling migrants as distinct from the majority population as well as cooperating with minority organizations was now abandoned. In fact, the backlash against multiculturalism that was particularly strong in the Netherlands focused in particular on the social-cultural reification of minority groups due to the Dutch group-specific measures in the past. Migrant integration was now primarily framed as promoting citizenship of newcomers in generic spheres like education, housing and labour. As far as the cultural dimension remained important within the recent Dutch policy approach, it involved promoting interaction and inter-ethnic contact rather than any form of group-specific bonding.
Policy coordination now became fragmented across the vertical as well as the horizontal dimension (interview with policy-makers). Since the end of the 2000s, there no longer is a special minister for migration nor for integration; since 2013 integration is just one of many dossiers of the Minister of Social Affairs. The role of the directorate for integration, in the Social Affairs Department, has also weakened considerably as most policy responsibilities shifted to generic departments, like Housing and Education. Furthermore, on the vertical level, local governments have become particularly entrepreneurial in terms of developing their own policy philosophies. Take for instance Amsterdam's anti-radicalization approach or Rotterdam's policies on residential desegregation for which it even managed to get a new national law for urban neighbourhoods adopted.
In the mainstreaming of integration policies, several arguments play a key role. First, in the political arena the argument emerged that rather than government investing in the integration of specific migrants, migrants should be made responsible for their integration themselves. In the context of the strong presence of first populist parties like the Fortuyn party and later the anti-immigrant Freedom Party of Wilders, policies that actively invested in integration had become less opportune. Especially when a government coalition came to power in 2010 that was led by the Liberals and received support from the Freedom Party, the term 'integration policy' was abandoned. Secondly, a powerful narrative had emerged in the context of the multiculturalism backlash, that policies from the past had put too much emphasis on minority groups, thus contributing to the reification of ethnic boundaries. Rather than stressing the existence of minority groups, policy should stress how migrants become part of mainstream society. Thirdly, especially at the local level in highly diverse cities, the argument was raised that diversity had deepened to such an extent with many different origins and growing discrepancies between generations also due to hyphenation with Dutch society, that it has become impossible to speak of specific migrant groups. Finally, austerity also clearly played a role in the Dutch policy changes. Whereas in the 1990s Dutch government organized and financed civic integration courses for newcomers, newcomers nowadays have to finance courses and to find relevant course providers themselves. This is framed as a means of promoting individual responsibility.
As a result of this mainstreaming, migrant integration almost entirely vanished from the Dutch policy agenda in the 2010s. There was no longer a central coordinating Minister (although integration remained a side-responsibility, now of the Minister of Social Affairs). Importantly, budgets specific to migrant integration policy almost vanished; even the budgets for the Civic Integration and the Civic Integration Abroad Acts were reduced to almost zero, by making migrants responsible for the financing of their courses and civic integration tests. This created a significant paradox with the continued relevancy of migrant integration in political debates. Also, the rapid increase of intra-EU mobility and the emergence of the European refugee crisis in 2015, clearly revealed the limits of the mainstreaming of migrant integration (interview with policy-makers). This created a new momentum for possibly yet another change in Dutch migrant integration policies. In this new momentum, especially local governments played a leading role in putting migrant integration on the agenda.

Ethnic statistics: learning for fundamental or incremental change?
In the context of this mainstreaming of integration policy, knowledge and expertise played a very specific role. This concerns in particular the role of ethnic statistics, or data about the societal position of ethnic minorities (De Zwart, 2012;Guiraudon, Phalet, & Ter Wal, 2005). Though the Netherlands does not have a census (like the UK), it does manage to collect a significant amount of data on specific ethnic groups. These data were generated by making use of government databases on nationality, those who are born outside the Netherlands (1st generation), and those who have at least one parent born outside the Netherlands (2nd generation).
In particular the annual integration reports from the authoritative Social and Cultural Planning Office, the Central Planning Bureau and the Central Bureau of Statistics play a key role in targeting Dutch generic policies in this way. These reports bring together data on various dimensions of the position of migrant groups, including socio-economic participation but also about crime, attitudes, housing and various others. This involves 'traditional' migrant minorities as well as relatively new groups such as refugees from Somalia, Iraq and Afghanistan as well as labour migrants from Eastern European countries like Poland, Bulgaria and Romania.
These reports played a role in several specific respects. First, as a soft-governance tool for the interdepartmental coordination of integration policy (interview with policymakers). This became particularly prominent in the 2010s in the context of mainstreaming, the de-institutionalization of the integration policy and government retrenchment. This reflects a broader pattern that in areas where no formal and binding measures for policy coordination are available and where multiple actors are involved, knowledge and information can provide efficient tools for policy coordination (Geddes & Scholten, 2014).
Secondly, these reports allow for the monitoring of the effects of 'mainstreamed' policies for specific groups. Without such data, it would be impossible to check whether mainstreamed policies effectively manage to address the concerns of specific migrant groups, even without targeting them directly. Rather than using ethnic statistics for legitimizing group specifically targeted measures, as had been the case in earlier episodes of policy development, ethnic statistics are now used to intensify or otherwise 'fine-tune' mainstream measures in domains like housing, education and labour wherever these data suggest it is needed (interview with policymakers).
Finally, these reports at various points also helped set the public agenda. Besides being able to speak directly to policy and to politics, institutes like the Social and Cultural Planning Office also have a strong coverage in the media. This way it can contribute significantly to raising problem awareness and agenda setting on integration-related issues. For instance, at several occasions integration reports contributed to the agenda setting of integration issues in relation to intra-EU migrants and refugees. Especially the collection of integration-related data regarding EU migrants has become salient over recent years; whereas Dutch government advocates a more integration-oriented approach towards EU migrants, the EU prohibits any mandatory integration measures for EU citizens. However, although institutes like the SCP do have an important say in the selection of research topics, as a government financed and associated institute its research agenda must be seen as a coproduction with government (interview with director of SCP). As such, the emphasis of the SCP over recent years on collecting data on EU mobile citizens can be equally seen as growing problem awareness of government on this issue, rather than the other way around.

Congruence analysis of policy learning
The empirical case study of policy learning and migrant integration policy-making in the Netherlands is aimed at establishing congruency with the two theoretical frameworks discussed above: the frame reflective approach and the discourse coalition framework. The two frameworks were compared via a congruency analysis on four key aspects of the two perspectives: type of actors involved, type of knowledge/expertise involved, organization of the learning process, and outcomes of policy learning (see Table 3 for a summary).
When analysing the case of Dutch immigrant integration policy-making from the perspective of frame reflection, there have been some efforts from advisory bodies to organize critical frame reflection. This applies in particular to the reports from the Scientific Council for Government Policy that explicitly addressed the level of policy framing, and did so from a relatively independent position. To some extent, this applies also to the Blok Commission, which had a clear multi-party composition but also held expert hearings and even public hearings to involve many perspectives in its analysis. However, central for the Blok Committee was its advocacy of a specific frame that defined education as the key domain of integration in a period where political debate concentrated on the cultural dimension.
However, such efforts did not lead to non-incremental policy change. In fact, as we have seen for the Blok Commission as well as the Scientific Council for Government Policy, the attempts to critically reflect on the prevailing assimilationist discourse as well as on possible alternative discourses of integration triggered significant criticism. In fact, reports from both institutes led to open questioning of the credibility of the experts involved. The roles and positions of these relatively independent bodies bringing in knowledge and expertise within a politicized setting, was actively put on the line. The frame-shift towards assimilationism, as well as the later turn towards mainstreaming, were clearly fed more by political developments in spite of these efforts of critical frame reflection. The political there was a non-incremental policy shift toward assimilationism, but learning does not seem to have been a key role in this policy shift there was a non-incremental policy shift toward mainstreaming, but learning only seems to have substantiated rather than triggered this shift context was so strong that even the involvement of the mentioned research institutes was politicized, rather than that a setting was created in which frame reflection was possible based on openness and trust. Indeed, our analysis shows much evidence of learning contributing to discourse institutionalization, and to some extent also to discourse structuration. In this context, especially the role of the SCP stands out. Rather than contributing to a shift in discourse, or a non-incremental policy change, its involvement in policy learning contributes rather to structuration and institutionalization of policy discourse. This has become particularly important in the context of the mainstreaming of integration policies, where the data produced on migrant integration helps to raise awareness of integration issues (discourse structuration) as well as to the institutionalization of a mainstreamed approach (discourse institutionalization).
In sum, the Dutch case of migrant integration policies reveals a match in terms of policy outcomes of policy learning with the Discourse coalition framework, which argued that learning is an unlikely cause of non-incremental change. What speaks against the frame reflective approach in this case is, in particular, that the type of actors and the type of knowledge required for frame reflection were clearly there, especially in the role of the WRR and the Blok Commission. Yet the organization of the learning process was obstructed to such an extent in the politicized context that rather than learning the consequence was contestation about these actors and their knowledge and expertise. What speaks in particular for the Discourse Coalition Framework is the type of knowledge that was utilized and the organization of the policy process, which clearly reveal a preference for knowledge and expertise that fitted within the developing policy discourse both in the 2000s (assimilationism) and in the 2010s (mainstreaming), and that helped structurate discourse as well as institutionalized these policy discourses. The key role of the ethnic statistics produced by the SCP is possibly the best example of discourse institutionalization in this regard, given its key role in the cross-sectoral coordination of a policy approach without specific funds.

Conclusions
This article takes a constructivist angle on the role of policy learning in fundamental or non-incremental policy change. A congruence analysis was made of the explanatory leverage of two constructivist frameworks that theorize the role of learning in change differently; the frame reflective approach (Schön & Rein, 1995) and the discourse coalition framework (Hajer, 1995). Whereas the frame reflective approach does hypothesize that under the right (institutional) conditions learning or 'frame reflection' can lead to fundamental change or 'policy frame shifts' , the discourse coalition framework hypothesizes that learning based on knowledge and expertise takes place within discourse coalitions rather than between discourse coalitions. The congruence analysis involves an in-depth case study of migrant integration policy-making in the Netherlands, which provides a case where several non-incremental policy changes have taken place over the last decades. As this also involves a field in which knowledge and expertise plays a key role, it provides a revelatory case study for the congruence analysis of the role of policy learning in non-incremental policy change.
The analysis provides support for the thesis that the role of policy learning in fundamental policy change is rather limited. There have been several attempts to critically reflect at the level of frames, through the powerful political instrument of a parliamentary investigative committee and by authoritative institutes such as the Netherlands Scientific Council for Government Policy. However, none of these attempts contributed to the frame shift from an integrationist to a more assimilationist approach. Rather, this fundamental policy change in the first half of the 2000s was triggered primarily by political developments and took shape in spite of efforts towards policy learning rather than thanks to these efforts.
The failed efforts of critical frame reflection showed in particular how difficult it is to achieve policy learning when the fundamental policy problem is politically contested. Schön and Rein define amongst others trust, empathy, introspection and willingness to change as conditions for frame reflection. However, what happened was that the credibility and the authority of the institutes and the persons involved in these efforts of frame reflection were put on the line in political and public debate. This case study suggests that it may be unlikely to assume that the conditions for frame reflection as mentioned by Schön and Rein will be met when fundamental problem definitions are at stake which is likely to trigger political contestation rather than reflection.
The congruence analysis showed that policy learning does indeed play an important role in substantiating specific discourses, as hypothesized by the discourse coalition framework. Policy learning based on knowledge and expertise played a key role in what is described as discourse structuration. This applies in particular when it comes to the use of data and reports produced by the Social and Cultural Planning Office that played a key role in defining the issue of migrant integration. Moreover, the data and reports by the SCP were also key to what is defined as discourse institutionalization. With the growing inter-departementalization of migrant integration, and in particular the mainstreaming of migrant integration since the late 2000s, research was the only remaining 'soft governance' tool for policy coordination.
The analysis of policy learning and fundamental policy change in Dutch migrant integration policies lends support to a rather sceptic conclusion about the potential for policy learning. Rather than knowledge and expertise being a key motor behind fundamental policy change, I observe a remarkable capacity of policy processes to ignore knowledge and expertise when it contradicts policy assumptions at the level of problem definition. Schön and Rein's conditions for frame reflection resemble Habermas' ideal speech situation in the fact that they may be too ideal and too little realistic to ever apply in contested settings that involve problem definitions.
The analysis also shows that learning is not just driven by a desire to change but also by a desire to develop further or to establish a specific frame. Hence, in the study of policy learning, it is important to not only analyse what knowledge is utilized, but also what knowledge is produced in the context of learning processes. It is this knowledge producing role that appeared very important in terms of discourse institutionalization and structuration in the case of Dutch migrant integration policies.

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