Frame ambiguity in policy controversies: critical frame analysis of migrant integration policies in Antwerp and Rotterdam

ABSTRACT Policy frames are understood as the outcome of a policy process in which multiple frames are contesting, but where one frame prevails and characterizes policies. Policy frames are therefore perceived and studied as coherent interpretations of a policy issue containing a problem definition and a matching strategy to solve it. This rather fixed understanding of policy frames contrasts with other interpretive approaches which recognize a more dynamic and sometimes ambiguous character of policy language. The aim of this article is to analyze whether policy frames may be ambiguous and if so, how this can be understood by the problem context and political context of the policy issue. This study conducts critical frame analysis of local migrant integration policies in Antwerp and Rotterdam over the past 15 years. The analysis demonstrates presence of frame ambiguity in this controversial policy domain in the form of incomplete frames, solely focusing on the policy strategy while leaving the problem definition open to interpretation, and inconsistent frames in which the problem definition and policy strategy do not match. Ambiguous frames indicate a ‘strength of weak frames’: in a context of problem complexity and political contestation ambiguous frames can serve to overcome a deadlock in policy-making.


Introduction
Frame analysis has become a popular methodology in policy sciences following the linguistic or constructivist turn in policy sciences that took place during the 1990s (Fischer and Forester 1993). Most constructivist approachessuch as discourse analysis have a rather fluid conceptualization of policy as 'language in use'. They recognize ambiguity in policy language, for example, in the use of certain expressions or metaphors. Frame analysis, in contrast, assumes that policies communicate a rather coherent and singular frame the policy problem. Frames are considered to be interpretive packages providing a consistent causal story of how the problem came about and how it should be solved Modigliani 1987, 1989;Entman 1993). Policy frames are understood to resolve ambiguity in social reality rather than being ambiguous themselves. In a critical frame analysis of policies related to the policy domain of migrant integration, this study argues that policy frames can be much less coherent and unitary with regard to problem definition and proposed policy actions than what theory so far suggests. Furthermore, this study seeks an understanding of when frame ambiguity emerges.
The conceptualization of policy frames as internally coherent structures of which one at a time is present in policies is somewhat different from the notions of framing in other scientific disciplines such as social movement studies (cf. Benford and Snow 2000), public opinion and representation (cf. Chong and Druckman 2007) and media and communication (cf. Vliegenthart and Van Zoonen 2011). Frame analysis in these fields of study shares a more dynamic conceptualization of frames as it primarily focuses on the process of framing in which frames are jointly constructed and reconstructed by actors and their audiences (Polletta and Ho 2006, 189-192;Snow et al. 1986, 467). While it is understood that frames in interpersonal communication, politics or the media can be incomplete or ambiguous, this has been less described with regard to policy frames as they are considered to be the product of considerate political and bureaucratic processes.
This article aims to critically evaluate the assumptions of unicity and coherence with regard to policy frames in an in-depth case study of local migrant integration policies. Frame ambiguity is understood as the state of policy frames consisting of incoherent problem definitions and policy strategies. Ambiguous frames are therefore open to multiple interpretations. Subsequently, it aims to contextualize under what conditions frame ambiguity is likely to emerge. Studies recognizing ambiguous qualities of policy language direct us to a theoretical understanding of when policy frame ambiguity emerges (Jones 1999;Stone 1988). To these aims, the following research question is addressed: In what ways does frame ambiguity occur in local migrant integration policy frames and how can this be understood?
The analysis focuses on migrant integration policies in two city cases: Antwerp in Belgium and Rotterdam in the Netherlands. The policy field of migrant integration provides a suitable case to study frame ambiguity as it is generally perceived to be an intractable policy controversy characterized by uncertainties in the problem definition and political controversy. These qualities have been pointed at by interpretive policy analysts as circumstances under which ambiguity emerges in policy. Earlier studies already pointed at ambiguous language in immigration and integration policy texts (Teitelbaum 1992;Chock 1995;Boswell, Geddes, and Scholten 2011). Based on a typology of frames that are commonly distinguished in migrant integration policies, I study whether and how frame ambiguity may exist. I specifically focus on the unicity and coherency of frames in terms of problem definition and policy strategy. Municipal policies are studied because at the local level, problem definitions are combined with specific policy strategies.
The theoretical section of this paper first discusses the literature on policy frames that generally assume coherency and unicity of policy frames. In order to theorize the existence of frame analysis, we subsequently discuss interpretive studies that recognized ambiguity in policy language. This literature review provides us with theoretical expectations on the context in which frame ambiguity may emerge. In the methodological section of this paper, the data and methodology are further introduced, including a typology of integration policy frames that is used for critical frame analysis. The findings of this study indicate that policy frames are much less coherent and unitary with regard to problem definition and proposed policy actions than what theory so far suggests. Frame ambiguity was encountered in the form of incomplete and inconsistent policy frames. In both cities, similar complexities in the problem context and political context existed that according to interviews with key policy actors played a key role in the ambiguous framing of local integration policies. Frame ambiguity is a sensible policy outcome in a complex problem context and in a situation of political controversy.

Policy frames
Frame analysis made its way into policy science during the 1980s and 1990s following the 'linguistic', 'interpretive', 'argumentative' or 'constructivist' turn in policy sciences (Fischer and Forester 1993). This turn entailed a focus on the linguistic and nonlinguistic symbols in politics and policy (White 1992;Fischer and Forester 1993). Instead of departing from the notion of an inherent nature of policy problems, many policy scientists today acknowledge a (partly) socially constructed nature of policy issues (Edelman 1988, 13;Stone 1989, 299). These socially constructed problem definitions guide the policy actions that are taken. Different practices of policy analysis spring from this turn in policy scienceone of which is frame analysis.
Framing, in the interactionist tradition of Bateson (1955) and Goffman (1974), is about making a difference between what is important and what is not, regarding a particular issue or situation. Frames arise when people are encountered with a situation and ask themselves: 'What is going on here?' (Goffman 1974, 8). In policy sciences, this key characteristic of frames is applied in the collective setting of policy-making (Baumgartner and Mahoney 2008). Policy actors are making sense of a policy problem and hereby adhere to a certain perspective of reality. As Hajer and Laws (2006, 252) state, 'policy practitioners seek stability and act in a social world that is a kaleidoscope of potential realities'. Frames are interpretive schemata and ordering devices that are needed by policy-makers to structure the reality of a policy issue. Frames allow people to make sense of a reality and attach a meaning to it, besides possible alternative meanings: 'Whatever is said of a thing, denies something else of it' (Rein and Schön 1977, 239).
Apart from structuring our perception of reality, frames promote a certain course of action as well. They have normative-prescriptive implications by outlining not only what is but also what ought to be (Rein and Schön 1977, 240; see also Schön 1983, 40). Only when interpretation enables actors to make sense of a situation, they can imagine what should happen next (Rein and Schön 1977). The well-known definition of frames by Rein and Schön (1993, 146) acknowledges the normative leap that is taken in frames: policy frames are ways of 'selecting, organizing, interpreting and making sense of a complex reality to provide guideposts for knowing, analyzing, persuading and acting'. While discourses can take any form and are more dynamic, framesas Schön (1977, 1996) have arguedtake the form of a causal story with a normative component (Stone 1989(Stone , 300, 2006. They place responsibility with certain actors and call for a certain policy response, including a specific objective, type of solution and instruments that are considered most effective (Stone 1988;Gamson and Modigliani 1989;Weiss 1989).
A key assumption is thus that policy frames are coherent in terms of problem definition and proposed policy action or strategy (Hajer and Laws 2006, 257). This internal coherency contributes to the strength of a frame. 'Strong frames' are effective in communicating a set of beliefs to the public (Chong and Druckman 2007). Tewksbury and Scheufele (2009, 24) state that the most effective or powerful frames are the ones that communicate all four of the elements described by Entman (1993, 52): a problem definition, causal interpretation, moral evaluation and treatment recommendation. As the end result of lengthy political and bureaucratic processes, policy frames are expected to be well elaborated and less ambiguous than, for example, frames in day-to-day communication or media frames.
A second assumption about policy frames is that policies are shaped by one dominant frame at a time that directs policy action. This assumption is closely related to the assumed coherency of frames. In case of intractable policy controversieswhen there is debate over the existence and nature of the policy issuea multiplicity of frames exists in the policy process (Schön and Rein 1994, 240). Different (groups of) stakeholders in the policy process maintain a different issue frame and there are 'struggles over the naming and framing of a policy situation [. . .] (as well as) 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' (Schön and Rein 1994, 29). Actors involved in the policy process uphold certain frames and strive for their framing of the issue to become the dominant interpretation in policy. According to Baumgartner and Jones (1993), 'policy monopolies' act upon a unitary dominant frame. Dominant frames become vested in political and administrative institutions and the broader culture through both the instrumental effects of policiessuch as new rules and new organizationsas well as the rhetorical or symbolic effects (Ingram, Schneider, and DeLeon 1999). It is assumed that policy frames do not change incrementally and fluidly but they are only replaced in situations of policy change. This may be caused by political power shifts or to external events that provide new information on the policy issue at hand.
In contrast to political positions or interests, frames are not negotiable stands. Schön and Rein (1994) have outlined how frames are cognitive schemata that incorporate the worldviews of policy actors and become part of their identities. It is therefore hard to think outside frames and be reflexive of one's own frame. Van Hulst and Yanow (2016) claim that reframing a policy issue involves reconceptualizing not only vested interests, but also personal identities that are interwoven with beliefs that the world is or ought to be as one perceives it. Policy frames are therefore not considered to be the outcome of political negotiation, but in policies one frame has prevailed and is comprehensively elaborated. Policy frames are thus assumed to be singular and coherent.

Ambiguity in policies
Recognizing ambiguous elements in policies is not a novelty in the interpretive tradition of policy analysis. In fact, the interpretive approach recognizes ambiguity of policy language as a rule, rather than an exception (Yanow 1996, 228). However, while it is often considered a property of linguistic expressions, it has not been connected to the construct of frames. Scholars among whom Stone (1988), Yanow (1996) and Hajer and Wagenaar (2003), for example, pointed at ambiguity of policy narratives by use of certain expressions and metaphors. In order to gain a better understanding of why ambiguous frames emerge, literature on ambiguous elements in policy helps to theorize ambiguity of policy frames and to develop theoretical expectations on when frame ambiguity emerges. Two types of explanations for ambiguity are provided, related to the problem context and to the political context of the policy issue.
First, related to the problem context, scholars encountered ambiguity in policies as a result of 'bounded rationality' (Simon 1957;March 1978). Bounded rationality assumes that individual or collective decision makers are intendedly rational but deal with limited information on complex policy issues and cognitive limitations in processing this information (Jones 1999). Ambiguity may be a satisfactory solution when available information about a policy issue is limited and uncertain. Ambiguous framing is thus sensible when it reflects uncertainties in the future consequences and preferences related to the problem (March 1978). According to Baumgartner and Jones (1993) bounded rationality is an inherent quality of the policy-making process. It is not only an attribute of the policy problem, but also of the policy-makers' acts of meaning making. Policy-makers engage in disproportionate information processing as they attend selectively to information dependent on whether it complies with the current policy frame. Ambiguity is thus more likely to exist in policies related to issues on which information is limited, uncertain or contradictory.
Second, interpretive policy analysts recognize partisan drivers of ambiguity in policies. Stone (1988, 157), for example, describes how in the decision-making process ambiguity allows policy-makers to placate multiple political actors in a policy controversy. Yanow (1996, 228) as well shows that ambiguity is purposively used to resolve conflict and to accommodate political differences. Newman and Clarke (2009) consider ambiguitynext to articulation and assemblagesto be a key concept in understanding how publics and public services are constituted. Ambiguity results from contesting political understandings of citizenship and exposes alignments of political power (2009). Newman (2013, 105) demonstrates how the language of political contestation sometimes is reappropriated in policy texts with ambiguous meaning. This results in political opposition losing power. Ambiguity is thus likely to exist in policies related to politically contested issues. Based on this literature review, the analysis of understanding the emergence of frame ambiguity will be focused on the problem context, including complexity and bounded rationality, and the political context, specifically controversy over problem definition and policy solutions.

Methodology
Local migrant integration policies in Antwerp (Belgium) and Rotterdam (The Netherlands) were analyzed in order to study in what forms frame ambiguity may be present and how this can be understood. The policy field of migrant integration was chosen as a strategic case. Migrant integration has been described as an intractable policy controversy (Scholten 2013). How migrant integration is defined and what policy strategies are chosen varies and is often subject to frame shifts (Scholten 2011;Vertovec and Wessendorf 2010). It is a complex and politically contested issue characterized by discussion about the problem definition, how to measure it and the effectiveness of policy measures. The struggle over interpretation of the issue is not overcome by appeals to evidence or facts, and competing issue frames remain present in the policy process. The problem context and political context of this policy field thus makes a suitable case for studying frame ambiguity in policies. Indeed, scholars have already pointed at ambiguous language in immigration and integration policy texts (Teitelbaum 1992;Chock 1995;Boswell, Geddes, and Scholten 2011). This makes migrant integration a suitable policy field to study the ambiguity of frames as well.
The local level of policies is chosen because on that level policies often include concrete strategies. Antwerp and Rotterdam are cities with long histories of migrant integration policies. Both are large, industrial port cities that have attracted many immigrants over the past decades. The cities' integration policies often functioned as an example for other cities and their respective regional and national contexts. The time period from 1998 to 2013 was chosen as the integration policy field became controversial due to a multicultural backlash (Vertovec and Wessendorf 2010). Both cities departed from their multiculturalist policies during the early 2000s and several frame shifts took place in the years after. This enables us to study frame ambiguity in relation to different policy frames within a relatively short time period. Critical events during this time period became clear from the interviews and are addressed in the empirical section of this article. Important critical events in Antwerp include a racist murder in 2006 and the election victory by the Flemish nationalist party 'N-VA' in 2012. In Rotterdam, the election victory by right-wing nationalist 'Leefbaar Rotterdam' in 2002 and shortly thereafter the political murder of their party leader Pim Fortuyn was mentioned frequently as a turning point in migrant integration policies.
I conducted critical frame analysis of integration policy documents from the past 15 years, focusing on problem definitions and policy strategies. For this purpose, I operationalized a typology of commonly distinguished migrant integration policy frames in the literature. This includes an assimilationist, multiculturalist, differentialist and universalist policy frame of migrant integration (Castles and Miller [1993] 2009; Koopmans and Statham 2000;Scholten 2011, 38-42). Table 1 provides a short overview of the problem definitions and policy strategies that are elaborated in these four policy frames.
Assimilationism problematizes various aspects of the presence of minorities in society and aims for unidirectional integration of the immigrant in the host society while focusing primarily on the sociocultural domain of migrant integration. Assimilationist policies encourage adaptation of immigrants to dominant cultural norms, values and behaviors. Differentialism (also described as segregationism) facilitates the presence of immigrants in society by organizing minority group identities and needs outside mainstream society. They are excluded from mainstream arrangements and different arrangements are created for themwith fewer citizenship rights than the dominant population. Ethnic groups live alongside each other rather than with each other. Multiculturalism promotes cultural pluralism and encourages emancipation of immigrant groups while accommodating specific group identities. Finally, a 'universalist' frame concerns a more liberal egalitarian view on immigrant integration and aims for ethnic equality. It is closely related to the 'interculturalist' model of Wood (2009) and Cantle (2012). The policy strategy is not only targeting immigrants. Instead, it is 'mainstreamed' aimed at 'citizenship' or 'participation' of all citizens to overcome group thinking and engage in a two-way process of sociocultural adaptation. This goes along with mainstreaming of policies and policy institutions integration permeates all government's policies and activities as opposed to being a stand-alone policy field (Scholten, Collett, and Petrovic 2016). This means that governments adopt generic instead of targeted measuresa practice that has its roots in gender policies (Booth and Bennett 2002). I evaluated the coherence and unicity of local integration policy frames by coding different frame elements (including problem definitions and policy strategies) in policy documents and evaluating their fit with the theory-based typology of integration frames. This concerned an iterative process in which empirical data was confronted with our theoretical expectations. In addition, interviews with municipal policy-makers and civil society partners were conducted in order to gain a comprehensive understanding of the context in which the policies were developed. Six interviews were held in Rotterdam, five interviews in Antwerp. The interviews in Antwerp were conducted by the author and were taped and transcribed ad verbatim. Four of the interviews in Rotterdam were conducted by research assistants who provided interview summaries. Two of the interviews were held by the author and were transcribed ad verbatim. After the main argument of this article was developed, some interviewees in both cities were recontacted to check the validity of the argument and to get an update on the latest policy developments.
The following section describes first the findings of this analysis per city case. These case analyses reflect policy language that was used in policy documents and interviews. This language is not neutral: terms used to refer to ethnic groups and processes of migrant integration differs over time and in relation to different policy frames. This is made explicit at several points in the discussion of our findings, for example, by the use of quotation marks. Subsequently, instances of frame ambiguity are discussed more in general and understood from the context in which they emerged.

Integration policy frames Antwerp
Until the mid-2000s, Antwerp employed a multiculturalist migrant integration policy that was embedded in the Flemish integration decree of 1998. Migrant integration was framed as a lack of recognition and accommodation of 'diversity' (Flemish Parliament 1998). This issue definition was coupled with targeted measures to 'empower' minorities such as municipal interpretation services and adjustment of certain welfare provisions to fit the needs of ethnic minorities. The municipality actively safeguarded minorities' access to servicesan arrangement called the 'interculturalization' of municipal services. A designated representative organ for immigrants was installed, the 'minority forum' that still exists today (minderhedenforum.be). Issue definitions and policy strategies in Flanders were unequivocally multicultural. In this period, the city of Antwerp was under social democratic rule of the SP.a (Socialistische Partij Anders/Socialist Party Differently), a party that was a strong supporter of the multicultural policy frame.
Under the last social democratic city executive of mayor Janssens (2006Janssens ( -2012, the policy frame shifted remarkably. Instead of actively accommodating different minority groups, group provisions were revoked except for civic integration programs for recently arrived immigrants (cf. Van Puymbroeck 2011). The common denominator for integration policies from this period was 'equal opportunities' with regard to socioeconomic aspects of integration, and 'interculturalization' or a 'diversity policy' in terms of sociocultural integration (Antwerp City Executive 2007; Department of Social Cohesion 2009). The problem definition focused on unequal opportunities of citizens and intolerance toward minority groups, including the ethnic minorities. Policy goals were to create equal opportunities, encourage active citizenship, participation and shared responsibility among all citizens of Antwerp. Integration meant a mutual process of adaptation and development of cultural plurality (Department of Social Cohesion 2008). Diversity was not problematized, but promoted as beneficial (De8 Centre for Integration 2008). This fits a universalist issue definition.
The policy strategies of this period were framed genericallythat is, targeting all members of the urban society. This means that the government does not want to differentiate and institutionalize differences between ethnic minorities and the majority population by targeting ethnic minorities specifically. However, some group provision remained, as characteristic for 'targeting within universalism' (Flemish Parliament 2009, 3;cf. Skocpol 1991). The goal of 'interculturalization' from the previous policy period was upheld with regard to the municipality's own organization and services. However, instead of accommodation, these services were now framed in terms of 'cultural sensitivity' toward diversity. The municipality aimed to make their own services and personnel sensitive to and competent in dealing with culturally diverse citizens. Executive integration organization 'De8' was issued to pursue this in other organizations such as schools, elderly care, police and health care.
A mainstreaming strategy that fits the universalist approach was implemented in the coordinative structures of the integration policy (Flemish Parliament 2009). Until 2007/2008, the municipality of Antwerp had a department of integration services (DIA -Dienst Integratie Antwerpen). This was split up and different tasks of this organization were redistributed over other governmental departments. Minority integration became a horizontal goal of the entire organization. Some targeted departments for minorities remained (Interview policy-maker Antwerp). For example, the hiring of minority personnel in the city's administration by the 'diversity management' bureau of the municipality.
The approach to minority integration as a generic goal became more 'quid pro quo' (Janssens 2011; Interview policy-maker Antwerp). It stressed citizens' obligations as well as rights. Saeys et al. (2014) define this policy approach in Flanders as 'new assimilationism'. The political discourse in Flanders and Antwerp in particular with the emergence of right-wing nationalist parties as 'Vlaams Blok' (Flemish Block) , 'Vlaams Belang' (Flemish Interest) (2004-current) and N-VA (Nieuw-Vlaamse Alliantie/New Flemish Alliance) (2001-current), indeed showed assimilationist traits. However, these were not (yet) present in policy framing. The policy strategies fit a universalist policy frame: no dominant culture is promoted, but cultural diversity is treated as a fact and approached neutrally by the government.
Only recently, the definition of integration problems in Antwerp's integration policy gained more assimilationist traits. The onset of this was the electoral win of N-VA during the 2012 local elections. While promoting respect for diversity, the policies strive for a unitary urban community in sociocultural terms (Antwerp City Executive 2013, 3). Most prominent is the emphasis on sociocultural aspects of integration such as Dutch language proficiency and adherence to Antwerp's or Flemish history and cultural norms and values. Exemplary of this is a paragraph about education: 'The city executive sees diversity as enriching. However, in schools we accentuate common basic values: mutual respect, freedom, democracy, universal human rights such as free choice, equality of men and women, freedom of religion, separation of church and state and independent scientific research' (Antwerp City Executive 2013, 31). Dutch language proficiency is no longer optional: 'Who refuses to learn to speak Dutch, will be sanctioned' (Antwerp City Executive 2013, 12). Sociocultural integration is considered to be a necessary condition for integration in other domains of social life. The city executive specifies a common language and common values in which immigrants should integrate. The emphasis on sociocultural domain of integration is characteristic of an assimilationist policy frame.
The current policy strategy is not fully developed yetand according to our interviewee at the municipality, it is not clear yet whether this will happen at all: 'We know that they absolutely want attention for civic integration and language, but the remainder is a big question mark. It is left to the administrative level'. To date, many universalist strategies are maintained and are expected to remain present. For example, measures to promote equal labor market access and access to municipal services. Integration is a horizontal policy goal that is executed by generic measures. Even the city's diversity management bureau has been canceled. Within the municipal department of 'social cohesion' (Dienst Samenleven) there is no specific team working on integration. This universalist policy strategy fits the institutional arrangements that were decided upon in the latest decree (Flemish Parliament 2013). The Flemish government now regularizes integration via an 'external government agency' (EVA -Extern Verzelfstandigd Agentschap) in most Flemish cities. As a large city, Antwerp remains excluded from this arrangement. Integration is now managed by a new centralized municipal association of integration partners (VZW Integratie en Inburgering).
In conclusion, we can see how Antwerp's local integration policies initiated as targeted, multiculturalist policies. From 2006 onward, this multicultural approach was replaced by a universalist approach under the label of 'interculturalization'. However, the policy strategy of mainstreaming was much more pronounced than the problem definition of unequal opportunities as problematic. Today, under N-VA rule, mainstreamed policy strategies are further developed, but the problem definition and goals can be characterized as assimilationist.

Integration policy frames Rotterdam
From 1998 to 2002, Rotterdam's integration policies had a multicultural outlook. Cultural diversity was promoted as a strength (Rotterdam City Executive 1998;Project Multicolored City 2000) and the municipality aimed to decrease socioeconomic deprivation of minorities in terms of education, labor market participation and housing conditions. Rotterdam navigated between a generic and targeted policy strategy. On the one hand, they urged immigrants to make better use of existing welfare arrangements, on the other hand specific measures were developed to prevent downward social mobility and promote self-sufficiency (City of Rotterdam 1998). Priorities of the policy program 'Multicolored City' were (1) to enhance the participation of 'allochthonous' citizens in subsidized organizations and initiatives, (2) for the administration of Rotterdam to hire more allochthonous personnel, particularly in higher positions, (3) to change the cultural policies of Rotterdam in order to fit the new cultural diversity of the population and (4) to promote and encourage ethnic entrepreneurship and labor market participation (Rotterdam City Executive 1998, 12-13). The term 'allochtones' is commonly used in Dutch migrant integration jargon to refer to ethnic minorities.
In 2002, the right-wing nationalist party 'Leefbaar Rotterdam' ('Liveable Rotterdam') won the local elections and for the first time in many years, the social democrats were not part of the city executive. The new city executive announced a radical break with previous integration policies. An assimilationist issue definition of migrant integration was outlined in integration policy documents (Rotterdam City Executive 2003). The executive program stated that newcomers are not yet at home, while at the same time native citizens feel less at home in Rotterdam. Social cohesion in Rotterdam was lost over the past decennia when new immigrants arrived in the city (Rotterdam City Executive 2002, 33). Priority of the city executive was to enhance the identification of citizens with (a preexisting definition of) Rotterdam and thereby to reinforce 'social integration'. This indicates a monistic view on society of integration into a majority culture. Integration problems were often connected to safety issues: it was expected that more social integration would contribute to public safety and vice versa.
However, policy measures were not always as assimilationist as would be expected based on the problem definition of the policy frame (cf. Uitermark and Duyvendak 2008). Instead, the policy measures can be characterized as universalist rather than assimilationist. Many of the local measures aimed to increase social cohesion and stimulate interethnic contact without assimilationist emphasis. The program 'Mensen Maken de Stad' (People Make the City), for example, aimed to enhance 'social cohesion' and 'active citizenship' in specific neighborhoods of Rotterdam. Integration goals were developed bottom-up and in coproduction with citizens of these neighborhoods. It often entailed addressing other issues in the neighborhood as well such as litter, nuisance, youth delinquency and health issues in order to pave the way for social cohesion and integration. Representation of diversity in neighborhood activities was deemed important in these measures.
From 2006 onward, the assimilationist issue definition was abandoned after the social democrats regained a majority in the city council. We can observe a turn from integration policy to 'participation policy' which was targeting all citizens of Rotterdam. Integration was hardly mentioned as such in policy documents. The participation policy stressed that citizenship not only comes with rights but also with obligations and responsibilities for each citizen. A dual notion of citizenship was maintained: Whilst the government will support people who are 'willing but unable' to participate, the ones who are unwilling will be approached with repressive measures. The participation policy targeted all citizens: 'The policy has an inclusive character. No distinction is made based on ethnicity. Mono-ethnic activities are not eligible for subsidies, unless there are strong arguments for doing this' (Rotterdam City Executive 2011, 24). This indicates a mainstreaming of integration policies from a universalist paradigm.
The policy strategy of mainstreaming was however more pronounced than the problem definition to which it relates. The policy defined four focus areas in need of improvementwomen's emancipation, gay emancipation, discrimination and diversity but the policy did not specify what policy problems exist. Instead, the end goals of equal chances and participation of all citizens were emphasized. Four 'Kenniscentra' (centers of expertise) consisting of key civil society stakeholders in these areas were subsidized to develop and implement measures. Even though these policy measures target all citizens of Rotterdam, there were still some programs that were specifically targeting ethnic minorities. There was, for example, the policy program 'Participation through Language' (Department of Youth, Education and Society 2007) and a program that particularly focusses on ethnic minorities with a Muslim identity: 'Building Bridges' (Department of Youth, Education and Society 2008). Most notable is the continuation of a policy program particularly focusing on Moroccan and Antillean youth. Rotterdam as well as 21 other Dutch municipalities received funding from the national government to specifically target the overrepresentation of Antillean and Moroccan youth in welfare dependency, school dropout, unemployment and crime. Rotterdam had already started this policy before national funding was initiated from the viewpoint that ethnic background and cultural identitynext to age and socioeconomic positionwere important factors in the overrepresentation of Antillean and Moroccan youth in these four areas (Rotterdam City Executive 2010). That is why targeted measures were chosen despite the dominant approach of universalism and a fear of stigmatization. In 2012, the effectiveness of the policy was evaluated and it was concluded that on most aspects, the targeted policy did not achieve its goals. The policy was neither continued in Rotterdam, nor elsewhere.
In sum, we can see how in Rotterdam the universalist frame (2006-2013) was ambiguous in the sense that the strategy was more pronounced than the accompanying problem definition. During the period when the assimilationist frame was dominant in integration policies (2002)(2003)(2004)(2005)(2006), universalist strategies were combined with an assimilationist problem definition. These two instances of frame ambiguity are similar to the frame ambiguity that was encountered in Antwerp. In the following paragraph, I will further outline the characteristics of frame ambiguity and address the question how frame ambiguity can be understood from the political context and problem context in both cities.

Two forms of frame ambiguity and their context
Antwerp and Rotterdam were both characterized by a tradition of multiculturalist migrant integration policies. Coherent with the broader European backlash against multiculturalism (Vertovec and Wessendorf 2010), we observe that both cities departed from their multiculturalist policies during the early 2000s. Frame analysis of problem definitions and proposed strategies demonstrates that subsequent frames of migrant integration policies have been ambiguous in two ways (Table 2).
First, ambiguous framing of migrant integration was present in the period of 2002-2006in Rotterdam and 2012 in Antwerp when proponents of an assimilationist frame toward migrant integration were in power. Policy documents of these periods show an emphasis on an assimilationist problem definition. This problem definition marked a frame shift with previous policies. However, this problem definition was not combined with an assimilationist policy strategy. Instead, universalist policy strategies were proposed. In Rotterdam, the central policy program 'Mensen Maken de Stad' explicitly targeted all citizens of the city and their social cohesion while not promoting one culture over another. In Antwerp, targeted policies and organizations such as the new executive agency dealing with migrant integration increasingly focused on their core task of civic integration. Integration measures for ethnic minorities in general were explicitly mainstreamed and organized by other bodies of government. Policy strategies were not just a continuation of the previous multiculturalist policy period, but concern newly initiated universalist policy strategies despite the assimilationist problem definition. Frame ambiguity is thus recognizable here in the form of frame inconsistency in terms of problem definition and policy strategy.
A second form of frame ambiguity was encountered in Antwerp from 2006 to 2012 and in Rotterdam from 2006 to 2013. During these periods, policy documents showed an overemphasis on a universalist policy strategy of mainstreaming, while leaving the problem definition open to interpretation. A universalist problem definition would entail defining diversity as an asset of the urban society, problematizing ethnic inequalities and encouraging mutual cultural adaptation to a new urban culture. Diversity is recognized in both cities as a fact, but it is not defined in policy documents in a positive or negative sense. The policies almost exclusively focused on what ought to be while not defining what is. Targeted measures were avoided and also the coordinative administrative structures in both cities were mainstreamed: departments particularly focusing on migrant integration were sized down or disappeared and the responsibility for migrant integration policies was distributed over other departments. Without defining a problem context, the policy strategy of mainstreaming became a goal in itself. This form of frame ambiguity can be seen as frame incompleteness. How can these two forms of frame ambiguity be understood from the political context and problem context in both cities? Several similarities in both cities with regard to the problem context and political context were pointed out in the interviews with policy-makers. In the period in which frame ambiguity first emerged (20022006, both cities were coping with an increasingly complex problem context. Both cities have a relatively large and diverse immigrant population: Antwerp is home to citizens of 174 different nationalities, together making up a share of 45.8% of the population (Antwerpen Buurtmonitor 2015). Rotterdam inhabits over 170 nationalities that constitute 49.5% of the population (Centre for Research and Statistics Rotterdam 2015). The population has internally become increasingly differentiated, including new immigrant groups, immigrant descendants of the third or fourth generation and interethnic marriages (Thomson and Crul 2007). In terms of Vertovec (2007), Antwerp and Rotterdam cope with a situation of superdiversity in which targeted measuresas characteristic of multiculturalist or differentialist policy framesare hard to implement. It became increasingly difficult to demarcate the target group of integration policies (cf. Chock 1995). Frame ambiguity emerged as a sensible option to cope with such issue complexities (cf. March 1978).
At the same time both cities were confronted with a need for significant budget cuts during the late 2000s. No longer funding targeted integration measures became a compelling choice. Since the late 2000s, in both cities many provisions and services for immigrants and immigrant organizations started to disappear, just as the administrative departments that governed them. A universalist policy strategy was practically and financially advantageous for both cities at that time. Ambiguous policy frames are suitable in such a situation of bounded rationality and limited resources to manage the complex problem context (March 1978;Baumgartner and Jones 1993).
The political context in both cities has had an evident influence on the emergence of frame ambiguity as well. Both cities knew a long tradition of multiculturalist integration policies under social democratic rule. This tradition was criticized and challenged in the late 1990s and early 2000s when there was a broadly shared opinion of failed multiculturalist policies (Vertovec and Wessendorf 2010). Even within the social democratic majorities that traditionally promoted a multiculturalist integration policy frame, this led to a shift toward a universalist policy frame. In Antwerp, this happened under the second term of Mayor Janssens and in Rotterdam early universalist policies can be recognized in the policy program 'Multicolored city'. However, the political struggle over the replacement of the multicultural frame was thereby not yet decided.
In both cities the support for right-wing nationalist parties grew in the early 2000s. They became a powerful political player in both city councils. In Antwerp, the radical right-wing party Vlaams Blok was forced to disband in 2004 after a trial ruled that the party was repeatedly encouraging discrimination. The party was reinstated under a new name 'Vlaams Belang' with a revised statute. Vlaams Belang is a far right, Flemish nationalist party that advocates the independence of Flanders and strict limitations on immigration, while immigrants are obliged to adopt the Flemish culture and language. The party rejects multiculturalism and takes an openly assimilationist stance toward integration. Vlaams Belang gained a significant share of votes in the municipal and federal elections in the 2000s, but never had executive power. Like Vlaams Blok, Vlaams Belang was subjected to a cordon sanitaire when all other Flemish parties agreed to systematically exclude the party from participating in coalitions on the federal or municipal level. However, our interviewees note that the voice of Vlaams Belang was very strong in the city council and in the public debate of Antwerp (interviewees municipality and De8). A policy-maker states: 'Vlaams belang posed great pressure on integration policies until recently. Not a single piece of policy went unnoticed. Every week they asked questions in the city council about all kinds of subjects: How many mosques are there? Who are visiting them? You name it.' The party was a latent but strong influence on integration policies. As a result of this, the 2006-2012 social democratic city executive already put a large emphasis on the obligations of minority citizens to participate (Interview De8; Janssens 2011).
In Rotterdam, the right-wing nationalist party 'Leefbaar Rotterdam' started in 2001 under political leadership of Pim Fortuyn. The party won the 2002 municipal elections and led the city executive in the period 2002-2006 while making migrant integration a priority issue. Even before and after this political reign, the quick rise and popularity of Pim Fortuyn and his political views pressured the political majority in the city council into taking a more assimilationist approach toward migrant integration. In the elections of 2006 and 2010, Leefbaar Rotterdam remained the largest opposition party in the city council, but the social democrats again won the elections and led the city executive. Nevertheless, Leefbaar Rotterdam remained a powerful presence in the city regarding integration issues; in 2014 it would again win the local elections and lead the city executive.
The use of ambiguous universalist framing of migrant integration was useful in the sociodemocratic executive periods in Antwerp from 2006 to 2012 and in Rotterdam from 2006 to 2013. A mainstreaming strategy enabled announcing harsh measures related to participation, while focusing on all citizens and not immigrants specifically. This incomplete frame enabled policy-makers to accommodate a broad range of political opinions in the local city councils (cf. Stone 1988;Yanow 1996;Newman and Clarke 2009). This policy strategy appeased the right-wing nationalist opposition that pushed for assimilationist measures, as a multiculturalist or universalist policy definition was not elaborated. The problem definition was left ambiguous. A universalist policy strategy of 'mainstreaming' is a glove that fits a universalist as well as an assimilationist problem definition. Similar to what Stone (1988) and Yanow (1996) have argued, ambiguity allowed policy-makers to placate multiple political actors in a policy controversy. Frame ambiguity is a strategy to deal with and increasingly complex problem context and political controversy.

Conclusions
Frames in policy are the result of lengthy and thorough processes of political negotiation and bureaucratic labor. While in other disciplines studying, for example, social movement or media frames sometimes incomplete, fluent and ambiguous frames have been described, policy frames are generally perceived to be well elaborated and unambiguous: They consist of a compatible problem definition (what is) and policy solution (what ought to be), while one frame at a time guides policy action. Policy frames are assumed to be 'strong frames' in the sense that they consist of a uniform, explicit and coherent problem interpretation and policy strategy.
Our frame analysis of policy frames related to the intractable policy controversy of local migrant integration problems demonstrates that this is not always the case. I encountered two distinctive forms of frame ambiguity. First, an incoherent problem definition and policy strategy exists. In both cities this entailed an assimilationist problem definition combined with a universalist policy strategy of mainstreaming at a time when the cities were under rule of right-wing nationalist parties. Second, an incomplete frame was encountered consisting of only a universalist policy strategy of mainstreaming while lacking a problem definition. In Antwerp, this can be recognized from 2006and in Rotterdam between 2006. Other forms of frame ambiguity that were not encountered, are thinkable. For example, merely symbolic policies could entail providing a problem definition without elaborating a policy strategy.
Analysis of the local problem and policy context demonstrates that the frame ambiguity is not a result of poor policy-making, but a way out of a deadlock caused by problem complexity (cf. March 1978) and political controversy (cf. Stone 1988). In both cities, the problem context became increasingly complex with a situation of superdiversity and the need for budget cuts. Targeted policies were no longer possible according to the main policy actors, forcing reframing to a generic policy strategy without many benefitting subsidies. Ambiguous policy frames are suitable in such a situation of bounded rationality and limited resources to manage the complex problem context (March 1978;Baumgartner and Jones 1993). Also, the political context became increasingly contested with the emergence and popularity of right-wing nationalist parties and a broadly shared belief that multiculturalist policies had failed. When the right-wing parties came in power, in both cities an assimilationist problem definition was strongly put forward. However, this was combined with a universalist policy strategy of mainstreaming. There is unexpected alignment between these two frame elements. This was a sensible outcome in a situation of political polarization of integration issues (Stone 1988;Yanow 1996).
Overall, a universalist policy strategy was present in all instances of frame ambiguity. As our analysis has shown, this policy strategy became a goal in itself and could be paired with and legitimized from different problem definitions. It proved a glove that fits multiple problem interpretations. This warrants integration researchers that when a mainstreaming policy strategy is encountered, this does not always indicate a coherent universalist policy framing, including the definition of diversity as a benefit. As mainstreaming practices are also popular in other policy fieldsincluding gender policiespolicy analysts should be sensitive to frame ambiguity in those fields as well.
Ambiguous frames should not necessarily be interpreted as weak frames, as some literature would presume (cf. Tewksbury and Scheufele 2009). Our analysis shows that in complex policy realities and political controversy, we can speak of a 'strength of weak frames'. Similar to Granovetter's (1973) argument of the 'strength of weak ties', I find that strong and weak frames serve different purposes. Weak frames may not be the most efficient in steering policy action, but can be sensible in situations of uncertainty and successful in compromising between competing information and interests (cf. March 1978;Stone 1988;Yanow 1996). Weak frames enable policy-making in a deadlock. This conclusion with regard to frame ambiguity is similar to what scholars already described to ambiguity in policy language in general (March 1978;Stone 1988). This analysis shows that ambiguity may also exist in the construct of policy frames. These findings contribute to theorization about policy frames and methodologies of policy frame analysis. Sensitivity to frame ambiguity is needed, especially when studying intractable policy controversies.