Introduction

What is today known as the Social Construction Framework (SCF) emerged from the path-breaking contributions of Helen Ingram and Anne Schneider starting in the early 1990’s (Schneider and Ingram 1993, 1997, 2005; Ingram and Schneider 1991, 2015). In contrast to other popular policy theories suggesting policy designs were largely the result of differences in power and resources, Schneider and Ingram argued that who gets what, when, and how is also based on the social construction of target groups as either ‘deserving’ or ‘underserving’. In the last decades, the SCF became one of the competing policy process theories (cf. Cairney and Heikkila 2014; Pierce et al. 2014), and its establishment within the ‘mainstream’ of policy analysis certainly increased the visibility of constructivism as a relevant orientation in policy analysis more generally.

At the same time, however, Schneider and Ingram (1993, 335) promoted a rather ‘thin’ understanding of constructivism, for instance by arguing that “constructions of target populations are measurable, empirical, phenomena”, that target groups “have boundaries that are empirically verifiable”, and that these boundaries “exist within objective conditions”. Probably not least due to their explicit demarcation from “strict-constructionists” (ibid.), the perspective found its place within the ‘mainstream’ of positivist policy analysis.Footnote 1 Sabatier (2000, 138) for instance applauded that the work of Schneider and Ingram, in contrast to “the vast majority of ‘constructivist’, ‘post-positivist’, or ‘argumentative turn’ policy analysis”, meets the “criteria of scientific clarity” and succeeds in formulating “numerous falsifiable propositions”. In consequence, he invited them to contribute to the second and third edition of the canonical handbook Theories of the Policy Process (Ingram et al. 2007; Schneider et al. 2014) where the SCF was positioned alongside middle-range theories like the multiple streams framework or the punctuated equilibrium theory.Footnote 2 In empirical applications of the SCF, scholars regularly apply a language of variables, operationalisation, measurement, and hypotheses testing (cf. Pierce et al. 2014), and thus also speak to the positivist search for generalizable causal relations.

This article starts with the premise that this development is problematic in two respects. Firstly, the fact that Schneider and Ingram’s work has been transformed into the ‘Social Construction Framework’ and incorporated into the ‘mainstream’ of policy research as one of the competing middle-range theories draws a reductionist picture of constructivism. By equating constructivism with a ‘framework’ that shares with its counterparts a single set of positivist principles, it is obscured that constructivism is first and foremost a distinct ontological and epistemological position. To be sure, the SCF does not (claim to) have a monopoly on the notion of constructivism, and its mere existence does of course not exclude other constructivist approaches to emerge. However, by positioning the SCF as one framework among others, the ‘mainstream’ of policy research suggests that meta-theoretical controversies, which have been sparked by the argumentative or interpretive turn (cf. Fischer and Forester 1993; Yanow 2000), have now been settled, as today constructivism has its place within the field of policy theories. This in turn disqualifies other constructivist approaches as not yet having reached, to use Sabatier’s words again, a sufficient level of ‘scientific clarity’. Secondly, and regarding the relationship between target group constructions and policy designs more specifically, the reductionist architecture of the SCF is not able to capture the complexity of the social and political processes involved in target group constructions. With its focus on group evaluations and its foundationalist conception of power, the SCF, as will be shown in detail below, squanders much analytical potential for understanding and explaining policy processes.

This article addresses both of these aspects. It problematizes the constructivist shortcomings of the SCF by developing a strictly constructivist reading of Schneider and Ingram’s basic arguments, and it demonstrates the analytical value of such a perspective for investigating the relationship between target groups and policy designs. By doing so, the article makes a more general plea for the distinctive value of strict constructivism and against the attempt to align this specific ontology to rationales of positivist research. To be clear, the intention of this article is neither to question the relevance of positivist middle-range theories per se, nor to claim that there is one single way of analysing the interrelation between target groups and policy design. The article’s (at times simplistic) juxtaposition of constructivist and naturalist ontologies and of interpretive and positivist methodologies (cf. Bevir and Blakely 2016) is not to be read as an attempt to disqualify one of the two sides. Rather, the distinction is used as an argumentative strategy in order to demonstrate the specific rationales and analytical possibilities of strict constructivism. In this sense, we will argue for acknowledging the plurality of ontological and epistemological commitments in policy analysis, a plurality which is obscured when constructivism is equated with the SCF. Moreover, it should be noted that Schneider and Ingram did not explicitly claim that the SCF is the only way of doing constructivist research, and there certainly is research which refers to their ideas without adhering to a positivist logic of inquiry (cf. Sidney 2001, 2005; Østergaard Møller and Sommer Harrits 2013; Nedlund and Nordh 2018). However, it has also to be beard in mind that categorising is a performative act. Thus, the classification of the SCF as constructivist and the programmatic presentation of constructivism as simply another frameworkFootnote 3 send out a message about how constructivism should be understood and ‘applied’ in order to be recognised as a valuable part of the field of policy process theories.

In what follows, we will, on the one hand, adopt a meta-theoretical position in order to reveal three constructivist shortcomings that are inscribed into Schneider and Ingram’s argumentation and the SCF, respectively, and we will provide a strictly constructivist alternative to each of the three (Sect. 2). On the other hand, we will argue from an empirical point of view by substantiating our theoretical claims with an empirical analysis of how the ‘middle class’ becomes a meaningful target group in German welfare policy (Sect. 3). Firstly, we will show that categorisation is a complex social practice which could not be reduced to simple positive or negative valences. With regard to the middle class in welfare policy, we will show that the construction of this target group is not only about evaluations but more fundamentally about boundary-drawing, with different boundaries allowing for different policy designs (Sect. 4). Secondly, we will argue that in order to understand the social and political significance of target group constructions, particularly with regard to the (re)construction of patterns of inequality, we need to investigate how the construction of a group is entangled with the constitution of ‘the others’. With regard to the middle class, this means that by evoking the middle category, other social spheres are inevitably co-constructed, culminating in a normative order of ‘centre’, ‘above’, and ‘below’ (Sect. 5). Thirdly, we will argue that the second dimension of the SCF, power, should not be demarcated from ‘social construction’, but deserves an anti-foundationalist conceptualisation which is able to shed light on how power relations emerge from discursive practices. In this regard, we show how the middle class category is entangled with statistical knowledge which produces a scientifically authorised ‘truth’ about ‘normal’ and ‘deviant’ societal developments (Sect. 6). In the conclusion, we will summarise the main arguments and suggest how to move forward with Schneider and Ingram’s oeuvre in the field of policy process theories (Sect. 7).

The SCF’s theoretical architecture and its constructivist shortcomings

In the early 1990s, when constructivist thinking was by and large absent in policy analysis, Schneider and Ingram made a path-breaking contribution. They argued that public policies do not only address certain groups in a certain way due to these group’s ability to serve policy purposes, but also because policy-makers align to the characterisations and evaluations with which the groups are commonly associated. Schneider and Ingram (1993, 335) label the entirety of a group’s valuations its social construction, defined as “stereotypes […] that have been created by politics, culture, socialization, history, the media, literature, religion, and the like”. Elected officials, the argument proceeds, bring their policy proposals in accordance with these images in order to generate support for their agenda. At the same time, these policies “impart messages to target populations that inform them of their status as citizens and how they and people like themselves are likely to be treated by government” (Schneider and Ingram 1993, 340). This in turn has an impact on the group’s understanding of politics and its willingness to political participation, which ultimately (re)establishes structures of social and political inequality. Thus, Schneider and Ingram’s work includes a strong normative aspect which critically investigates the relationship between policy design and democratic values (particularly Schneider and Ingram 1997).

These arguments constitute the basis of what today is known as the SCF. The analytical heart of the framework is a matrix with two dimensions (Schneider and Ingram 1993, 336; Ingram and Schneider 2015, 262). The first dimension is labelled social construction which is built as a continuum from ‘positive’ or ‘deserving’ to ‘negative’ or ‘undeserving’. It captures “(1) the recognition of the shared characteristics that distinguish a target population as socially meaningful, and (2) the attribution of specific, valence-oriented values, symbols, and images to the characteristics”; social constructions, Schneider and Ingram (1993, 335) summarise, “are stereotypes about particular groups of people”. The second dimension is power, again as a continuum from ‘low’ to ‘high’. With regard to its conceptualisation, Ingram and Schneider (2015, 261) refer to a range of “conventional measures of political power”, like the size of the group, its resources, or its access to political institutions. The two dimensions result in four ideal-types of target groups: the advantaged (powerful and deserving), the contenders (powerful and undeserving), the dependents (powerless and deserving), and the deviants (powerless and undeserving). These types are then used to derive assumptions regarding the allocation of benefits and burdens, the use of different policy tools, and the strategies of public justification (Schneider and Ingram 1993, 337–340; Ingram and Schneider 2015, 261–266).Footnote 4

In the following, we will problematize the SCF from a strictly constructivist point of view. By this, we mean an ontological and epistemological position which assumes that our knowledge of the world ‘out there’ is necessarily mediated, or constructed, by symbolic orders which are both resulting from and guiding practices of observation and interpretation. Thus, the meaning of social and natural phenomena does not reside with the phenomena themselves but is the product of contingent meaning-making practices (cf. Hacking 1999, 32–34; Moses and Knutsen 2007, 190–195). More specifically, we will follow a discourse-theoretical perspective which focuses on systemic orders of discourse and their significance for the constitution of meaning, knowledge, and truth (Foucault 1981; Laclau and Mouffe 2001; Bazerman 1990). To investigate these processes, an interpretive methodology is needed which enables one to reconstruct how ‘reality’ emerges in discursive practices (Dreyfus and Rabinow 1983, 118–125).Footnote 5 From this perspective, there are three shortcomings inscribed into the SCF: the premise that constructing a target group is a mere matter of evaluation and stereotyping, the tendency to neglect how the construction of a target group involves the construction of its outside, and the conceptualisation of power. We do not intend to say that all research that refers to Schneider and Ingram’s ideas reveals these shortcomings (see for instance Sidney 2001, 2005; Østergaard Møller and Sommer Harrits 2013; Nedlund and Nordh 2018). Rather, our aim is to show in what ways and with what analytical consequences the SCF refrains from strict constructivism, and what we can gain from articulating the basic ideas within the realm of strict constructivism and interpretive methodologies.

The first shortcoming is that the SCF equates ‘social construction’ with positive or negative evaluation, and thus ignores how a group is actually distinguished as a meaningful entity in the first place. This limitation corresponds to Schneider and Ingram’s (1993, 335) dissociation from “strict-constructionists” who would assume “that there is no objective reality”. However, the point of strict constructivism is not to deny the existence of ‘reality’, but that the classifications we use to describe reality are social artefacts. With regard to social groups, there certainly are “real distinctions, as in differences in skin color”, and constructivists would not claim that these were simply “made up” (Ingram and Schneider 2005, 3). However, it is neither inevitable nor a ‘natural’ perspective to use skin colour as a meaningful trait for categorising individuals, and thus categorising has to be seen as an active practice of “kind-making” (Hacking 1999, 128–131).

Starting with the premise that differences are ‘real’, the SCF goes on with formulating “propositions” (Schneider et al. 2014, 109–131) related to a group’s evaluation as deserving or undeserving (and to its power, see the third critique below). By this, the framework fails to recognise that the selection of certain social differences to construct a target group, while disregarding other possibilities, is not determined by any law of nature, and thus already a political act (cf. Yanow 2003). This more fundamental notion of group construction indicates that policy designs are influenced well before evaluative characterisations are made, and thus the relationship between policy design and target groups is much more complex than the framework’s dichotomy of positive versus negative evaluations suggests. In short, the SCF runs risk of missing much of the politics involved in practices of kind-making. In contrast, strict constructivism allows for a much more fine-grained analysis as it sheds light on how a group is constituted as a meaningful entity in social and political practices, and how its boundaries are legitimised (cf. Lamont and Molnár 2002). With regard to the middle class, this perspective invites us to investigate which social differences are actually mobilised to construct the middle of society, and how different boundaries facilitate specific policy designs (see Sect. 4). It also enables us to critically investigate how certain categorisations are stabilised by the production of ‘knowledge’ and ‘truth’ (cf. Foucault 1981, 54–55), as is the case with population research which disposes of the authority to tell ‘objectively’ what happens in society (see Sect. 6). Put bluntly, strict constructivism starts ‘earlier’ than an analysis of evaluative practices (although construction and evaluation could never be neatly separated) in order to examine how routes of policy-making are predetermined in the moment society is categorised in certain ways.

The second shortcoming results from the SCF’s focus on the relationship between policy design and those social groups explicitly addressed. By confining the analysis this way, the framework tends to disregard the fact that the construction of a group necessarily involves the construction of its outside. In contrast, a strict constructivist ontology is based on the premise that any meaning-making operates with a logic of difference. With regard to the discourse-theoretical perspective applied here, this assumption is rooted in the (post-)structuralist argument that meaning does not reside in the link between a designating sign and designated object, but that it emerges from differences between signs (Derrida 2001; De Saussure 2006). This fundamental premise is echoed in relational conceptualisations of identity according to which the recognition of the self is bound to an encounter with significant others (Mead 1959), and the same holds true for collective identities which necessarily build on a distinction between ‘us’ and ‘them’ (Tajfel 1981; Laclau and Mouffe 2001, 113–120). These perspectives share the conviction that group formation is intrinsically bound to practices of differentiation, so that constructing an entity is both dependent on and resulting in the construction of its constitutive outside.

Thus, in the moment a policy constructs its target group, the collectivities from which this group is (implicitly) distinguished from inevitably co-emerge. For instance, an allocation of benefits to ‘the deserving’ does not only establish a collective knowledge of who is ‘deserving’ (and why), but simultaneously evokes the realm of the ‘undeserving’ (DiAlto 2005, 81; see also Sidney 2005, 115, for the distinction between ‘black middle class’ and ‘low-income blacks’). Moreover, the policy design addressing a particular group is as much influenced by the boundaries and image of this very group as it is by all those groups which are not mentioned: For instance, policies aiming at ‘activating’ the ‘underclass’ are rendered possible not only because the ‘underclass’ is constructed as ‘inactive’, but also because there is the (implicit) assumption that there is a social group which already is ‘active’, i.e. the ‘middle class’, to which the ‘underclass’ should align to (cf. Schram 2012, 242; see Sect. 5). Thus, policy formulation operates against the background of a constantly moving totality of group constructions that never only encompasses the target group explicitly addressed by the policy at hand.

Moreover, distinguishing between groups typically comes along with the construction of hierarchies, for instance when a certain group is imagined as ‘the normal’, so that all others are by definition ‘unusual’ or even ‘abnormal’ (cf. Rosenblum and Travis 1996, 27–31). Confining the analysis to the explicitly addressed target group would thus only draw an incomplete picture of how policy establishes and recreates common stocks of knowledge about (the legitimacy of) social boundaries and divisions. In order to return to the normative questions that once were at the heart of Schneider and Ingram’s work, it seems more suitable to start from a strictly constructivist viewpoint that highlights how a specific configuration of target group and policy design constitutes ideas about the composition of society as a whole. A discourse-theoretical perspective is particularly suitable for this endeavour as it, firstly, builds on the premise that ‘othering’ is not only a negligible side-effect but a constitutive facet of any meaning-making practice, and, secondly, urges one to critically approach the (symbolic) orders of social and political inequalities that result from group constructions.

This leads to the third constructivist shortcoming which involves the notion of power. In this regard, Schneider and Ingram (1997, 75) evoke the well-established ‘three faces of power’: “the power to make policy decisions and therefore decide issues directly; the power to influence through indirect means such as by determining the policy agenda or shaping the rules and norms of institutions; and the most subtle form of power which is to influence the perceptions, wants, and needs of others”. Power is thus understood as the ability of an actor to accomplish their interests, whether through overt enforcement or through subtle manipulation. To empirically assess power, Schneider, Ingram, and DeLeon (2014, 109–110) relate the “political power of a target group” to “the extent of its political resources, such as whether it is large, united, easy to mobilize, wealthy, skilled, well positioned, focused on issues of concern to it, accustomed to voting and contacting public officials, and so on”. By identifying power with resources and abilities, power is depicted as a distinct and explicitly non-constructivist dimension of the SCF’s twofold typology.

What remains neglected here is an anti-foundationalist perspective, or what Digeser (1992) has called the “fourth face of power”. This perspective builds on the premise that “power cannot be viewed as a causal effect of either agency or structure since these are themselves constructed in and through power strategies that are operating at the level of discourse” (Torfing 2009, 111–112). As Foucault (1990, 93) has famously put it, “power is not an institution, and not a structure; neither is it a certain strength we are endowed with; it is the name that one attributes to a complex strategical situation in a particular society”. A discourse-theoretical perspective, as a representative of anti-foundationalist thinking, is able to capture these ‘complex situations’ by exposing, for instance, how material resources, located on the surface of the social, are substantiated by discursive orders which distinguish between ‘right’ and ‘wrong’, ‘legitimate’, and ‘illegitimate’, or ‘normal’ and ‘abnormal’ (Foucault 1981; cf. Howarth 2009). These “relations of power–knowledge” (Foucault 1990, 99) could only be revealed from an interpretive perspective which highlights how power structures emerge from symbolic orders (and vice versa). All this is not to say that material resources are unimportant, but rather that these have to be analysed in their entanglement with socially constructed stocks of knowledge. By differentiating analytically between power and social construction, the SCF suggests that these are separate dimensions, and thus their mutual constitution tends to fade into the background.

In the remaining part of the article, we will substantiate these three theoretical claims empirically. We will do so by making use of original empirical research into the construction and political targeting of the ‘middle class’ in German welfare policy. Before presenting our empirical insights, a few notes on the methodological basis are needed.

Middle class discourse and welfare policy: notes on data and method

The empirical insights we will present below stem from a larger research project on the discursive construction of the middle class and how these constructions leave their marks on welfare policies in Germany, Sweden, and the United Kingdom (cf. Barbehön and Geugjes 2019; Barbehön et al. 2020). The project builds on the premise that the ‘middle class’ is a socially constructed category (cf. Wahrman 1995) through which societies observe, order, evaluate, and govern themselves. More specifically, we assume that the middle class is a major political reference point when constructing and legitimising welfare policy, as the emergence and consolidation of (Western) welfare states could be read as the ongoing attempt to include the middle class in the welfare architecture (Goodin and Le Grand 1987; Baldwin 1990; Esping-Andersen 1990). However, due to the multiple ways of how the middle class can be imagined (see below), the boundaries of this category and its relation to welfare policies are not objectively given, and thus the middle class’ meaning is contingent upon the discursive practices through which it is articulated. Therefore, we maintain that the middle class is both a central and an ambivalent policy target group, which calls for an interpretive reconstruction of the (nationally specific) discursive practices that specifically realise this relationship. As we will not be able to give a comprehensive picture of all the three countries in this article, we will confine the analysis to Germany which is particularly instructive for exemplifying how different constructions of the middle class enable different policy designs, how the middle class and (other parts of) ‘society’ are co-constructed, and how the power of the middle class is entangled with the production of statistical knowledge.

To empirically investigate middle class discourses, we build on two empirical sources: mass media and qualitative interviews with key political decision-makers. We use mass media as a way of approaching the collectively held notions of what the middle class is, wants, needs, and deserves. It goes without saying that mass media does not simply mirror a ‘reality’ outside its own boundaries, but is instead actively engaged in the construction of that which it claims to report on. At the same time, the plausibility of media coverage depends on its resonance with the knowledge that is cultivated in the collective memory of a society. It is exactly this mutual relationship which makes media important for analysing “the construction of relations and identities” (Fairclough 1995, 126). The role of mass media is particularly essential in the case of ‘society’ which is hardly accessible without the complexity reducing techniques of mass media (cf. Anderson 1991, 37–46). Against this background, we take mass media as a proxy to approach the collective knowledge that societies have about themselves, but without claiming that this is the whole of the story.

Within the spectrum of mass media, we made use of newspapers. We considered those newspapers as candidates for data collection which appear on a daily basis, which have a nation-wide distribution, which belong to the mainstream (as compared to tabloids), and which represent different poles of the political spectrum. For each country, we finally selected those two newspapers which fulfil these criteria and which additionally have the highest circulation. We restricted the data collection to two newspapers in order to gather an amount of articles manageable for qualitative analysis. For Germany, this led to Frankfurter Allgemeine Zeitung (FAZ) and Süddeutsche Zeitung (SZ); two newspapers which are regularly chosen in discourse-analytical studies based on newspaper coverage (e.g. Gerhards et al. 2007, 110).

The definition of the time period for our data collection was driven by two rationales. On the one hand, we aimed at covering a broad period of time, instead of a specific point in time, in order to draw a general picture of the middle class discourse beyond temporal trends. On the other hand, we had to confine the amount of data in order to be able to conduct a detailed interpretive analysis of the newspaper articles. To meet both (conflicting) requirements, we conducted a comprehensive data collection for two distant points in time (for the year 2015 as the most recent year at the moment of data collection, and for the year 2005 as the earliest year covered by the digital archives of all newspapers), and a more narrow one for the period in between. For the years 2005 and 2015, we searched the newspaper archives (in the case of Germany) for Mitt* (as in Mittelschicht or Mitte der Gesellschaft) and picked all articles which used the term to refer to society, politics, or the economy, like ‘middle class’, ‘middle of society’, ‘political middle’, ‘mid-sized companies’ etc. This resulted in 2367 articles. For the period from 2006 to 2014, we restricted the data collection, due to the vast amount of texts, to articles which used either the term ‘middle class’ or ‘middle of society’, and which explicitly related the respective term to the welfare state in general or certain social policies in particular. We hereby gathered another 301 articles, adding up to a corpus of 2668 articles in total. This twofold strategy allows us to give both a comprehensive picture of the significance of different middle categories for societal self-descriptions (2005 and 2015), and a more detailed account of how the relationship between middle class and welfare state is constructed (2005–2015). The data also allow us to trace discursive shifts over time. In this regard, the German discourse shows a remarkable temporal consistency as regards overarching logics of classification and argumentation, which could be explained by the fact that the middle category is deeply anchored in the cultural repertoire of Western societies (cf. Wahrman 1995; Münkler 2010).

For the data analysis, we organised and coded all articles using the software Maxqda. In this article, we will focus on two basic dimensions with which we started the coding of the material, the phenomenal structure and the argumentative function of the middle category. The first dimension results from “the fact that discourses, in the constitution of their referential relation (and so, their ‘theme’), designate different elements or dimensions of their topic and link them to a specific form or to a specific phenomenal constellation” (Keller 2011, 58). In the present case, we therefore analysed how in the articles ‘the middle class’ (as the ‘theme’ of our research) is defined and attributed as a social group. The sub-codes of this dimension, which emerged inductively during the coding process, comprise of aspects like income, educational level, occupational status, or certain values. With the second analytical dimension, we captured how the middle class category is used to construct arguments that refer to the status of society, politics, or the welfare state. The focus on practices of argumentation results from the assumption that the constitutive aspect of a political statement is to argue what ought (not) to be done. Consequently, political discourse analysis should “give primacy to analysis of deliberation, which will narrow down to analysis and evaluation of argumentation” (Fairclough and Fairclough 2016, 188).Footnote 6 As was the case with the phenomenal structure, we coded the different types of arguments inductively, resulting in sub-codes like ‘We need to address the middle class as its contributions are essential for the welfare state’ or ‘We should forget about the middle class because the welfare state is only for the deserving’. The coding was carried out by two researchers. In order to reconcile the interpretations of the material, about one-third of the coded articles were counter-checked by the respective other researcher. Starting with these analytical distinctions, more nuanced analyses allow us to ask, for instance, how different arguments are linked to different notions of the middle class, or how certain arguments are constructed with the help of metaphors or semantics (see Sects. 4–6). In sum, our discourse-analytical investigation aimed at identifying recurring patterns of how the middle class category is used in societal self-descriptions in order to make sense of societal and political conditions. Thus, we conceive of the middle class discourse as “an abstract product of a society’s collective mind” (Parker 2014, 16) which constitutes the interpretive repertoire of a society to observe and talk about itself.

The second empirical source is qualitative interviews with political decision-makers from different fields of German welfare policy. The selection of interviewees was driven by the aim to cover all political parties represented in the federal parliament and to talk to actors who are majorly involved in issues of social policy in their respective party. We therefore requested interviews with the parties’ spokespersons in the fields of labour market, pension, family, and tax policy and with long-standing members of the respective parliamentary committees. In the case of the Social Democrats and the Christian Democrats as the two largest fractions in parliament, we conducted three interviews each; in the case of the Green Party, the Left Party, and the Liberal Party, we held two interviews each (we did not succeed in arranging an interview with a member from the Alternative für Deutschland), adding up to twelve interviews in total. We used the interviews to access the ways in which political actors imagine the composition of society in general, the role of the middle class in particular, and how they relate the middle class to welfare policies. Thus, the interviews provided us with insights regarding both the common knowledge about ‘the middle class’ and how this knowledge is taken up and processed in policy practices. The aim was not to reconstruct single policy reforms, but rather to draw an overarching picture of how imaginations of the middle class guide political actors’ meaning-making and argumentation. With regard to the analytical strategy, we again distinguished between how the middle class is imagined as a social group and how arguments are constructed that relate the middle class to the welfare state.

In the following, we will report on those findings that are related to the three abovementioned arguments for a strict constructivist reading of Schneider and Ingram’s ideas. We do not aim to deliver an in-depth examination of current German welfare policy, nor are we able to do justice to the complexity of the empirical material. Rather, our aim is to provide empirical insights that substantiate the theoretical criticisms and arguments presented above. To showcase the analytical value of a strict constructivist perspective, each of the following three chapters will briefly contrast our insights with those that would (probably) emerge when the SCF would be applied in its programmatic form.

From evaluation to categorisation: drawing the boundaries of the middle class

The observation of society on the basis of social classes is the business of social structure analysis which knows a plethora of different techniques to localise the middle class. Among the most common criteria are income, educational level, occupational status, values, or habitus (cf. Atkinson and Brandolini 2013), and depending on the definition chosen, the middle class emerges in different shape. Although all of these criteria, particularly income, educational level, and occupational status, could be said to reflect, in the words of Schneider and Ingram, ‘objective conditions’ and ‘real distinctions’, it is not determined which of them is actually used for drawing boundaries within the social realm. Rather, it is a methodological and thus contingent decision which income level, for instance, classifies for middle class membership. The common definition according to which middle class members are all households with a net income between 0.75 and 2 times the median (OECD 2016, 2) is ultimately arbitrary, although the research community might have agreed that this range captures the middle class. Thus, the fact that social differences exist ‘out there’ does not imply that social groups defined on the basis of these differences are also objectively given, as the transformation of observable differences into boundaries used for the construction of groups is a social act that is not determined by any law of nature (cf. Wahrman 1995; Rosenblum and Travis 1996; Parker 2014). In this sense, the middle class is not ‘real’ the same way the factors used to define it are, as there is neither an automatism between social differences and the category of class, nor a necessity to distinguish between an under, a middle, and an upper class (as illustrated by the fact that in other cultures it is not a meaningful practice to classify society along social classes). In this sense, a strict constructivist perspective maintains that the middle class is brought into being in the very moment it is ‘measured’.Footnote 7

With regard to the political realm, this implies that arguments referring to the middle class have to be seen as performative practices which constitute the object they are making statements about (cf. Foucault 1972, 40–49). Political arguments are not simply evaluating an objectively given social group in different ways, but are constructing different middle classes. In the German discourse, the middle class is constructed as the “relatively educated” (SZ, 21 June 2005), as the “well-qualified” (FAZ, 16 January 2005; SZ, 21 June 2005), or simply referred to as “the educated middle class” (FAZ, 16 January 2005; SZ, 13 August 2007). Elsewhere, the middle class is named “the working middle class” (e.g. FAZ, 5 May 2013; SZ, 6 May 2014), and its members are described as the “high achievers” (Leistungsträger) of society (e.g. SZ, 11 May 2009; FAZ, 10 January 2011) who are committed to “industrious, hard work” (FAZ, 3 May 2013). Middle class members are said to be “state-supportive” (FAZ, 11 September 2007; SZ, 7 May 2008) in that they “abide by the rules and pay taxes” (FAZ, 23 March 2010) and thus contribute “the lion’s share of what can be distributed” (FAZ, 29 December 2007). It is particularly this latter notion of the ‘hard-working’ and ‘rule-abiding’ which also features prominently in all of our interviews with policy-makers. Finally, the middle class is associated with a certain way of life, with its members defined as “the acting people” (FAZ, 13 November 2015) who are “autonomous” (FAZ, 25 January 2015), “self-confident”, and striving for “self-realisation” (SZ, 31 December 2005). At the same time, the current and future status of the societal middle is a constant political concern: It is said that the middle class is getting “more and more nervous” (SZ, 10 May 2008) due to large-scale processes of globalisation, individualisation, and privatisation, and that it is “increasingly afraid of changes and losses” (interview #3) and of “social relegation” (interview #1).

These different ways of constructing the middle class are not openly competing with one another, and they are not an indication for the fact that target group evaluations are in a phase of transition (cf. Schneider and Ingram 1997, 145–149). Rather, the different middle classes exist side by side in discourse, and each of them provides specific connecting points for claiming certain policy designs. In the case of Germany, two configurations stand out in particular: one related to tax policy and another related to systems of social security. The claim related to tax policy speaks to the notion of the middle class as the ‘hard-working’ and ‘autonomous’ core of society which secures the functioning of the welfare state. The Liberal Democrats for instance regularly long for a “tax relief for the middle class” in order to re-establish “justice of performance” (Leistungsgerechtigkeit) in German society (SZ, 17 August 2010). A similar claim is made by a Christian Democrat who argues that it is particularly problematic that “the middle class is bossed around by German tax law”, as “the greatest part of wealth” is generated by this social stratum (SZ, 7 May 2008). The bourgeois parties should not, the argument continues, “narrow their view to those depending on the welfare state”, but should focus on the “diligent members of the hard-working middle class” (SZ, 10 May 2008). In our interviews, even the representative of the Left Party evokes this interpretation by insisting that “when it is about the middle class, the Left is a tax reduction party” (interview #5).

A second prominent configuration of middle class construction and policy claims settles on the notion of the ‘endangered’ and ‘scared’ middle class. This argument articulates the middle class not primarily as the hard-working people who deserve a fair share of what they have earned, but rather as an increasingly threatened species which needs careful treatment or else it will be ‘the underclass of tomorrow’. This in turn renders plausible claims for an adjustment of the systems of social security in different fields of welfare policy. For instance, the head of the Federation of German Trade Unions argues that there is a need for an “interim payment” between short-term and long-term unemployment benefits, as the risk increases that “people from the middle of society slide into poverty” (SZ, 1 October 2009). A similar argument is made with regard to the current pension scheme which in the long run will lead to old age poverty, and “not only for fringe groups but also for the middle class” (SZ, 20 June 2012). As an interviewee told us, in pension insurance “the middle class is increasingly merging with the underclass” (interview #5). Although these claims likewise refer to the middle class as the main target group of welfare policies, they imagine a considerably different middle class which obtains its plausibility from a vital sociological literature identifying an increasing precarisation of this social stratum (Heinze 2011; Mau 2012; Koppetsch 2013). The tax relief argument could of course also be justified by way of relating it to the diagnosis of a shrinking middle class, though it does not evoke the notion of a fragile middle class which is constantly under threat of social relegation.

These examples demonstrate how policy claims are justified by way of relating them to specific components of the discursively constructed knowledge about what the middle class is, wants, needs, and deserves. The two configurations are not openly competing with one another, although the two stories differ regarding who is assumed to belong to the middle class (and one could assume that it is particularly this indeterminacy of the category that explains its symbolic power; see Sect. 6). Moreover, references to ‘the middle class’ might also be seen as a conscious strategy (cf. Sidney 2005, 115) in order to back unnamed interests or policy preferences, and our interview material suggests that political actors are well aware that the middle class could be a powerful source in political deliberation. In this sense, the middle class does not only function as a target of a specific policy, but more fundamentally as a symbolic source for legitimising overarching political objectives. Whatever the rationale for evoking the middle class, the differences of the argumentative configurations reveal that the social construction of a target group goes beyond the SCF’s dichotomy of positive versus negative evaluation and the resulting degree of ‘deservingness’. In the examples above, all types of middle classes are depicted as deserving, and thus the explanatory power of this category alone is quite limited. By starting with the claim that group boundaries reflect ‘objective conditions’ and ‘real distinctions’, the SCF in its programmatic form misses the fact that the middle class is not only evaluated but constructed specifically, as the ‘real distinctions’ themselves are contingent and, at times, a matter of political controversy. Thus, the analysis cannot start with presuming the existence of a target group, but has to reconstruct how it is brought into existence through practices of differentiation. In this regard, an interpretive perspective can reveal that policy claims very much depend on how the boundaries of the middle class are drawn in the first place.

From targeting groups to constructing society: the middle class and its outside

As explicated above, the construction of a (collective) identity always involves the construction of an outside from which the entity at hand is differentiated from. For instance, when referring to the ‘deserving’, this simultaneously constructs the realm of the ‘undeserving’ which is, either explicitly or implicitly, pictured as something different or even opposite. While this logic is a constitutive feature of any practice of meaning-making, with regard to the discursive construction of the middle class there are two specific features which go beyond this general logic of difference.

The first specificity stems from the logic of the middle category itself, i.e. its relational character: it only makes sense to speak of the ‘middle’ (class, of society, ground of politics, etc.) in as much as it is assumed that there is also a ‘below’ and a ‘above’, a ‘left’ and a ‘right’, a ‘margin’ or a ‘periphery’. The ‘middle’ cannot exist on its own as it implies an ‘in-between’ or ‘in the centre of’, and thus the term constructs also that which it is not. While one can say that any meaningful category articulates a difference by evoking its counterpart, the notion of the middle goes further than this, as the difference the category builds on could not be overcome without eliminating the meaning of the category altogether: whereas one can at least hypothetically imagine a situation in which everyone in society is ‘equally active’, it is not plausible to think of a society in which everyone belongs to the middle, as then there would no longer be a space in-between to occupy. One should thus pay close attention to how imagining the middle class as the realm of the educated, hard-working, and active simultaneously constructs both a group which is excessively rich, irresponsible, and detached from ‘normal’ society, and a group which is uneducated, lethargic, and dependent (van Dyk 2018): “in comparison to the under and upper classes, the middle class’ constitutive feature is its self-responsibility” (interview #4).

With regard to policy design, this inevitable co-construction of the middle class and its outside has two implications. Firstly, the (implicit) imaginations of those ‘below’ and ‘above’ can be used for constructing policies directed to these non-middle spheres—without the need to explicitly justify why they have to be addressed in a specific way. For instance, both the Christian Democrats and the Free Democrats claim that the possibilities for “social upward mobility” have to be strengthened by ensuring that “performance is being rewarded again” (e.g. FAZ, 25 February 2009, 10 November 2009). Similarly, a Social Democrat argues that in child care facilities a sense of “performance motivation” has to be taught so that “children from socially deprived backgrounds have the chance and the will to become middle class” (FAZ, 8 October 2006). In these instances, the middle class is not the target group of policies, but with its implicit characteristics nevertheless functions as a legitimisation for specifically addressing ‘the others’ (cf. Schram 2012). The second, and intrinsically linked, implication is that policies which speak to ‘the underclass’ also re-establish a specific imagination of the middle class. When policies are geared towards strengthening ‘performance motivation’ in order to allow ‘social upward mobility’, they do not only imply that the addressed target group is currently ‘inactive’ and ‘lazy’. They also articulate the belief, as rising the social ladder from below by definition means reaching the middle, that the middle class already is ‘active’ and ‘performance-oriented’. The relational character described above takes effect again: as the notion of ‘below’ implies that there is something above which can (and should) be reached, talk about the ‘underclass’ does not need to explicitly mention the middle class for confirming the latter’s status as a normative point of orientation.

The second specificity of the middle category is the fact that it brings the different spheres into a normative order: the middle is by definition not only one entity beside others, but the (mediating, balancing) centre of the entirety of entities. This notion goes back at least to Aristotelian thinking and its idea of the middle as the core of society (Münkler 2010) which is cultivated for instance in the German saying ‘the middle and the moderate’ (Mitte und Maß). The idea that ‘the middle’ is most valuable is deeply anchored particularly in Germany which has a long sociological tradition of describing society as a “levelled middle-class society” (Schelsky 1953), and this imaginative still haunts contemporary social theory (e.g. Reckwitz 2017). One could assume that the normative appeal of the middle also features in ‘third way’ discourses which claim to occupy a ‘middle ground’ between supposedly outdated oppositions (cf. Laclau and Mouffe 2001, xiv–xv; Bastow and Martin 2003). It comes as no surprise then that in public discourse the middle class is closely related to the common good, so that its status is essential for society as a whole. Based on a metaphorically rich language, the middle class is pictured as the “heart” (SZ, 20 November 2006), the “backbone” (SZ, 5 January 2010), the “mainstay” (FAZ, 26 September 2005), the “fundament” (interview #3), or the “main pillar” (interview #2) of society and the welfare state. Against this background, the attributes that are ascribed to the middle class, like a certain work ethos, autonomy, self-dependency, etc., do not have to be explicitly named as universal values for society as a whole—they become universal automatically as soon as they are described as features of the ‘middle’.

This logic has severe implications for social and political inequalities. Since constructions of the middle class simultaneously articulate what the ‘healthy core’ of society is, the middle class discourse constitutes an image of the whole of society and how its different parts relate to one another. According to a widely told story, middle class members are said to have earned their societal position as they are hard-working, motivated, and following the rules. In turn, this implies that those not belonging to the middle of society could be accused of simply not being committed to these values (cf. Schram 2012, 242), what in turn legitimises inequalities: “when the middle class and its principle of merit are ignored, then one paves the way for late Roman decadence”, a former chairman of the Free Democrats has warned in a widely noticed statement (SZ, 12 February 2010). These implications do not only result from a, in the words of the SCF, ‘positive’ construction of the middle class. In the German discourse, the middle class certainly is pictured ‘positively’, but its hegemonic position for describing and evaluating social differences derives just as much from the abovementioned relational logic of the middle category and how it is entangled with its constitutive outside. To understand how the middle class and society as a whole co-emerge in relation to one another and how these dynamics result in a common knowledge about ‘legitimate’ or ‘normal’ levels of inequality, which is a central concern of Schneider and Ingram’s work (see particularly Schneider and Ingram 1997), one has to go beyond investigating the linkage of a certain policy with an explicitly addressed target group. This holds also true for questions of policy design as the construction of the middle class simultaneously enables politics to address other groups in a certain way, and by addressing these others, the middle class is confirmed as the normatively desired social space, and so on and so forth. In this sense, the middle class discourse is not only about constructing a target group for a certain policy, but also, and more fundamentally, about establishing a certain idea of society. These facets can only be grasped from a strict constructivist perspective which goes beyond the analytical components offered by the SCF.

From resources to discourse (and back): the power of the middle class

According to a post-foundationalist conception, power is not only a matter of resources that could be invested to uphold specific interests (Schneider et al. 2014, 109–110), but more fundamentally a complex societal constellation that emerges from specific orders of knowledge and discourse. Seen from this perspective, there are two central aspects relating to the power of the middle class. Firstly, the notion of the middle class is surrounded by an aura of normality (Link 1991; van Dyk 2018). This feature results from the symbolic quality of the middle category as described above (in Sect. 5). The middle does not only build on an outside, but simultaneously puts itself in the centre of the imagined totality. As the middle is, at least in Western thinking, associated with balance, moderateness, and compromise as compared to the marginal, biased, or extreme (Münkler 2010), the use of the middle as a classificatory principle inevitably conveys a message about what is (not) normal and desirable. As regards the German middle class discourse, this is indicated by the regular appearance of phrasings like “the perfectly normal middle of society” (FAZ, 3 September 2005), “the decent middle class society” (FAZ, 18 November 2015), “the ordinary citizens from the middle class” (SZ, 1 December 2015), or “the normal middle class families” (SZ, 10 October 2009). Moreover, the middle is also the social place the vast majority of individuals identify with: “if respondents are given a choice between low, middle, and upper class, 90 percent identify themselves as middle class” (Stone 2012, 191). The middle class discourse is thus a prime example for how discourses construct the seemingly undisputable realm of the ‘normal’ which can, with regard to the power it exercises, hardly be overestimated (cf. Foucault 1981).

Secondly, the middle class is specifically entangled with scientific knowledge. In this regard, social structure analysis plays a crucial role. As a ‘hard social science’, it is endowed with the ability to produce unambiguous numbers that picture both the current composition of society and changes over time. In the case of Germany, this statistical knowledge regularly finds its way into larger public and political debates. Particularly visible are the income statistics the German Institute for Economic Research (DIW) publishes periodically. These do not only estimate the size of the middle income bracket and whether it increased or decreased over time, but they are also utilised, both by journalists and by DIW researchers themselves, to argue, for instance, that there is a “growing gulf between rich and poor” which “particularly causes fears within the middle class” (SZ, 15 June 2010), or that “a new downwards mobility” emerged which causes “the feeling among middle class members that they have to move faster and faster in order to secure their precarious status” (SZ, 4 July 2013). Thus, quantitative findings of income distribution are translated into alarming assessments about how the middle class and society (should) feel. The point here is not to question the validity of social structure analysis, but to highlight that it is performative: On the one hand, income statistics do not only picture what is happening ‘out there’, but they make ‘reality’ comprehensible by choosing certain definitions over others, by translating these definitions into measurable indicators, and by defining certain thresholds. These practices of complexity reduction are ultimately contingent, as defining the middle class by means of income is only one of several possibilities (see Sect. 4). On the other hand, these quantitative assessments are used to draw conclusions as regards the relevance of changes in numbers for the well-being and mood of society. As these normative interpretations are interleaved with quantitative findings which convey a notion of ‘objectivity’ and ‘neutrality’ (cf. Hacking 1999, 143–145), they tend to appear, as several statements of our interviewees suggest, as ‘secured knowledge’ that goes beyond ‘mere opinion’.

Taken together, these two aspects illuminate the power of the middle class. It is seen as the ‘normal’ societal realm and as the vital core of society, and at the same time there is a scientifically authorised knowledge attributed with the ability to capture the current status, feelings, and desires of the middle class in a seemingly unambiguous manner. Thus, the middle class discourse could be seen as a manifestation of Foucault’s (1990, 99) notion of “relations of power-knowledge”: on the one hand, population statistics do not simply gather knowledge about a society that exists ‘out there’, but they construct society through the production of a specific knowledge (cf. Foucault 2009, 273–275; Barlösius 2007), while on the other hand, this knowledge is particularly powerful as it represents an infallible indicator of the well-being of society and the welfare state. Sociological research also plays its role here, for instance when it observes and classifies societal dynamics as ‘normal’ or ‘deviant’ against the background of the idea of a society centred around a middle class (cf. Heinze 2011; Mau 2012; Koppetsch 2013; Reckwitz 2017).

As for the political realm, this constellation helps understanding why the middle class is regularly positioned as the main target group of (welfare) policies. Inasmuch as the middle class is interpreted as the core of society, it is only ‘logical’ that politics, in one way or another, has to address the issues identified ‘scientifically’: “the SPD party executive passed a concept which includes political conclusions stemming from a DIW study on the problems of the middle class” (FAZ, 29 January 2013). The power effects inscribed into and emerging from the middle class discourse are also indicated by the fact that the claim that the middle class should be the major concern of politics appears as a self-evident statement not in need of an explicit justification: “politics needs to be about the middle of society”, a representative of the Christian Democrats contends (FAZ, 26 January 2015). Or as a Social Democrat puts it plainly: “politics should not focus on the needs of a single social class but has to be a broad reform movement, in the centre of which stands the modern, social middle class” (SZ, 28 February 2005). Similar all-embracing claims are made by our interviewees, for instance when it is argued that “politics in general has to be done for the middle of society and not for the margins” (interview #2). In these instances, the middle class category no longer designates a particular social group, but rather, in the sense of an ‘empty signifier’ (cf. Laclau 1996, 36–46), the ‘common good’, resulting in the conviction that caring for the middle class lies in the interest of all: “prosperity for all”, a Christian Democrat maintains, “could only be realised in a middle class society” (FAZ, 25 November 2006).

To be sure, this does not rule out political controversies about how exactly the middle class should be addressed by welfare policies (see Sect. 4). However, these controversies unfold on the common belief that the middle class is decisive for the stability of society and the welfare state, and thus voices which fundamentally criticise a middle class bias in German politics and society are rare. There are only a few instances in the media in which the “privileges” (FAZ, 14 September 2005) or “subsidies” (SZ, 20 October 2015) of the middle class are being criticised, but these statements do not fundamentally question the idea that a prosperous middle class is a sign of success and thus generally desirable. It is here where the relation of power/knowledge has an effect on what can be said (cf. Foucault 1972, 28), i.e. what appears as a ‘plausible’ and ‘legitimate’ political argument and what is disqualified as ‘irrational’ or even ‘false’. However, this does not mean that welfare policies are necessarily designed in favour of the middle class. While it is assumed that middle class members, as compared to those ‘below’ and ‘above’, possess of resources “they have worked hard for” (FAZ, 25 January 2015) and that it is thus “a matter of fairness” (SZ, 6 February 2012) that the welfare state is particularly attentive to its high-performing members, it is also a constitutive part of the story that the middle class bears the main costs of the welfare state. For the middle class to occupy its normative function in discourse, middle class politics is thus a continuing balancing act in which a ‘moderate’ and ‘reasonable’ middle ground has to be found. It is this ambivalent configuration which puts the middle class at centre stage of welfare controversies as the welfare state’s stability is seen as being dependent on both the material contributions and the material well-being of the middle class.

In sum, a post-fundationalist perspective is able to highlight how the middle class discourse realises and reproduces a specific power structure on which political processes unfold. In contrast, the SCF’s focus on “conventional measures of political power” (Ingram and Schneider 2015, 261), like the size of a group, its material resources, or its ability for collective action, tends to miss the fact that these aspects are related to underlying orders of knowledge which provide them with meaning. For instance, the size of a group as such does not bear a specific meaning, but obtains its meaning from relations of power/knowledge operating at the level of discourse, with size in the case of the middle class representing an indicator for the general health of society. This is not to say that asking for ‘conventional measures’ is irrelevant for or inconsistent with a constructivist ontology, but rather that this question has to be entangled with an interpretive reconstruction of orders of discourse. Thus, a post-fundationalist perspective does not necessarily come to a radically different conclusion regarding the power of a target group, but to a more nuanced one which highlights the mutual constitution and stabilisation of knowledge and power. Interestingly, Schneider and Ingram (1997, 150–188) have also, at least in their early work, extensively reflected on how knowledge is socially constructed and how it enables certain political rationales to become dominant. However, the SCF in its paradigmatic form treats the dimension of power as something opposed to ‘social constructions’, and thus their interrelation fades into the background.

Conclusions

This article started with the observation that during the last decades the extensive work of Schneider and Ingram on the social construction of target populations has gradually been transformed into the so-called Social Construction Framework (SCF). Within the ‘mainstream’ of policy research, the SCF has been positioned as one of the competing middle-range theories, and it has been aligned to an overarching set of positivist principles. This article problematized this process from a strictly constructivist point of view by highlighting three theoretical shortcomings of the SCF and by developing a discourse-theoretical alternative for each of them.

One could object that the SCF and the alternative developed here are simply different perspectives which make up the conceptual plurality of policy research. There is certainly some truth in this argument, not least as the SCF does not have a monopoly on the notion of constructivism. In this sense, there is nothing to be said against the existence of the SCF, and the purpose of this article has not been to question its analytical value per se. However, by incorporating constructivism into the repertoire of middle-range theories, the ‘mainstream’ of policy research actually obscures the plurality of ontological and epistemological commitments, what in turn undermines the recognition of the distinctiveness of strict constructivist and interpretivist research. This feedback effect is illustrated by the explanation why in the current fourth edition of the already mentioned handbook Theories of the Policy Process the SCF has not been included again: “this particular theory is supported by a less vibrant and active research community without much advancement in knowledge beyond the ideas in its original publication” (Weible 2018, viii). While it is debatable whether there really is no ‘active research community’ (the literature review from Pierce et al. 2014 suggests otherwise), the more important point here is the claim that there is a lack of ‘advancement in knowledge’, as this argument implies that the value of constructivism is to be judged on the basis of its ability to generate ever more “testable propositions” (Pierce et al. 2014, 2). This position ignores that constructivism features specific scientific rationales, like giving detailed and context-specific explanations of political dynamics in specific social settings (cf. Schwartz-Shea and Yanow 2012, 113), instead of testing causal hypotheses and developing generalised models.

Against this background, the purpose of this article was not only to present a strict constructivist reading of Schneider and Ingram’s arguments, but also to reclaim the notion of constructivism. This is not only about semantics, but about acknowledging the value of different ontological, epistemological, and methodological perspectives. It is thus to be welcomed that Cairney and Heikkila (2014, 377–378), in their comparative discussion of policy process theories, use the notion of framing instead of ‘social construction’ when describing how the SCF approaches the relevance of target groups. While the concept of framing is heterogeneous and also anchored in interpretive perspectives (cf. van Hulst and Yanow 2016), it does not by itself evoke a constructivist ontology, and therefore, it seems better suited to characterise positivist investigations of the relationship between group stereotypes and policy design choices. In this sense, we should take a theory’s identification with constructivism seriously, what ultimately calls, as Ingram and Schneider themselves have pointed out, for “interpretative methodologies” (Ingram and Schneider 2008, 184) and “interpretive research methods” (Schneider and Sidney 2009, 115).

Moreover, a further aim of this article was to demonstrate the specific analytical value of a strict constructivist and interpretive perspective on the relationship of target groups and policy designs. To do so, we presented an empirical analysis of the discursive construction of the ‘middle class’ in German welfare policy. Our analysis revealed that, firstly, constructing the middle class is not only a matter of positive or negative evaluation, but more fundamentally an act of boundary-drawing. We showed that there are different features referred to in discourse when the middle class is defined, and each of them provides distinct connecting points for designing and legitimising welfare policies. Thus, evaluations (positive versus negative) are not sufficient to understand the emergence of specific policy designs. Secondly, our analysis demonstrated that the construction of a target group is always accompanied by the constitution of a whole system of groups which are (implicitly) brought into an order. In this regard, the middle class is positioned as the ‘core’ of society, which in turn legitimises a certain political treatment not only of the middle class, but also of ‘the others’ as the counterparts of the ‘active’ and ‘autonomous’. Research needs to take these complex symbolic configurations into account in order not to risk reproducing the hierarchy inscribed into the relationship between the middle and its outside. Thirdly, we adopted an anti-foundationalist conceptualisation of power which is not restricted to an analysis of the distribution of resources, but which sheds light on the entanglement of social groups with relations of power/knowledge. This perspective indicates that the discursive omnipresence of the middle class and the obsession with its status are linked to the (German) conviction that the middle represents ‘the normal’ and ‘the common good’. Moreover, there is an elaborated system of socio-structural knowledge which is assumed to be able to capture the well-being of the middle class, and thus of society as such, in an objective manner, which ultimately stabilises the idea that political problems can be identified by an unambiguous statistical knowledge about the middle of society.

To abstract from these findings, we can infer two overarching aspects strict constructivist investigations of target groups and policy design should take into account. On the one hand, it should focus on practices of categorisation, i.e. carefully investigate how boundaries within the social are drawn, how they are provided with a notion of validity and adequacy (e.g. by way of ‘science’), and how they are arranged into a complex configuration of multiple groups that constitute ‘society’. In this regard, evaluations of social groups are also important; however, the value of a group is not solely dependent on explicit evaluations, but also, and more fundamentally, on the way group boundaries are drawn and arranged with other boundaries. On the other hand, a strict constructivist perspective should not discern ‘social construction’ as a neatly separated facet of political practices which simply adds up to ‘objective’ facets, as the SCF suggests when it differentiates group constructions from power resources. As a consequence, a strict constructivist perspective needs to refrain from the twofold matrix of the SCF and the “propositions” (Schneider et al. 2014, 109–131) derived from it. Again, the point is not to question the analytical value of the matrix as such, but to point out that it is too reductionist for a strict constructivist perspective which needs to start well before group evaluations are made. These meta-theoretical differences should not be levelled out by constructing an allegedly coherent set of policy theories, as these differences constitute the plurality of policy research.