Resistances to scientific knowledge production of comparative measurements of dropout and completion in European higher education

Abstract

The article is a critical sociological analysis of current transnational practices on creating comparable measurements of dropout and completion in higher education and the consequences for the conditions of scientific knowledge production on the topic. The analysis revolves around questions of epistemological, methodological and symbolic types and, in particular, how the social processes in the creation and use of different measures offer researchers different positions in the knowledge production. Descriptions of statistics and measurements from statistics agencies in Sweden and the UK and from the OECD, EUROSTAT and Eurydice, as well as policy texts and data collection manuals from European Union bodies, have been compared and analysed. Particular interest is directed towards examples of measures used in Sweden and the UK. The results suggest that available data on student completion offers only a very limited basis for research-driven comparative analysis. It offers also a problematisation of the notions of researchers seen as users or producers of data and different position takings in statistical reasoning in using statistics as for example different types of evidence for policymaking.

Introduction

This article is about discovering road bumps, pitfalls and dead ends and has its origin in a comparative research endeavour, which became increasingly difficult to pursue. As a researcher with an interest in the boundary areas of students’ irregular pathways within higher education (HE) there is a more or less natural quest to compare different ways of using university education in different settings. The comparative approach is a promising method and it seemed a reasonable beginning to try to use international student completion statistics. However, comparing a very large number of rather disparate studies on the topic and domestic statistics on student completion in different national contexts raised more questions than it answered.

There have been good examples of international comparative research, such as the European Union (EU) funded project ‘Access and Retention: Experiences of Non-traditional Learners in HE (RANLHE)’, which was mainly a project using a qualitative approach, but nonetheless published a report on the state of the quantitative aspects of access and retention. However, it made rather disappointing reading as the authors concluded that there were severe obstacles in comparing data and the data they relied on was by then out of date (RANLHE, 2008). The same disappointing message about comparability was formulated by the EU project HEDOCE in 2015 (HEDOCE, 2015) and by Quinn in 2013. She reported on an extensive European comparative analysis on dropouts in which profound and multiple difficulties in comparing different countries’ completion and dropout rates was discussed (Quinn, 2013). What started as an attempt to compare statistics and research on student completion transformed into an analysis of the measures and the question of availability, quality and reusability of existing data and its manner of production and value for research. Ultimately, to question the role of the researcher in quantitative data production and in scientific knowledge production on the topic became a viable way to start analysing it.

Aim of the article

The aim of the article is to analyse and discuss the striving for comparative measures of student dropout and completion in European HE policy and the consequences of this for the ways that scientific knowledge is produced and how the social processes in the creation and use of different measures offer researchers different positions in the current knowledge production. The analysis theme ‘Resistances’ is associated with the critical sociological perspective, which further enables symbolic, epistemological and methodological questioning of the knowledge production of dropout and student completion.

The European Union and the striving for comparative measures

In order to analyse different positions in the social process of knowledge production on the international comparability of student dropout and completion in HE, one has to take into account the context of educational policy and practice. Of great importance are the EU and the EU policies and practices that work towards a unitary European Higher Education Area (EHEA) through the Bologna process, which is considered as a top-down politically driven process (Trow, 2006) that lies at the core of European policy for education: harmonisation, Europeanisation and globalisation (De Wit, 2006). The vision of EHEA is fuelling a broad range of practices in developing new ways of measuring the success of policy interventions made by different EU bodies (see the Appendix – European Commission, 2012). The new educational governance contextualised within the processes of EU-ification implies that education is mainly meant to produce competitive human capital (Daun, 2011), which is in line with the EU’s goal to become the world’s most competitive knowledge economy with competitiveness as a master discourse (Gornitzka, 2006). By pushing forward education and science as a ‘silver bullet in Europe’s quest for “economic sustainability” the higher education policy was reinvented and the higher education area is now a key priority’ (Serrano-Velarde, 2015).

Hence, there is no doubt that EU policy affects the national HE agendas in different ways, especially if the EU should be considered as a new supra-bureaucratic field in Bourdieu’s notion, for example, a field of Eurocracy, which some studies suggests (Georgakakis and Rowell, 2013), or a neo-medieval empire (Zielonka, 2006), while others, also with inspiration from Bourdieu, suggest the emergence of a global education policy field (Rawolle and Lingard, 2008). Recent studies have contributed to the understanding of the governance relation between EU and domestic bodies (e.g. Gornitzka, 2006; Gornitzka and Sverdrup, 2008; Serrano-Velarde, 2015; Trondal, 2010), which demonstrates that processes of routinisation are apparent in the multilevel interaction between EU bodies and the domestic sector ministries and subordinate agencies (Trondal, 2010). Other studies point out the combined effects of data production and transnational networking and pertinent constant comparisons between educational systems, which are open and receptive to different degrees, as a sort of new form of governance in the Europeanisation processes in education (Grek et al., 2009).

Nonetheless, it is not important at this stage of the argument to prove whether the policy and context originate from a social field in a Bourdieuan sense or not, but just to conclude on the multilevel and complex character of the context of the object in study – the striving for comparative measures of student progress in European HE and its consequences for knowledge production and knowledge producers on the topic.

Again, the strategic framework ‘Education and Training 2020’ (ET 2020; see the Appendix – European Commission, 2009) states that the principles behind achieving needed to achieve the common strategic objectives and selected benchmarks in prioritised areas, such as student completion, which is one of the goals. The headline targets, a 40% completion rate of tertiary level education among 30–34-year-olds and that the proportion of early leavers from education and training should be under 10% by 2020, were formulated in the report ‘Rethinking education: Investing in skills for better socio-economic outcomes’ (see the Appendix – European Commission, 2012).

Furthermore, in the key area ‘modernising higher education’, where the Bologna process is the main vision, the Bucharest Communiqué stated in 2012 that further efforts in enhancing the ‘social dimension’ would be to ‘strengthen policies of widening access and raising completion rates’ (see the Appendix – European Commission/EACEA/Eurydice, 2015: 25). In 2013 the EU council agreed upon the social dimension of HE, and saw it as one of the things to strive for, which also included securing more equitable access to, participation in and completion of HE. Furthermore, they invited

…the member states, with due to regard for subsidiarity and the autonomy of higher education institutions to…adopt national objectives which are aimed at increasing the access, participation and completion rates of underrepresented and disadvantaged groups in higher education, with a view to progressing towards the Bologna process goal that the student body entering, participating in and completing higher education at all levels should reflect the diversity of Member States’ populations. ¹

Hence, there are goals that on the one hand promote the unity of the EHEA and on the other call for convergence in national objectives on progress in these issues, despite careful use of the word ‘invites’. Moreover, earlier studies conclude that there are two fundamental competing ideas in the EU’s institutional architecture: either supranational rule and national sovereignty (Heidbreder, 2013) or the trend towards convergence in the educational area – this despite the differences in domestic educational systems, the differences in both their way of organising them and in their ‘cultural homogeneity-heterogeneity, religious demographics, type of state, political party constellations, type of response to and interaction with global processes’ (Daun, 2011).

Regarding information and progress of the interventions and mechanisms that aim to increase student completion, Eurydice is a key agent. In 2014 they concluded that there were vast differences between countries in how dropout rates and completion incentives were associated with funding mechanisms. In most European HE systems there are financial incentives for students to complete their studies in time, for example a restricted student financial support associated with insufficient level of study progress, as in Sweden. The HE system in Sweden also has incentives for HE institutions to enhance students’ completion rates, that is, a combined funding formula and a performance-based mechanism, which is common for the Nordic countries, while half of the European countries do not use either one of them. In the other half of Europe there are countries that do not have any institution funding connected to completion or dropout rates at all. Even if most countries in Europe have a general policy on widening participation and general attainment targets, many countries do not measure policy output in terms of completion rates or dropouts or for different target groups, such as underrepresented social groups (see the Appendix – European Commission/EACEA/Eurydice, 2014). This is, of course, problematic primarily in the attempt to create solid knowledge of the state of the emerging EHEA because it would demand severe reforms in the domestic funding design of HE, which is of course rather problematic since the EU uses soft policy pressure. On the other hand, the ‘normative pressure [being] placed on countries to look good and fear embarrassment’ when the benchmarks are not achieved (Gornitzka, 2006: 46) is clearly an instrument being used in the Eurydice report (see the Appendix – European Commission/EACEA/Eurydice, 2014).

Returning to the EU council’s meeting in 2013 and the document ‘Council conclusions on the social dimension of higher education’, the invitations to the member states to develop national objectives are clearly directed to a widening participation theme with focus on equal opportunities for underrepresented groups to access, complete and become employed. ² Regarding the content and processes of the national objectives, it mentions cooperation between educational providers at all levels, including those that provide non-formal and informal learning, ensuring widening participation also within the teacher profession at all levels, developing more outreach activities and guidance and permeability in completion routes from non-formal education and identifying underrepresented groups at all levels of education. It also mentions flexible learning, such as part-time studies, distance education, the financial aspects of supporting access and completion among underrepresented groups, as well as post entry support and quality of teaching. The notion of underrepresentation of groups is also expanded by disparities between regions within the member states. Finally, there is the question of data. The member states are invited to ‘engage in the systematic collection of relevant comparable data - while making optimum use of existing resources - in order to enhance the evidence base for policy development and to enable the effective monitoring of national objectives on access, participation and completion rates among underrepresented and disadvantaged groups in HE’. In the document the Council also ‘welcomes the commission’s intentions’ to achieve the following:

• embark on a mapping study of policies on access and dropout and completion rates in HE with a view to analysing the effectiveness of different national and institutional approaches and how structural, institutional, personal, socio-cultural and socioeconomic factors influence dropout and completion;

• pursue joint work with Eurostat on a feasibility study to improve the methodology for collecting administrative data on the duration of studies and completion rates in HE;

• develop a study on the influence of different models of funding, or cost-sharing, on the effectiveness, efficiency and equity of HE in line with commitments in the 2011 agenda for the modernisation of HE. ³

Clearly, the objectives are related to the creation of an evidence-based policy in the area of widening participation, and access, dropout and completion in particular. The question is what kind of ‘studies’ they mean and what kind of evidence is needed when creating evidence-based policy.

There are some crucial aspects to bear in mind regarding a quest for evidence-based knowledge production. This is not primarily about promoting scientific knowledge; this is about promoting aggregated statistics to monitor the development in line with certain goals from an administrative perspective. It means that there are some important aspects to discuss from a critical sociological perspective. One aspect is connected to the dominant data production, that is, who is producing data and for what purpose? Another aspect is the quality of data, that is, the advantages and disadvantages of aggregated data and administrative data. A third aspect is about the nature and necessities of comparative research and the fourth and final aspect is about the epistemology of the knowledge production in this area. These aspects are discussed within the following themes in the article – symbolic resistance, epistemological resistance and methodological resistance – in the meaning of resistances in the scientific knowledge production on student completion.

Notes on theory and methodology

This article is based in critical sociology and uses theoretical and methodological tools mainly from Pierre Bourdieu’s (1989) perspective on social space and symbolic power and draws on Alain Desrosières’ (1998) work on controversies in statistical reasoning. In Bourdieu’s sociology a social space is full of ‘conflicts between symbolic powers that aim at imposing the vision of legitimate divisions’. This means that, ‘symbolic power, in this sense, is a power of “world making”’ (Bourdieu, 1989: 23). When a social space is highly structured it is conceptualised as a social field, in terms of a field of symbolic forces (beliefs), which are created by formation of specific groups with different amounts of cultural, economic and social capital, following their interests to monopolise expertise in the field. The success depends on the social authority acquired in previous struggles; thus, conservatives and their descendants in the dominant positions have an advantage through the accumulated symbolic capital, for example in an institutionalisation process. A social field has its own doxa, which is constantly challenged in the struggles of legitimacy in field-specific expertise. The performative struggles are seen as classification struggles in which an economy of symbolic goods are shaped and thereby contribute to a polarisation, that is, into an orthodox sphere, where the dominant agents are positioned as preserving the ‘natural order’ and into a heterodox sphere where opposition towards the current dominant order is positioned (Bourdieu et al., 1994).

Hence, a doxic object – Student completion – is here constructed as a product of the dominant position of state bureaucracies, that is, a problem category, which the domestic HE authorities should attend to. By researching dropout, student completion and continuation, etc., at face value it means researching administrative terms in a bureaucratic state logic, for example in a national agency. It means these concepts are ready-thought, impinged with meanings from an administrative doxa (Bourdieu, 2014), hereby denoted by the term ‘doxic object’. Hence, the sociologist has to distance herself/himself from the spontaneous knowledge about the social universe in which the research object and researcher are nested (Bourdieu et al., 1991). Bourdieu et al. (1994: 1) wrote: ‘one of the major powers of the state is to produce and impose (especially through the school system) categories of thought that we spontaneously apply to all things of the social world-including the state itself’. In our everyday knowledge ‘the state’ is a natural system because we are taught that it is; it is like the ‘repository of common sense’ (Bourdieu, 1989). The state is producing symbolic goods, that is, social problems and social categories, to which state authorities through different state bodies aim to support/punish/educate, etc. If the researcher unreflectingly borrows these categories (i.e. borrows the thought of the state of ‘state problems’) and makes it into a sociological problem, she/he takes part in a reproduction and reinforcement of the symbolic order. A critical sociologist cannot do that since the mission is to ‘question all the presuppositions and preconstructions inscribed in the reality under analysis as well as in the very thoughts of the analyst’ (Bourdieu et al., 1994: 1). Thus, it is a neutral and disinterested stance that is needed, which could be obtained through a reformulation or reframing of the research object or through an epistemological break, as Bourdieu et al. wrote in 1991. As an example of such reformulation, in a study inspired by Bourdieu’s work on epistemological breaks (Bourdieu et al., 1991), the research interests of ‘dropout’ and its causes were reframed and reconstructed as one component of use or non-use of HE in a social space of educational strategies. Rather than solely focusing on how student characteristics were associated with high dropout rates, which is a concern for policy makers and administrative authorities, a much broader scope was adapted in order to understand how students were using HE or not and how cultural and social resources were at play in these use patterns. The dropout pattern was analyzed as a non-use of HE in relation to other educational strategies of students’ enrollment, efficiency and completion. In that way, the researcher did not uncritically apply the dropout issue framed as an administrative and state bureaucratic concern (Bourdieu, 2014). It was rather to reconstruct the social space of educational strategies in which dropout is one way for student to deal with participating in HE or not. The administrative notion that dropping out from HE is bad, is then neutralized and seen as one possible educational strategy among others, while a study on dropout rates taken at face value would have reinforced the deficit view of the phenomenon and the state administrative symbolic power in the research field of education. To explore the different ways of using HE in such research design it also contributes to shed light on the consequences of paradoxical policy views in which the educational strategies is valued and thereby deliver a profound and advanced understanding of the dropout phenomenon. The dropout pattern was analysed as a non-use of HE in relation to other educational strategies of students’ enrollment, efficiency and completion. In that way the analysis of the social space of educational strategies and its objective relations reflects consequences of paradoxical policy processes and a relatively open and flexible system at the time (Carlhed, 2016).

Returning to the striving of comparable measures and the bureaucratic state logic, one of the main tasks of a welfare state is to deliver services, and one of the services is to offer HE and accordingly also monitor the services, where an administrative terminology is created and used nationally and transnationally. Thinking with Bourdieu, reframing the research object could be done by analysing the discourse and different position takings and their social origins in short terms of who says/does what and why. A way to research the polarisation within a social space or field is to study the discourse. In Bourdieu’s terminology it is not possible to study the doxa itself, since it is silent and undisputed. The dominant agents are generally silent until the doxa is questioned, through a heterodox positioning and as responses to orthodox position takings in defence of the doxa appear (Bourdieu, 1977, 2014). The struggle can then be reconstructed and studied. In addition, by analysing what the position takings say, from which position in the social space they are pronounced and by which capital composition and weight the agents have, it is possible to reconstruct the structure of either a social space or a social field.

Hence, the point of departure for this analysis concerns the contextualisation of the common administrative terms – student completion, dropout and continuation – which have in this article been put into a widened social context, that is, the striving for comparative measures, which is set to be pursued both in the national bureaucratic fields and within an emerging transnational social space of EU policy practices for evidence-based policy reasons. However, the possibility of reconstructing a full picture of an emerging social field of European higher education is far out of the reach of this article due to its complexity, but some knowledge about the available positions for researchers in the social space of knowledge production on student completion in a European policy context would be reasonable to achieve.

Now, in reconstructing the social processes in the striving for comparable measures and thereby being able to visualise the heterodox and orthodox positions, Desrosières’ study (1998) on statistical reasoning among agents in different European countries and the USA is of very valuable assistance. His study borrows much of its basic assumptions of the social world from Bourdieu’s work. The analysis revolves around controversies of statistics as either results of conventions in the actual world or as subjects for debating. The statistics are seen as core vehicles in the knowledge production and decision-making in state construction processes, ‘combining the norms of the scientific world with those from a modern rational state, which are centred on the general interest and efficiency’ (Desrosières, 1998: 8).

A result from the analysis is a typology of statistical reasoning. Arguments with statistics are used either in the language of science or of action, in which Desrosières identifies four ways of reasoning. In the scientific language of description one attitude is that statistics (1) approach reality in the language of description and mathematical tools. It could be seen as ‘the statistical discourse’. This position is also a reference point for the other attitudes. Furthermore, in science it is (2) also possible to reconstruct social and historical geneses of the statistical objects and how mental schemes in unification and equivalence processes are shaped through social struggles, law or customs. This is the position of the constructivist sociology of knowledge. The political and administrative language of action, which derives support from scientific positions and is characterised by its normativeness and pragmatism, argues either in (3) an objective fashion, by talking about the real objects described and the action needed upon them, or (4) a relativist version that uses several modalities. It could be accusative and polemic or ideological and aims to open up the statistical objects to see what is hidden (Desrosières, 1998).

The analysis clarifies different available knowledge production positions using the typology of Desrosières and thereby ways to analyse possible consequences for the knowledge production of the student completion complex, particularly from a heterodox constructivist researcher’s perspective versus the orthodox bureaucratic state perspective. In regard to the critical theoretical perspective and following the tradition to also put science itself on the social research topic agenda (Bourdieu, 2004) this article could also be seen as a contribution to the heterodox position takings about the struggle of legitimate knowledge of student completion. Hence, the pertinent perspective in the text relies on the second position in Desrosières’ typology, the constructivist sociology of knowledge.

Empirical material

The empirical foundation in the article is based on descriptions of different kind of statistics (indicators and measures) using documents, data collection manuals, etc., from the European Statistical System (ESS) and the European Statistical System Committee (ESSC), The Swedish Higher Education Authority and Statistics Sweden, and from the Higher Education Statistics Agency (HESA) and the Higher Education Funding Council for England (HEFCE) in the UK, and statistics from OECD documents, as well as from Eurostat and Eurydice. Besides analysing definitions of indicators and how measures of student completion and dropouts have been used in both research-driven and policy-driven studies, attention has also been given to how the statistics are used in domestic and European policy discourse, that is, in policy texts from the EU/European Commission (EC), ET (2020 and the EC recommendations to national reform policy in Sweden and the UK (see list in the Appendix). For analyses of epistemological and methodological aspects, particular interest has been directed towards comparability between different indicators and measures in different countries, where examples of measures of student completion used in Sweden and the UK have been the main cases.

Symbolic resistance

By placing the search for comparative measures of student completion within a social space, that is, in an emerging space of EU policy practices, struggles of classification and of expertise are acknowledged. Hence, the symbolic resistance becomes visible for analysis. Now, whether the EU could be considered an autonomous field is yet to be researched, but some studies have come a considerable way in understanding European integration (Georgakakis and Rowell, 2013; Kauppi, 2003; Rawolle and Lingard, 2008; Zielonka, 2006). Nevertheless, the more open concept of social space could be used as a tool for analysing policy and practices in the area of interest. By using social space in a Bourdieuan sense it focuses attention on symbolic power. Earlier, the term ‘doxic object’ was introduced and its associated links to a dominant bureaucratic administrative doxa. The doxa, or rather the orthodox position within the social space of EU policy practices, could be connected to the centre of gravity, where Georgakakis (2011) suggests that the Commission is in a dominant position. Accordingly, the position takings of the Commission are agents of the dominant doxa. In the orthodox view of the EC and its aim to create evidence-based policy, there is a need for data that is aggregated and comparable across the EU. In Desrosières’ (1998) terminology, this is the position of using statistics that are real objects that need (political) action and that lend arguments from the scientific statistical discourse. By doing so, the combination of aggregation and comparison is possible, combined with a strong belief in one variable measure holding the status of ‘truth’ like the OECD-data and so on (Thomas and Hovdhaugen, 2014: 458).

How does the striving for comparable data affect the research conditions for knowledge production on HE and student completion? Currently, there are signs of acts of domination or imperialistic movements by EU bodies, such as the ESS, Eurostat and EACEA through Eurydice, in attempts to monopolise quantitative data collection. Clearly, the statistical production has its main receivers – the EU policymakers – but if researchers are reduced to users of the statistics, as could be observed in Eurostat’s documents, such as the Annual Statistical Work Programme (see the Appendix – Eurostat, 2015b), ⁴ it has implications that researchers are consumers of data and not producers. It could also impact on how the need for quantitative research data is defined, such as high-level aggregation data in terms of benchmarks (one variable) regarded as valuable, but not middle range data or data that is designed for analysis of causes and effects. Due to the orthodox discourse dominance of the transnational scope of comparable data and the EU’s own dominance in the production of data, it makes EU-produced data ‘all you need’. In this way, in research policies and funding schemes, it may also de-prioritise researchers’ opportunities to collect their own data. If so, it would connect to and reinforce a trend in policies of creating research infrastructures, that is, promoting reuse and open access of research data that has been pertinent for a time, which was initiated in the UK in the mid-1990s and reinforced by the OECD’s Declaration on Access to Research Data from Public Funding in 2004 and the OECD’s Principles and Guidelines for Access to Research Data from Public Funding in 2006 (Carlhed and Alfredsson, 2009; Gustafsson et al., 2008; Slavnic, 2011) and manifested at the EU level in 2010 (see the Appendix – High Level Expert Group on Scientific Data, 2010).

Thus, when dominant performative discourses and practices of data production align they impose legitimate forms of truths (symbolic power) that create orthodox fundamental principles of what data really are, who is the legitimate producer and who they are supposed to serve. Hence, the pertinent policy-derived belief that highly aggregated statistics from one country compared to another would be sufficient to draw conclusions on what is the state-of-the-art in a certain area is delusive and may transform the function of educational science into a policy tool rather than an independent site for scientific knowledge production. Failure to regard power relations in the knowledge production and evidence-based policy leads to ideological support for a set of power configurations instead of explaining them (Fischer, 1998; Shortall, 2012).

However, other signs that point in the same direction have been discussed in relation to the development of international studies, such as the Programme for International Student Assessment (PISA) and the Third International Mathematics and Science Study (TIMMS) (Gustafsson, 2008). From the start of the International Association for the Evaluation of Educational Achievement (IEA) in 1959 and the new organisation in 1990, the presence of researchers has declined while the policy and administrative presence have increased. The ambition has changed from conducting studies with explanatory purposes towards descriptive purposes, building a base for domestic educational policy. In 2010 when PISA was launched, the emphasis of the evaluation of educational equality in the service of educational policy became even more emphasised (OECD, 2001). While educational researchers more or less abandoned research problems involving causality in the early 1980s, economic researchers have shown interest in the international datasets, which are made available to the research community for secondary analyses. However, as Gustafsson (2008) points out, the data easily invites misuse and misinterpretation as regards possible causes and effects in spite of the fact that the international data are cross-sectional and are not designed for causal analysis. There is a risk that when educational researchers do not offer explanations that others will; if not economic researchers, then the media and politicians.

Hence, the questions of the nature of useful data, who is the legitimate producer and who they are supposed to serve are connected to the doxa in the social space of European policy practices. Thus, useful data in the eyes of policy makers is probably not equivalent to research utility; the institutionalisation of practices and statistical techniques invested in international studies by Eurostat, PISA, TIMMS, OECD Education at a glance, etc., is hard to compete with for researchers with an interest in comparative studies but with their own (heterodox) ideas of how to design robust and valid data collections for causal analysis, particularly since the OECD, UNESCO and Eurostat in 1995 joined in creating a common data bank of key areas of education. In the production of indicators and benchmarks, the OECD has been an important reference in the transfer of ‘metrological technology from one supranational organisation to another’, into the European configuration of expertise on education and in supporting the Lisbon Strategy in Education and Training in 2007 (Normand, 2010: 415). One of the EU’s core working modes, the open method of coordination, which engages expertise at the domestic HE authority level beyond the legal framework, is facilitating the institutionalisation of instruments and methods and application of benchmarks and indicators (Ertl, 2006; Normand, 2010). Other studies show how the decision-making processes within the national central administrations are ‘strongly sector-penetrated by the Commission’, engaging sector experts within sector agencies and sector ministries and thereby fostering a rift in responsibilities between the domestic government and core executives (Trondal, 2010).

Furthermore, linking the usefulness of data from a sociological perspective to an epistemological perspective is another aspect of usefulness for different positions of statistical reasoning in the striving for comparable measures that is relevant to analyse and discuss.

Epistemological resistance

Epistemology is commonly known as the theory of knowledge, especially with regard to its methods, validity and scope. A relevant question to ask is what the nature of knowledge is and how is the relationship between the knower and the would-be known. One vital aspect of the analysis is based on the epistemology of social science in contrast to the epistemology of natural science. Thus, resuming the discussion on what it means to research something at face value and the suggestion of a reasonable way to reframe administrative terms into sociological researchable units or projects, reframing or reconstruction of a research object, is what Bourdieu et al. (1991) call an epistemological break. If the researcher mimics the method of exact sciences (positivism) without reaching an exact epistemology of the social sciences, no social science is done, a statement that can be related to the constructivist sociology position in Desrosières’ (1998) thinking. The mimicking implies producing research that is detached from all connections with both place and time and therefore universally valid, exemplifying the first position in Desrosières’ (1998) typology, the scientific statistical discourse.

Hence, following the constructivist sociology stance, comparative research is most challenging and difficult, at least if we talk about comparative social science. Trow (2006) gives an example of profound differences in the structure and traditions of different domestic systems (Sweden and UK) in spite of the resemblance, which is superficial. In addition, there are different HE contexts concerning the diversity of stakeholders and the amount of co-existence of universities and a strong non-university sector (De Wit, 2006). Välimaa and Nokkala conclude that the researcher in comparative studies:

…very soon faces the need to try to explain, or even to give causal explanations to, social phenomena examined because comparative research settings normally reveal both common characteristics and differences between the cases studied, or between the units of analysis (whether individual, departmental, institutional of system). This need for explanations is soon followed by the notion that one cannot make general and generalising remarks unless one understands deeply the social phenomenon examined in its contexts. The tension between generalising arguments and contextual understanding of the cases makes comparative studies very different from single country or HEI case studies, because in comparative research settings there is a real and concrete need to find common categories and concepts to study the cases and to explain differences and similarities between them. (Välimaa and Nokkala, 2014: 425)

Hence, according to the constructivist position, it is important to recognise what kind of knowledge production is needed to give solid ground for drawing conclusions about the European HE systems and all kinds of phenomena therein. The EC’s policy framework on HE aims to use comparative studies to evaluate political interventions and advance their agenda, in line with the pragmatic position of action towards real objects in Desrosières’ (1998) typology. The question is whether the EU’s attempt to find comparative measures acknowledges this or uses a light version of the comparative ideal. Despite the declarations made in ‘Council conclusions on the social dimension of higher education’ in 2013, ⁵ it is unclear what ‘studies’ means and how ‘comparative’ is interpreted and used in the knowledge production. According to a constructivist position (Desrosières, 1998), the knowledge aims and claims made in EU-produced/commissioned studies have weaknesses due to a desire for an all embracing and comprehensive perspective. Thus, the (higher) education systems need to be seen as comparable entities, as equals, regardless of the social dynamics. It can only be done using structural functionalism as an underlying logic, where the differences in cultural and geographical contexts and the relation between various actors are neglected (Välimaa and Nokkala, 2014). A manageable solution is through the transformation of contextual information into numbers and statistics, which can offer snapshots of differences between countries, systems or regions following the statistical discourse (Desrosières, 1998). But how valid are the numbers and how can one make sense of them?

In the ‘Eurostat regional yearbook 2015’, the guidelines state that each country is responsible for the domestic data infrastructure. In relation to the headline targets in ET 2020, which were mentioned earlier, there is a note that in spite of the fact that ‘objectives have been set across the whole of the EU, they do not specifically apply at a national or a regional level.’ Indeed, each Europe 2020 benchmark has been translated into national (and sometimes regional) targets, which reflect the different situations and circumstances of each EU Member State (see the Appendix – Eurostat, 2015a). However, when looking into the EC recommendations of reform programmes in Sweden and the UK, both governments have adopted the original benchmarks of the ‘Early school leaving target’ and ‘Tertiary education target’ in their reform plans. ⁶ The figures are based on measures in the EU’s labour force survey (LFS), where ‘early leavers’ means young people 18–24 years old having no upper-secondary education. The next target involves the tertiary educational attainment and the indicator is defined as the percentage of the population aged 30–34 who have completed tertiary studies (above International Standard Classification of Education [ISCED 5]). The evaluation of progress is based on data that mainly come from the ESS, which includes a range of surveys and international cross-sectional data collections including the LFS mentioned above. ⁷ The standardisation process of international statistics is a huge and complex endeavour, while the Eurostat, the statistical office to the EU, ESS and ESSC have key roles in the processes to develop and produce comparable statistics.

So far so good with the headline targets, at least on the discourse level, that is, the words in the definition of the measure are the same for all countries and they are connected to another set of combined words in a linguistically symmetric fashion. In the pragmatic position in statistical reasoning this is unproblematic (Desrosières, 1998). However, in the eyes from the constructivist position this could contribute to policy making with an air of scientific facts, neutral and universal, and on which politicians act rationally and methodologically. It creates a cognitive space of equivalence and comparability made for administrative practice (Desrosières, 1998). However, what early school leaving means is another issue, as well as the explanations of the level of young people leaving early from school. In summary, we can obtain a satellite perspective by mapping percentages of dropout and completion in schools or HE, which gives us statistical knowledge, but can the statistics be relied upon? The following section will address methodological issues, such as validity and reliability of the comparable measures on student completion.

Methodological resistance

Now turning to the construction of measurement on student completion, the following section will discuss the variety of measurement procedures and quality aspects of the transnational comparative approach in the views of the statistical discourse position and the constructivist position (Desrosières, 1998). For example, completion and dropouts are measured in a variety of ways for a variety of purposes. Concerning dropout rates, France, Italy, Portugal, the UK and Iceland measure dropout rates at the end of the first year, while some countries also measure it every year, such as in parts of Belgium, Bulgaria, Denmark, Estonia, France, Italy, Finland and Norway. Sweden does not specifically measure dropout rates systematically (see the Appendix – European Commission/EACEA/Eurydice, 2014); instead the regular measures used in Sweden are completion rates, retention rates, time to degree and performance indicators (UKÄ, 2015). ⁸

Hence, there are profound and multiple difficulties in comparing different countries’ completion and dropout rates. Validity and reliability problems occur when the figures are based on general assumptions that students always intend to pursue a degree, or on different design of registration data procedures or the students’ opportunities to transfer, without being registered as a dropout from HE. In addition, the analyses are based on either cross-sectional or true cohort follow ups or surveys and are conducted at different periods of time between education systems, which are designed with a different amount of flexibility for the students’ pathways through the system. Some systems have fees, such as in the UK, while others do not (such as in Sweden), which also could affects students’ determination to pursue a degree. In some countries students who transfer to another programme are considered in the statistics as dropouts (Hovdhaugen, 2009). This is also true in Sweden.

Completion measures

Due to the complexity and different domestic HE contexts, the international ‘comparable’ figures are by their aggregate nature superficial and coarse and either over- or underestimated. For example, one of the few international datasets available for comparative studies is the OECD report ‘Education at a glance 2013’ (see the Appendix – OECD, 2013). However, the Swedish figures seem to be very low at around 48% completion rate. ⁹ Yet, included in these calculations are students entering single courses who may never intend to study all courses needed for a degree (freestanding courses), which is quite misleading. In fact, when excluding the freestanding courses and focusing on the programme data, the overall mean degree frequency rate in Sweden is 68% (see the Appendix – UKÄ, 2014). The consequences of comparisons like this become problematic for both positions, statistical and constructivist (Desrosières, 1998).

Furthermore, the completion rates in Swedish statistics are underestimated due to the fact that it only counts if the student has requested the award of a degree certificate. A completed education in terms of earned full credits of the programme is not enough to result in a ‘completion’ in these measurement terms (see the Appendix – UKÄ and SCB, 2013). In a national follow up of dropouts conducted in 2010, it became apparent that 10% of those who did not request the certificate had completed all their courses (see the Appendix – HSV, 2010) and a study of dropouts from the teacher programmes showed that a year after the follow up date 65% had stopped permanently, while 35% had returned to their studies or had taken temporary leave (Carlhed, 2015). Moreover, in the UK the term ‘completion’ is not widely used and it is not a direct measure of actual students completing their degree but rather projected outcomes (see the Appendix – HESA, 2015).

Another measure of completion is the synthesis made by the OECD on the tertiary ‘survival rate’ in the report ‘Education at a Glance 2007’ (see the Appendix – OECD, 2007a: 65). The survival rate at the tertiary level is defined as the proportion of new entrants to the specified level of education that successfully complete a first qualification. It remains unclear, however, if the Swedish students on the freestanding courses are included here as well, since the country-specific information is omitted. Comparing the Swedish figures on completion rates of students on Type A programmes (37.7%) and the survival rates on Type A (60%) it could be puzzling in terms of the gap, but a qualified guess would be that the students from the freestanding courses are included in the completion rates in this publication as well as ‘Education at a Glance’ in the 2013 version, but it remains unclear due to the omission of information in the manual (see the Appendix – OECD, 2007a, 2007b).

Moreover, it is clear that there are a variety of sources that the data comes from and there are variations in data collection (see also Thomas and Hovdhaugen, 2014). Some countries have chosen to not use the UOE data collection which is a joint collection of education data by the United Nations Educational, Scientific and Cultural Organisation (UNESCO) Institute for Statistics (UIS), the Organisation for Economic Cooperation and Development (OECD) and which is based on cross-sectional design and used longitudinal data based on a true cohort method instead, and some countries only have data on a part of the population, that is, Belgium and data only on the Flemish population (see the Appendix – OECD, 2007b, 2013). Other international data collections that include graduation or completion, such as UNESCO’s data collection on education, is done by the use of UNESCO Institute for Statistics (UIS) education questionnaires, in which completion means the number of graduates at certain levels and not the proportion of student completion. In UOE surveys UNESCO-UIS, the OECD and Eurostat (UOE) have, since 1993, jointly collected more detailed education statistics. They report figures on completion in terms of ‘Gross graduation ratio’, which is defined as the ‘number of graduates regardless of age in a given level or programme, expressed as a percentage of the population at the theoretical graduation age for that level or programme’. ¹⁰ In the OECD publication ‘Education at a Glance 2014’, completion rates are only mentioned in relation to upper-secondary school. In the section where tertiary education is dealt with, the relative measure of completion rate has been replaced by the absolute measure of graduation rates (number of graduates) (see the Appendix – OECD, 2014). The Eurostudent surveys contain detailed information about students’ living conditions but have no information about completion or success (Orr et al., 2011; Quinn, 2013).

Retention measures

When reviewing statistics on completion and retention there are apparent similarities between general patterns of non-completion in different countries, for example, different rates in relation to the type of programmes and their accessibility, students’ social background and their educational capital, gender, ethnicity, etc. To make sense of the data, some caution and data management skills are needed because the construction of the measures differs. For example, in comparing data on retention from statistics agencies in Sweden and the UK, an expected active registration in HE is missing in the UK for the third semester (non-continuation) (see the Appendix – HESA, 2015; UKÄ, 2015). The latest figures show that 5.7% of UK domiciled, young, full-time, first-degree entrants in 2012/2013 did not continue in HE in 2013/2014 (see the Appendix – HEFCE, 2015). In Sweden, the official measure is called retention (kvarvaro), that is, the proportion of HE entrants who are still registered in HE in their second semester (see the Appendix – UKÄ, 2014). Generally, student status is collected in October. In the autumn semester of 2011, the most recent semester for which there is statistics, there were a total of 68,500 HE entrants. Of these, 71% were registered in the following semester, which means a 29% non-continuation rate (see the Appendix – UKÄ, 2015). In spite of the large difference between UK and Swedish figures, if the Swedish statistics were constructed in the same way as in the UK, that is, one semester later, the non-continuation rate would probably be larger. Hence, immediately sets of questions about how this could be understood appear, questions that need specific fine-tuned profound comparative research to offer explanations of why these levels differ between countries or at least data that are designed for causal analysis. A recent study commissioned by the EC, which aligns to the chorus calling for more research on dropout and retention, is reported in ‘Dropout and completion in higher education in Europe among students from underrepresented groups’ in which the issue of non-existing coherent datasets on access or on completion and dropout in Europe is discussed (Quinn, 2013). Quinn uses the OECD’s report ‘Education at a Glance’ as a source of quantitative data but remains sceptical as to the value of using it. The report also mentions and supports the results from the RANLHE project, which was a large EU-funded research project that had already concluded in 2008 that it was ‘difficult, if not impossible, to make a meaningful comparative analysis of retention statistics gathered in different EU countries for several reasons’ (RANLHE, 2008: 1).

In Eurydice’s report from 2015 on the implementation of the Bologna process, the attainment of tertiary education is shown as odds ratios and is analysed in relation to parental educational levels and native- and foreign-born students, but the definition of attainment is not explicit. In Desrosières’ (1998) terminology, this is typical of the pragmatic position, which aims to act on the statistics as if they were real objects. Validity issues are not relevant to discuss. In the conclusion it mentions attainment as having obtained a HE qualification, and in the appendices it clarifies that it means having completed ISCED levels 5 or 6 (see the Appendix – European Commission/EACEA/Eurydice, 2015). The choice of a self-reported completion measure and data from individualised surveys instead of register data is probably wise. It seems reasonable and easy to compare. No conflicting or overlapping definitions or measures appear or are problematised. While referring to the Eurydice report on access and retention ‘Modernising higher education’ from 2014, which goes deeper into the student completion topic, a messy reality is revealed with multiple approaches to measuring student completion. On the one hand, for the social researcher and the position of constructivist sociology of knowledge, the messiness would call for theory-driven research. On the other hand, the shift in measurement and discourse of measurement between the European Commission/EACEA/Eurydice reports of 2014 and 2015 could be seen as an action taken in terms of Desrosières (1998) by the political and administrative language of action and a pragmatic stance towards an easier way to bring clarity to objects in the need of political action. Thus, the messiness is not an object for a knowledge challenge; rather, it is concealed and narrated as harmonised comparisons.

Dealing with numbers only in a theoretical comparative analysis, such as comparing odds ratios of logistic regression analyses or dropout or continuation percentages in different countries, is not a research mission and for scientific knowledge production it is a dead end. Research into the sociology of education is about understanding social phenomena in educational contexts, while numbers without theory and context is a pure mathematical exercise.

Conclusion

To summarise, it has been apparent that the available data on student completion offers only a very limited basis for research with international comparative analysis, because there are huge differences in the European HE systems. For example, terms such as ‘retention’, ‘dropout rates’ and ‘continuation’ and ‘completion’ vary enormously. It is difficult to quantitatively research retention and dropout within these different systems because we might not necessarily compare the same phenomena. Different national agencies use diverse terms, (i.e. non-completion, non-continuation and dropouts) or even identical terms that are defined differently. The picture is further complicated by the fact that a wide range of methodologies is employed for gathering data and calculating student completion rates in the various countries. In the international datasets on completion there is a mix of sources coming from both cross-sectional studies and longitudinal data and hybrid datasets in which both sources apply (c.f. Thomas and Hovdhaugen, 2014). However, how should we interpret the shift from relative completion rates to absolute graduation rates by the OECD (see the Appendix – 2007a, 2013, 2014)? Should we see it as retrogression to simpler and towards more ‘true’ measures? Or is it more of a refinement and unification of vision of the political objects and the action needed (raise participation rates but do not mention completion measures)?

Nevertheless, it is apparent that the studies made by the EC, ESS and OECD are more evaluation studies than research shaping cognitive space of equivalence and comparability of ‘evidence’ made for administrative practice (Desrosières, 1998). However, recent studies suggest that the overall aims of the Bologna process have become less relevant for participating countries in terms of a decrease in rank and size of national delegations in official attendance at all ministerial conferences of the Bologna process. The Bologna process may not be necessary to legitimise domestic reforms anymore (Vukasovic et al., 2015). It would suggest that the on-going data standardisation processes could be at risk of receiving less domestic attention or support. Maybe it was sufficient for the EC to conclude that it was possible to obtain key indicators in the strategic framework ET 2020 that worked, that is, the headline targets of 40% completion of tertiary-level education among 30–34-year-olds, and that the proportion of early leavers from education and training should be below 10% by 2020 (see the Appendix – European Commission, 2012). However, in the latest report on implementation of the Bologna process there is clearly further attention paid to measuring member states’ success in enhancing the ‘social dimension’ and to ‘strengthen policies of widening access and raising completion rates’ (see the Appendix – European Commission/EACEA/Eurydice, 2015: 25). Recent evidence for a continued interest in the topic is an unpublished report on the on-going work on computing and collecting data on completion rates and average duration in HE, which is referred to in the HEDOCE project report ’Dropout and completion in higher education in Europe’ which is the latest publication on surveying student success policies and related data status within EU member states (ICON and QUANTOS, 2015, in HEDOCE, 2015).

Thus, there is reason to believe that the data standardisation processes of domestic statistical nomenclatures and cross-national surveys will continue. The rise of general interest in big data and data integration ¹¹ and the content in the ESS’ ‘Annual Statistical Work programme 2015’ (see the Appendix – Eurostat, 2015b) and in the ‘Multi Annual Work Programme 2013–2017’ ¹² shows that there is increasing attention towards the quest for how to handle and compare big data and this path is far more probable and irreversible than retrogression into refinements of domestic statistics.

Nonetheless, to gain more scientific knowledge of student completion in HE there is a clear need for comparison between different HE systems, specific educational programmes and student characteristics. Many education organisers pose questions about whether their study programmes are exceptional when it comes to dropout rates or if it is a part of some transnational trend in Western society. There are also concerns about how dropouts can be contextualised and understood and whether dropouts are something that should be addressed or not. Quinn (2013) made an effort to create an overview of the current knowledge on the dropout situation in EU member states, but concluded in the report that there is a need for more research on how many students do not complete and who they are, based on comparable information and definitions with regards to the national contexts. In the report it was also pointed out that the Eurostudent surveys would have promising potential if variables of dropout or completion were added.

Hence, if we need to design comparative research projects on student completion in EU countries, social science researchers cannot rely on the collections of existing domestic statistics on ‘ready-thought’ measures and concepts unless these are problematised in close relation to the domestic context and the role and function of HE in that particular setting. If the research is supposed to answer why there are differences, there is a strong reason for knowledge production that produces critical social research. Not only consensus driven statistics – but research in which researchers have absolute control of the construct validity in the tools of measurement. It is also important to be able to rely on multiple measures to complement the indicators as a multifaceted measure and not composed as single indicators/variables.

Reusing ready-made data – a bad idea?

Resuming the issue of researchers as users or producers of data, which was initiated above, and the user position, which was framed as negative for researchers, it is interesting to refer to other studies, which have problematised ‘the knowing capitalism’. They suggest that social scientists should more or less abandon causal analysis and instead strive for excellence in description and classification. In their view researchers should seek access to use the huge amounts of existing data collections collected by, for example, companies within the private sector, which means engaging with the community outside academia (Savage and Burrows, 2007) and dealing with the apparent ‘knowing capitalism’ (Thrift, 2005). However, from the research community perspective the question of being a user of statistics, that is, using figures in the OECD’s reports or reuse of data obtained in the European Social Survey, ESS, the EU’s LFS, etc., could be unfolded in several different and deviant directions. One aspect has to do with quality aspects of data and how to handle them; the other aspect involves power relations in knowledge production between the research community and the on-going data harmonisations in the EU context and also their emergent position in data production. Instead of relying on different domestic registers, the collection of comparable data is currently provided by transnational surveys as mentioned earlier, such as the LFS and ESS, which researchers are encouraged to (re)use.

However, following Desrosières’ (1998) typology of positions in statistical reasoning, the statistical discourse in science could call for a critical stance as regards validity and reliability aspects, which are prompted when researchers refer to and use measures and data produced in the surveys, especially when there is a comparative interest. Moreover, the non-response bias increases when fewer people participate in surveys due to the increasing trend of survey fatigue, which also contributed to issues of reliability. ¹³ Regarding the comparative prospect and data quality, Gustafsson (2008) has disentangled the risks of misuse and pitfalls when adapting causal analysis aims on the cross-sectional international datasets, which are not designed for purposes other than descriptive use. The international data production, such as that of the OECD, PISA and ESS, is also generally policy driven and not research driven or theory driven. Which is important to bear in mind. It may be possible to conduct country-specific analysis of causes and effects, but cross-country analyses are hazardous. Thus, even with newly developed advanced statistical and compensatory techniques in researching a complex phenomenon, a theoretical perspective is necessary in order to apply stringency in fundamental assumptions in the research design and analysis of phenomena (Gustafsson, 2008).

In contrast, while referring to the statistical discourse, the position of constructivist sociology of knowledge sees statistics as tools in processes of unification and equivalence in societal structures (Desrosières, 1998). Due to the serious validity issues being raised, it also implies distrust in the political transnational visions of large-scale comparisons of isolated indicators, at least as tools for researchers. In that view there is not necessarily a sufficient foundation when comparing variables with the same name attached with similar sequences of words in the definitions. If the researcher does not take into account the complexity and time/place data in the social reality then, as regards this scientific position, it is not social research that is produced, but rather a pure political exercise.

On the opposite side of the researcher being a user of ready-made data, the researcher is seen as being a producer of data. Instead of being confined by the quantitative quasi-research data that exists by proxy, the research community would be considered as expert in knowledge production and consequently on a mission to produce robust and valid research data. However, science can be disputed and it is not evident that researchers are the only experts who are allowed to provide policy makers with evidence nowadays. There are many professionals that can use scientific methods in collection, piling and analysing data of all kinds (Savage and Burrows, 2007; Shortall, 2013).

Nonetheless, in the third position of Desrosières’ typology, the objective voice of the pragmatic political and administrative language of action argues that researchers should reuse data instead of reinventing the wheel. This position, that is, in terms of the EC, produces grey literature reports that mix policy discourse with terms such as ‘large-scale’, ‘studies’, ‘data’ and ‘comparable’, which lend an air of scientific practice but in consequence blur the boundaries between follow ups, evaluation and research practice. Thus, in order to gain fast delivery of easy access to comparable data, the EC produces its own data and, consequently, monopolises and excludes the research community from the comparative data production, because data already exists. There is more to come, since the data harmonisation processes have just begun the construction of a European digital internal market, in which data integration within Europe is the core aim (see the Appendix – EDF, 2015). There is a severe and acute need to research the quantitative survey data production in general and its consequences for the knowledge production. For example, a recent study of the Eurobarometer surveys between 1995 and 2010 showed how the Eurobarometer selects and frames questions in ways that systematically produce ‘integrationist’ outcomes and that, in general, has significant methodological anomalies and this kind of production blurs the line between research and propaganda (Höpner and Jurczyk, 2015). The fourth position in the typology – the relativist or the questioning accusative (Desrosières, 1998) is not yet seen in these contexts, because it would, for example, necessitate political mobilisation of the social categories that statistics are constructed upon (see Carlhed, 2007).

Probable and serious consequences of the apparent scenario of the striving for comparable measures on student completion on scientific knowledge production are that research practices and specific research interest that competes with topics of interest for EU policy will be repressed. Policy makers on different levels seem to be satisfied with evidence from grey literature reports with easily digested facts and univariate tables of isolated indicators. Broadly, the research funding schemes are tuned towards policy issues and a large number of policy-driven ‘studies’ are funded from EU programmes. It seems also that the national funding agencies also follow this trend. In addition, international research journals demand, to a greater extent, comparative studies rather than accounts on single countries, but there is scarce funding for researcher-driven comparative studies in terms of opportunities to pay international researchers as the workforce within projects. Between funding schemes with national scope and large EU projects there is a middle range level of collaborative and comparative research that needs funding. Hence, there is a serious risk that knowledge production on student completion in HE is drifting, unnoticed, towards superficial and atheoretical analyses or compliance silently in the wasteland of univariate indicators.

In this regard, it is crucial to point out the nature of academic specialist knowledge compared to generalist knowledge and how the knowledge is used and mediated through social institutions, civil servants status and authority. Shortall (2013) points out that critical knowledge could criticise civil servants responsible for a policy area and, by the characteristics of academic knowledge being open to debate and illuminating multi-perspective views of the world, research could hamper the legitimacy of existing policy, while non-critical knowledge or consensus-driven knowledge production would reinforce valid (and existing) policy choices and therefore be more attractive to policy makers, simply because certainty and social stability is preferable. These priorities shape the social conditions in the use of evidence in designing policy (Shortall, 2013). However, the question is whether only social research will be accepted as the means of informing policy in the future and the critical basic research was a short parenthesis in social science’s history.

However, the strongest resistance to scientific knowledge production on student completion and similar doxic objects lies perhaps in the task of explaining how the research is policy-relevant by being critical to the current natural order of framing social categories, and not to bring solely statistical evidence for the dominant preferable policy choice as a ‘policy engineer’ or a ‘policy entrepreneur’ (as Ball wrote back in 1995), but scientific evidence on how knowledge production and differential modes of acceptance are socially conditioned. This means shedding light on symbolic power that seems to undermine the conditions of critical social science in favour of quasi-scientific reports on the success of policy interventions.