ESL in the EU: Learning from differences and common trends

Eurostat (2017) reports that since 2000 ESL in the EU has been stably decreasing and approaching the agreed target (10% of ESLers in 2020). Yet, greater insight reveals that, despite this common trend, differences among member states remain large not only in the shares of ESL (e.g. 19% in Spain, 4.9% in Slovenia in 2016), but also in the approaches (policies and practices) used to tackle it. Another explanation of these big differences is that EU cooperation in the area of education (including ESL) is based on the open method of coordination (OMC), a specific, non-binding form of cooperation that does not require national laws to be harmonised, and thus enables very unique national responses to be developed for a widely recognised problem and for common EU goals to be achieved in the field. In this environment, differences should be seen as a lucky situation and a precondition for policy learning (Radaelli, 2003). [1] The emergence of policy learning as a central mode of governance [2] has been explained and justified from various perspectives (Gornitzka, 2006; Hartlapp, 2009; Kan, 2005; Radaelli, 2008), while policy learning in official EU documents is also presented as a key component of EU governance (e.g. Council of the EU, 2009; 2015).

Radaelli (2008) defines policy learning as a potential mechanism for deliberative collective problem-solving since it encourages member states and other actors to exchange information, draw comparisons and evaluate existing practices. A closer look at the existing literature reveals there is no uniform definition of policy learning and that several different classifications of policy learning exist, mainly varying in terms of who the actors are that provide the source of the knowledge exchange (Etheredge & Short, 1983; Hemerijck & Visser, 2003; Grin & Loeber, 2007; Hartlapp, 2009; Gilardi, 2010); the types of policy transfer (Dolowitz & Marsh, 1996) and the type of changes brought about by it (Argyris & Schön, 1978; Hall, 1993; Olsen & Peters, 1996; Eising, 2002). All of these discussions and classifications may be summarised in the following questions: ’Who learns?’, ’What do they learn?’ and ’With what effect do they learn?’ which, as part of classifying the original policy learning theories, were asked by Bennet and Howlett (1992). [3] Fink Hafner et al. (2012) supplemented these questions with two additional ones: ’How do they learn?’ and ’Why do they learn?’. As evident from Table 1 below, these questions provided the basis for a definition that attempts to encompass all of the diverse forms of policy learning, while simultaneously also enabling any exclusionary or partial definitions to be used for explaining policy learning in a specific context. We argue this broad definition (aggregation of various forms and academic definitions of policy learning) presented in Table 1 can in addition provide a valuable framework for understanding policy learning processes in the EU in the area of tackling ESL.

Table 1: Definition of policy learning. Source: Fink Hafner et al. (2012).

The aim of this article is to present, based on an in-depth review of existing literature and practical examples, how ESL policies at the EU level are formed, implemented and evaluated within the policy learning framework. The article shows that policy learning in the EU context is not just a set of various different activities but is organised in a well-thought-out policy learning architecture. [4] These insights are not only scientifically relevant but also practically useful for relevant and other interested actors to understand and exploit the various policy learning opportunities available in the field. Policy learning has a stronger impact if actors consciously enter the policy learning process and are aware of its potential.

After introducing the topic (EU framework for addressing ESL) in policy learning theory, the article systematically elaborates the topic by asking ’Who learns?’, ’How do they learn?’, ’What do they learn?’, ’Why do they learn?’, and ’With what effect do they learn?’. For each question, answers are provided through a theoretical review and highlighted with specific examples from ESL policy in the EU and some practical insights from the TITA project (marked in italics).

To address the article’s aims, we employ the following methods: (a) an analysis of relevant literature and secondary sources. In this framework, we conducted a literature search of the scientific EBSCOhost and Web of Science online research databases. The main key words used in the search for relevant scientific literature were: early school leaving, education, European Union, policy learning, policy borrowing, policy lending, peer learning, lesson drawing; [5] (b) an analysis of formal documents and legal sources at the EU and national levels (EU official documents in the field of educational policy, non-official documents, press releases etc. Policy documents taken into account had to meet the following criterion: to have been published between 2000 and 2017.

Authors (e.g. Larionova, 2007; Radaelli, 2004) explain that in its ideal-typical and most abstract form, policy learning in the area of education in the EU takes place between institutions found at multiple levels, and runs in at least four directions:

• EU-level learning within communities of policymakers engaged in EU policy processes (or “learning at the top”);
• hierarchical learning from the EU level down to the domestic and local level (or “learning from the top”)
• “learning from below” (i.e. social actors, regions, local governments) to the top (or “bottom-up learning”); and
• horizontal learning (learning between the member states, or regions).

Policy learning is a social process, meaning that individual policy actors have to had acquired knowledge somewhere (from either other individual actors or within organisations, networks and communities labelled as collective actors) (Holzinger & Knill, 2005). As seen in its ideal form, policy learning in the EU involves various state and non-state actors [6] who rely on variable partnership forms, which is also congruent with the Lisbon Strategy architecture (European Council, 2000, para 38).

May (1992) believes there is a gap between normative assumptions (who is supposed to learn) and behaviourist assumptions (who actually learns). Early assessment of policy learning in the EU in the area of education policies painted quite a disappointing picture. For example, empirical evidence (Radaelli, 2004) indicated that it was especially learning at the top (or EU-level learning) among a highly professionalised technocratic elite that was significant, there was embryonic evidence of cognitive convergence from the top, and almost no evidence of learning from below (“bottom-up learning” or “social learning”). Another deficiency of describing policy learning in a multi-level context was that insufficient attention was placed on the flow between all three levels (EU, national, regional). Structures to ensure the multi-level coordination of actors were indispensable and in practice the learning process occurring between them was often interrupted for various reasons (e.g. weak institutionalisation).

More than a decade later, it seems that the first quite disappointing evidence has become more positive. An important contribution to this is the new E&T 2020 framework and especially the Erasmus+ projects within Key action 3 devoted to providing support for policy reform, which the TITA project also forms part of. They provide a well-thought-out framework for the formation, implementation and evaluation of local policy experimentations. [7] Based on the contextualisation of their results, specific policy recommendations are prepared for the EU level. E.g. in the TITA project framework, local multi-professional teams for addressing ESL have been experimented with in France, Luxembourg and Switzerland. The TITA project thus represents a well-considered balance between top-down, bottom-up and horizontal learning approaches, with the involvement and close cooperation of various state and non-state actors from all levels of multi-level governance. Although the local experiments’ effectiveness has been evaluated through a strictly prescribed protocol (see EACEA, 2015), their actual exploitation at the EU level and dissemination in the form of mainstream practice across the EU remains open and calls for monitoring and evaluation over a longer period of time once the project formally ends.

In a more or less institutionalised manner

Policy learning is most likely to take place when a suitable institutional infrastructure is in place (absence of any uniform or rigid rules, enabling sufficient informal communication and knowledge exchange). It should include a set of policy instruments that build up the participating actors’ institutional learning capability (Stone, 1999; Gornitzka, 2006; Radaelli, 2008).

On a voluntary or coercive basis

Dolowitz and Marsh (1996) distinguish three policy transfer types: 1) voluntary transfer, which occurs when policymakers decide on the policy transfer on their own and on a voluntary basis due to their dissatisfaction with the existing national policy; 2) direct imposition of the transfer, which occurs when one actor forces another to adopt a certain policy; and 3) indirect imposition of the transfer, which is based on certain (extrinsic) factors or a functional dependence, that encourages actors to transfer a policy voluntarily.

With various motives and strategies

While some accept new knowledge and lessons for purely symbolic or strategic reasons to muster political support, others apply the new knowledge and lessons to actually make changes and improve their understanding of policy problems and the current situation. It is widely believed that policy learning is more common during a period of failure and crisis, when actors have a stronger desire and need to change and improve the current situation (Levitt & March, 1988; Hemerijck & Visser, 2001, 7; Radaelli, 2004).

Taking the complexity of public policies and the institutional context into consideration

Some authors (e.g. Rose, 2002; Hsueh, 2009) point out that superficial imitation alone is far from sufficient to successfully transfer ideas, policies and practices from one institutional framework to another environment. Here it is necessary to understand and consider the complexity of public policies, as well as the varying institutional contexts entailed. In circumstances of increasing complexity, it is thus essential to: a) recognise one’s own policy problem; b) decide where to look for a solution; c) research how a certain policy and practice work in another institutional framework; d) develop a causal model of the way the policy and practice work in that institutional framework; e) (re)formulate lessons for the transfer of the policy and practice into one’s own institutional context; f) decide whether in fact to transfer the policy and practice; g) provide suitable sources for the policy and practice to be implemented; h) look at any additional special risk that might arise when adopting the lesson in one’s own institutional context; i) reduce the special risk by conducting a prior evaluation; and j) apply others’ (positive or negative) lessons in one’s own context. These phases are widely recognised as 10 steps to follow in seeking to learn lessons from abroad.

On the basis of their own experiences, from others or with others

Hemerijck and Visser (2003), Radaelli (2004), Volden et al. (2008), Hartlapp (2009) identify three different types of learning:

1. Learning from one’s experiences, which helps actors prevent or reduce future mistakes. Reducing mistakes provides a motivational basis for looking at one’s own institutional and organisational past and thus for learning based on monitoring and evaluating one’s own systems.
2. Learning from others, which occurs when actors use information about the consequences of political decisions in other countries, without them being affected themselves.
3. Learning with others, which includes consultations, experimentation, exchange and the collection of information and experience in pursuing shared goals.

They form, give and/or take

Actors within policy learning are never simply passive recipients of the newly acquired knowledge. They can suitably contextualise the newly acquired knowledge and use it to alter their conviction (Dunlop & Radaelli, 2010). Based on the premises of policy (Heclo, 1972), imperialistic and coercive learning (Lange & Alexiadou, 2010), the dominant actors in policy learning processes are those who attempt to impose their convictions, policies and practices on other actors (in)directly. In doing so, they assume the role of a teacher, whereas other actors who accept the convictions, policies and practices may be seen as learners. Accordingly, the question should not only be ’Who learns?’ but ’Who teaches?’ (Grek, 2009).

The TITA project is a clear example of institutionalised, voluntary policy learning with the aim to find a solution to ESL. Irrespective of whether its results will provide a policy learning example of failure or success, learning from its experiences (in the French, Luxembourgish and Swiss Francophonic local contexts) will require in-depth contextualisation. Research (e.g. Jugović & Doolan, 2013) shows that not only specific local contexts but also specific EU regions can share some similarities regarding ESL, which distinguish them from other parts of the EU. By learning lessons from the TITA experiences, Rose (2002) steps for learning lessons from abroad, presented above, could be taken into consideration.

No standard definition of knowledge in terms of what is supposed to constitute the core of policy learning can be found in the literature. Knowledge (as a subject of policy learning) can (regardless of whether it refers to new policy instruments, frameworks or paradigms) [8] be formulated rationally (the positivist approach to understanding knowledge dominates here) or socially [9] and politically [10] (where the constructivist approach dominates).

Taking research findings showing that tackling ESL in the EU is not only policy but also political exercise (e.g. Macedo, Araújo, Magalhães & Rocha, 2015) into consideration, how we understand benchmarking charts, which reveal differences and common trends in EU anti-ESL efforts, is important. A quick and superficial overview of them may lead to an impression about which are good and bad education systems/practices worth learning from. These are facts which can easily trigger social and political learning. Making benchmarking charts a reliable source for policy learning calls for their in-depth contextualisation. Well-developed national information systems for gathering information on ESLers are a precondition for this. They (should) not only consist of comparable statistical data but also other (qualitative and quantitative) information that helps to understand the phenomenon in detail (e.g. Thematic Working Group, n.d.). Hordosy (2014) reports that the big differences in national information systems (e.g. student registers, national or regional surveys) have so far allowed little room for making comparisons across Europe. Given that “first experiences in countries applying more advanced data collection systems show that high-quality monitoring is very useful in maximising the reduction in ESL” (European Commission, 2016), these could provide an unexploited source of more in-depth policy learning and an even more positive trend of reducing it at the EU level.

Policy learning helps in controlling uncertainty by enabling changes to be understood and to thereby construct scenarios and predictions in an ever changing environment. In terms of the factors that catalyse policy learning and explain why policy learning occurs in the first place, two sets of factors can be identified: a) factors at the individual level, which include looking for solutions to policy problems, controlling uncertainty, achieving individual and collective goals and searching for the truth; b) structural factors, which include time-related factors that stimulate policy learning (an election period, economic crisis, regime transformation, a period of lack of success, a period of uncertainty), political culture, the institutional structure of a country, special characteristics of public policy and organisational competence (e.g. Bennet & Howlett, 1992).

The motives for policy learning in the form of policy experimentation for dealing with ESL within the TITA project can be explained by the great complexity of the ESL policy problem and the consequences it holds for individuals, nation states and EU society as a whole, along with the already established infrastructure for multi-professional cooperation in participating countries (France, Luxembourg, Switzerland). Experiences of the leading partner (France) with social experimentations in the ESL field clearly also represent an important factor explaining the EU’s interest in policy learning in the TITA setting.

Certain authors (e.g. Stone, 1999; Radaelli, 2008; Hartlapp, 2009; Gilardi, 2012) believe it is extremely difficult to explain the reasons for policy learning, mainly due to the difficulties in causality and the fact that several factors may trigger policy learning. These connections become all the more complicated when seeking to explain policy changes as the key consequences of policy learning. In relation to this, a big divide is seen in the literature in answers to the question of whether a policy change is a prerequisite for policy learning. While some authors (Sabatier, 1993; Grin & Van de Graaf, 1996; Pemberton, 2003) argue it is only possible to talk about policy learning when it results in a policy change, others (Dolowitz & Marsh, 1996; Hemerijck & Visser, 2001; 2003; Lange & Alexiadou, 2010) believe a policy change is not a prerequisite for the policy learning process. Hartlapp (2009) adds that policy learning, per se, is not a goal, but is aimed at the implementation of political programmes and achieving of policy goals. When policy learning is regarded as learning that brings about policy changes, this encompasses a change in ideas, values, interests, instruments, goals, programmes and the institutional structure of (some of) the actors involved (Rose, 2005; Dolowitz & Marsh 2000). It can also provide the preconditions for ideational convergence, that is, the convergence of policymakers around a set of criteria that define good policy.

At least the following intended changes should be brought about by the TITA project: a) ideas and values concerning how ESL is understood and should be resolved; b) instruments for solving the problem (from sectoral to cross-sectoral collaboration); c) an institutional structure in terms of establishing local multi-professional teams for addressing ESL across the EU. The precondition that policy learning within the TITA project in the form of policy experimentation will actually bring any substantial changes in educational policies and practices at the EU level and across the EU is that internal and external actors show interest and engagement. Follow-up activities for monitoring and evaluating its long-term impact are also necessary. According to Vilpišauskas (2011), a decade would be an appropriate time frame for such a review.

This article shows that the EU’s policy framework on ESL establishes a unique architecture for policy learning that is not simply limited to top-down learning (»listening recommendations« from EU institutions), but encompasses a wide variety of bottom-up learning (local experimentation) and bottom-bottom learning (comparing lessons from among the states involved). The article reveals that this is also the case within the TITA project.

The question of whether TITA is just a 3-year policy experiment (learning) or will actually trigger any policy change (at local, regional, national or EU level) remains unanswered. Given the broadness of policy learning theory, lessons will in any case be learnt: how (not) to implement multi-sectoral approaches to ESL. It is currently still uncertain how they will be transferred to wider (EU and member states) environments. Considerable further effort at all levels is needed to ensure that the local experiments do not remain strictly local, completely experimental and limited in time and funding (Berthet & Simon, 2012).