Evading the Burden of Proof in European Union Soft Law Instruments: The Case of Commission Recommendations

The European Union is making increased efforts to find simpler and more effective ways to function adequately in the eyes of its citizens by using ‘soft law’ instruments such as recommendations. Although they have no legally binding force, recommendations have practical and legal effects occurring partly due to their normative content in which a course of action is prescribed and further supported by arguments intended to persuade the addressees of a political position. Although recommendations function as persuasive instruments due to their argumentation, the characteristics of argumentation and how it is employed to convince the addressee to comply with certain measures have not been investigated at all. The main goal of the paper is to explain how arguments are used by the European Commission when recommending Member States to follow a new course of action. First, we will unravel the justificatory reasons employed by the Commission in order to make Member States comply with new measures and we will show how these reasons are combined into an argumentative pattern (van Eemeren in Argumentation 30(1): 1–23, 2016; J Argum Context 6(1): 3–26, 2017). This pattern basically prescribes a course of action to Member States, which is further supported by arguments in which the necessity and advantages of following the proposed course of action are justified. Second, we will explain how and why the way in which the arguments are combined in this complex pattern could be potentially persuasive for the Member States despite the legally non-binding character of recommendations. We will show that the European Commission tries to persuade the Member States to take new measures by evading the burden of proof imposed by the legislative framework. At a more specific level of analysis, we will delve into the implicit premises in the argumentation, which enable us to identify cases of evasion of the burden of proof due to the Commission’s use of implicit starting points which might not be accepted by the Member States.


Introduction
It is now more than ever before clear that the European Union (EU) is in crisis and must make increased efforts to find simpler and more effective ways to function adequately. So-called 'hard law' instruments, such as directives and regulations, 1 adopted in order to attain the Union's objective of a common legal framework in various policy areas, do not seem to ensure the envisaged efficiency in the EU [26]. In this regard, it is striking that simplification proposals call for using 'soft law' instruments, such as communications, resolutions, recommendations, currently accounting for 12% of the EU law [20]. Take, for instance, the Recommendation issued in 2015 requesting EU countries' law authorities to improve their use of videoconferencing to help judicial services work better across borders.
Vital for an effective functioning of the soft-law instruments in the EU are the arguments advanced in order to encourage the addressee to follow a course of action [3]. Arguments showing how the proposed measures will effectively solve problems at EU and national level (e.g., by ensuring equitable tax systems in Member States) are employed to support a standpoint advocating the implementation of a new course of action at Member State level (e.g., reducing tax wedge on labour for lowearners). Being legally non-binding instruments, compliance with such instruments largely depends on the persuasive force of the arguments, but the problem is how we can make sure that these instruments devoid of formal legally binding force can function as effective tools for convincing the addressee to comply. Argumentation plays an elusive but fundamental role to ensure the practical effectiveness of soft law instruments, but so far little if anything is known concerning the characteristics of argumentation in this context and how it can exactly contribute to their effectiveness and intrinsic quality. The very fact that there is yet no insight into the argumentative characteristics of soft law instruments and into their quality, while it is recognized that they are influential due to their argumentation, makes it imperative to explain how arguments are employed in such documents and for what purposes.
In this paper, we will deal with these important issues in the case of recommendations enacted by the European Commission and addressed to the Member States in the last 5 years (2012-2016). Our focus on recommendations enacted by the Commission stems from two reasons. One reason is that such 1 European Union directives are legislative instruments employed to attain the Union's objective to adopt a common legal framework in various policy areas. They are binding with regard to the aim to be achieved, but leave discretion to the Member States as to how they are to be implemented, and are not directly applicable. Regulations are comparable legal instruments which are binding in their entirety and directly applicable in all Member States (see [1] for a detailed discussion of EU directives). recommendations are part and parcel of the current efforts to ensure better governance, being also more and more commonly acknowledged in litigation [21]. The other reason for studying Commission recommendations is connected to the important role played by these instruments in the EU legal order. Recommendations issued by the Commission have far-reaching objectives, but little if anything is known about the way in which they function as argumentative instruments aimed at convincing Member States to follow a course of action.
The main goal of the paper is to explain how argumentation is employed by the European Commission to persuade Member States to follow a certain course of action. First, we will unravel the justificatory reasons employed by the Commission in order to make Member States comply with new measures and we will show how these reasons are combined into an argumentative pattern [23,24]. This pattern basically prescribes a course of action to Member States, which is further supported by arguments in which the necessity and advantages of following the proposed course of action are justified. Second, we will explain how and why the way in which the arguments are combined in this complex pattern could be potentially persuasive for the Member States despite the legally non-binding character of recommendations. We will show that the European Commission tries to persuade the Member States to take new measures by evading the burden of proof 2 imposed by the legislative framework. At a more specific level of analysis, we will delve into the implicit premises in the argumentation, which enable us to identify cases of evasion of the burden of proof due to the Commission's use of implicit starting points which might not be accepted by the Member States.
Methodologically, our analysis will involve a conceptualization of the recommendations in which insights from political and legal scholarship and argumentation theory will be combined. In argumentation theory, our analysis will be guided by the pragma-dialectical perspective on argumentation [23], which provides us with the tools necessary for reconstructing arguments and their combinations into various argumentative patterns, as well as with the tools enabling a proper evaluation of the arguments as evasions of the burden of proof. In addition, we will make use of insights from the Argumentum Model of Topics [6,13] which provides us with knowledge of the inferential configurations of arguments. This will enable us to identify evasions of the burden of proof at a different level. The combination of pragma-dialectics and AMT provides us with progressively more fine-grained analyses of the quality of EU arguments, both at the level of how the argumentation is structured and at the level of how specific inferences are legitimated. In particular, through an AMT analysis, we can shed light on the EU's implicit starting points on which the Commission relies, on the extent to which they might be shared by the addressees and eventually on how functional these starting points might be for persuading Member States to comply with the prescribed course of action. By doing so, we provide an explanation of the various cases of evasion of the burden of proof which occurs by making use of implicit starting points that might not be accepted by 2 Evading the burden of proof is interpreted here as not fully accepting the commitment to justify one's standpoint in front of the addressee (in this case, Member States). Evading the burden of proof can be strategically exploited to persuade the addressee without mentioning those aspects that could potentially weaken one's case.
Evading the Burden of Proof in European Union Soft Law… 81 the Member States. Our analysis and evaluation of the arguments employed in Commission recommendations will be guided by the characteristics of the institutional framework within which recommendations function and the legal guiding principles. Moreover, we will take into account the special status of recommendations as policy instruments standing between law and politics. The remainder of this paper will be organized as follows. In Sect. 2, we will explain the legal and political context affecting the argumentation in EU recommendations enacted by the Commission. We will account for Commission recommendations as instruments that have gained a particular significance in the EU legal landscape, and we will outline their content, purpose and function within EU law. In this section we will also explain the types of obligations that the Commission takes upon itself as well as the types of obligations it imposes on the addressees, and how these obligations can be enforced. This framework is necessary to understand the function of argumentation within recommendations. In Sect. 3, we will outline the argumentative pattern in Commission recommendations and interpret it for its strategic potential by closely examining its inferential structure. In the Conclusion, we will discuss the most important insights that our argumentative analysis provides for understanding Commission recommendations and we will outline some further research directions.

EU Recommendations as Steering Instruments
Recommendations have gained a particular significance in EU law. They are qualified as soft law instruments, more precisely as steering instruments enacted with the goal of guiding action in a legally non-binding way [16,20]. This characterization of recommendations follows closely Article 288 from the Treaty for the Functioning of the European Union (TFEU), which briefly specifies that ''recommendations […] shall have no binding force''. In line with the definition of soft law instruments, recommendations prescribe ''rules of conduct which, in principle, have no legally binding force but which nevertheless may have practical effects'' [19: 64] and also legal effects [20: 11].
Legal effects of EU soft law instruments such as recommendations can consist of ''a normative framework for future negotiations […], concretizing the duty of institutional cooperation […], producing a stand-still effect on the non-conforming conduct of a State or institutions […]'' [20: 16]. Recommendations can influence national and European legislation, because they may help in interpreting hard law provisions of directives and regulations, and also serve as the basis for enacting national legislation. Moreover, as Stefan [20] convincingly shows in her analysis of the role of soft law in courts, recommendations provide the basis for judicial review, they have been used by the parties in litigation, and have often served as an aid in the interpretation of hard law provisions.
Practical effects of recommendations include the policy changes in the practices of the Member States whom they address, going as far as changing policy principles [8]. They become politically and socially binding by means other than the legal force of the instrument. As Stefan illustrates [21: 17], in the employment field, for example, putting social and time pressure on the Member States can be effective in fostering cooperation. Most often, the effectiveness of soft law instruments is the consequence of the arguments employed to convince an addressee of the necessity of changing policies. Notably, it is by now generally recognized [3,20,22] that these instruments produce practical effects due to their argumentation in which positions are taken with regard to a future course of action which is then supported by arguments.
The last decades have witnessed a significant increase in the use of soft law instruments in general and recommendations in particular in a variety of policy areas ranging from competition and state aid to employment, immigration, education and social policy [20: 12]. They cover also policy areas such as the environment in which traditionally legislative instruments, such as directives and regulations, have been preferred [2: 193]. Arguably, the Commission issues nowadays recommendations in every policy area that falls within its competences as established in TFEU and that ultimately connects with the establishment of the internal market. In this way, recommendations are favored instruments for ensuring the free movement of goods, persons and services. 3 In all policy areas, recommendations have become an instrument of 'governance' lacking legally binding force but focusing instead on persuasion and guidance rather than enforcement [20: 2]. Terpan explains [22: 88] that ''recommendations […] have given rise to 'new forms of governance' based on the desire of the participants to agree, through collective deliberation, on procedural norms, forms of regulation and shared political objectives, while preserving a diversity of solutions and local measures''. 4 Paradoxically, these characteristics of recommendations indicate that instruments of soft governance may have political impact, a view which goes against the commonly accepted opinion that only legally binding instruments may exercise such influence [22: 89]. This is exactly how the European Commission has justified their use in its White Paper on European Governance (COM(2001), 428final).
Although creating a paradoxical situation in which legally non-binding instruments have a significant impact on the addressees, their use as instruments of governance is not that surprising. In fact, the search for other modes of governance besides legislation has been representing a priority on the EU agenda 3 EU recommendations are favoured for ensuring the free movement of goods, persons and services because they are flexible, easy to adopt and put less pressure on the Member States in comparison with legislative instruments such as directives. More often than not, directives are not complied with or are complied with only partly because they oblige Member States to adjust national legislation in a way which requires many financial, institutional and social resources [25]. EU recommendations, however, do not involve such resources, which arguably makes them more suitable instruments for the establishment of the internal market. 4 There are various reasons for this new reality in the EU legislative landscape. Not only are soft law instruments easier to enact, because they involve less bureaucracy, but they are favored particularly because Member States do not want to put considerable interests at stake by avoiding the risk sanctions and reputation costs which are usually associated with hard law instruments (White Paper of the European Commission, COM(2001), 428final, [22: 89]).
Evading the Burden of Proof in European Union Soft Law… 83 already for several decades, and it is all the more important at a time of crisis such as the present one. Guided by the fundamental principles of subsidiarity and proportionality, 5 if European measures are deemed necessary, recommendations should be preferred with a view to cooperation between Member States, because they involve simple action leaving as much scope as necessary for national decisionmaking [17]. While in itself the use of recommendations may ensure effectiveness of EU policy-making and decision-making through the employment of a simple and flexible means, questions have often been raised regarding the legitimacy of such soft law instruments [5,9,16]. Issues such as how the European Commission can be held accountable for the decisions while not being democratically elected, or how soft law instruments can be seen as legitimate if the European Parliament has not been involved in the decision-making, remain open. It is possible to argue in this case for the legitimacy of soft law instruments by pointing out that they complement rather than replace existing legislation and in this way they do not jeopardize the democratic control of the legislative process [5: 161]. Nevertheless, practice has shown that in numerous cases recommendations are a substitute to legislation and are therefore perceived as potentially illegitimate [16].

Content, Purpose and Function of the European Commission's Recommendations
From the viewpoint of their contents, European Commission recommendations are prescriptive documents laying down rules aimed at influencing the behavior of Member States [16]. Recommendation 2002/590/EC, prescribing ''rules, standards and/or regulations on statutory auditors' independence in the Member States'', is a good example in this regard. While sometimes the reference to 'rules' is explicit as in this recommendation, in other cases terms such as 'guidelines', 'minimum principles' or 'common principles' are employed to point at the normative content of recommendations. Recommendation 1405final/EEA, for instance, ''provides guidelines to Member States to assist them in a better and more effective implementation of the equal pay principle in order to combat pay discrimination and to contribute to tackling the persistent gender pay gap''. The prescribed rules are usually of a general nature, in that they prescribe a course of action in an area of general applicability at Member State level, rather than on a specific issue concerning one Member State [16]. The recommendations are most often addressed to the Member States, but they can also be addressed to national economic agents such as banks. The addressees are expected to make legislative, administrative or other kinds of changes in a way that accommodates the proposed course of action.
The function of recommendations remains a debatable matter. On the one hand, some authors [5] view soft law instruments in general as complementary to legislation. It is not very clear from their account whether this view is informed by empirical observation or it is rather a normative view on this type of instruments, which, should they play other functions than complementing legislation, would be seen as illegitimate. The latter seems to be the case, because this view is nowhere supported with evidence from practice. On the other hand, empirical research on recommendations [16] indicates clearly that they have three functions in the EU legislative landscape. First, they have a pre-law function [16: 181-182] by being adopted with a view to facilitating and accelerating future binding legislation. This function is obvious from the Commission's explicit mention of support for an upcoming directive or regulation. Second, recommendations have a para-law function, particularly in those cases in which there is no need, desire or room for legislation [16: 182]. This may be a temporary situation in which recommendations function as temporary substitutes for legislation until binding legislation, already in the process of being adopted, is actually enacted. Sometimes, recommendations are a permanent alternative to legislation when they are adopted on a certain matter which cannot be dealt with in binding legislation, since there is no legal basis for it there. Additionally, sometimes no agreement can be reached on legislation, which imposes on the Commission to issue recommendations with a para-law function (16: 183]. In this case, issues of legitimacy arise. Third, recommendations have a postlaw function when recourse is made to them in order to supplement existing legislation in order to facilitate uniform compliance with it, particularly when technical guidelines are provided.
The main purpose for which recommendations are enacted is twofold [16: 179-181]. This purpose involves both cooperation and coordination at EU level and harmonization of European policies. Senden explains that coordination is not aimed ''so much at establishing certain national legislation'' [16: 179], whereas harmonization aims at ''aligning national legislative and administrative provisions, particularly with a view to removing obstacles that impede the realization of the internal market and economic and monetary union'' [16: 179]. Due to their harmonizing role, recommendations seem to be close to directives, which have a similar purpose [1], except that recommendations have no legally binding force. 6

Obligations and Enforcement in Commission Recommendations
Theoretical and empirical legal and political insights confirm that soft law instruments such as recommendations impose constraints on the addressees, even if these constraints are softer than the hard obligations imposed by the Court of Justice of the European Union (CJEU) [15,20]. Member States, as the main addressees of these instruments, need to position themselves with regard to the norms that have been set up at EU level. This is all the more necessary as Member States have been 6 The first recommendations as early as the 60 s were seen as a legal instrument similar to directives. They were binding with regard to the result to be achieved upon those to whom it was addressed, leaving open the choice of form and methods [16: 170]. This initial characterization of recommendations can partly explain their current resemblance with directives.
Evading the Burden of Proof in European Union Soft Law… 85 involved in the negotiation process of adopting recommendations. This entails either complying with the norm and implement it, or resist the norm and take action deviating from the norm [15: 3]. In some cases, the Commission actually requires Member States to explicitly explain why they do not comply. This reality, which in itself is identical to the case of binding legislation, points at the importance of understanding the kinds of obligations that recommendations can and do entail on the addressees. In fact, in order to understand the function of argumentation in recommendations, it is important to frame it within the obligations that these documents impose on their addressees; obligations and enforcement help to understand the type of prescriptive standpoint that is contained in the recommendations. Some authors [9,11,16] view soft law instruments as limited to non-binding norms with legal relevance. Other authors [2,14] define them as involving binding norms with a soft dimension. Others [22] combine the two views in an attempt at doing justice to all types of obligations which can be distinguished for soft law instruments, as well as the rights that the actors involved can enjoy. This last view seems to us particularly useful for interpreting the argumentation at the heart of the recommendations. Not only does this view capture the complexity of the existing situation, but it also enables explaining various situations occurring in practice.
Terpan [22] provides a comprehensive mapping of the types of obligations imposed by EU soft law instruments by envisaging a continuum running from nonlegal positions (in which no legal obligations are imposed) to legally binding and judicially controlled obligations (such as those imposed by case law of the Court of Justice of the European Union). This nuanced view in which various levels of legal obligations are envisaged, ''helps describe with precision the different ways of adopting and enforcing norms in the EU'' [22: 72]. Terpan explains that hard and soft law can be distinguished not so much on the basis of the existence of an obligation, which applies to both instruments since they both prescribe action, but rather on the basis of how the obligation is enforced [22: 73].
Enforcement is defined as going from ''monitoring to more coercive mechanisms, including judicial control and sanctions'' [22: 73]. It can vary from hard enforcement in the case of judicial control, to soft enforcement relating to procedures aimed at ensuring compliance without constraint, to no enforcement. Terpan situates soft law between hard law and non-legal norms. This characterization views soft law instruments as quasi-legal, a view which is perfectly in line with the opinion according to which the EU is ''both a polity and a governance system producing different kinds of policies'' [22: 76].
Four types of situations are distinguished depending on the kinds of obligations imposed and the kind of enforcement [22: 77]. 7 First, hard obligations can go hand in hand with soft enforcement. Drawing on examples from fiscal and 7 Terpan [22: 77] distinguishes a total of six situations, but for the purposes of this article only four such situations are relevant. We excluded the case of ''soft obligation/no enforcement'', because this case is not illustrated by Terpan and it does not seem to occur in recommendations. We also left aside the case of ''no obligation/soft enforcement'', because this case does not apply to recommendations. In these instruments, there is always some kind of obligation at issue, whether soft or hard. The case in which no obligation is at issue applies to declarations and strategic documents. macroeconomic surveillance, Terpan convincingly illustrates the case in which the obligation imposed by the soft law instrument is hard (similar to directives), but enforced by ''a rather weak mechanism relying upon the Member States' willingness to make it effective'' [22: 79]. Second, hard obligations can also occur without enforcement, as in some cases of the Common Foreign and Security Policy. All actions and positions in this case are meant to be legally binding in spite of a total lack of enforcement by the European Commission. Whereas Member States are expected to align their national policies to the Union positions, the Commission as the guardian of EU law cannot impose sanctions in case of non-compliance.
Third, soft obligations combine with soft enforcement. Weak forms of obligations are at issue when Member States have to engage in cooperation mechanisms, but not in achieving specific objectives. In this way, there is no direct application into national legislation, as in the case of regulations, but only an agreement to take the obligations into account when forming new policies. Forth, soft obligations can be combined with hard enforcement. This situation occurs when the instrument is not normatively formulated (as any piece of legislation is) or contains loose obligations but still needs to be enforced.

Argumentative Patterns in Commission Recommendations
The rules laid down in European Commission recommendations and the obligations imposed to Member States are in all cases supported by arguments. A prescriptive standpoint is employed advocating the implementation of certain measures (such as Measures 1…Measures n should be implemented) by the Member States, which doubt whether doing so is necessary. Characteristically, this standpoint is supported by pragmatic argumentation indicating that a certain desirable result will be achieved (for example, Certain problems at Member State level will be solved). This way of arguing results in an 'argumentative pattern' [23,24] determined by the specific institutional constraints of the EU legal framework which imposes providing argumentation for an advocated course of action. This way of arguing in which pragmatic arguments support a prescriptive standpoint covers only the first level of defense and therefore constitutes a 'basic' argumentative pattern [24: 6], which in argumentative reality becomes much more elaborate and includes further levels of defense which are much more complex. However, identifying a basic argumentative pattern permits to find commonalities among different concrete instances of recommendations, as well as to identify regularities in terms of the main arguments employed in these documents.
Our close argumentative analysis of recommendations enacted by the European Commission 8 indicates indeed that a course of action is proposed in the form of measures M1…Mn. Two main arguments are advanced to support the main position proposed by the European Commission (standpoint 1). Besides arguing that the proposed measures are the right way to solve an existing or potential problem (argument 1.1a), a coordinatively related argument is advanced (argument 1.1b) pointing at the advantages of following the proposed course of action (argument 1.1b). The first argument (argument 1.1a) is in turn supported by four interrelated arguments which together argue why the proposed measures will solve an existing problem. First, the proposed measures deal with issues which contradict EU fundamental values (stated and agreed upon in the Treaties) concerning as varied aspects as the citizens' well-being, consumer protection, equal pay between men and women, protection of children, etc. (argument 1.1a.1a). Second, they help remove the variations in the way in which the Member States deal with the identified issues (argument 1.1a.1b). Third, the proposed measures have been adopted in accordance with the right legal basis (argument 1.1a.1c) as mentioned in the Treaty for the Functioning of the European Union (argument 1.1a.1c.1a) and they do not contradict the principle of subsidiarity (argument 1.1a.1d). This argumentative pattern is represented here in a nutshell. Three arguments in the outlined argumentative pattern giving certain rights to and creating particular obligations for the Commission and the Member States are particularly striking and require further clarification and discussion, namely Footnote 8 continued Since the study does not investigate specific words and phrases, but concentrates instead on the reconstruction of arguments and standpoints in context, as well as on the specific connection between arguments, the qualitative analysis is based on manually retrieved data. 9 The square brackets in the argumentative pattern indicate that this argument is sometimes not presented in the argumentation. See more elaborate discussion of this argument in this section. argument 1.1a.1c, argument 1.1a.1c.1 and argument 1.1a.1d. These arguments are connected to the three fundamental principles governing decision-making in the European Union. First, the principle of conferral states that the EU is a union of Member States and all its competences are voluntarily conferred upon it by its Member States (Article 5(1) and 5(2) TEU). Second, the principle of subsidiarity defines the circumstances in which it is preferable for action to be taken by the Union rather than the Member States (Article 5(3) TEU). Third, the principle of proportionality specifies that the content and form of the proposed action does not exceed what is necessary to achieve the objectives set by the Treaties (Article 5(4) TEU).
The principle of conferral is pointed at in the argumentative pattern characterizing Commission recommendations by indicating that the Commission has adopted a recommendation in accordance with the right legal basis (argument 1.1a.1c). In legal terms, the EU has the competence to propose a course of action explicitly agreed upon in the Founding Treaties 10 by all Member States only when there is a legal basis provided for it. This means that the Union has competence only when it can point to having power in a policy area cited in the Treaties. Related to this issue, Commission recommendations indicate a picture that reveals both cases in which the legal basis is not mentioned at all (and the argument is therefore not employed at all) and cases in which a specific legal basis is mentioned, as well as the procedure prescribed therein, such as consulting the European Parliament.
The case in which the legal basis is not mentioned at all is not viewed as surprising by legal scholars [17: 88] who recognize the function of recommendations as para-law instruments. As explained in Sect. 2, recommendations are adopted sometimes as alternative instruments to legislation because there is no legal basis on which legislation can be adopted. This raises, however, concerns whether the fact that no legal basis is available for legislation implies automatically that soft law instruments, such as recommendations, can be enacted as alternatives to legislation without being based on a legal basis. The question is whether enacting recommendations, and soft law instruments more generally, as para-law instruments, does not endanger the legitimacy of the European Commission as an institution and does not possibly lead to abuses of power. According to Article 288 TFEU, the Commission has the right to adopt non-binding instruments, but this does not imply that it may do so at its own discretion without relying on the conferred power to do so. 11 The Commission has been assigned a certain role and power in the EU institutional framework, and making decisions outside this framework can endanger its legitimacy. In argumentative terms, not mentioning the legal basis on 10 The Treaties is used in EU law to refer to the Treaties of the European Union on which every action taken by the EU needs to be founded: Treaty of Lisbon, Maastricht Treaty, Treaty of Nice, Treaty of Amsterdam, Single European Act, Brussels Treaty, Treaties of Rome and the Treaty establishing the European Coal and Steel Industry. For a full explanation of the different treaties and their content, see website of the European Union: www.europa.eu. 11 On the background of a discussion on the implications of institutional balance in the EU, [17: 88-89] supports the idea that Commission recommendations should always mention the legal basis in order to show that its role in the institutional system as initiator of legislation and non-binding acts is not abusively broadened.
Evading the Burden of Proof in European Union Soft Law… 89 which the recommendations have been enacted amounts to an evasion of the burden of proof: the Commission should argue that it has a power as conferred in the Treaties, but instead falls sometimes short of doing so. Our evaluation of the Commission's avoidance to state the legal basis as an evasion of the burden of proof is reinforced by the existence of many cases in which the legal basis is mentioned. The recent recommendations mentioning the legal basis on which they were adopted make reference in all cases to Article 292 TFEU according to which ''the Commission […] shall adopt recommendations''. This is the only provision in TFEU that explicitly confers on the Commission the power to adopt recommendations when necessary. Although the content of this provision is rather vague, because it does not explicitly indicate the adoption procedure, it makes it possible to control, at least on the surface level, whether the Commission has made appropriate use of the exercise of powers conferred upon it. What is intriguing in the case of recommendations is that they cannot be annulled in case the wrong legal basis is selected. Unlike in the case of directives [1], Article 263 TFEU excludes recommendations from CJEU's jurisdiction, making it impossible to refer an action of annulment to this Court.
Turning to the principle of subsidiarity, it seeks to protect the capacity of the Member States to take decisions and authorizes intervention by the Union when the objectives of an action cannot be sufficiently achieved by the Member States, but can better be achieved at Union level (Article 5(3)-(4) TEU). The principle applies in areas in which the EU does not have exclusive competence as mentioned in Article 3 TFEU. In areas such as customs union, common commercial policy, monetary policy and competition, the Commission has exclusive competence, as well as in those areas in which action on the part of the Member States is not by definition possible, such as rules for access to documents of EU. In all other areas, the EU does not enjoy exclusive competence. In the Commission recommendations we examined, this principle is sometimes mentioned explicitly (argument 1.1a.1d) 12 and sometimes implied 13 in arguments which recognize that the competence on a certain matter rests with the Member States but a common European framework is preferred. 14 The same principle is implied when arguing for the necessity of taking the proposed measures (argument 1.1b), because Member States cannot deal effectively with an issue on their own (argument 1.1b.1b). Notice that recommendations do not refer to the principle of subsidiarity even in those cases in which the EU does not have exclusive competence. The arguments in which the principle of subsidiarity is mentioned justify the choice of the recommendation over hard law, particularly in those cases in which the Treaty leaves open the choice of instrument. The argument is presented most often as if the recommendation is not only necessary but also sufficient for solving a certain issue, suggesting that there is no need for proposing and adopting legislation. Instead of arguing why the different approaches taken at national level (argument 1.1a.1b) do not and cannot solve a certain issue, the Commission merely outlines a problem which requires common standards at European level in the way prescribed in the recommendation. Instead of showing that the recommendation is necessary because the issue cannot be solved satisfactorily at Member State level and the Union's intervention is therefore justified, more often than not it simply mentions that taking different approaches to an issue is problematic. This way of arguing is again indicative of an evasion the burden of proof with respect to the principle of subsidiarity. The Commission should present in the recommendation both the advantages and the disadvantages that Union action may have relative to Member States' action and should outline the situations in which EU action produces clear benefits compared to action at Member State level by reason of its scale or effectiveness, but fails short of doing so. 15 In addition to the principle of conferral and the principle of subsidiarity, EU decision-making is governed by the principle of proportionality (Article 5(4) TFEU) which ensures that the policy approach and its intensity match the identified problem. Commission recommendations are by definition less interfering instruments as compared to hard law, perhaps ''the least interfering means of regulation'' [17: 87]. This might be the main reason for the Commission not to mention this principle at all in the recommendations that we have examined. While indeed it might be the case that by choosing a recommendation, the Commission does not go beyond what is necessary to solve a problem satisfactorily, because it is one of the simplest form of imposing measures on Member States, it remains unclear whether the scope of the recommendation is limited to those goals which the Member States cannot achieve on their own and where the Union can do better. Moreover, in some cases recommendations can be too soft for solving a certain issue, which raises concerns about whether the recommendation is to be preferred over legislation. The Commission seems to prefer an alternative to legislation which does not exceed what is necessary to achieve its purposes, as implied by enacting recommendations, but this alternative is not necessarily appropriate for solving the identified issue effectively. Also in this case, the Commission systematically evades the burden of proof by not arguing for the principle of proportionality.
From the discussion of the three fundamental principles for EU decision-making, we can conclude that the Commission evades its burden of proof by avoiding to argue for the duties which have been institutionally imposed on it. The Commission either does not provide any arguments at all or provides arguments which do not support the standpoint related to the principle at issue.
While we can conclude from this that the Commission does not fulfill some of its own institutional obligations, the argumentation in the recommendations indicates at the same time that the Commission imposes certain obligations on the Member States. The prescriptive standpoint recommending Member States to take certain measures imposes such obligations on the addressee. Most legal scholars [11,16,18] and political scientists [9] conclude that by imposing obligations recommendations resemble legislative provisions and have ''an intended legally binding force'' [17: 94]. This criticism, going as far as attributing to the Commission intentions to legislate by the backdoor, is supported by ''the imperative wording'' [17: 95] and the fact that Member States are expected to implement and comply with the prescribed course of actions whose application is controlled by the Commission.
Nevertheless, a close look at the way in which the arguments are presented in Commission recommendations points towards a different conclusion. In our analysis, we have at all times indeed come across directive speech acts by means of which the Commission makes an attempt to get the Member States to adopt some measures. How else could a recommendation be made if not by requesting a course of action by means of imperative wording? But, unlike in legislative provisions, the directive speech acts advanced in recommendations have characteristically a reduced degree of strength. The modal ''should'' is in all cases employed to prescribe a certain behavior on the part of the Member States, but without the strong obligatory force of ''must'', ''will'' and ''shall'' characteristic of legislation [4]. In Terpan's typology of types of obligations related to soft law, we can place this usage under ''soft obligations'' occurring in situations in which Member States are invited to engage in cooperation activities. The existence of soft obligations in recent Commission recommendations corresponds in fact perfectly with their goal of achieving cooperation among Member States [16: 179], aimed at promoting better coordination of national policies rather than at establishing certain national legislation. Contrary to what most scholars are tempted to believe, having no legally binding force does not imply that recommendations cannot impose certain obligations. After all, recommendations produce practical and legal effects, as unanimously recognized in the literature, and these effects cannot be produced other than by imposing some obligations. 16

Implicit Material Premises as Shared Starting Points?
The argumentative pattern outlined in the previous subsection represents a complex structure made of ''a constellation of argumentative moves [23: 14], constituted by ''a particular combination of argument schemes in a particular kind of argumentation structure'' [23: 14]. In Sect. 3.1, we outlined all the arguments present in the pattern and, subsequently, we discussed if and how those arguments that are closely related to the EU principles of conferral, subsidiarity and proportionality, the kind of rights and obligations this creates and how this sometimes amounts to an evasion of the burden of proof by the Commission. It emerged that EU recommendations often evade the burden of proof.
Such evasions of the burden of proof depend on the omission of arguments that would be legitimately expected. In this section, we wish to make a further step and identify violations of the burden of proof that can be found in those arguments that the EU actually provides. In order to do so, we need to characterize the inferential configuration of the single argumentations that are present in the identified argumentative pattern according to the Argumentum Model of Topics (AMT [13]). The analysis of implicit premises within the inferential configuration of arguments permits to identify further cases in which the Commission evades the burden of proof by selecting implicit starting points that might not be accepted by the Member States. This analysis concerns a level of further detail and, therefore, can be combined with the reconstruction of an argumentative pattern; in fact, it specifies how the different arguments are connected one to another at an inferential level.
The AMT distinguishes between two types of (explicit or implicit) premises, i.e. starting points that are present within arguments. Premises of an inferentialprocedural nature define the type of principle of support on which the argument is based; these represent the sources of the inference (for example: analogy, meansend, cause-effect, effect-cause, etc.). Premises of a material nature define the contextual or ''cultural'' starting points, assumed to be shared between the interlocutors, on which the whole discourse is based. Material premises are based on interlocutors' shared starting points at different levels: culture, factual knowledge, customs, shared goals. Their acceptance depends on the specific audience to which an argument is addressed; in this sense, they are contextually embedded and dependent on the specific situation of the audience. According to the AMT, both procedural and material premises are always present in argumentation. What this distinction adds to the analysis of specific arguments in recommendations will become clear in the following of this section; violations of the burden of proof will be identified in relation to material premises.
Our analysis shows that the two main coordinative 17 arguments in the pattern identified (arguments 1.1a and 1.1b) justify the adoption of the recommendation by the Member States in different ways at the level of inferential-procedural premises. We found that each of these arguments includes a different type of evasion of the burden of proof at the level of material premises.
Before addressing such cases of evasion, we will briefly reconstruct the inferential configurations of both arguments. Argument 1.1a presents the recommendation as a necessary means to reach the important goal to solve a problem at the EU level; argument 1.1b argues that the recommendation should be accepted because of its positive consequences. Not surprisingly, as the standpoint is prescriptive and recommends a specific course of action, both arguments build on the relationship between an action and its goals and effects. Both are based on the locus of the final cause as a principle of support [12]. Loci are defined as the sources of arguments; more precisely, they instantiate a relationship between two extremes (means-end, cause-effect, analogous-analogous, etc.). Such a relationship is at the basis of the inferential-procedural component that is present in arguments. More in particular, in the case of argument 1.1a, the recommendation is presented as a means to solve a problem; in this sense, argument 1.1a represents a typical case of argumentation based on a means-end relation. The implicit inferential-procedural starting point at work in this argument is the following: ''if a means is the right way to reach an important goal, that means should be adopted''. This inferentialprocedural starting point, called maxim, represents one of the possible inferential rules connected to the locus from the final cause. In argument 1.1b, instead, the focus is on the relationship between an action and its positive consequences; in this case, the implicit maxim is: ''If an action brings about positive consequences for the agent, it should be accomplished''. Thus, we see that the two arguments advanced by the European Commission for the adoption of the recommendation, although both based on the same locus (the relationship between an action, its goals, and its effects), involve different inferential-procedural starting points. Both inferentialprocedural starting points are acceptable in abstracto; as we will detail in what follows, what is problematic is how they have been applied by the Commission to the specific arguments advanced.
As a whole, the inferential-procedural part of argument 1.1a (see Sect. 3.1) can be reconstructed as follows: ''If a means is the right way to reach an important goal, that means should be adopted; and measures M1…Mn proposed in the recommendation are the right way to reach an important goal; therefore, measures M1…Mn should be adopted''. The inferential-procedural part, however, only describes a part of the implicit starting points that constitute this argument. Two further implicit material premises are necessary in order to state that ''measures M1…Mn proposed in the recommendation are the right way to reach an important goal''. First, a general premise (endoxon) is needed, which defines that ''solving existing/potential problems at the level of Member States is an important goal''; second, it is necessary to reconstruct a factual premise (datum), stating that ''measures M1…Mn are the right way to solve existing/potential problems at the level of Member States''. Endoxon and datum are of a different nature from inferential-procedural starting points, 18 because they are not abstract inferential rules; they describe the material starting points on which the whole argument is based, which should be accepted by the interlocutors in order for the argument to have a hold on them. The endoxon is a general premise about an allegedly shared goal (solving existing problems), while the datum consists of factual information about the specific measures M1…Mn proposed in the recommendation.
The problem with argument 1.1a is at the level of the material starting points. In fact, if these material starting points are not really shared-if, for example, it is not accepted by the Member States that solving existing problems is an important goal, or if it is not accepted that there is a problem to be solved-then the whole argument will not adequately support the prescriptive standpoint contained in the recommendation. This is not because of how the argument is constructed at the inferentialprocedural level, but because of the material starting points: the latter must be shared in order for the argument to be acceptable. Therefore, it becomes very important for the Commission to argue for the acceptance of material starting points in order to assure compliance by the Member States.
In Sect. 3.1, we have shown that the Commission does provide subordinative arguments to give further backing to argument 1.1a; however, these are not sufficient to grant the addressee's acceptance of material starting points of 1.1a. In fact, subordinative arguments advanced by the EU (1.1a.1a and 1.1a.1b, see Sect. 3.1) only support the factual premise (datum): they support the fact that a problem (in the EU) exists. Both 1.1a.1a and 1.1a.1b are based on a locus from definition: the Commission, in fact, is using these subordinative arguments to define 19 which problem needs to be solved at the EU level. Both 1.1a.1a and 1.1a.1b support the datum of 1.1a, namely they show that there is a problem; but the endoxon of 1.1a, i.e. the fact that ''solving existing problems is an important goal'', remains unsupported. Basing a whole argument addressed to Member States on an endoxon that is not necessarily accepted counts as a kind of evasion of the burden of proof. It is different from the evasion that we discussed in Sect. 3.1. In this case, in fact, the Commission does not omit an argument, but they advance an argument as if the implicit material starting points (in particular, the endoxon of 1.1a) were accepted by Member States, while this might not be the case. In addition, the Commission does not provide subordinative arguments to show that the endoxon should be accepted.
A similar situation has been found when analyzing further layers of subordinative argumentation: more often than not, endoxa are left unjustified. In the documents that we have examined, whilst 1.1a.1a is supported by further subordinative argumentation, 20 1.1a.1b is rarely further supported. The endoxon in 1.1a.1b can be 19 The endoxa of these two arguments contain the definitions of what the Commission means by ''existing problem''. In 1.1a.1a, the Commission defines a problem as ''something that goes against agreed upon principles''; in 1.1a.1b, the problem is defined as ''variations in the way in which Member States deal with the existing issues''. 20 In such subordinative argumentation, the Commission explains what kind of principles have been agreed upon and why they say that there is a key issue that goes against these agreed upon principles. As an example, take Commission Recommendation 2013/112/EU on investing in children: breaking the cycle of disadvantage. In the opening paragraphs of this text, the Commission states that: ''Respect for human dignity is a founding value of the European Union, whose aims include promoting the well-being of its people; the Union must protect the rights of the child, combat social exclusion and discrimination, promote social justice and protection''. Immediately after, they note that ''Children are more at risk of poverty or social exclusion than the overall population in a large majority of EU countries''. It is clear that reconstructed as: ''variations in how Member States deal with existing issues are a problem''. But why variations should be a problem, in general, is not made explicit by the Commission. This endoxon is not self-evident. In fact, variations in how Member States deal with specific issues exist in several areas of policy in the EU (there are, indeed, many variations across Member States on a range of issues). Therefore, it is not upfront clear why variations in dealing with the specific issues considered in the recommendations should be necessarily deemed as negative. In this sense, the endoxon of argument 1.1a.1b often remains unjustified.
The analysis of the second main argument identified in the pattern (1.1b, see Sect. 3.1) confirms this problem. In this case, our close analysis of implicit premises reveals a problem due to a misalignment between maxim and endoxon. The maxim, in fact, is the following: ''if an action brings about positive consequences for the agent, it should be accomplished''. The endoxon can be reconstructed as: ''Consequences a, b, c, are positive'', whereby the specific contents of a, b, and c vary depending on each specific recommendation. The misalignment between endoxon and maxim lies in the fact that the agents who should accept the recommendation are Member States; but the positive consequences that are normally mentioned in Recommendations are positive for the EU and its Member States. The alignment between the interests of the EU and those of the Member States is generally presupposed. 21 It is true that, at least to some extent, the interests of the EU and those of the Member States should be aligned. For one, Member States might participate in the writing of Recommendations. Moreover, they are generally committed to be part of the EU. Nevertheless, this does not imply that a positive consequence for the EU will necessarily have a positive impact on each and every Member State; conflicts of interests are likely to occur in some cases. 22 In this sense, the positive consequences mentioned in argument 1.1a might not be felt as such by the Member States that are asked to accept a recommendation. Also in this case, an evasion of the burden of Footnote 20 continued they are showing how a state of affairs (children being more at risk than the rest of the population) contradicts an agreed upon principle (protection of the rights of the child, etc.). In a similar vein, Commission Recommendation 2014/1405/EU on strengthening the principle of equal pay between men and women through transparency at first lists all the documents and legislation in which the principle of equal pay is accepted; and then provides evidence that: ''Women in the Union still earn an average of 16.2% less than men for each hour worked (Eurostat 2011), in spite of their significant progress in terms of educational achievements and work experience''. 21 For example, in the Commission Recommendation 2013/112/EU on investing in children: breaking the cycle of disadvantage, the positive consequences mentioned include the following: ''The Europe 2020 Strategy has given new impetus to efforts to address poverty and social exclusion in the EU, by setting a common European target to reduce the number of people at risk of poverty and social exclusion by at least 20 million by 2020''. It is not clear if this target is the Member States' target or if it is imposed onto them. In the Commission Recommendation 2014/1405/EU, it is said that action at the EU level ''would help national authorities and relevant stakeholders to step up their efforts in tackling the gender pay gap and pay discrimination through better implementation of current legal requirements''. Although here an argument is given to Member States, it is still not explained how better implementation will occur; also, the Member States' efforts and their efforts to step them up are taken for granted. 22 This idea is confirmed by the existence of numerous litigation cases before the Court of Justice of the European Union (see [20] for a full list of such cases until 2013). proof might be identified, though not in absolutely in all cases, if one considers that the Commission gives an argument that is focusing on advantages for the EU rather than advantages for the intended audience of this document. This again might impact on how arguments are received by the Member States.
With respect to Sect. 3.1, the AMT analysis qualifies a different type of evasion of the burden of proof, relative to the management of implicit starting points. In the case of argument 1.1a and its subordinative arguments, this depends on the fact that premises that are not self-evident (notably, endoxa) are nevertheless not further supported. In argument 1.1b, it depends on an assumption about the coincidence of interests between the EU and Member States, which is an unexpressed starting point that might not hold in all cases. In both cases, supporting an argument on the basis of material starting points that might not be really shared by the addressees and that remain unjustified could affect compliance. In fact, if the ultimate material premises were not accepted by Member States, the whole argument would become less persuasive. Based on the principle that persuasion in soft law instruments is due to arguments (see Sect. 1), these aspects might reveal weaknesses concerning the ultimate grounds on whose basis the Commission asks Member States to comply with the recommendations.

Conclusion
In this paper, we analyzed and evaluated the arguments that are central in the recommendations enacted by the European Commission. Our detailed explanation of the specific content, purpose and function of recommendations in the EU legal framework has provided the institutional framework against which we dealt with the recommendations as argumentative instruments. We made clear the main arguments employed by the Commission when proposing certain measures to the Member States, as well as the way in which these arguments connect to each other. We have shown that more often than not the Commission evades the burden of proof by ignoring to fulfill its fundamental obligations imposed by the legislative framework. We also delved into the implicit premises in the argumentation, which enabled us to identify further cases of evasion of the burden of proof at a different level; these are due to the Commission's use of implicit starting points which might not be accepted by the Member States.
This argumentative analysis and evaluation could be further extended. The framework proposed in this paper can be applied to all types of recommendations, including instruments complementing, replacing or preceding legislation. In view of the specific function of these instruments, the analysis and evaluation of the argumentation could bring to light slightly different argumentative patterns based on other implicit premises than those we identified. If this intuition is correct, then judging the quality of the arguments will also lead to different results. In any case, examining more recommendations has the advantage of increasing the generalizability of our conclusions. Going one step further, similar studies can be carried out for all other types of soft law instruments, each of which has its own characteristics affecting the argumentation.
Evading the Burden of Proof in European Union Soft Law… 97 Apart from these future practical applications, our study can be further extended on a theoretical level. In this paper, we have integrated the pragma-dialectical notion of argumentative pattern with the reconstruction of inference as developed in the Argumentum Model of Topics. Our analysis has shown that these two theoretical instruments might profitably be combined in the analysis of argumentation. A new theoretical issue that has emerged as we were working at this combination is the interpretation of ''coordinative argumentation''. Coordinative arguments, in fact, might conceal different inferential configurations and should therefore be interpreted in various different ways at the level of loci and maxims. This could help explain more specifically in what ways coordinative arguments work together to support a standpoint. Proceeding with the analysis of other cases, we might be able to give a more refined definition of coordinative argumentation and the different inferential configurations associated to it.