Are integrated impact assessments the way forward for mainstreaming in the European Union?

This article assesses the utility of impact assessments (IAs) as an effective tool for mainstreaming. Speci ﬁ cally, it analyses whether the European Union ’ s (EU ’ s) system of integrated impact assessments (IIAs) contributes to the realization of six mainstreaming objectives de ﬁ ned in the EU treaties. The article ﬁ rst studies whether the legal framework for the EU ’ s system of IIAs makes it a viable tool for mainstreaming. It then proceeds with an empirical analysis of 35 IAs to assess the extent to which mainstreaming objectives are taken into account in practice. The analysis shows that all six mainstreaming objectives have a place within the IIA system, but the system does not ensure systematic consideration of them. There is considerable variation across mainstreaming agendas, as well as across Directorates-General. The article concludes that the IIA system is no panacea for mainstreaming, and proposes changes to the institutional framework to improve its mainstreaming potential.


Introduction
One of the main institutional challenges for modern governance is the increased differentiation of modern society, which requires the coordination of policy interventions across many areas of social reality.Sectoral differentiation is a challenge for governance at all levels, but particularly at the European Union (EU) level.The functional nature of European integration led to the EU's development as a patchwork of sectoral integration initiatives, which is reflected in the EU's institutional design.The Council of the EU is organized along sectoral lines, and the European Commission is divided into strong sectoral Directorates-General (DGs) with often opposing policy agendas and frameworks (Pollack & Hafner-Burton 2010;Hartlapp 2011).As the EU has intervened in ever more policy areas, the need for policy coordination has also increased.As a result, "mainstreaming" has become popular in EU policymaking, particularly since the 1990s.When a political objective is "mainstreamed", it becomes horizontally applicable across all policy areas.In concrete terms, this requires that policy actors working beyond the policy area of the objective must also account for the protection or promotion of the mainstreamed objective in question.First introduced in relation to environmental policy concerns, other horizontal (i.e.cross-sectoral) mainstreaming objectives have subsequently been formulated, such as gender equality, non-discrimination, social concerns, consumer protection, and fundamental rights.Literature on mainstreaming in the EU has focused mainly on gender and the environment, identifying an array of different tools for mainstreaming.Among these tools, impact assessment (IA) is considered one of the more potentially effective; however, there is little scholarship on the actual functioning of IAs as part of a mainstreaming strategy.In addition, literature on the EU's system of integrated impact assessments (IIAs) has not addressed the use of this tool in the function of mainstreaming agendas.
This article confronts this gap in the literature by analyzing to what extent the European system of IIAs can and does function as an effective mainstreaming tool.Mainstreaming is not necessarily an explicit objective of an IA system, and is not likely to be the primary objective; however, in theory, the IA, and, particularly, a system of IIAs, may be an especially useful tool with which to pursue mainstreaming objectives.Indeed, IAs aim to define the objectives of a particular policy initiative, to set out the different policy options, and to provide an impact assessment of these different options.By definition, such ex ante screening of both objectives and impacts of new policy initiatives requires that there is assessment of the impacts of the initiative beyond the sector in which it originated.Therefore, this systematic, horizontal screening of all main policy initiatives is a potentially useful tool for mainstreaming, as it enables the assessment of whether the policy initiative is likely to have a positive, negative, or negligible impact on the mainstreaming objectives.In this article, we analyze whether the European system of IIAs is, indeed, a useful tool for mainstreaming and whether it can overcome several of the weaknesses the literature has highlighted in relation to other mainstreaming tools.
We focus our analysis on the six mainstreaming objectives that have been constitutionalized in the EU treaties.Following the Lisbon Treaty, the Treaty on the Functioning of the European Union (TFEU) includes a new Title II, "Provisions having general application", with a set of horizontal mainstreaming clauses.Introduced by the provision that the Union shall ensure consistency between its policies and activities (Article 7 TFEU), five mainstreaming objectives are presented to be taken into account in all EU policies: gender equality (Article 8); the horizontal social clause, which includes promoting a high level of employment, adequate social protection, the fight against social exclusion, and a high level of education, training, and protection of human health (Article 9); non-discrimination on the basis of gender, racial or ethnic origin, religion or belief, disability, age or sexual orientation (Article 10); environmental policy integration (EPI) for sustainable development (Article 11); and consumer protection (Article 12).1 Most of these mainstreaming agendas predate the Lisbon Treaty, often initially set out in soft law documents, and subsequently "constitutionalized" at different moments in time; but it is the Lisbon Treaty which brought them together at the start of the TFEU as a common set of horizontal provisions.We add to our analysis a sixth constitutionalized mainstreaming objective (although not presented with the other horizontal clauses), namely, securing fundamental rights.Article 6 of the Treaty on European Union (TEU) affords the Charter of Fundamental Rights of the EU ("the Charter") the same legal status as the EU treaties themselves, thus, making all EU action clearly subject to the respect of fundamental rights.Moreover, the Charter itself (Article 51) requires all EU institutions and bodies to respect and promote the rights of the Charter.In addition to this, the Commission has adopted soft law documents to ensure that fundamental rights inspire EU action throughout the policy cycle.The text of the horizontal clauses and Charter does not suggest that there is any hierarchy between these mainstreaming objectives.Hence, the constitutional expectation is that all of them are taken into account to a similar extent.
To assess the extent to which the European system of IIAs can and does function as an effective mainstreaming tool, our article proceeds as follows.Based on the literature on both mainstreaming and IAs, Section 2 formulates four hypotheses about how the European system of IIAs may provide an answer to the main shortcomings that have been identified in the literature in relation to other mainstreaming tools.More precisely, we argue that, in theory, the IIA system has the potential to ensure screening of new policy initiatives in function of the mainstreaming agendas, both at the level of defining the objectives and the impacts of these initiatives; and could address concerns about the soft character, the issues of overload, and the overly bureaucratic nature expressed in relation to mainstreaming.Whether the European system of IIAs lives up to these expectations is subsequently assessed at two levels, namely, by analyzing both the institutional framework and practice.In section 3, we begin by analyzing the institutional architecture of the IIA system, which is set out in the guidelines on IAs.Through a legal hermeneutical interpretation of the official texts, taking into account the relationship between them and questions around the hierarchy of norms, we assess whether the guidelines provide a suitable framework for the IAs to function as an effective mainstreaming tool for the EU's six mainstreaming agendas.We argue that, in contrast to the "constitutional expectation," the six mainstreaming objectives are not given equal attention.While social and environmental concerns are primary objectives of assessment of the IIA system, fundamental rights constitute a more ad hoc horizontal category; and consumer protection, gender, and nondiscrimination constitute subcategories that are even less systematically addressed.Moreover, while the six mainstreaming objectives receive attention in the IIA institutional set-up, other objectives receive at least as much attention.Indeed, both the assessment of economic impacts and of regulatory burdens are predominant in the set-up of the IIA system, although neither of these are set out in the treaties as constitutional horizontal objectives.
Section 4 then provides an empirical analysis of how IAs are used in practice.Analyzing a selection of 35 IAs (from seven DGs) we explore how and the extent to which the mainstreaming objectives have been incorporated into the IIA process.In addition to the four hypotheses based on the literature (as set out in section 2), the empirical analysis makes use of a set of additional hypotheses relating to the institutional framework studied in section 3 (e.g.whether being a primary or subcategory of assessment in the IIA guidelines makes a difference to whether a mainstreaming objective is taken into account).Finally, we formulate two extra hypotheses to explain the differences in mainstreaming practice between DGs.
We conclude that the IIA system does not ensure the systematic screening of the six mainstreaming objectives.Only three out of the 35 IAs we analyzed referred to all mainstreaming objectives; and some DGs completely ignored several mainstreaming objectives in all their IAs (i.e. of the 35 analyzed).As primary categories of the IIA system, the social and environmental concerns fared better than other mainstreaming objectives.However, being such a primary category of the IIA system does not ensure that an objective is always taken into account.Gender and non-discrimination fared most poorly, with gender showing the biggest gap between institutional prescription and practice.We conclude that the system of IIAs is no panacea for mainstreaming, but we are able to provide some policy recommendations to improve its use as a tool to support the constitutionalized mainstreaming objectives.
2. The critique of mainstreaming tools and how a system of integrated impact assessments (IIA) might provide a solution Three critiques are regularly made in the literature on mainstreaming.Firstly, mainstreaming is usually implemented using soft and persuasive policy instruments, such as the training of officials, internal guidelines, the appointment of officials with "mainstreaming responsibilities" (often weakly defined), the creation of meeting groups (often low-profile and with weakly structured agendas), and soft communications intended for policy actors involved in implementation.Furthermore, there is little to encourage policy actors to employ or engage with these soft policy instruments, by way of either "carrots" or "sticks" (Hafner-Burton & Pollack 2009, p. 115).This mixture of soft policy and weak incentives has proven a challenge in terms of implementation, and is surely in large part responsible for the patchy success of the longer standing mainstreaming agendas.Indeed, empirical research into both gender mainstreaming (e.g.Kantola 2010) and EPI (e.g.European Environment Agency 2005) has shown that, at best, implementation has been mixed.
Secondly, over recent years there has been quite a proliferation of mainstreaming agendas, as illustrated by the EU's six constitutionalized agendas.Moreover, additional or partially overlapping mainstreaming agendas have been promoted by the Commission in soft law documents, such as "children's rights mainstreaming" (Commission of the European Communities 2006) and "climate policy mainstreaming" (European Commission 2014a).Unsurprisingly, then, it has been said that in "EU circles" reference is made to the "mainstreaming of mainstreaming" (Pollack & Hafner-Burton 2010, p. 309), and scholarship has cautioned the risk of "mainstreaming overload" (see e.g.Geyer & Lightfoot 2010;Allwood 2013).The impact upon decisionmaking and decisionmakers in terms of accommodating this increasing array of agendas must not be overlooked.In particular, there is a risk that older mainstreaming agendas lose their novelty value and risk being eclipsed by newer mainstreaming agendas.
Thirdly, the literature on mainstreaming, and on gender mainstreaming in particular, has often criticized existing tools as merely technocratic internal coordination instruments.Instead, in its more transformative manifestation, over time, mainstreaming changes the nature of mainstream decisionmaking itself, as consideration of mainstreamed objectives becomes embedded within the culture of an organisation: "gendering" and "greening," policymaking, and so on.In order for such transformative mainstreaming to occur, bureaucratic coordination mechanisms are not sufficient and broad participatory processes are required (e.g.Jahan 1995;Beveridge & Nott 2002;Rees 2005;Walby 2005).A more participative, "agenda setting" interpretation of mainstreaming will "reorient the nature of the mainstream" (Jahan 1995, p. 13), as opposed to an "integrationist" interpretation, which sees the mainstream objective simply integrated within (and subsumed by) other policy priorities.Successful institutionalization of a transformative and participatory mainstreaming agenda protects the corresponding policy objectives from both the short-term agenda of bureaucracies and the ebbs and flows of political currents, which push them in and out of favor.However, scholarship highlights that there is a tendency for mainstreaming to be understood as a technocratic exercise that resembles a more integrationist (as opposed to a more transformative, agenda setting) interpretation (e.g.Mazey 2002;Lombardo & Meier 2006).
In light of these common critiques of mainstreaming (soft policy and weak incentives; mainstreaming overload, and prioritization of technocratic over more participatory tools), how could a system of IIAs contribute to more efficient mainstreaming?Regulatory impact assessments (RIAs) have been introduced in many countries over the last decade.They may be used for many reasons, such as providing sound evidence for policymaking, ensuring cost-effective policymaking, reducing the regulatory burden, and facilitating participation (Rowe 2006;Bäcklund 2009;Dunlop et al. 2012).IAs have often been introduced within a broader policy agenda to reduce the regulatory burden on companies and society (Hertin et al. 2009, p. 415).In that context, mainstreaming may not be the primary objective of IA.Yet most IA systems do include more or less developed aspects of policy coordination, which provide opportunities for it to be used as an instrument for mainstreaming.Although the scope of IAs can be narrowly defined, that is, within the confines of the policy area in question, most of the time policy measures have (potential) impacts across different policy sectors.Therefore, assessing impacts includes at least some level of scrutiny across policy areas.This is certainly the case with the EU system of IAs, which aims to provide "integrated impact assessments." Previously, the Commission used separate IAs at a sectoral level, on an ad hoc basis, including business, gender, environmental, small and medium enterprises (SMEs), and trade IAs (Commission of the European Communities 2002, p. 3).In 2002, the Commission introduced a system of IIAs, which became fully operational in 2004.The sector specific IAs were mainstreamed within one IIA: a horizontally applied, ex ante policy instrument for all legislative and major policy measures, to analyze both benefits and costs, and to address all significant economic, social, and environmental impacts of prospective initiatives.As the TFEU has constitutionalized the mainstreaming objectives as objectives to be taken into account in all areas of EU intervention, it is reasonable to expect that the IIA'ssystematic"integrated" screening of all main policy initiatives would acknowledge this constitutional requirement.
Such a proceduralized, ex ante screening of all new legislative and main policy initiatives can function in support of mainstreaming in two ways.Firstly, European IAs must start with the definition of the policy problem and objectives of the new policy initiative.This provides an opportunity for the systematic consideration of the mainstreaming objectives at this initial stage of policymaking in order to steer new policy initiatives in the direction of the mainstreamed objectives.Secondly, IAs assess the impacts of different policy options.Impacts can be assessed on the basis of many criteria.If mainstreaming is taken seriously, IAs systematically assess new policy initiatives on their potential impacts upon the realization of mainstreaming goals.A proactive approach to mainstreaming would take into account the mainstreaming objectives both at the level of defining objectives of a new policy initiative and of assessing its impacts.
Looking more closely, the use of IIAs may potentially address the three main shortcomings formulated in relation to mainstreaming instruments in the literature, namely soft policy and weak incentives, mainstreaming overload, and lack of broader participatory capacity.First of all, the system of IIAs is strongly institutionalized and applied systematically.This contrasts with the predominantly soft and ad hoc nature of many mainstreaming instruments.Indeed, in the literature on mainstreaming, IAs are addressed as a promising "hard" instrument, compared to other "soft" mainstreaming instruments (Pollack & Hafner-Burton 2010).The IIAs are systematically applied to all policy areas, at least for all legislative initiatives and for all main policy initiatives (although some discretion is left to the Commission here).All Commission DGs have either specific units or identified staff to deal with IAs.The procedural requirements for the European IIAs are set out in the 2009 Commission Guidelines on IIAs.These were drafted by the Secretariat General of the Commission and create procedural expectations for the Commission DGs when undertaking IAs.An IA, for instance, has to be structured in a particular way (procedural issues, problem definition, objectives, policy options, impact analysis of options, comparing options, and evaluation provisions), must follow particular procedures (such as setting up an IA Steering Group among DGs concerned with the topic, or ensuring some form of consultation and use of relevant expertise), and must take into account substantive concerns (economic, environment, social, and regulatory burdens).Strictly speaking, the European system of IIA is not "hard law," as its procedural framework is set out in "guidelines" that are not legally binding upon either the Commission or its officials.Indeed, the chances that the Court of Justice will annul a European regulatory act because of a failure to respect procedural requirements in relation to the IIA preceding that act are extremely low, although, hypothetically, it is not entirely out of the question (Alemanno 2011).Nevertheless, Commission DGs are under high pressure to respect the (main) procedural aspects set out in the guidelines, and sticks are in place to ensure such proceduralization.For instance, non-respect of the guidelines can become subject to a European Ombudsman inquiry for bad administrative practice.Most importantly, the Commission has set up its own procedural watchdog to control respect of the guidelines, by way of the Impact Assessment Board (IAB).The IAB is composed of senior Commission officials, at the level of the Secretariat General, and, thus, in a superior position to the DGs drafting the IAs.All IAs are screened by the IAB, and DGs are often asked to redraft part of their IA in light of procedural comments made by the IAB.Hence, compared to other mainstreaming instruments, the IA is a systematically applied and highly proceduralized tool.Whether IAs also function as a strong proceduralized tool for mainstreaming depends upon the extent to which mainstreaming objectives are taken up in the guidelines and are part of the proceduralization process.
Secondly, the EU's IA system may respond to the critique of "mainstreaming overload" as, compared to IA systems in other countries, the EU's system is characterized by its "integrated" nature.European IAs do not simply provide an economic cost-benefit analysis, but require a broad assessment on economic, social, and environmental grounds, at least.A broad IIA system, which is systematically applied at the start of all legislative and main policy actions, has the potential to address different policy priorities simultaneously, such as the constitutionalized mainstreaming objectives.
Thirdly, in response to the need for participatory practices to realize more transformative mainstreaming, another feature that sets the European system of IIA apart from IA systems in other countries is its relatively strong attention to participation within the IIA procedure.Although the Better Regulation agenda was established in the context of the EU's Lisbon Strategy (which focused particularly upon increasing competitiveness) this happened at a time when the Commission was also aiming to increase the legitimacy of its decisionmaking by paying more attention to the role of civil society and participatory procedures, particularly visible in the 2001 White Paper on European Governance (Radaelli 2007).The guidelines on IAs present several provisions regarding the types of actors to be involved.An IA Steering Group has to be set up for the drafting of each IA, bringing together the DGs concerned with the topic at hand.While the Steering Group reflects a more bureaucratic understanding of mainstreaming, the guidelines also include a two-page section on the consultation of interested parties.Such consultation is considered "an obligation for every IA" and "must follow the Commission's minimum standards of consultation" (European Commission 2009, p. 19).This allows for a more participatory approach to IAs and, therefore, a more transformative approach to mainstreaming.
Hence, based on the literature on both IAs and mainstreaming, four broad hypotheses inform our analysis: H1: Mainstreaming objectives are expected to be taken into account in IAs, both at the level of defining policy objectives and assessing the impacts of new policy initiatives.
H2: Compared with other mainstreaming instruments, the strong proceduralization of the IA system is expected to ensure more systematic mainstreaming.
H3: The integrated approach of the IIA system is expected to provide the opportunity to "streamline mainstreaming" and avoid overload.
H4: The participatory requirements of the IIA system are expected to secure high levels of participation, contributing to more transformative mainstreaming.This is not to say that a system of IIA is the complete solution to effective mainstreaming.It is worth remembering that, at best, IAs can only be one tool among others to ensure mainstreaming.IAs are focused on the initial coordination stage of new policy initiatives, while mainstreaming relates to both the definition and implementation of policy.Whether IAs function as a useful tool for mainstreaming at the stage of drafting new policy depends on whether mainstreaming objectives are given appropriate attention within the IIA institutional framework, which we assess in section 3, and how this is implemented in practice, which we assess in section 4.
3. The institutional architecture of the IIA system: A viable tool for mainstreaming?3.1.Overview of the soft law framework: No explicit mainstreaming objective The institutional architecture of the IIA system is presented in a set of three types of documents, available on the Commission's IA website (European Commission 2014b); namely, the General IA Guidelines of 2009, the Annexes to these guidelines, and a set of operational guidelines.The general guidelines and its annexes are drafted by the Commission's Secretariat-General.The operational guidelines provide guidance in relation to a particular type of impact and are drafted by the DG most involved in that field.Operational guidelines exist for social impacts, fundamental rights, competitiveness, territorial impacts, and impacts on micro enterprises.
From a legal perspective, the general guidelines, its annexes, and the operational guidelines are all soft law documents, and there is no hierarchy between them in terms of which are more legally binding.However, our contacts with Commission officials drafting the IAs (i.e. at the level of the DGs)2 suggest that they consider the general guidelines to be more "binding," as they are drafted by the Secretariat-General who has hierarchical control over the DGs.The Commission officials seem to focus primarily on the core text of the general guidelines, which is already quite detailed, rather than following all of the detailed suggestions set out in the annexes.This may also wellbethecasewiththeoperationalguidelines,drafted"merely" at DG level.The IAB considers the operational guidelines as complementary to the main guidelines.Indeed, "[t]heir use is left to the discretion of services preparing the IAs" (European Commission 2012, p. 30).The screening practice of the IAB also confirms this, as, to date, it has focused on the main structure and principles advised in the general guidelines.
To assess whether IAs are a useful tool for mainstreaming, it is worth acknowledging that the IA guidelines do not refer to mainstreaming as an objective of the IIA system, neither do they refer to the horizontal clauses in clarifying the raison d'être of the IIA system.Nevertheless, in addition to objectives such as better informed decisionmaking and taking stakeholder views into account, the explicit aim of the IIA system is also policy coordination.The guidelines state that the IIA system "helps to ensure coherence of Commission policies and consistency with Treaty objectives such as the respect for Fundamental Rights and high level objectives such as the Lisbon or Sustainable Development strategies" (European Commission 2009, p.6).This statement suggests that there may be a partial overlap with the mainstreaming objectives (such as fundamental rights and sustainable development).Yet it is also clear that the IA is not presented as a mainstreaming tool as such.In fact, the concept of mainstreaming and the horizontal clauses of the TFEU are hardly mentioned in the different sets of guidelines.Only the operational guidelines on social impacts mention mainstreaming and the corresponding treaty articles in relation to gender and health mainstreaming, without, however, identifying a particular role for IAs in contributing to such mainstreaming.Moreover, while the IIA system may pay attention to mainstreaming objectives (without acknowledging the concept or the horizontal clauses), the system is clearly (also) geared toward other policy objectives, such as the Lisbon Strategy.
To assess to what extent the IIA institutional framework acknowledges mainstreaming objectives, the following sections analyze whether the guidelines advise taking into account the mainstreaming objectives when identifying the objectives of new policy initiatives (3.2.); how the mainstreaming objectives should be taken into account when assessing impacts (3.3); and the attention to "counter objectives", that is, other IIA objectives that may receive greater attention than mainstreaming objectives (3.4).

Mainstreaming objectives at the level of identifying objectives for new policy initiatives
According to the guidelines, IAs are supposed to identify general, specific, and operational objectives of new policy initiatives (European Commission 2009, p. 27).General objectives are linked to treaty-based goals; specific objectives refer to how the specific policy contributes to certain aspects of a treaty-based goal; while operational objectives are defined in terms of deliverables of the specific policies.This stage of identifying objectives of new policy initiatives provides an excellent opportunity to take the constitutionalized mainstreaming objectives into account.If the system of IIAs were to be used as an optimal tool for mainstreaming, the guidelines should refer to the horizontal clauses in the treaty and suggest that all IAs have to systematically address whether the new policy initiative can contribute to (one of) these constitutionalized mainstreaming objectives.However, the guidelines fail to refer to the horizontal clauses or to identify the potential use of an IA as a way to proactively promote these mainstreaming objectives.
While the IA guidelines fail to acknowledge the mainstreaming objectives at the level of the IA'sroletoidentify the objectives of new policy initiatives, the annexes to the guidelines do mention the mainstreaming objectives in relation to the IA'sroleofdefining the problem that justifies new policy intervention.Defining the problem and defining the objectives are two separate but related aspects that each IA must include.Annex 6.3 refers to the potential discrepancy between the fundamental goals of the EU and the existing situation as an argument to justify new policy intervention.The annex sets out a list of EU objectives, which is described as "non-exhaustive".All of the constitutionalized mainstreaming objectives appear in this list, among many other objectives; however, the annex does not indicate that these mainstreaming objectives are, according to the treaty, to be taken into account in all EU actions.

Mainstreaming objectives when assessing impacts
While the guidelines fail to guarantee the role of IAs to proactively address the mainstreaming objectives at the level of defining the objectives of new policy initiatives, they do take into account all of the constitutionalized mainstreaming objectives at the level of assessing the impacts of new policy measures.However, not all mainstreaming objectives are given the same attention.Annex 1 provides a brief overview of how the mainstreaming objectives are taken into account in the different guidelines.
The general guidelines clearly state that all IAs need to assess the economic, environmental, and social impacts of new policy initiatives, which function as the primary categories of assessment in the IIA system.More particularly, the likely economic, social, and environmental impacts need to be addressed for each policy option considered in the IA (European Commission 2009, p. 31).For each of these three categories, the guidelines provide a detailed table with questions relating to different subcategories of these topics.The guidelines make clear that these questions are not to be answered with a simple "yes" or "no" but are designed to help to develop the analysis (European Commission 2009, p. 33).Yet while it is more than a tick box list, there appears no obligation for Commission officials to address all of the subcategories systematically.Instead, it is at the discretion of Commission officials to identify which are the appropriate (sub)questions to be addressed in the particular case, in order to provide an IA of the three main categories.Hence, while social and environmental impacts are supposed to be addressed in all IAs, other mainstreaming objectives, which only appear as a subsection of these wider categories, are less likely to receive similar systematic attention.There is also a difference in the extent to which the guidance on assessing the impact of the mainstreaming objectives is further worked out in the annexes and operational guidelines.
Comparing the different mainstreaming objectives, social impacts seem to be best addressed within the institutional framework for IIAs.Being one of the three main impact categories, there is a detailed table for social impacts in the general guidelines.The five social mainstreaming objectives set out in the treaty's horizontal clauses (employment, social protection, social inclusion, education/training, and human health) all constitute a subcategoryinthissocialtable.Moreover,separate operational guidelines for social impacts exist, paying equally detailed attention to all five social mainstreaming categories.Finally, Annex 8.2 to the general guidelines deals briefly with employment, social protection, social exclusion, and training under the title "quantity and quality of jobs," while Annex 8.3 provides a more detailed tool for quantitative assessments of health impacts.Thus, Commission officials drafting an IA have a clear package of guidance on social impacts at their disposal as general guidelines, annexes, and operational guidelines are drafted in a complementary way.
Environmental impacts are equally a key category with a main detailed table in the general guidelines, but there is little further elaboration on them beyond this.There are no separate operational guidelines on environmental impacts, and such impacts are only briefly mentioned in Annex 9 of the general guidelines.
Gender, non-discrimination, and consumer protection do not constitute key impact categories in the IIA system, making their systematic assessment less likely, but they feature as subcategories of either the social or economic impact categories.Gender appears as a subsection in the social impact table in the general guidelines.While gender is not dealt with in the Annexes, it figures prominently in the operational guidelines on social impactsnot only is there a section on equality of treatment and opportunities, gender is also mentioned in all other subsections of these operational guidelines, thus, calling for attention to gender when dealing with impacts on job quality, employment, social exclusion, etc.Like gender, non-discrimination (for all types of discrimination mentioned in the horizontal clause) is a subcategory of the social impact table of the general guidelines and also appears in the operational guidelines on social impacts, but is much less detailed than gender.It is also briefly mentioned in Annex 8.2 dealing with quality of jobs.Consumer impacts figure as a subcategory within the economic impact assessment table of the general guidelines, and appear as a section in the operational guidelines on competitiveness.Moreover Annex 5.3 includes specific guidelines on consulting consumers, whereas Annex 8.3 deals further with impacts on consumers.Of the three subcategories, gender is dealt with most extensively in the guidelines, followed by consumer protection and non-discrimination.
Fundamental rights constitute a separate type of impact in the guidelines system; they are neither one of the three main impact categories (social, economic, environment), nor a subcategory of them.The general guidelines (p.39) conceive fundamental rights as a horizontal concern that should be assessed if relevant, throughout the IA, that is, when dealing with economic, environmental, or social impacts.Annex 8.1 provides a list of Charter rights, but does not clarify what officials are supposed to do with them.More importantly, there are separate operational guidelines that provide detailed advice on assessing impacts in relation to fundamental rights, although it remains up to the judgment of the official to consider whether fundamental rights issues are at stake.

"Counter objectives"
While the mainstreaming objectives have a place in the IIA system, they exist alongside other objectives that appear at least as equal in importance.Economic impacts constitute a key category of the IIA system, and economic costs and regulatory burden appear as the predominant focus overall.
A detailed table for economic impacts is provided in the general guidelines, and most economic impact categories are very well developed in the annexes.In fact, in the main, the annexes frame IA from an economic perspective from the outset.Annex 6, which deals with how to address the "problem definition" in IAs, first deals extensively with "market failures" (which includes a reference to negative externalities of markets, but is then focused on the problems of incomplete markets, weak competition, and information imbalances), and, subsequently, with regulatory failures (which are either economically framed, such as inadequate defense of property rights and barriers to entry; or presented in general regulatory terms, mentioning regulatory capture, enforcement failure, and poorly defined targets).Only as a third category of potential problem definitions is reference made to the objectives set out in the Treaty (including the mainstreaming objectives).
Also in terms of types of impacts to be assessed, the Annexes prioritize economic impacts.Although some mainstreaming objectives receive attention in the Annexes (see above), the list is dominated by economic concerns.There are separate sections paying attention to impacts on: SMEs; technical development and innovation; firms, in terms of investment and operating costs; international trade; developing countries; public authorities; as well as macro-economic impacts.Moreover, additional operational guidelines exist for impacts on micro-enterprises, competitiveness, and territorial impacts.None of these are listed in the treaties as horizontal mainstreaming objectives or as main objectives of policy coordination.Yet Commission officials are advised to pay particular attention to them across policy areas when assessing impacts.
Most importantly, the general guidelines also provide for separate rules on "assessing administrative burdens" and on "assessing simplification potential".Both elements receive similar attention to the category of "economic, social and environmental assessments", which suggests their overall centrality for consideration in each IA.Assessing administrative burdens and simplification potential (which are central to the conceptual framework of reducing regulatory burden) are often in tension with mainstreaming objectives, which generally require more public intervention.

Conclusions on the institutional framework
The overview of the institutional framework makes clear that mainstreaming objectives do have a place within the IIA system.At the same time, the institutional framework of the IIA system also shows weaknesses as a tool to ensure the full consideration of the constitutionalized mainstreaming objectives.In contrast to H1, the institutional framework does not ensure appropriate attention to mainstreaming objectives at the level of defining policy objectives but only really considers them at the level of assessing impacts.Moreover, even at that level, and in contrast to the constitutional expectation that gives equal weight to the different mainstreaming objectives, the institutional framework of the IA guidelines prioritizes some mainstreaming objectives over others.Social and environmental concerns are recognized as primary categories of assessment, with the first the most elaborated.Fundamental rights are less systematically addressed but are still expected to receive more attention than the mere 'sub-categories,' consumer protection, gender, and non-discrimination.Moreover, the institutional framework for IIAs is patchy; with different types of guidelines (general guidelines, annexes, and operational guidelines), of which the exact status and relative hierarchy is not entirely clear.As a result, much is left to the interpretation and discretion of the Commission officials drafting the IA.Hence, although the IIA system itself is strongly proceduralized, as a tool for mainstreaming, it does not guarantee a systematic application, living only modestly up to the expectations set out in H2.
The institutional framework also tells us something about the extent to which IAs are really "integrated" and can overcome the problem of the multiplication of mainstreaming agendas (H3).Applied systematically to all main new policy initiatives, the IIA does, indeed, provide a more systematic integration of several mainstreaming agendas compared with previous ad hoc, sectoral IAs on these topics.Even for mainstreaming objectives which are not treated as main categories in the general guidelines (such as gender or non-discrimination), they are relatively well integrated within one of these main categories (namely, social impacts) and one could, therefore, argue that they are still more likely to be taken into account as part of such an integrated, systematic process rather than apart from it.The downside is that the IIA system pays attention to, and is even biased toward, other objectives that are not mainstreaming objectives.Concerns and benchmarks about economic impact and regulatory burden are more prominent in the IIA guidelines than (most of) the mainstreaming objectives.
While our analysis of the institutional framework provides some indications of the strengths and weaknesses of the IIA system as a tool for mainstreaming, the extent to which IAs have taken up mainstreaming objectives is a question of empirical assessment, which is pursued in the following section.At the same time, the analysis of the institutional framework allows the formulation of some hypotheses on the relationship between institutional setting and institutional practice: H5: Mirroring the institutional framework, we expect IAs to take into account mainstreaming objectives mainly at the level of assessing impacts and very little at the stage of defining policy objectives.
H6: We expect that mainstreaming objectives constituting a primary category of the IIA system are taken into account more systematically than those only mentioned as a subcategory or horizontal category.
H7: We expect the existence of separate detailed operational guidelines for a mainstreaming objective to increase its chances of being taken into account.
4. Empirical analysis of the practice of impact assessments (IAs) 4.1.Methodology In order to gauge the extent to which the Commission's IAs do indeed contribute to the EU's constitutionalized mainstreaming agendas, we analyzed a selection of IAs in relation to the number of references to mainstreaming objectives.Data was taken from 35 IAs, adopted between 6 May 2011 and 20 February 2014.An overview of the selected IAs is provided in Annex 2. The starting point of 6 May 2011 was chosen as the most recent operational guidelines relevant to our research (namely those on fundamental rights) were published on that date.Five IAs were taken from seven DGs.We first selected the five DGs with primary responsibility for one of the mainstreaming objectives: DG Justice (DG JUST, for gender equality, non-discrimination and fundamental rights); DG Employment and Social Affairs (DG EMPL, for the social clause); DG Education (DG EAC, for the education provision of the social clause); DG Environment (DG ENVI, for EPI); and DG Health and Consumer Affairs (DG SANCO, for human health under the social clause, and consumer protection).At first sight, it might make more sense to select DGs that do not have primary responsibility for a mainstreaming objective, for example, in order to assess environmental mainstreaming, it might be more sensible to assess whether DG Competition (as opposed to DG Environment) has taken environmental concerns into account.However, most of the mainstreaming objectives are divided into sub-objectives or are narrower in scope than the wider remit of the DG that has responsibility for it.Hence, by choosing DGs that have primary responsibility for the mainstreaming objectives, one can assess whether they live up to their own mainstreaming guidance in all their actions.Moreover, each of these DGs obviously acts as a non-lead for the other mainstreaming objectives.Hence, by selecting these five, one can compare whether there is a difference in respecting mainstreaming objectives depending on whether a DG has a primary responsibility in the area or not.In addition, we selected two DGs without any primary mainstreaming responsibility, namely DG Competition (COMP) and DG Market (\MARKT).This is based on our understanding that the five mainstreaming objectives set out in the horizontal clauses are a counterbalance to the predominant marketoriented character of the EU constitution.DG MARKT and DG COMP are key actors in relation to these economic objectives.
The analysis is based on the coding of keywords related to each mainstreaming objective, using the software NVivo; for example, for EPI, keywords include "environment,"" pollution," and "climate change."See Annex 3 for the full list of keywords per mainstreaming objective.Coding keywords enabled the highlighting of paragraphs relevant for each mainstreaming objective in the 35 IAs.An interpretative reading was performed on all these paragraphs and the surrounding text to ensure the coded paragraph did indeed refer to a mainstreaming objective.Equally, paragraphs including more than one keyword for the same mainstreaming objective were counted as a single reference.
The analysis allowed the identification of the number of paragraphs attending to each mainstreaming objective in each IA, across the seven DGs.The coding and interpretative reading, complemented by an analysis of the IAB annual reports, allows us to assess the following: 1.While mainstreaming objectives are supposed to be taken into account in all policy areas, we identify IAs that do not consider them at all.Moreover, there is variation in the extent to which different DGs take the mainstreaming objectives into account.2. Identifying the number of paragraphs in which mainstreaming concerns are dealt with gives an indication of the weight given to particular concerns.3.By counting the number of references in both the "objectives/problem" and the "impact" parts of each IA, we can assess whether mainstreaming happens at the level of identifying the objective of new policy initiatives or rather at the level of assessing impacts of several policy options.As it is not always easy to distinguish between objectives and problem definition of new policy initiatives, we included the problem definition section of IAs as part of our calculation on references relating to the objectives of new initiatives.

Overview.
Table 1 shows the number of paragraphs for each mainstreaming objective for all IAs, ordered according to DG. Please note, we have not counted mainstreaming objectives in IAs for policy initiatives that addressed the mainstreaming objective as a key objective.Hence, none of the IAs from DG ENVI were coded for environmental mainstreaming; none of the IAs from DG EMPL were coded for social3; and none of the IAs from DG SANCO were coded for consumer protection.A single IA from DG JUST dealing with a gender initiative was not coded for that mainstreaming objective.Moreover, some IAs were not coded for a subcategory of a mainstreaming objective if that was the key objective of the initiative.Namely, none of the IAs from DG EAC were coded for education (but they were still coded for other subcategories of the social clause); and an IA from DG JUST and DG EAC were not coded for a subcategory of discrimination, but were still coded for its other subcategories.Annex 2 provides more detail on when IAs were excluded from the counting.
It is clear from Table 1 that the five constitutional mainstreaming objectives are not considered systematically.There are only three IAs (EMPL 2, EMPL 3, and EAC2) out of 35 that include considerations on all mainstreaming objectives.Some mainstreaming objectives have been ignored completely by several DGs in all of the IAs we assessed.This is particularly marked with respect to gender concerns, which did not receive a single mention in any of the IAs from DG COMP, ENVI, MARKT, and SANCO, and were only considered in 10 out of 34 IAs.Discrimination also fared poorly, not being considered at all by DG ENVI, MARKT, or SANCO, and only appearing in 13 out of 35 IAs.
At the other end of the spectrum, social concerns were the only mainstreaming category to be taken into account in all IAs.In terms of the number of paragraphs, social concerns rank considerably higher (at 766 paragraphs) than the other mainstreaming objectives.
Environment scored second best with all but five of the IAs taking it into account.The relative success of social and environmental mainstreaming should not come as a surprise as both are one of the three primary categories (together with economic impacts) that each IA has to assess, thus confirming H6.Interestingly enough, even then, five IAs did not consider the environment at all.Moreover, in terms of the number of paragraphs attributed to it (204), environment is at a similar level to consumer protection (148), which is not a key category for IA, but only a subcategory of economic impacts.Indeed, consumer protection fares relatively well as a subcategory of the IIA system, with 22 out of 30 IAs.In this respect, it not only does much better than gender and discrimination, but also better than fundamental rights, which appear in 19 out of 35 IAs.
There is certainly a difference in scope among the mainstreaming objectives and, therefore, it could hardly be expected that they would be taken into account to the same extent.Discrimination, for instance, is much narrower in scope than the broadly defined category of the social clause.Hence, if we split up the broad social clause into its subcategories as set out in the treaty, the picture of the "success story" of social mainstreaming becomes somewhat more blurred.Table 2 provides an abbreviated overview (for reasons of space) of the number of paragraphs dedicated to the subcategories of the social clause by DG (merging the assessed IAs per DG).
Table 2 shows that both social exclusion and social protection fare poorly.Three DGs do not refer to social protection in any of their IAs, while social exclusion is never dealt with by DG MARKT and is only referred to in one or two IAs by other DGs.Employment (29 out of 33 IAs), education (19 out of 30), and health (22 out of 34) are taken into account much more systematically.
As some mainstreaming objectives are narrower in scope than others, any comparison between the mainstreaming objectives should be undertaken with caution.It is beyond the scope of this article to make normative arguments on substantive grounds about why some IAs could reasonably have been expected to take account of a particular mainstreaming objective.However, our data allow several conclusions to be drawn, particularly of a more procedural nature.

4.2.2.
No systematic procedural consideration of mainstreaming objectives.The mainstreaming objectives are not taken up in any systematic way from a procedural perspective.Even if the different mainstreaming objectives are not equally relevant for all types of policy intervention, the expectation of mainstreaming is precisely that policymakers will at least ask systematically whether these mainstreaming objectives can be addressed in all their new policy initiatives, whatever the policy area.If a mainstreaming objective is considered not relevant in a particular case, an IA should at least provide a brief acknowledgement that the mainstreaming objective has been considered, ideally providing a justification if it is concluded not to be relevant.In fact, this happened in eight of the IAs we analyzed.These "negative justifications" all related to environmental impact.It was argued that environmental impact was not relevant or significant, or that the proposed policy options did not differ in environmental impact; therefore, no environmental concerns were taken up Percentages to nearest whole number.†The impact assessment (IA) was not coded for this mainstreaming objective because the mainstreaming objective coincided with the main objective of the policy initiative.‡ The IA was not coded for one subcategory of this mainstreaming objective because that objective coincided with the main objective of the policy initiative.The IA was still coded for the other subcategories.See Annex 2 for list of codes excluded for each IA.COMP, competition; EAC, education and culture; EMPL, employment, social affairs and inclusion; ENV, environment; IIA, integrated impact assessment; JUST, justice, fundamental rights and citizenship; MARKT, internal market and services; SANCO, health and consumer policy.
in the IAs at all.It is not surprising that these "negative statements" occurred in relation to the environment, given that it is one of the three key impact categories (economic, social, environment) that each IA should assess.DGs, thus, felt compelled to justify the absence of any environmental assessment.However, most of these justifications are very succinct.Moreover, while environment is a key category of the IA system, we still found five IAs that did not include any reference to the environment and provided no justification for this.
In relation to other mainstreaming objectives, DGs never felt the need to justify absences.As all of the IAs we assessed dealt with social concerns, no justification was needed for this key category.However, IAs that failed to address fundamental rights, gender, discrimination, or consumer protection never provided any justification for this (with the single exception of an IA justifying the absence of consumer concerns in part of its analysis).Classification as a subcategory, rather than a key category of the IIA system clearly affects how DGs consider the importance of mainstreaming agendas, confirming H6.

The relationship between guidelines and practice.
As the previous point illustrates, the way in which mainstreaming objectives are dealt with in the guidelines impacts how they are dealt with in practice.DGs are inclined to explain themselves if not considering one of the main IIA categories (thus confirming H6).However, being a key category of the IIA system does not completely guarantee consideration.This relates to the lack of consistency in the way in which IAs are structured.Some provide separate sections to deal with economic, social, and environmental impacts.In others, the impacts are organized around different policy options; however, not all of these then systematically address economic, social, and environmental considerations.
Beyond the distinction between main impact category and subcategory of the IIA system, it is difficult to discern clear patterns between the detail of guidance provided on mainstreaming objectives in the general guidelines, annexes, and operational guidelines, and the attention paid to these mainstreaming concerns in practice.
As discussed, the social clause is treated most comprehensively within the guidelines, with a clear table in the general guidelines, explanations in the annexes, and detailed separate operational guidelines.This corresponds to social considerations faring best among the mainstreaming objectives in the practice of IAs, both in terms of the number of IAs dealing with them and the number of paragraphs attributed to them.
The existence, however, of separate operational guidelines for a mainstreaming objective (H7) is no guarantee for the systematic consideration of that objective, as illustrated by the case of fundamental rights.Moreover, consumer protection, for which there are no separate operational guidelines, is more systematically assessed.However, fundamental rights (which is a horizontal category in the general guidelines, and for which there are operational guidelines) do score much better than gender and discrimination, which are simply treated as subsections of social considerations in the general guidelines and for which there are no separate operational guidelines, thus providing partial support for H7.
The widest gap between guidance and practice occurs with gender mainstreaming.The poor fate of gender mainstreaming is interesting as the topic is rather comprehensively included as a subcategory of social considerations, both in the general guidelines and in the social operational guidelines.4. 2.4.Mainstreaming objectives at the level of defining policy objectives and assessing impacts.We have argued that mainstreaming objectives can be considered both at the level of defining the objectives of new policy initiatives and of assessing impacts.We also identified that the IA guidelines fail to provide clear guidance to Commission officials to consider mainstreaming objectives at the level of defining objectives of new policy initiatives.Therefore, we expected to see a similar pattern when it comes to the practice of IAs (H5).The analysis of the IIA practice, however, shows a different picture.Mainstreaming considerations were addressed almost as often at the level of defining objectives (including problem definition) as at the level of assessing impacts.
Table 3 shows that one mainstreaming objective (non-discrimination) was taken into account more at the level of defining the policy objectives than of assessing impacts.For two others (the social clause and consumer protection) considerations are at an equal level for defining objectives and assessing impacts, while gender is not far from that balance.Considerations are clearly more developed at the level of assessing impacts only for fundamental rights and environment.This overall picture clearly contrasts with what would be expected on the basis of the guidelines, which give very little attention to mainstreaming objectives at the level of defining policy objectives in the IA.
Nevertheless, it is worth noting that although IAs frequently refer to mainstreaming objectives at the level of defining policy objectives, none of the 35 IAs we assessed refers to the horizontal clauses (Articles 8-12 TFEU) as a reason for policy intervention.While the definition of objectives of a policy initiative requires an indication of a legal basis in the treaties, in the case of mainstreaming objectives, reference is made to the core legal basis of policy competences (e.g. the TFEU'sarticles on social policy when considering social objectives, or, in some cases, the Charter) and never the mainstreaming clauses.
Regarding the different balance between objectives and assessment of impacts among the mainstreaming objectives, it is not surprising that environment is the one that gives comparatively more attention to impact assessment than to a definition of the objectives, given that in this sector there is a well-established tradition of IAs with an appropriate methodology attached.
The high score of fundamental rights for impact assessment rather than the definition of objectives may come as more of a surprise.If fundamental rights are considered a normative framework that guides policy intervention, it is expected that they would be taken into account when defining the objectives of new initiatives.However, the EU's fundamental rights regime is mainly conceived as a negative guarantee, intended to ensure that the EU should not negatively impact on fundamental rights, rather than as a positive regime promoting these values in a proactive way at policy level (De Schutter 2005).The operational guidelines on fundamental rights in the IA are, thus, steered to set off a warning light whenever policy intervention would negatively impact on fundamental rights, while failing to use IAs actively to define the objectives of new policy initiatives that positively promote fundamental rights.4.2.5.Different mainstreaming practices of Directorates-General. We selected the DGs to enable a comparison between those with primary responsibility for a (market-correcting) mainstreaming objective and those that have no such responsibility and are primarily market-oriented.This allows two hypotheses to be tested:H8:DGs that are primarily market-oriented are less likely to take into account market-correcting mainstreaming objectives.H9:DGs that have a primary responsibility for one of the mainstreaming objectives are likely to be more familiar with, and positively inclined toward, mainstreaming more generally, so they are also more likely to take into account mainstreaming objectives other than the one for which they have primary responsibility.
H8 is confirmed by our evidence.Table 4 highlights that DG MARKT and DG COMP show the least interest in the mainstreaming objectives.This is partially a result of the nature of their activity.However, they show the least inclination to address the mainstreaming objectives systematically (even in a procedural way).As one may expect, they have the least concern for social considerations, but each of them also completely ignores two other mainstreaming objectives.DG MARKT is clearly the weakest mainstreamer, giving only significant attention to consumer protection.In addition to consumer protection, DG COMP also pays considerable attention to environmental concerns.However, being the lead responsible for a mainstreaming objective does not imply that a DG also acknowledges the importance of mainstreaming other objectives, which tends to contradict H9.DG ENV pays more attention to social concerns than DG MARKT and DG COMP but fares at very similar levels to them with respect to the other mainstreaming objectives.DG SANCO is only slightly better.DG EMPL, and, in particular, DG EAC4 and DG JUST are much more systematic mainstreamers.
Finally, for some DGs we were able to assess whether they follow up their own guidelines and mainstreaming objectives for all their policy initiatives.We tested DG EMPL on the subcategories of the social clause, and DG JUST on fundamental rights, gender, and discrimination.A similar test could not be performed for DGs ENVI, SANCO, and EAC, as, by definition, they address their mainstreaming objective in all their initiatives (environmental protection, consumer protection, and education, respectively).
Directorate-General JUST pays considerable attention to fundamental rights, even for policy initiatives that are not primarily focused on this.However, gender and discrimination are not systematically addressed in all policy initiatives, despite being the lead DG on these objectives.Hence, DG JUST follows its own guidance, set out in the operational guidelines on fundamental rights, but is less committed to gender and discrimination, which appear only as subsections of social concerns.The existence of separate operational guidelines does not seem to have the same positive effect on DG EMPL.Although all aspects of the social clause are dealt with in detail in the operational guidelines on social impacts, DG EMPL does not always address them.In particular, social exclusion is far from systematically addressed (see Table 2).4.  objectives.As already shown in the analysis of the IA guidelines, the mainstreaming objectives are just some of the policy concerns taken into account in IAs.In particular, economic objectives and concerns about the regulatory burden appear to dominate the IIA system.This is confirmed by our analysis of 35 IAs.Although not providing any comparative, hard, statistical measurement, the data on the number of mainstreaming paragraphs (as listed in Table 1) give some indication of the weight of mainstreaming objectives in relation to other concerns in IAs, if one acknowledges that the average number of pages in an IA is c.53.The text of an IA that does not relate to mainstreaming objectives was found normally to deal with economic arguments (economic impacts and regulatory burden) and procedural issues (e.g. who has been consulted and evaluation provisions).The predominance of market concerns in the IIA system is also illustrated in the annual reports of the IAB.Each year, the IAB adopts a report with an overview of its activities of that year, identifying trends in the IIA process.Since its first Annual Report in 2007, the IAB has indicated on which topics it has given most recommendations.Most IAB comments concern economic impacts.In 2013, more than 80% of IAB reports included comments on economic impacts, which is a similar percentage to previous years.Second are recommendations on social impacts (about 60%), followed by impacts on administrative costs, and on SMEs.Recommendations in relation to environmental impacts appear fifth (about 30% of reports include them), followed by regional impacts, impacts on the EU budget, and recommendations in relation to fundamental rights and simplification (both less than 10%; European Commission 2013, p. 7, fig.6).While this overview suggests a reviewing process in which economic concerns are central, it is worth noting that the number of recommendations in relation to social impact has gradually increased, from about 30% in 2010 to about 60% in 2013.Yet while the IAB has paid increased attention to social impacts, the annual reports repeatedly express the IAB's frustration that DGs do not seem to change their practice.In contrast, while our analysis shows that IAs include more paragraphs of social than environmental considerations, the IAB appears more satisfied with how DGs deal with environmental impacts, considering improvements were needed in only 30% of cases.The IAB has also expressed satisfaction with DGs' increased attention to administrative burdens (in half of the IAs in 2013, compared to a third in 2010), and the considerable increase of IAs including a competitiveness impact (an increase of 30% from 2012 to 2013; European Commission 2013).

The practice of participation
Space prohibits detailed attention to participation patterns in IAs.However, previous work (Smismans and Bozzini 2015) has analyzed such patterns in all EU IAs over a period of 10 years, concluding that although the use of general online consultations by the Commission has increased over the last decade, this is still not a general practice.Consultation still more often takes the form of targeted consultations.Analyzing the pattern of actors involved in IIAs, this analysis showed that IAs are first used as coordination tools, ensuring, in particular, the participation of other DGs via the Steering Group, rather than as instruments to ensure wide participation.This pattern of participation corresponds with a technocratic rather than transformative participatory approach to mainstreaming.While it is not ruled out that mainstreaming concerns emerge during consultation, the targeted consultations are unlikely to reach actors beyondthetraditionalcircleofthepolicyarea.Atthesametime,evengeneralonlineconsultationsaremostoftenalready framed by the questions set by the Commission.The "upstream" participatory transformative dimension of mainstreaming is, therefore, likely to be modest."Downstream" it is even less likely that the system of IIA would guide a multitude of actors to reflect on mainstreaming objectives, as it is primarily a tool to gather and coordinate evidence at the Commission level.This contrasts to H4 regarding the high levels of participation that the IIA system would secure in practice.

Conclusion
Analysis of both the institutional framework and practice shows that the system of IIAs is not a panacea for mainstreaming.In contrast to the constitutional expectation of mainstreaming clauses, the IIA system does not ensure a systematic screening of new policy initiatives in light of these objectives.In the IAs analyzed, some DGs completely ignored several mainstreaming objectives.We found that DGs with a primary market-oriented responsibility take less account of mainstreaming objectives.At the same time, having a lead responsibility for one of the mainstreaming objectives does not ensure that a DG is a "good mainstreamer" on other mainstreaming objectives.The mainstreaming objectives were also screened to a different extent.Social and environmental concerns were most often taken into account, although the "success story" of the social clause is mitigated when one assesses its several subcategories, with social exclusion, for instance, receiving very little consideration.Gender and non-discrimination fared the worst; this is particularly striking in relation to gender, given its potential relevance across all policy areas and the detail with which the concern has been set out in the IIA framework (although not as a main category), thus, showing the widest gap between "institutional prescription" and practice.Moreover, the mainstreaming objectives are overshadowed in the IIA system by concerns about economic impact and regulatory burden, although these have not been constitutionalized in the treaties as horizontal objectives.
One can conclude that the IIA system does not live up to the expectations set out at the start of this article.Although the IIA system as such is strongly proceduralized, this has not led to a comparable hard proceduralization and systematic consideration of the mainstreaming objectives (H2); the capacity of the IIA system to tackle the problem of mainstreaming overload, thanks to its integrated nature (H3), is left in doubt, given the risk of all mainstreaming agendas simply being subordinated to other (market-oriented) priorities; and its participatory practice is too weak to ensure a more transformative approach to mainstreaming (H4).
Nevertheless, it would be wrong to conclude that the IIA system has no added value for mainstreaming.As primary categories of the IIA system, social and environmental concerns are addressed in nearly all IAs, and it is reasonable to argue that fundamental rights and consumer protection also profit from more consideration in policy drafting than if there were no IIA system.Moreover, one has to acknowledge that not all mainstreaming objectives are equally relevant for all policy initiatives.Non-discrimination (in terms of gender, religion or belief, etc.), for instance, may well not be an issue in regulating emissions from diesel engines.A low total number of paragraphs on this topic in an IA is, therefore, not a problem as such.However, if the constitutionalized nature of the mainstreaming objectives is taken seriously, the IIA system would be expected to provide at least a procedural guarantee that these objectives are taken into consideration, even if only to justify the objective as irrelevant to the case in hand.Our main critique is that the current IIA system does not provide such a procedural guarantee.
Therefore, we suggest some changes that can improve the proceduralized nature of the IIA system in the function of mainstreaming (H2), and more effective "mainstreaming of mainstreaming" (H3).Firstly, the IIA guidelines should set out more clearly that IAs have to assess mainstreaming objectivesnot only at the level of assessing impacts of new policy initiatives, but also at the level of setting their policy objectives (H1).While the empirical assessment suggests that IAs, in contrast to the guidelines, do take mainstreaming objectives into account at the level of defining policy objectives to a certain extent, a more proactive use of mainstreaming would be encouraged further by clarifying this requirement in the guidelines.Secondly, as the analysis shows, being a primary category rather than a subcategory of the IIA system has a positive impact on whether a mainstreaming objective is taken into account.Given the constitutional nature of the six mainstreaming objectives, it could be argued that all of them ought to be included as primary categories of the IIA system.Alternatively, some of them could still remain as a subcategory, as long as there is an explicit requirement to address them.Indeed, there should be a minimum requirement for each IA to justify explicitly if a mainstreaming objective is considered not relevant to the case in hand.To date, this has only occurred in relation to environmental mainstreaming.Thirdly, separate operational guidelines, such as they exist on fundamental rights and social mainstreaming, provide useful detail, but the main guidelines should make more explicit reference to them in order for Commission officials to use them more actively.Fourthly, clarification of the constitutional nature of the mainstreaming objectives in the guidelines (suggesting their priority over the mushrooming of other categories of assessment in the IIA system) should go hand in hand with an increased willingness of the IAB to screen the IA practice in function of these requirements.Finally, a more participatory approach to the IIA system, beyond its current focus on coordination, is required to contribute toward a more transformative approach to mainstreaming.subcategories, as not all of DG EMPL's initiatives by definition deal with them.We have, therefore, coded all IAs from DG EMPL on social concerns for Table 2

Table 1
Number of paragraphs for each mainstreaming objective for all IIAs, ordered according to DG

Table 2
Overview of the number of paragraphs dedicated to the subcategories of the social clause by DG (merging the assessed IAs per DG) Mean average to nearest whole number.COMP, competition; EAC, education and culture; EMPL, employment, social affairs and inclusion; ENV, environment; IA, impact assessment; JUST, justice, fundamental rights and citizenship; MARKT, internal market and services; SANCO, health and consumer policy.

Table 3
Location of mainstreaming objectives in the IIA reports

Table 4
Attention to mainstreaming objectives according to DG One of the IAs from this DG was not coded for the objective at all.§ At least one IA from this DG was not coded for one of the subcategories of this mainstreaming objective.See Annex 2 for list of codes excluded for each IA.COMP, competition; EAC, education and culture; EMPL, employment, social affairs and inclusion; ENV, environment; HSC, horizontal social clause; IA, impact assessment; JUST, justice, fundamental rights and citizenship; MARKT, internal market and services; SANCO, health and consumer policy. ‡

Table 1 .
but not for Table 1. 4. The comparatively high number of mainstreaming paragraphs by DG EAC should be nuanced, as one of its IAs was considerably longer than the rest.Taking the length of IAs into account, DG EAC fares at similar levels to DG JUST.Annex 1: Overview of how the mainstreaming objectives are taken into account in the different guidelines.(Continued) Proposal for a directive of the European Parliament and of the Council amending Directive 2004/40/EC on minimum health and safety requirements regarding the exposure of workers to the risks arising from physical agents (electromagnetic fields) (18th individual Directive within the meaning of Article 16(1) of Directive 89/391/EEC) Council Regulation on the exercise of the right to take collective action within the context of the freedom of establishment and the freedom to provide services Regulation on the Fund for European Aid to the Most Deprived Directive on seafarers amending Directives 2008/94/EC, 2009/38/EC, 2002/14/EC, 98/59/EC and 2001/23/EC

Table 1 .
(Continued) Regulation on entrusting the Office for Harmonisation in the Internal Market (Trade Marks and Designs) with certain tasks related to the protection of intellectual property rights, including the assembling of public and private sector representatives as a European Directive amending Directive 2009/65/EC on the coordination of laws, regulations, and administrative provisions relating to undertakings of collective investment in transferable securities (UCITS) and Directive 2011/61/EU on Alternative Investment Funds Managers in respect of the excessive reliance on