Comparative Social Policy Analysis in the EU at the Brink of a New Era

Abstract Comparative policy analysis plays an important role in the EU: European social policy, as it developed from the 2000s, is intrinsically interwoven with comparative social policy analysis, notably in the context of the Open Method of Coordination (OMC). A review of the role of comparative social policy analysis in the OMC highlights a number of long-standing challenges, which are discussed in the first section of the paper. Do EU member states commit themselves to a real “burden sharing” with regard to the agreed objectives? How can we effectively mainstream social policy objectives into the EU’s economic and budgetary governance? How can we assess the comparative efficiency of EU welfare states? These challenges have a political, a scientific and a technical (data-related) nature. They underscore the necessity of in-depth, shared and validated comparative social policy analysis of member states’ policies. The second section of the paper explains that we may be entering a new era, given the reform agenda for the Economic and Monetary Union set out by the so-called Five Presidents’ Report; these reforms, if implemented, may lead to important spill-over effects in the domain of European social policy, with new demands for comparative social policy analysis. The third section zooms in on the related challenges for the statistical capacity of the EU and the need to invest in the development of reliable and comparable data.


Introduction
At the time of writing this contribution, there is considerable uncertainty about the future course of the European Union. Europe was hit by a chain of economic, social and political crises, during which the break-up of the Eurozone even became a real possibility. By now, the most acute threats to the Eurozone seem to have disappeared and the "Five Presidents" of the European institutions have put badly needed reforms of the Economic and Monetary Union (EMU) on the agenda (Five Presidents' Report 2015).
Potentially, these reforms will lead to important spill-overs in the governance of social policies. Although no decisions have been taken and different options remain open, we are at the brink of a new era. My contention in this article is that comparative social policy analysis, if it is to play a significant role in the development of European policies and politics, has to face up to this new challenge. In fact, this has become clear since the aftermath of the 2008 financial crisis, and the prolonged period of recession and low growth it produced: in terms of both European policy challenges and the type of comparative social policy analysis they require, we are at the brink of a new era. In the first and main section of this contribution, I draw lessons from past experience. In the second section, I sketch the new departure that we may witness. In the third section, I zoom in on some challenges for the statistical capacity of the EU.

From Lisbon to Europe 2020 and Beyond: Three Lessons We Learned
This special issue of the JCPA is timely, a fortiori when we consider the state and role of comparative social policy analysis in Europe. As a matter of fact, in the context of the EU, comparative social policy analysis finds itself at a moment of transition from one era to another. The era that we leave behind us is marked by the Lisbon strategy (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) and the first five years of its successor strategy, Europe 2020. In the social domain in particular, this era was dominated by varieties of the Open Method of Coordination (OMC). Although there have been regular changes within the concept and the way of its application, open coordination generally implied that the EU defined common objectives. Given that framework of common objectives, national governments and regional governments within nation states developed action plans and strategies, which were subsequently submitted at the EU level for peer review, with the aim of mutual learning. In the transition from the Lisbon Strategy to Europe 2020, there was a promise of a stronger and more integrated form of coordination. For the first time, a common quantitative target was adopted with regard to poverty and social exclusion: the EU set the ambition to reduce the number of people in poverty and social exclusion by 20 million. Poverty and social exclusion is a mathematical union of three subgroups: first, people who have an income below their national poverty threshold; secondly, people living in households with a very low work intensity, which signals a very fragile attachment to the labour market; and, thirdly, a subgroup of people facing severe material deprivation. The common target was cleverly designed, with degrees of freedom for countries to decide whether their priority is income poverty, material deprivation or employment. There has been scholarly criticism on the consistency of this target (e.g. De Graaf-Zijl and Nolan 2011), but I consider it satisfactory as a sensible political compromise.
Yet, unfortunately, when we consider the outcomes, we must recognize that this new approach has been largely a failure. Poverty and social exclusion have increased in the EU, rather than decreased. What can we learn from this experience?
Since the launching of the OMC in Lisbon, social policy researchers, social NGOs and political and social actors have criticized this new methodology for its "soft"i.e. nonbindingcharacter. When compared with traditional hard legislation, the soft law approach was seen as intrinsically too weak to achieve its stated ambitions. Space forbids elaboration upon this debate here. I, for one, think that one should not deny the fact that soft law is "weak" compared to hard law; maybe some of my fellow travellers in the Comparative Social Policy Analysis in the EU 391 campaign to get the OMC off the ground during the Lisbon era were overoptimistic in this respect. However, one cannot simply replace the existing soft guidelines with binding directives with the same substantive content, given their broad outcome-oriented character. Hence it seems that this line of criticism misses some of the key issues, and does not always formulate practical alternatives. This line of criticism may also overlook the fact that the institutions and processes set in motion in the context of the OMC played a modest but increasingly important role "under the radar", even when the much "harder" economic and budgetary governance seemed to be in the driving seat of the so-called European Semester (Zeitlin and Vanhercke 2015). Therefore, I first want to sketch what I consider to be three key lessons we learned during the first half of the Europe 2020 strategy, from a perspective of comparative social policy analysis.

True Coordination and Its Prerequisites
First, the establishment of common EU targets was not complemented by real coordination on these targets. Real coordination, at least as it is commonly understood, implies a notion of "burden sharing" to reach an agreed, common target. The OMC is not based on such a notion of coordination. Sadly, the new target on poverty and social exclusion illustrates this problem well. Member states were invited to propose their own ambitions, but they were not coordinated from the perspective of the common target. If one adds the national targets as they were submitted to the EU, the collective ambition only amounted to a reduction by 12 million people in poverty and social exclusion, and not a reduction by 20 million. Obviously, if one uses the label "coordination", but there is no real coordination in the standard understanding of what "coordination" means, the legitimacy of the exercise is threatened from the start. A proliferation of sub-targets emerged without a common understanding of their fitting together in the overall ambition. In addition, some member states simply indicated that they would diminish the aggregate number of people in poverty and social exclusion by amount x, but they did not disaggregate what they would do, and on which of the three targets of the menu they would focus. This problem was compounded by the fact that member states could select a different target, other than the abovementioned three elements of the menu. For instance, Germany used this extra liberty to formulate a target with regard to long-term unemployment; Sweden also formulated a target with regard to long-term inactivity; the United Kingdom put the focus on child poverty; Ireland preferred to proceed on the basis of a notion of "combined poverty".
Here, comparative social policy analysis should enter the scene, in two different acts. The first act is about efficient and effective burden sharing: if the collective ambition really is to reduce the number of people in poverty and social exclusion by 20 million, how should the burden of that ambition be distributed among the member states? The question is eminently political; an obvious and simple solution would be to calculate the reduction that is needed to reach the target at the EU level as a percentage of the total European population in poverty or social exclusion, and to ask each member state to reduce its overall number of individuals living in poverty or social exclusion by the same percentage. Subsequently, in such an ideal scenario of "burden sharing", each member state should make a choice concerning the sub-target it focuses on, depending both on its policy preferences and on the potential for progress it sees. To make that exercise credible, member states should establish a relation between policy inputs they would deploy and the policy outcomes they promise; this is exactly what is absent in a number of the national programmes currently submitted in the context of Europe 2020. In order to evaluate, at the European level, the credibility of the claimed relations between policy inputs and policy outputs, and to engage in relevant mutual learning, a common analytical understanding is necessary: in this respect, comparative social policy analysis should provide key support.
The second act concerns targets that deviate from the Europe 2020 menu. If a country prefers to focus on a different target, it should be able to justify this choice by explaining that its preferred target is better able to contribute to the collective European ambition. This requires ex ante comparative policy analysis: I label it "ex ante" because the analysis must show that potential progress on the chosen (deviating) target will, ultimately, be a more efficient way to achieve progress with regard to the collective European ambition. The setting of the specific Irish target can be related to thorough examination and criticism by Irish academics of the official European target (Nolan and Whelan 2011). But, apart from that, insofar as I know, member states with different targets did not present elaborate justifications. If such justification were to be presented, its assessment at EU level would have to be based on a common, mutually agreed analytical capacity. This implies a need for shared and validated comparative social policy analysis.

Establishing Credible Links between Economic and Social Policy
The second lesson we learned during the first years of the Europe 2020 strategy concerns the crucial importance of economic policy: one cannot build a sustainable social policy on unsustainable financial and economic policies. When the financial crisis erupted, it soon became clear that the spectacular (and much applauded) increase in employment in a country like Spain was based on a low-skilled construction workforce, fuelled by an unsustainable real estate boom. Across the whole EU, much of the progress made in terms of employment during the "good" economic years of the Lisbon era was wiped out by the financial crisis. In the first years of the financial and economic crisis, the traditional automatic stabilizers of European welfare states prevented financial poverty from increasing, but soon these automatic stabilizers had to be switched off in the context of fiscal consolidation packages, and poverty started to increase.
The lesson learned, in a hard way, is that we need a much stronger link between economic, budgetary, employment, education and social policies, in two senses: success in social, employment and education outcomes requires sustainable economic and budgetary policies; simultaneously, economic and budgetary policies should be assessed on their employment, education and social impact. The latter concern is related to the need to "mainstream" social policies into economic and budgetary policies, as it is expressed in the European jargon. Although the notion of mainstreaming seems straightforward, how exactly it should be done and which institutional actors should take the lead in it is a matter of internal debate in European policy circles and some ambiguities need clarification (Vandenbroucke with Vanhercke, 2014). If that does not happen, there is a real risk that mainstreaming will remain an empty concept and will even be discredited in the end. The issues are both political and technical. Today we are confronted with prima facie contradictions between Europe 2020 targets and actual budgetary priorities. Education is a telling case: in 2013, the volume of public spending on education was lower than over Comparative Social Policy Analysis in the EU 393 the years 2006-2008 in at least ten EU member states (Vandenbroucke and Rinaldi, 2015). 1 Obviously, this is first and foremost a political issue: do the Europe 2020 targets carry sufficient weight at the highest level of decision making in the EU and the member states? But it also highlights the lack of comprehensive and encompassing policy analysis in the so-called "European Semester". Below, we return to some of the technical preconditions for successful mainstreaming: the use of reliable and timely quantitative indicators, and the development of analytical capacity at the EU level. Again, all this needs support by validated comparative social policy analysis.

Understanding the Comparative "Efficiency" of Welfare States
A third conclusion which I would draw from recent experience and academic research concerns our understanding of the large disparities in the actual performance of European welfare states. It is by now generally accepted that high levels of social spending can be compatible with economic competitiveness. However, some countries seem to be more efficient in the organization of their social spending than others, with high disparities between national welfare states. A comparison of key outcomes suggests that countries like Sweden, the Netherlands and Denmark achieve low levels of poverty, high employment rates, a high share of people with secondary or higher education and a good score in generally accepted standards of competiveness, while a country like Italy, which has more or less the same level of social spending, performs worse on all these counts (Vandenbroucke 2014, notably Figures 2, 3 and 4). Hence, the concern that some social expenditure is used inefficiently should carry much more weight than the idea that high levels of social spending and competitiveness contradict each other.
The strong record of northern welfare states with regard to both employment and poverty has been linked to their long-term orientation towards "social investment", i.e. activation, investment in human capital, and capacitating social services such as child care (Vandenbroucke et al. 2011;Hemerijck 2013). Obviously, investment in education and child care is no panacea; welfare states also differ with regard to the effectiveness of their social protection systems. For instance, Greece does not have a system of minimum income assistance, and minimum income protection in Italy is generally considered to be inadequate. Cash transfer systems are highly fragmented in a number of welfare states. In some welfare states, pensions play a considerable role in the social protection of families with children, a phenomenon that seems associated with the inadequacy of their cash benefits other than pensions and with the presence of multi-generational "extended families" (Vandenbroucke et al. 2013).
Welfare state performance depends on the complementarity of effective investment in human capitalby means of education, training and child careand effective protection of human capitalby means of adequate transfer systems and health care. The redistributive role of social protection remains important per se (Cantillon and Vandenbroucke 2014). All this means that there is still considerable scope for reform in a number of member states, with a view to improving the performance of their welfare states; but tackling inadequacies in welfare states requires a holistic approach to their functioning. Inspired by the academic research on "social investment", the European Commission has developed a more holistic vision and translated it into a bundle of initiatives, called the Social Investment Package (European Commission 2013). Related to this, the Commission services have developed analyses of significant differences between member states in the efficiency of their social protection spending (e.g. European Commission 2014, Chapter 6).
This recent emphasis on the overall efficiency of welfare states is welcome, but there are important caveats. For a start, one should be clear about the notion of "efficiency". Pareto's well-known definition of efficiency holds if, with the available resources and given a set of objectives, it is impossible to improve on one objective without diminishing the performance on another objective. Prima facie, Pareto efficiency seems very appropriate to assess welfare state performance, since social policy typically involves many objectives simultaneously, which are strongly interdependent: social inclusion, income insurance, education, health, child care. Therefore, Lefebvre and Pestieau (2012) argue that one must not benchmark the performance of welfare states with regard to one single objective, but take into account that welfare states may display different preferences with regard to the objectives to be achieved. Hence, they propose that we benchmark the overall performance of welfare states by constructing a "best practice frontier" for a set of objectives (a best practice frontier for "outputs" of policy), whereby the "input" of policy is not "euros spent" but, for each welfare state, "one unit of government". They then benchmark the performance of each welfare state on the basis of its distance to that frontier. This technique respects diversity with regard to national social objectives and yields interesting observations with regard to the global social performance of EU member states. However, Lefebvre and Pestieau reject an application of this technique to measure "efficiency per euro spent" on this level of aggregation; hence, they do not test Pareto efficiency. One of their arguments is that the relation between inputs and outputs on this level of aggregation is too complex, notably with regard to education and health outcomes. This implies that a true assessment of efficiency of welfare states in a Paretian sense is, for the time being, far beyond our conceptual and empirical means. Whilst accepting this argument, Vandenbroucke et al. (2013) propose a set of "efficiency benchmarks" with reference to a specific, single objective (child poverty), differentiated on the basis of the factors taken into account to explain performance. Rather than assessing Pareto efficiency, these benchmarks measure the "productivity" of spending, conditional on other "inputs" such as employment and human capital. The conclusion of this econometric research is sobering, in the sense that there is no "silver bullet" to "efficiency", so conceived. The observation that state-of-the-art comparative social policy analysis leaves a substantial puzzle in explaining why countries perform so differently signals the need for further research. But it also suggests that EU governance should be far removed from a top-down, "one size fits all" approach to social policy making in the member states, at least when it comes to assessing and promoting the overall performance of welfare states. What is needed, in this respect, according to Vandenbroucke et al. (2013), is a combination of (i) greater room for manoeuvre and support for member states that opt for a social investment strategy, and (ii) policy guidance based on clear and sufficiently stringent and constraining objectives with regard to well-defined social outcomes on the one hand, and genuine scope for exploration and mutual learning on the ways and means to achieve those outcomes on the other. Comparative social policy analysis should define its role and develop its methodologies and databases in this perspective.
Comparative Social Policy Analysis in the EU 395

From Past Experience to New Perspectives
Wrapping up the lessons we learned, I would say that true coordination on social objectives requires commonly accepted and validated comparative policy analysis (next to political agreement on "burden sharing"); and, that much stronger reciprocal links should be established between economic, budgetary, employment, education and social governance at the EU level. If "mainstreaming" is experienced as a tangible reality, the idea of "mutual learning" on welfare state performance, which was at the heart of OMC, may also regain its credibility among larger audiences. Whilst there are no "silver bullets" and "one size fits all" recipes for welfare state performance, all this implies that we need more thorough, rigorous and incisive comparative social policy analysis than what has been tabled so far in the context of European governance. To be sure, there is no denying the remarkable (and steadily increasing) quality of the yearly reviews by the European Commission, Employment and Social Developments in Europe, and the sophistication already reached in the "Joint Assessment Framework" applied to the member states' employment policies (European Commission, EMCO, SPC 2010). But an even bigger effort is necessary. The depth of the Eurozone crisis sheds new light on this urgent necessity, but also suggests the need for a new departure and a new perspective. When we officially launched the OMC in 2000, my argument was that it would be both a cognitive and a normative tool. I considered OMC a "cognitive tool" because it would allow us to learn from each other's actual policies. In addition, OMC is also a "normative tool", because common objectives embody substantive views on social justice. Thus "open coordination gradually creates a European social policy paradigm" (Vandenbroucke 2002, p. xxi). Whereas 15 years ago the definition of a European social policy paradigm, motivated by common social objectives and based on a shared conception of "the European social model" might have been dismissed as interesting but not strictly necessary, today it is no less than an existential conundrum for the Union. Space does not allow for elaboration upon this argument here, but let me summarize it as follows: the short-term financial and economic emergencies with which the EU has to deal cannot be isolated from the imperative to develop a social policy concept for the EU, i.e. a basic consensus on the role the EU should play and the role it should not play in the domain of social policy (Vandenbroucke 2015).
My argument for a European social policy paradigm is certainly not that the EU should become a welfare state in its own right. What is needed, rather, is the development of a European Social Union. I use the notion "social union" deliberately, for three reasons. First, it is a clear-cut concept, in contrast to the vague and elusive notion of "a social Europe", which often surfaces in discussions on the EU. Second, it signals that we should go beyond the conventional call for "a social dimension" to the EU. It would indeed be wrong to assert that the EU has no social dimension today. The coordination of social security rights for mobile workers, standards for health and safety in the workplace, and some directives on workers' rights constitute a non-trivial acquis of 50 years of piecemeal progress. The EU also developed a solid legal foundation for enforcing nondiscrimination among EU citizens. The notion of a "European Social Union" is not premised on a denial of that positive acquis. However, although the next steps we have to take can build on that acquis, their nature and rationale respond to a new challenge. We have to understand the novelty of that challenge, which is about more than adding "a social dimension". Third, the emphasis on a Social Union is not a coincidence. A European Social Union is not a European welfare state: it is a union of national welfare states.
The core idea can be summarized as follows; a social union would support national welfare states on a systemic level in some (not all) of their key functions, such as macroeconomic stabilization; and it would guide the substantive development of national welfare statesvia general social standards and objectives, leaving ways and means of social policy to member stateson the basis of an operational definition of "the European social model". In other words, European countries would cooperate in a union with an explicit social purposehence, the expression "European Social Union". A social union, so conceived, is not only desirable, but necessary. The idea of a Social Union raises complex issues of reciprocityi.e. a conception of what member states can demand from each other and how they should support each other with a view to delivering on commonly agreed objectives. 2 Developing such a sense of reciprocity requires a shared understanding of the relationships between policy inputs and policy outcomes, taking on board all the complexities, nuances and uncertainties involved in these relationships. This underscores, again, the need for in-depth and comprehensive comparative social policy analysis.
Is the concept of a European Social Union unduly idealistic, and therefore a bridge too far for today's Europe? As a matter of fact, the Five President's Report on the future of EMU, to which I referred in the introduction, signals an increasing consciousness of the need to develop a "social union"although that label is not explicitly used. That is, it signals an acute awareness that we need to organize more solidarity in the Eurozone, and that reconnecting with the aim of economic convergence (in contrast to the current experience of divergence across the Eurozone) also requires the definition of common social standards. Following the Five President's Report, the European Commission has launched, on one hand, the idea of a European "Pillar of Social Rights" and, on the other hand, a revamped effort for social benchmarking. The defining feature of this new mode of social benchmarking would be its focus on "policy levers", which marks a departure from the emphasis on outcomes which dominated in the social dimension of the Europe 2020 strategy (and which quite naturally characterizes any overall analysis of welfare state performance; cf. my argument in section 1.3, supra). This means that the need for solid comparative evidence with regard to the relationship between "policy inputs" and "policy outcomes" comes ever more to the fore, as the Commission rightly emphasizes: Benchmarking, if appropriately used, can be a truly powerful lever for action. In particular, benchmark indicators need to meet two requirements. First, they need to closely relate to the policy levers, such that they can lead to actual and meaningful policy implications. Second, there needs to be robust evidence and enough consensus that they contribute significantly to higher level objectives such as jobs, growth, competitiveness, social inclusion and fairness or financial stability. (European Commission 2015, p. 6) At the time of writing, it is not clear which direction will be taken and what the exact relationship will be between (i) a "Pillar of Social Rights", (ii) the existing processes of benchmarking in EU governance (such as the "Joint Assessment Framework" for Comparative Social Policy Analysis in the EU 397 employment and social policies), and (iii) the new approach to social benchmarking.

Many options are still open.
Admittedly, there is a tension between this new emphasis on "policy levers" (in the Commission's plans, which I support) and my earlier conclusion that one-size-fits-all approaches are to be avoided and member states should have room to explore the best ways and means to reach certain outcomes. However, in light of our recent experience and the need to make EMU sustainable, focusing on a selective set of policy instruments definitely makes sense with a view to achieving more convergence across Eurozone member states in specific policy domains. A salient example is the quality of unemployment insurance schemes: if macro-economic stabilization is a matter of common concern (and if, one day, the EU may have to play a supporting role in this respect), sufficient coverage and generosity of unemployment benefits also becomes a matter of common concern, just as the quality of activation policies is a (related) matter of common concern. A similar argument can be applied to minimum wages and minimum income protection, which may be seen as a matter of common concern per se in an ever more integrated union. However, even if the focus shifts to a set of "policy levers", there must be sufficient room for diversity; in the domain of minimum income protection, national policies should be analysed in terms of "policy packages" rather than in terms of single instruments, thus leaving a useful space of subsidiarity (see Cantillon et al. 2015 for an assessment of policies in terms of "policy packages", including minimum income assistance, minimum wages and the tax-and-benefit system). The analytical challenge then is to relate diverse national "policy packages" to national policy outcomes. There is in fact no contradiction between a governance process in which the overall performance of EU welfare states is scrutinized, on the basis of overall outcomes, and a governance process that focuses on a subset of specific policy levers: these may be complementary approaches.

Challenges for the Statistical Capacity of the EU
The challenges for comparative social policy analysis which I indicated in the previous sections imply related challenges for the statistical capacity of the EU. "Statistical capacity" refers not only to the capacity to collect data, but also to the quality of the monitoring processes organized on the basis of those data and the analytical capacity of institutions such as the European Commission.
First of all, a solid statistical infrastructure is vital. During the last decade, tremendous progress has been made in this respect thanks to the European Union Statistics on Income and Living Condition (EU-SILC), which replaced in 2004 the European Community Household Panel (ECHP) as the common European source for data on income and social inclusion. The EU-SILC database is a cross-sectional and longitudinal sample survey, coordinated by Eurostat, based on data from the EU member states. EU-SILC provides data on income, poverty, social exclusion and living conditions in the EU. It was established to provide data on structural indicators of social cohesion and to provide relevant data for two "open methods of coordination" in the field of social inclusion and pensions in Europe.
Common guidelines for EU-SILC ensure output harmonization of the survey results. Within these guidelines, national statistics offices have a certain degree of discretion to implement the guidelines according to the national conditions. Although basic rules on definitions, time reference, minimum effective sample sizes, etc. are legally binding, considerable differences remain among participating countries in terms of sample design, data collection and post-collection processing, with varying impact on the comparability of the results (see Decancq et al., 2014, pp. 78-79 for a short discussion of these aspects and references for further reading).
Tackling a number of rather obvious problems with EU-SILC simply requires more investment in the instrument. The issue of sample sizes is one example. On the basis of EU-SILC we can be quite sure that the increase in child poverty is related to migration in a number of countries: in Belgium, for instance, the quantitative impact of poverty in migrant households with children on the overall evolution of child poverty is very important, but, taking into account statistical confidence intervals, the Belgian sample size is too small to pursue a fine-grained analysis of the relationship between migration and poverty. The need to link survey data with administrative data is another example: in a number of countries, Belgium again being an example, huge progress would be possible by linking EU-SILC survey data with administrative data; but this requires an extra budgetary effort. Insufficient investment in EU-SILC may also contribute to differences between estimates based on EU-SILC and those based on elaborate national surveys (with much larger sample sizes and more extended longitudinal series) which remain largely unexplainedan example is the striking difference between the evolution of German poverty estimated on the basis of EU-SILC and German poverty estimated on the basis of the German Socio-Economic Panel (SOEP). Instead of the rather spectacular increase in German poverty between 2005 and 2007 registered in EU-SILC, the German SOEP is indicative of relative stability in the poverty headcount from 2004 onward, which comes after a considerable increase between 1999 and 2004 (Vandenbroucke and Diris 2014, pp. 6, 53). Obviously, an increase in poverty before 2004 rather than after 2004 suggests a very different explanation with regard to the causal role of German policies. Admittedly, asking member states to invest more money in what is, basically, common European data, is a thorny issue in times of budgetary austerity. But it is nonetheless a necessity: without sufficient national investment in the statistical infrastructure, the EU cannot develop a credible social policy monitoring system. Other issues with EU-SILC, such as the need to increase both the uniformity and the transparency of certain procedures for the users of SILC, are maybe less matters of money than matters of principle.
Space forbids to elaborate on yet other challenges, such as the need to increase the duration of the panels in EU-SILC (currently four years), and the need for good data on the use of social services. 3 Let me stress one final point: efforts to improve EU-SILC must be reconciled with continuity of the data series. For an intelligent discussion on the comparative performance of European welfare states, one needs both cross-sectional (cross-national) and longitudinal (cross-time) analysis. One should be aware that analysing changes in poverty risks in single countries, from an intertemporal perspective, and making a cross-national analysis of poverty levels are two quite different exercises. Much confusion, both in the academic and the policy debate, is caused by the unwarranted conflation of these two perspectives. By way of example, perhaps we ought to explain how contemporary Sweden can continue to rank quite well in terms of poverty in a cross-national perspective and yet, as a country, has clearly followed an inegalitarian Comparative Social Policy Analysis in the EU 399 course with increasing poverty in the second half of the past decade. Factors accounting for the Scandinavian "superiority" in terms of social inclusion and employment (relative to performance levels across Europe) are not preventing some Scandinavian welfare states from shifting to less outstanding outcomes (more so than is the case in other European welfare states). In order to develop a meaningful analysis of these trends, one needs data that span a longer time period than the EU-SILC can offer today; hence, we must not create unmanageable breaks in EU-SILC. It may even be better to resist the temptation to make marginal improvements to the database if such action will result in breaks in series. The French phrase "le mieux est l'ennemi du bien" is appropriate in this respect.
We not only need a leap forward in the statistical capacity of the EU in a narrow sense, but also a leap forward in its analytical capacity. Compared to the analytical capacity which the European Commission continuously uses in its economic surveillance, the analytical capacity applied in social policy monitoring is limited. Analytical capacity implies the capacity to model the relevant relationships between various economic and social variables. In this respect, a major problem is that the analysis of social developments is informed by realities that are commonly two years old; we need to develop realtime analysis of social developments and the social impact of economic developments. "Now-casting", a methodology that is currently studied, may be a promising development (Rastrigina et al. 2015).
Finally, I should stress that we need interdisciplinary research and a productive combination of qualitative and quantitative research. A holistic approach to the performance of welfare states must study the complementarities between interventions in the domains of education, child care, activation and social protection. It must understand the spill-overs from poverty to education, from poverty to health, from education to health and poverty, etc. This requires an interdisciplinary perspective in academia and a crossdepartmental approach within the European Commission. Understanding these complex complementarities and spill-overs also requires the combination of qualitative research and quantitative analysis. Even if survey data were impeccable, comparative social policy analysts cannot content themselves with just quantitative analysis. They would still be unable on this basis alone to make a detailed assessment of the quality of social and employment policies pursued in Europe's welfare states. Without recourse to contextual evidence, other data and in-depth analysis of policy, one cannot assess with great certainty the adequacy of policy trajectories. In general, one should resist making sweeping generalizations about such complex questions simply on the basis of statistical analysis of survey data. Not only microsimulation, policy indicators and in-depth studies of national policy, but also a broader perspective capturing discourse, social and demographic transitions, institutional change and macroeconomic policies are needed for an appropriate understanding of welfare state performance (Cantillon and Vandenbroucke 2014, p. 320).

Conclusion: Increasing Demands for In-Depth Comparative Social Policy Analysis
Comparative policy analysis has come to play an increasingly important role in the policies and politics of the EU. European social policy, as it developed since the 2000s, is intrinsically interwoven with comparative social policy analysis, notably in the context of the OMC. A review of the role of comparative social policy analysis in the OMC highlights a number of long-standing challenges. Do EU member states commit themselves to a real "burden sharing" with regard to the agreed objectives? How can we effectively mainstream social policy objectives into the EU's economic and budgetary governance? How can we assess the comparative efficiency of EU welfare states? These challenges have a political, a scientific and a technical (data-related) nature. They underscore the necessity for in-depth, shared and validated comparative social policy analysis of member states' policies. In the second section of this contribution, I indicated the need for and the possibility of the new departure in EU social policy, which we may now witness and which entails new demands for comparative social policy analysis. In the third section, I zoom in on the related challenges for the statistical capacity of the EU: the bottom line is that a huge investment is required in the development of reliable and comparable data.