Social investment as a conceptual framework for analysing well-being returns and reforms in 21st century welfare states

Welfare provision is often conceived through the lens of decommodification and analysed in (re)distributive terms. This article argues that a distributive approach does not sufficiently capture the complexity of 21st century welfare state dynamics. It proposes re-conceptualizing provision as a mix of three policy functions: raising and maintaining human capital stock; easing the flow of gendered life-course and labour-market transitions; guaranteeing social safety-net buffers. This analytical perspective allows theorizing life-course multiplier effects and policy (non-)complemen-tarities, both at the level of individual objective and subjective well-being and in terms of aggregate employment, poverty and fiscal sustainability. This perspective also enables us to extend the temporal horizon of welfare politics beyond short-term electoral logics for explaining welfare reform. The article underscores how methodological pluralism remains key for understanding contemporary welfare states, and for grasping welfare outcomes and institutional change in a research endeavour that involves both generalization and contextualization.


Introduction
Starting from Karl Polanyi's Great Transformation, political economy scholarship has acknowledged the key role of the welfare state in reconciling market capitalism with liberal democracy. In the face of economic and demographic challenges of the post-industrial age, advanced welfare states have experienced a groundswell of reform over the past decades V C The Author(s) 2022. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https:// creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. (Taylor-Gooby, 2004;Armingeon and Bonoli, 2007;Hemerijck, 2013;Garritzmann et al., 2022). Since the early 2000s, the notion of 'social investment' as welfare policy has become appreciated at the level of international organizations (European Commission, 2013;OECD, 2014;World Bank, 2016), where it is increasingly associated with 'inclusive and sustainable growth'. Beyond the early interventions of Esping-Andersen et al. (2002) and Morel et al. (2012), the reception of social investment has remained somewhat hamstrung in academia. A common practice has been to juxtapose activating social investment policies to more traditional social compensation policies, thereby suggesting inherent tensions between the categories of social investment and social compensation (Cantillon, 2011;Beramendi et al., 2015;Ronchi, 2018;Ferragina, 2022). Scholarly efforts considering social investments as complementary to social protection have been made (e.g. Hemerijck, 2013Hemerijck, , 2017Garritzmann et al., 2022), but agreement is lacking on the theoretical logic behind social investment welfare provision.
Breakthroughs in the study of the 20th-century welfare state find their roots in the joint effort of sociology and political science to reduce real-world complexity based on shared conceptual heuristics. Over 30 years ago, Esping-Andersen's (1990) Three Worlds of Welfare Capitalism epitomized a paradigm revolution in comparative welfare state research. By systematically highlighting decommodification as the gist of industrial-age social security, The Three Worlds became the keynote reference point in welfare state analysis (Kersbergen and Vis, 2015). Today, large swaths of welfare scholars continue to employ decommodification as a key yardstick to assess and interpret welfare state change and outcomes (among others, Scruggs and Ramalho Tafoya, 2022). However, transformative reform in welfare provision over recent decades is increasingly recognized as making extant concepts and categories difficult to apply (Esping-Andersen, 1999;Hemerijck, 2013;Ferragina, 2022).
Theoretical progress in welfare state research is further hindered by sectoral and methodological specialization, which more than before inhibits multidisciplinary engagement, at a time when interdisciplinary research is held in high regard. In sociology, there is a strong advocacy to specialize as a 'population science', privileging a focus on population regularities as main explananda, whose micro-level mechanisms are to be tested through rigorous statistical analysis aimed at generalizability (Goldthorpe, 2016). Many social policy scholars, aiming for isolating causal inference, limit welfare policy analysis to the study of single purpose policies (e.g. Card, 1999;Havnes and Mogstad, 2011) or of short-term redistributive effects (e.g. Cantillon and Vandenbroucke, 2014). Partly due to the methodological challenges that such assessments pose, the study of longer-term effects of policy mixes, their synergies, their second-order effects on other life-course outcomes beyond disposable household income, as well as their indirect effects on the overall carrying capacity of the welfare state thereby receive less attention . Political science scholars are increasingly reorienting to bottom-up electoral competition over welfare reform agendas by concentrating on the micro-level demand-side of the political process (Beramendi et al., 2015;Abou-Chadi and Immergut, 2019;Manow et al., 2018). With few exceptions (e.g. Garritzmann, Hä usermann and Palier, 2022), the throughput and output sides of the policy process have taken a back seat. Comparative political economy has similarly shifted attention to how external and internal pressures unleash dynamics of distributive policy drift, understood in terms of a convergent erosion of social and labour rights developed over the industrial era (Baccaro and Howell, 2011;Emmenegger et al., 2012;Amable et al., 2019). Alternative trajectories of 'embedded' liberalization are generally considered exceptions to the general rule of social retrenchment and labour market deregulation (Thelen, 2014). All of the above highly diverse theoretical perspectives and methodological approaches share an understanding of welfare provision as principally centred on redistribution. It is our contention that a redistribution-only conceptual bias makes it difficult to grasp contemporary welfare transformation. Twenty-first century welfare provision can neither be reduced to mere products of partisan political behaviour or downward liberalization reforms, nor wished away as spurious noise between aggregate sociodemographic patterns and individual characteristics.
In terms of policy analysis, the welfare state can be studied from the perspective of either problem-or interaction-oriented questions (Scharpf, 2000, pp. 762-763). Problem-oriented research is about the effectiveness and efficiency of public policies at solving societal problems, while interaction research is more concerned with the institutional conditions that favour or impede the adoption of given policies. In problem-oriented research, substantive knowledge is brought forth by discipline(s) closely related to the policy under scrutiny, which in the case of welfare state research are sociology and economics. In interaction oriented research, knowledge is provided by disciplines that study processes of policy selection, implementation, and delivery, ranging from political science to public administration. We maintain that in order to identify general causal mechanisms while also accounting for context-specific factors, welfare provision needs to be understood through the lenses of both problem-oriented and interaction-oriented policy research.
This article proposes a novel conceptual and methodological framework to re-engage sociological and political science in the empirical study of contemporary welfare state dynamics, centred around four contributions. First, we theorize a 'social investment life-course multiplier' whereby policy complementarities potentially fuel a virtuous cycle of well-being returns over people's life courses, while at the same time fostering employment levels so as to grant the macroeconomic sustainability of welfare systems in the aggregate. Second, we move away from treating different policies in terms of distributive trade-offs and conceptualize the welfare state as a bundle of interdependent policy provisions, which include income redistribution, but also capacitating welfare policies that can reap long-term social returns beyond redistribution. Third, we recognize the growing importance of subjective perceptions of the quality of welfare provision, alongside objective welfare outcomes such as poverty reduction and employment. Fourth, we underline the imperative to explore the political-institutional prerequisites of effective and durable social investment reform across countries and at different levels of governance and delivery. Within this overarching framework, theoretical advancement is accompanied by principles of methodological pluralism, from which we derive our 'multi-layered' methodology. Different quantitative and qualitative research efforts should be triangulated to best grasp problem-oriented welfare outcomes and interaction-oriented institutional dynamics. A continuous interaction between generalization and contextualization-we argue-is imperative to advance contemporary welfare state research in a multi-disciplinary endeavour.
The next section takes stock of the recent developments in welfare state research and provides an operational definition of social investment welfare provision. This serves as a lynchpin for developing a comprehensive policy theory to analyse today's welfare state dynamics. Section 3 spells out our theoretical and normative contribution. Section 4 makes the case for methodological pluralism in comparative welfare research by way of suggesting a multilayered methodological platform to both test and contextualize the life-course multiplier conjecture across different layers of inference. The last section concludes and elaborates on how comparative welfare research can further exploit emergent theoretical and methodological innovation.

2.
A case for theoretical reconceptualization in comparative welfare state research 2.1 Beyond the equality-efficiency trade-off Towards the end of the 20th century, welfare states were considered to be fraught with tensions between equality and efficiency. According to the so called 'big trade-off' argument (Okun, 1975), while on the one hand welfare spending reduces inequality, on the other it harms economic growth due to market distortions that social protection generates. In the mid-1990s, Okun's trade-off seemed to apply particularly to generous and expensive European welfare states, whose economic sustainability was called into question. Welfare retrenchment and labour market deregulation was required to overcome economic stagnation, albeit at the cost of greater inequality (OECD, 1994).
As described in Hemerijck and Ronchi (2021) and summarized in Figure 1, this predicament no longer holds. In the 21st century, the trade-off between economic efficiency and social equality does not apply to all advanced welfare states. Only the United States (US) and the United Kingdom (UK) seem to reach high employment levels to the detriment of equality, given relatively low levels of welfare spending (upper-left corner of Figure 1; the size of country markers is proportional to welfare expenditure). Other welfare states, particularly those characterized by high spending in continental and northern Europe, have high employment rates and comparatively low levels of inequality (upper-right corner of Figure 1).
Social spending levels have remained relatively stable over the past two decades. Nevertheless, since the 1970s European welfare states have been recalibrating their social policy mixes to address new social risks (Taylor-Gooby, 2004). In addition to the new social risks associated with deindustrialization, globalization, population ageing and the feminization of employment, governments also had to cope with the growing costs of established welfare programmes and tighter financial constraints. By the late 1990s, the concept of social investment emerged as a welfare strategy to reconcile the economic and social goals of advanced capitalist democracies (Esping-Andersen et al., 2002). The countries that went beyond Okun's trade-off are those that pursued 'social investment' reform strategies (Hemerijck and Ronchi, 2021), as reflected by the relatively high shares of welfare spending on in-kind benefits (which include human capital-and family-oriented welfare services) in the upper-right part of Figure 1.
As intimated above, Esping-Andersen (1990) anchored his conceptualization of welfare proficiency on levels of decommodification from market forces. Ever since, welfare research settled on a view of social policy as boiling down to redistributive tussle, notably in political science. For example, Paul Pierson's (2001) conception of the 'new politics of the welfare state' was based on the idea that welfare reform was difficult as this would require cuts in popular welfare programmes, to be enacted by politicians seeking re-election. Distributive politics struggles to make sense of widespread-gradual but nonetheless transformativereforms in labour market and family policies observed since the millennium (Ferragina, 2022). This shortcoming, we argue, is inherently related to understanding of welfare (non-)reform in terms of decommodification and redistribution. Today, welfare reform seems to hovers largely around (embedded) labour-market commodification by, for example, increasing female employment, reconciling work and family life, and active ageing, for the purpose of rendering existing and future welfare commitments such as pensions more sustainable (Hemerijck, 2013).

A working definition of social investment in terms of complementary welfare policy functions
The social investment perspective extends the focus of welfare state provision from ex-post income compensation to risk prevention and capacitation. Policies such as early childhood education and care (ECEC), education and training, lifelong learning (LLL), active labour market policies (ALMP), parental leave, other work-life balance (WLB) policies and longterm care (LTC) all share objectives that go beyond here-and-now redistribution. The unifying objective of social investment is to enhance people's capabilities and opportunities in knowledge-based labour markets, so as to promote high levels of employment, inclusive social cohesion, individual agency, and overall life satisfaction. In addition to their direct effects on programme beneficiaries, social investment policies have the capacity to increase the number and productivity of (today's and tomorrow's) workforce. This in turn bolsters tax revenues necessary to guarantee the financial sustainability of generous welfare states (Myles, 2002;Hemerijck, 2017). The 'carrying capacity' of the welfare state can be expressed in a simplified equation (adapted from Myles, 2002) presenting social cost allocation:  In welfare state research, the focus is habitually put on the distributive side of the equation (the numerator), referring to the number of welfare recipients to be supported by the welfare state and the level of benefits. However, the long-term sustainability of social welfare largely depends on the denominator of the above formula, which becomes especially important at a time when much financial resources are needed to fulfil established welfare commitments in the face of population ageing. A wide range of policy interventions and investments across different life stages of individuals is required to boost the quantity and quality of employment so as to reinforce the productive side of advanced welfare states.
Since the beginning of the 2000s, various attempts have been made to qualitatively trace and quantitatively assess the social investment turn across countries (Huo, 2009;Vandenbroucke and Vleminckx, 2011;Morel et al., 2012;Hemerijck, 2013;Kuitto, 2016;Ronchi, 2018). The first wave of social investment research (and its critics) harboured a conscripted understanding of the new welfare paradigm. In empirical terms, this was often reduced to a policy-based operationalization centred on the ambiguous dichotomy between 'social investment' policies (also referred to as 'enabling' or 'activating') and 'social protection' policies (also labelled 'passive', 'compensatory' or 'consumption-oriented'). Although this provided a convenient heuristic to trace macro-trends of welfare recalibration, it encountered many difficulties in policy categorization (De Deken, 2017). Moreover, it contributed to the creation of an artificial juxtaposition of investment versus protection policies, which also fuelled the idea of social investment being a substitute for de-commodifying social protection programmes. Such juxtaposition, however, is at odds with the original notion of minimum income guarantees as a precondition for effective future-oriented social investments (Esping-Andersen et al., 2002, p. 5). Moreover, in operational terms, early advances in social investment research followed the same point-intime redistributive conceptualization as with traditional cash transfer programmes without considering that social investment policies can generate returns over different time horizons and may be designed to fulfil also preventive (ex-ante) welfare functions. This distribution-only operationalization has partially guided the identification of the so-called 'Matthew Effects', which refer to situations when the middle classes disproportionately benefit from capacitating services at the expense of lower income groups (Cantillon, 2011;Cantillon and Vandenbroucke, 2014). This delineation falls short of grasping the very engine of the social investment edifice, that is, potential positive-sum policy complementarities rather than redistributive trade-offs.
More recently, in their comprehensive work on the World Politics of Social Investment, Garritzmann, Hä usermann and Palier (2022) seek to go beyond the policy-based definition of social investment by elaborating a new one that distinguishes between social investment goals, functions and policies. From a political science perspective that takes welfare change as explanandum, this definition promises to be useful for identifying coalitions behind welfare reforms following diverse political rationales. Given that our aim is to provide theoretical concepts and methodological tools for welfare state research in both political science and sociology, we rely on a working definition of social investment based on policy functions and their interactions, thus leaving the goals (i.e. outcomes) open to be subjected to empirical enquiry.
As argued by Hemerijck (2013Hemerijck ( , 2017, the social investment edifice is built on three complementary policy functions 1 : (a) raising and maintaining the 'stock' of human capital and capabilities throughout the life course (lifelong human capital stocks); (b) easing the 'flow' of gendered labour market and life-course transitions (work-life-balanced flows); and (c) granting inclusive safety nets (inclusive buffers). In day-to-day policy practice these three welfare functions overlap and perform in an interconnected fashion. Inclusive buffers, for example, aim at reducing poverty, but can also ease the labour market flow by improving job matching and reducing skill depletion (Gangl, 2006). Childcare services are aimed at boosting human capital stock, but can also facilitate labour market flows Nieuwenhuis, 2022). However, it should be recognized that stock, flow and buffer policy considerations are subject to highly diverse socio-economic dynamics. As measures of income compensation, buffers originate from business-cycle Keynesian macroeconomics and demand management, serving to secure household economic security over periods of employment inactivity. Flow policies, especially parental leave and family policies, are inherently related to a post-industrial dual-earner family demography, whilst stock policies find their origins in the (changing) skill requirements on the supply side of the economy.
From these combinatorial intricacies, it follows that a welfare state cannot be characterized by looking only at specific policies nor by presumed singular policymaker goals. The ideal-typical social investment welfare state is best understood in terms of an optimal configuration of stock-flow-buffer policies. It then requires further theorisation and empirical enquiry to assess the joint effects that existing varieties of policy mixes produce both at the micro level of individual life courses and at the macro level of post-industrial knowledge economies and aging societies. By implication, not all configurations of social policies are necessarily conducive to realising the policy objectives of (full) employment, social inclusion and welfare sustainability. Empirically, the comparative inquiry should therefore be shifted towards questions of whether specific policy mixes realize social investment returns, how policy complementarities unfold, and why non-complementarities between varying policy mixes can emerge in real-world welfare provision.

Theoretical contribution
The focus on policy functions opens the door to a research perspective for an improvement in understanding and analysing contemporary welfare state dynamics. In particular, it 1 We acknowledge other conceptualizations of social investment as equally valuable. For example, the three categories proposed by Garritzmann et al. (2022) (creating, mobilizing, and preserving human capital and capabilities), resemble the logic behind what Hemerijck (2013) refers to as stocks, flows and buffers, with the exception of the 'buffer' function which, in Hemerijck's conceptualization, is more encompassing, as inclusive social protection is understood as a key ingredient to social investment. Overall, the most common approach has been to exclude the so-called 'passive' or 'compensatory' social policies from the analysis of social investment, which is useful from an analytical point of view. Our argument is to include them as part and parcel of welfare policy complementarities.
Social investment as a conceptual framework allows for the integration of four research objectives. The first concerns the theoretical conjecture of a 'life-course multiplier', which forms the backbone of the social investment paradigm. The second pertains to the identification and evaluation of the complex interactions behind policy complementarities (or the lack thereof), which is needed to assess welfare policy functions effectively. The third adds the normative dimension of capabilities and subjective well-being returns to the empirical analysis of social investment. The fourth considers the political-institutional pre-conditions that facilitate the long-term political commitment and effective delivery of social investment across countries.

The life-course multiplier: the logic of social investment
For social investment to work in practice much depends on how inclusive buffers, work-life balanced flows and lifelong stocks can be made to work together to reap synergetic wellbeing returns. The growing evidence on how policies reinforce each other has inspired the formulation of a social investment 'life-course multiplier', which theorizes that complementary social investment policies over the life course of individuals and at a macro level create a cycle of well-being in terms of skill acquisition, employment and productivity, gender equality and poverty alleviation (Hemerijck, 2017). As represented in Figure 2, the virtuous cycle begins with early investments in children, boosting cognitive development and skill acquisition in later life (Cunha and Heckman, 2007). These lead to better educational outcomes, which, together with training, produce higher and more productive employment (Card, 1999) and increased ability to shape one's life (Sen, 1999;Wolff and De-Shalit, 2007). Effective ALMPs and work-care reconciliation policies are expected to produce higher levels of employment and lower gender inequality in the labour market (Nieuwenhuis et al., 2012;Korpi et al., 2013). More opportunities to combine paid labour with family commitments can protect households against worklessness and poverty (Cantillon and Vandenbroucke, 2014) and can have a positive effect on perceived life satisfaction and met aspirations, for example with regards to work-life balance (Hobson, 2014) and family formation (Sobotka and Beaujouan, 2014). Finally, active ageing and lifelong learning policies can lead to a higher effective retirement age (Walker, 2002). This in turn implies a larger tax base to invest in poverty reduction and social inclusion programmes, squaring the circle of the social investment multiplier function. The social investment multiplier captures both micro-level and macro-level dynamics. At the micro level it captures the material and subjective well-being returns for individuals and households. At the macro level, it theorizes cumulative societal benefits in terms of productivity, employment, equality, social inclusion, economic growth and fiscal sustainability (in italics, Figure 2). This new welfare state logic offers a theoretical basis for hypothesis generation.

Policy complementarities: the modus operandi of social investment
Notwithstanding conceptual headway, the conjecture of a social investment life-course multiplier, however plausible through the lens of extant macro-level evidence (OECD, 2014;Hemerijck, 2017;Plavgo and Hemerijck, 2021;Nieuwenhuis, 2022), continues to be short of micro-empirical foundations: we still do not know whether and how bundles of stockflow-buffer policies enhance citizens' social and labour market inclusion over the life course, and how these interact at a macro level. Empirical research efforts need to be explicitly geared towards a better understanding and assessment of the interaction effects between different policy provisions across space and time. The concept of 'institutional complementarities' from the Varieties of Capitalism literature (Hall and Soskice, 2001) serves to clarify this point. According to this literature, two institutions are complementary when they mutually reinforce each other's returns (Hall and Gingerich, 2009). This definition goes along with the ideas elaborated by OECD economists, who recognize that the outcomes of given policies can depend on interactions with other policies, including the socio-economic context within which such interactions unfold (Bassanini and Duval, 2009;Thévenon, 2016). On the other hand, complementarity also refers to situations whereby different component parts of complex bundles of policies 'compensate for each other's deficiencies' (Crouch et al., 2005). Both delineations of institutional complementarity are relevant when assessing wellbeing returns on various welfare state policies. For example, the positive impact on female employment of childcare services and leave policies is maximized when both policies are implemented in tandem (Thévenon, 2016). Also, quality ALMP can serve to mitigate the potential moral hazard predicament attached to out-of-work income buffers, improving instead of hindering job search and job matching (Ronchi, 2019). The effectiveness of any given policy, like ALMP, often depends on the availability of other social policies, such as childcare services in combination with other work-family reconciliation policies Nieuwenhuis, 2022).
Large swaths of welfare state research today are preoccupied with single policies, either focusing on their outcomes (e.g. Havnes and Mogstad, 2011), or their relationship with individual preferences (e.g. Gingrich and Hä usermann, 2015). We, however, propose to remain

Early Childhood EducaƟon and Care
Children's cogniƟve and social development; Parents' employment inserƟon/preservaƟon

Investments in educaƟon
Access to quality educaƟon, inclusion, agency Higher ecomomic producƟvity, social mobility

Labour market policies
Higher employability, skills, income support SubjecƟvely percieved job saƟsfacƟon, work-life balance Higher, more robust labour-market parƟcipaƟon; Higher economic producƟvity  Source: Authors' illustration true to the notion of configurational rationales that characterized the literature on welfare regimes during the 1990s and 2000s (e.g. Esping-Andersen, 1990;Crouch et al., 2005;Hemerijck, 2013). Welfare regime theory implicitly assumed rather strong regime-specific institutional complementarities across the different policy provisions that make up modern welfare states. The main strength of this literature was to expose the structuring power of broad policy configurations shaping welfare outcomes and developmental trajectories, crucial for the understanding and explanation of policy change and institutional continuity in a path-dependent temporally ordered fashion. Yet, by holistically presupposing 'tightly-coupled' integrative policy repertoires, mutual interaction effects across the policies that make up welfare regimes remained theoretically underspecified and, by implication, empirically understudied. We need a better understanding of how different compensatory, activating and regulatory social policies work together and this requires us to disentangle configurational complementarities. Recognizing that the welfare regime literature struggles to effectively grasp within-regime differences (Ferragina, 2022), we propose that future research should inspect the policies making up welfare states as more 'loosely-coupled' bundles or policy-mixes. Policy complementarities can manifest themselves as mutual reinforcement of desirable welfare outcomes. Conversely, less fortuitous outcomes can be attributed to specific institutional non-complementarities of different welfare provisions at the country level and/or in lacking service delivery at the local level.

Work-life balance policies
Considering policy complementarities from a life-course multiplier perspective is beneficial to advance our understanding also in relation to the 'Matthew Effect' critique of social investment mentioned above. As anticipated, this strand of research considers capacitating and activating public services as substitutes for social cash transfers, while, according to our theoretical proposition, stock-flow-buffer policies are to be considered as complements rather than substitutes. Although assessing the immediate redistributive effects of capacitating and activating policies is important, disregarding the longer-term perspective of their intended outcomes can lead to an underestimation of the potentially positive returns on equalizing opportunities over life courses. By the same token, tracking household disposable income, which is easily done due to ample data availability, cannot easily capture the long term effects of social inclusion (especially those concerning employment inclusion). Consistent with the lengthier dynamic perspective suggested by the life-course multiplier conjecture, social investments such as childcare are to be measured not only in short-term redistributive terms, but also over a longer time period and with respect to intended outcomes. Moreover, policy effects should be considered not only at a micro but also at a macro level in order to see how they affect the carrying capacity of the welfare state (for example, whether activation policies increase the fiscal space to finance inclusive social transfers for risk groups not directly benefiting from them). We discuss this further in Section 4.

Social investment returns: bringing capabilities into the welfare equation
Thus far, the bulk of empirical attempts to assess social investment returns have focused on objective material socio-economic conditions, such as employment and poverty (for example Cantillon and Vandenbroucke, 2014;Rovny, 2014;Van Vliet and Wang, 2015;Bakker and Van Vliet, 2021). Although such outcomes are quintessential for assessing social investment returns, an exclusive focus on material welfare outcomes fails to capture the gist of the social investment paradigm in terms of a promise of increased agency and secure capabilities to realize life-course aspirations. Responding to the call of a number of scholars and international organizations to go beyond material conditions when measuring social policy outcomes (e.g. Hobson, 2014;Morel and Palme, 2017;OECD, 2017), 21st century social investment research should take the subjective capabilities dimension on board as a litmus test of the 'goodness of fit' of social investment for well-being progress, alongside objective socio-economic outcomes.
A growing interest in sociology for non-economic outcomes of social policy related to 'quality of life'-including such aspects of subjective well-being as life satisfaction and WLB-is taking root (Costanza et al., 2007;Hobson, 2014;OECD, 2017). From the perspective of political theory, there is a striking normative connection between social investment and the Nobel laureate Amartya Sen's 'capability approach' (Sen, 1999) with its strong emphasis on 'human flourishing', development and autonomy, signalling a shift from the Beveridgean notion of 'freedom from want' towards 'freedom to act' (for more on the link between social investment and capability approach, see Hobson 2014; Morel and Palme 2017). In the last decades, employability has come to play a large role in relieving poverty. But this surely begs the question of employment at what cost in relation to other life goals and their subjective appreciation. The capability approach of Amartya Sen and Martha Nussbaum capability approach provides a normative-theoretical benchmark for capturing the tension between people's aspirational preferences and ultimate 'achieved functionings' in terms of the quality of life they experience (Sen, 1999;Nussbaum, 2011;Robeyns, 2017). It opens up a new understanding on how welfare provision helps reduce 'capability gaps'-namely gaps between desired and attained life-course social, employment and family-related objectives (Hobson, 2014). The political philosophers Jonathan Wolff and Avner De-Shalit (2007, p. 159) add the notion of 'secure capabilities', underlining that the activation of capabilities requires an expectation that a certain level of 'achieved functionings' can be sustained over time. Drawing on these normative theories and applying them to welfare state research, we are interested in how the policy mixes of capacitation, activation and income protection (and demand stabilization) affect people's material conditions and subjectively perceived well-being.
By way of expanding Barbara Hobson's study on WLB capabilities to the broader context of comparative welfare state research, we propose a multidimensional set of indicators of 'social investment returns', outlined in Table 1, which includes both objective indicators of economic and social conditions, and the appreciation of the quality of social policy in terms of subjectively perceived life satisfaction and agency. As a first step to bridge the (secure) capabilities approach with empirical social investment research, we suggest a focus on 'quantity' and 'quality' returns in two core domains: citizens' sense of autonomy to act and satisfaction with their family and working lives and with how various welfare support policies allow them to reconcile work and family life while feeling financially secure. For different phases of the life course, social investment policy synergies can be assessed not only against outcomes in objective material conditions, but also against their proficiency in fostering people's capabilities to steadily fulfil their personal aspirations.

From generalization to contextualization: the political-institutional prerequisites of social investment
Any functional operation of social investment welfare states critically relies on managing policy synergies at the political and administrative level. The redistribution-only bias that we exemplified for sociological welfare research is even more prevalent in political science Social investment as a conceptual framework following its conception of politics in terms of a struggle over 'who gets what, when and how' (Lasswell, 1936). Consequently, the politics of welfare reform is increasingly studied from the supply-side of politics, mapping how since the mid-1990s some political parties started to promote education and activating policies to meet the demands of new middleclass constituencies (Beramendi et al., 2015). Following our theoretical propositions, research on welfare state development should not merely limit itself to the short-term (re)distributive conflicts between partisan actors, but must also look into longer-term policy-choice processes, such as the objectives of improving employment opportunities, social inclusion, pooling insurance, and fiscal sustainability. The recent evidence surveyed in the first sections of this article suggests that joint problem solving is not at all utopian. True enough, success in raising employment without trampling on redistribution depends on favourable institutional preconditions under which diverse political actors can achieve directional consensus over welfare reform and the administrative capacities for effective delivery.
Indeed, the alignment of stock-flow-buffer policies across time and place requires a considerable degree of cooperation across ministries and positive coordination between national and sub-national levels of government. While the traditional social protection bufferoriented welfare state mainly involves an effective (central) tax-benefit system, implicating at the administrative level the ministries of social affairs and finance, a social investment service-oriented welfare state also implicates the ministries of health, education and economic affairs. Moreover, as the social investment state is service-intensive, and since services in the areas of labour market policy and family support are provided locally, high-quality sub-national delivery capacities are quintessential. Altogether, two administrative prerequisites align on this score: (a) national political-administrative backing for sub-national social investment policy discretion and (b) implementation network quality surrounding local social investment delivery. Finally, the alignment of stock-flow-buffer policies across time and place requires an element of administrative continuity. Given the overarching life-course time-horizon, the effective delivery of social investment requires successive governments to adapt social policy provisions to newly emerging life-course social needs. Intuitively, 'consociational' democracies may be institutionally more prone to pursue consolidated social investment reform than majoritarian or presidential systems, because they allow for more continuity over time across legislatures. In party systems with proportional representation, alternation in government tends to be more partial whereas in majoritarian systems government turnover tends to be sweeping (Iversen and Soskice, 2006). Beyond these broad patterns, any reform reality is bound to be far more complex.

Methodological pluralism: layering causal adequacy
Doing justice to the complexity of 21st century welfare provision is challenging. Many methodological predicaments arise, which concern the multidimensional and complementary nature of social investment, its diverse implications for different aspects of well-being, the temporal complexity of such implications, and the goodness of fit between national policy repertoires and local delivery. To address this challenge, we need to bring together existing approaches by combining different methods and levels of analysis into an integrated research endeavour. For analytical clarity, we spell out four methodological 'layers' which should underlie a comprehensive and consistent agenda for contemporary comparative welfare state research. Each layer addresses specific issues that have been discussed above. Although the layers are independent in terms of tasks and methods, when taken together, all in all, the multilayered methodology is greater than the sum of its parts, as it can achieve 'causal adequacy' in a cumulative fashion (Scharpf, 2000;Goldthorpe, 2016). The proposed methodological framework seeks to integrate different quantitative and qualitative methods of data collection and analysis into a collective research endeavour whereby large-N statistical analyses (macro and micro) and small-N comparative historical and case-study analyses complement each other. While large-N analyses highlight general patterns that transcend idiosyncratic cases, studies based on small-N analyses uncover the underlying causal mechanisms, which can be time-and context-specific, while also detailing actual policy contents and institutional contexts that are hard to grasp through aggregate and internationally comparable quantitative indicators.

Establishing macro regularities
A fundamental first step in the multi-layered methodology is to establish regularities at the macro level. This research layer sets the scene for the more in-depth analyses conducted in other layers by tracking developments of stock-flow-buffer policies over time and by shedding light on their association with the economic and social performance at the country level. This can be done by examining variation in social policy configurations and societal outcomes between countries, showing correlation between, for instance, aggregate social investment expenditures and national employment and poverty rates. Another method is to perform country fixed-effects regression analyses on pooled time-series cross-sectional data to exploit over-time variation in policy performance within countries. Pooled time-series and panel data research designs are particularly useful in cases of sufficient over-time variation in the policies of interest.
Some examples of establishing macro regularities in the social investment literature, although not directly aimed at testing the life-course multiplier conjecture, can be found in Van Vliet and Wang (2015), who study poverty outcomes, and Bakker and Van Vliet (2021) and Nieuwenhuis (2022), focusing on employment outcomes. Typically, such crosscountry analyses are based on population aggregates. However, given the differences in target populations across various social investment interventions, macro regularities related to social investment returns should preferably be analysed also among crucial population sub-groups that are especially vulnerable to the new social risks, such as youth, mothers with young children and older low-skilled workers. Furthermore, when methodologically feasible, the time-frame of measuring social investment returns should be adjusted to account for the time needed for different policies to have an effect. An example by Plavgo and Hemerijck (2021) focuses on employment and poverty outcomes among families with children across OECD countries. Findings show that in countries with higher social investments for this population sub-group, the share of dual breadwinner families and female employment is higher, and child poverty rates are lower. Following the life-course multiplier theory (Section 3.1), it is plausible that national-level investment in stock and flow policies enlarges fiscal space for poverty-reducing buffer policies also for risk groups that do not directly benefit from such policies. Research by Hemerijck et al. (2016) shows, for example, that higher social spending on ALMP and ECEC policies is associated with lower poverty rates among older low-skilled males, suggesting positive externalities for population sub-groups not directly targeted by social investment programmes like childcare.
Potential institutional spillover effects at the macro level need to be investigated further to test the life-course multiplier theory. Methodologically, examining complementarity among policies can be done by introducing interaction terms in regression analyses. In this way, it is possible to study if the effect of a given policy on a given socio-economic outcome is moderated (strengthened/weakened) by investments in (or specific characteristics of) another policy. Today, there is little empirical research on social investment complementarities. Hemerijck et al. (2016) showed that the positive association between investments in ALMPs and employment is stronger when investment in ECEC services is higher, especially for specific risk groups. Similar findings have more recently been exposed by Nieuwenhuis (2022) on the complementary nature of ALMP and ECEC policies on female employment. Bakker and Van Vliet (2021) found diminishing marginal returns between the same two policies. These authors, however, do not distinguish between risk groups. These mixed results illustrate that further research is imperative to better gauge macro regularities in terms of (conditional) effects of specific social investment policies on aggregate socio-economic outcomes, preferably by risk group.

Putting macro regularities to the micro-foundational test
On top of establishing regularities at the macro level, it is important to corroborate country level correlations by putting their micro-level foundations to the test. The second research layer thus includes empirical research that aims at exploring the micro-mechanisms beneath observed macro-regularities by studying whether and how social investment policy mixes affect individuals' life-course transitions. For this purpose, it is important to move to the individual level as the unit of analysis. This allows assessing the direct effect of policy (mixes) on individual labour-market and family-formation transitions, and enables to identify potential heterogeneity in social investment returns depending on individual and household characteristics. The latter point is especially important for detecting possible inequalities in the extent to which different population sub-groups benefit from social investment policies, not only by income (the focus of the 'Matthew Effect' literature), but also by gender, age, parenthood status, occupation, contract type, among others.
One method that allows for this is multilevel modelling. This approach helps to analytically disentangle individual-level (sociodemographic) mechanisms and macro-level 'effects' of policies, which can be conditional on both individual characteristics and other policies or contextual/institutional factors. One example of such research is the work already mentioned by Hemerijck et al. (2016) which, besides aggregate data, also uses longitudinal EU-SILC micro-data for most European countries and applies multilevel modelling to study the direct and indirect policy effects on individuals' employment and poverty status. Another example by Rovny (2014), drawing on the Luxembourg Income Study micro-data in combination with macro data, applies multilevel modelling to analyse the effects of various social policies on the likelihood of being poor for specific risk groups. Findings of both studies show that family policies and ALMPs are important predictors of a decrease in the probability of being poor among vulnerable risk groups, as also found more recently by Zagel and Van Lancker (2022). These findings corroborate the macro regularities described above and shed further light on the mechanisms behind them.
A more fine-grained methodological approach to dig deeper into context-specific causal mechanisms is the use of country-specific long-term longitudinal or panel data. This approach follows individuals' lives over a longer period of time, assessing the (causal) relationships between (changes in) policy mixes and individual well-being outcomes at different lifecourse transitions. While some social investment policies have rather direct, immediate effects (e.g. the income-support function of cash-transfer buffers), others can be expected to have a measurable effect on people's life transitions over a longer time horizon. The prime example of this temporality issue in social investment is education. Investments in education are expected to affect people's skill development, educational pathways and transition to the labour market. Capturing such processes requires multiple time points. Thus, a limited time dynamic perspective can lead to underestimating the true over-time effects. Since in particular stock and flow policies tend to have effects over longer time-periods, snapshot comparison of their outcomes with those of buffer policies measuring immediate redistributive effects may lead to an underestimation of the 'true' effects of such policies and the overestimation of their 'Matthew effects'.
To address these issues, longitudinal household survey data and register data can be used to study the effect of time-varying indicators of stock-flow-buffer policies on critical lifecourse transitions. One such example is research by Havnes and Mogstad (2011) who draw on population register and administrative register data in Norway using difference-indifferences estimates to study long-term effects of subsidized childcare on educational attainment and labour market participation when children reach adulthood. Another example by Gangl and Ziefle (2015) exploits survey data from the German Socio-Economic Panel to identify family policy effects on mothers' work commitment. Another type of data, particularly useful for the study of flow policy effect on labour market outcomes, is firm-level micro-data. Petersen et al. (2014), for example, use longitudinal matched employer-employee data from Norway to examine changes in the gender wage gap during a period of significant reforms in family policies. All three studies find that the analysed childcare and family policy reforms had substantial effects on later life outcomes in education, employment and/or wages. These and other studies of this kind, however, analyse one policy at a time. Further research is needed to better understand policy (non)complementarities and how these affect life transitions of different population sub-groups exposed to different social and economic risks.

From quantity to quality returns
While the methodological approaches of policy analyses described in Section 4.2 are commonly applied to study objective socio-economic outcomes, it is equally important to study how policy mixes affect individuals' subjectively perceived well-being. As discussed above (Section 3.2), the aim of the welfare state broadening from ex-post income compensation to ex-ante risk prevention and capacitation has been to ensure not only high levels of employment, but also to promote equal opportunities, autonomy and agency, with a view to enabling citizens to achieve lives they value worth living. This normative reorientation shifts the focus of social investment returns beyond economic outcomes to include subjectively perceived wellbeing and capabilities to fulfil aspirations at different life transitions. It is therefore important to incorporate the subjective dimension into the framework of (social investment) welfare state research to investigate whether and how policies are appreciated by citizens and how their (interdependent) effects influence the quality of people's well-being.
As an example, previous research has established that parenthood can negatively affect not only one's probability of being employed (especially among women), but also one's subjective well-being. Lehmus-Sun (2021) studies whether social investment policies targeted at families with children mitigate the negative association between parenthood and 'subjective capacity', operationalized as individual agency, resilience, potentiality and meaning in life. Drawing on Eurofund's European Quality of Life Survey data, the author finds that while parenthood is generally associated with lower levels of 'subjective capacity', in countries with significant social investment, especially with respect to service provision, subjective well-being among parents is higher. These preliminary findings are in line with the notion that under certain circumstances, social investment policy mixes are capable of improving citizens' subjective well-being and reducing the 'capability gaps' between citizens' aspirations and realizations during critical life-course transitions (Hobson, 2014). More research is needed to better apprehend the relationship between social policy mixes and citizens' life satisfaction, sense of autonomy and ability to meet life course-specific aspirations.
Putting capabilities and subjective well-being alongside the objective outcomes, moreover, enables us to check whether and to what extent the capacitating potential of social investment interventions is appreciated by citizens. It may well be that policy synergies that show positive objective outcomes are subjectively felt as intrusive by relevant risk groups. By the same token, it may be that policies showing no positive objective outcomes (such as no insertion in the labour market despite participation in ALMPs), are valued by vulnerable groups and bring positive subjective well-being outcomes. There are some indications that available social policy mixes matter for subjective well-being. As an example, a recent study by Hadjar and Kotitschke (2021) finds that the gap in subjective well-being among people with and without disabilities varies across welfare regimes, with Nordic countries performing best in generating equal perceived life satisfaction. Social investment policy mixes as a potential explanation of such country differences is yet to be explored. Considering subjectively perceived well-being in parallel to objective socio-economic outcomes can reveal a well-being potential that goes beyond distributive outcomes.

The institutional support bases for social investment
A crucial layer of comparative welfare research to uncover is the political-administrative institutional conditions that matter for the effective implementation of social investment policies. This firstly entails tracking sequences of reforms in stock-flow-buffer policies, and assessing the extent to which the sequence of reforms is functional at achieving long-term socio-economic objectives. This can be done for instance by analysing publicly available documents that provide governments' socio-economic and financial priorities, and codifying the rationale behind governments' welfare reform efforts (Karremans, 2021). The next step is to trace the decision-making processes that lead to such policy sequences in order to determine what facilitates or hinders welfare continuity in governments' agendas. Methodologically, this can be done through process tracing with a focus on political (dis)continuity in the partisan composition of government, and the extent to which social investment objectives have been established in ministerial agendas. Elite interviews with policymakers and state bureaucrats help explore deeper under which conditions, policy legacies and administrative traditions, by and through which institutional channels, social investment priorities are placed on the political agenda. Separate attention should also be given to recommendations governments receive from international organizations such as the OECD and the European Union. To the extent that these organizations promote social investment recalibration, they can be conceptualized as a factor of policy continuity.
Finally, as already intimated above, the evolution towards more service-intensive social investment welfare states requires positive interdepartmental coordination across several ministries (social affairs, education, health and finance) and also effective national administrative backing of sub-national social service delivery. Based on a comparative case analysis of three post-industrial municipalities (Gothenburg, Barcelona and Lyon), Scalise and Hemerijck (forthcoming) have been able to expose how regional and local governance discretion is crucial for the activation of social investment policy synergies. Given the high reliance on sub-national governments in social investment policy delivery, additional research should address an important blind spot in extant welfare enquiry, which often neglects 'local contextualization'. In this respect, the objective should be that of seeking out the proficiency of vertical coordination between national administration and sub-national levels of government, alongside the critical role of horizontal policy discretion at the local level to align social benefits and capacitating services.

Conclusion: the 'double bind' of generalization and contextualization in 21st century welfare state research
Welfare state research is inter-disciplinary par excellence, bringing together different research traditions from sociology, economics and demography on the one hand, and political science, public administration and political economy on the other. Yet, after decades of methodological refinement and specialization, these disciplines risk losing sight of one another. Working in silos theoretically and empirically can result in not seeing the forest for the trees when it comes to understanding the new modus operandi and political reform dynamics of 21st century welfare states. In this article, we have proposed an overarching theoretical and methodological platform to re-engage political science, sociological and economic research in a common endeavour to better understand how welfare reform can work towards social investment and its effect on citizens' life-course chances and well-being.
The conceptual and theoretical contribution of our comparative policy research design consists of identifying potential synergies across complementary social policy functions that jointly ensure even-keeled economic and social progress in contemporary advanced capitalist democracies. This entails analysing the interplay across the three welfare functions of: (a) raising human capital stocks over the life course; (b) easing work-life-balanced flows in gendered post-industrial labour markets; (c) granting inclusive social protection buffers, and to explore how they develop and reinforce each other in the following terms: (a) objectively, at the macro level of national material socio-economic outcomes and (b) at the micro level across individuals' life-course transitions; (c) subjectively in terms of citizens' well-being and secure capabilities; (d) in terms of political processes of welfare recalibration at the national level, and, finally, (e) in sub-national contexts of capacitating service delivery. Ideal typically, these various dimensions of mutual reinforcement between social policy functions add up into a virtuous cycle of cumulative returns both at the individual level of citizens' well-being and at the societal level of macroeconomic welfare sustainability-a mechanism that we term the 'social investment life-course multiplier'.
Methodologically, in order to study whether, how, when and under what institutional preconditions social investment well-being returns are reaped, we need to persistently iterate between (problem-oriented) generalization and (interaction-oriented) institutional contextualization, induction and deduction, quantitative and qualitative methods, with a view to better recognise and appreciate emerging general patterns and their life-course effects against a background of ongoing reform dynamics and welfare delivery and governance. In other words, the various disciplines engaged in the study of welfare state change should re-unite and combine their efforts and methodological approaches under a multi-layered methodology. The recognition of a joint shared methodological platform should therefore incorporate large-N cross-sectional and longitudinal quantitative analyses as well as smaller-N qualitative methods and acknowledge the strengths and weaknesses of both approaches to answer different research questions.
The pluralist theoretical perspective and the multi-layered methodology we have outlined in this article are aimed at gaining a more comprehensive understanding of 21st century welfare provision, once again, as a configurational Gestalt, in order to reach generalizable knowledge about the mechanisms by which governments generate well-being returns in contemporary societies. Coming back to the Polanyian concept of social embeddedness of economic structures, we believe that today, after three-quarters of a century of the social market economy, the role of the welfare state in advanced capitalist democracies is best appreciated in terms aligning 'politics with markets', normatively and institutionally, in search of positive-sum wellbeing solutions for the knowledge economy in an ageing society, rather than as the product of mobilising 'politics against markets' in a zero-sum distributive conquest. As ever, the proof of the pudding is in the eating-namely empirical welfare research. For this, we hope, to indicate a way forward.