Introduction
Narratives about the pathways, stages and dynamics of welfare state development are important themes in social policy scholarship. For instance, descriptions about European welfare states pass through ages of emergence and take-off rooted in industrialization and urbanization, a golden age of expansion in the context of post-WWII Keynesianism, and a more recent stage of retrenchment and new social risks often associated with neoliberalism, globalization, and demographic change. Similarly, research on social policy in the global South has identified its origins in models of capitalist development emerging at critical junctures, traced its exhaustion in the context of the global debt crisis and finally pinpointed a new model of expansion triggered by democratization and new types of social policy schemes. While existing narratives have provided crucial insight into welfare state developments for various places, times and policy fields, a truly global account of social policy dynamics is missing so far. This relates not only to geographically limited coverage of the various analyses but also the extent of historical coverage and the – often overlooked – fact that the social policy areas under study are mostly conceived narrowly as protection measures for social risks of (male) workers. This is specifically relevant as we expect that what may seem like a high degree of internal coherence within a specific field of social policy among a set of countries in a specific period, may wane once more countries, longer time periods, and/or additional policy fields are considered.
Addressing these shortcomings, with the present article we offer a broader picture of social policy development and discuss our findings in light of existing notions about the role of time in comparative social policy research. In doing so, we address three research questions: Firstly, which types of temporal narratives are present in studies about international social policy development? Secondly, what are the global patterns and pathways of social policy introduction and how do they relate to existing narratives about time in social policy? Thirdly, how do social policy introduction trajectories vary between countries and regions?
To answer these questions, the article first takes stock of the rich and varied literature that has ascribed a substantive role to time and timing in welfare state development. We argue that existing studies can be distinguished into three main types: Stage models focus on the successive accumulation of institutions, which are implicitly directed towards some end-state. Timed orders uncover patterned variation in the order and timing of adoption. Periodizations emphasize self-contained episodes ideally marked by clear ruptures situated in historical time. Uncovering these different types of theorization about timing also shows that the understanding of time in social policy is a multifaceted subject necessitating a multidimensional approach. Therefore, in our subsequent empirical analysis, we draw on the concept of sequences to simultaneously incorporate three characteristics: The order of events, the timing of events relative to each other, and the occurrence of events in relation to historical time. We thus employ sequence analysis as a method to novel data for eight distinct social policies in 164 countries over 140 years (1880–2019), which, for simplification, we later group into four social policy areas: work-related protection, education mandates, health care systems, and care-related services.
While our inductive methods are primarily descriptive, our theorization of temporal narratives allows us to relate the empirically identified patterns and pathways to the three existing types present in the literature. Even though we do not test any claims directly, our systematic exploration of pathways and trajectories serves to highlight the plausibility of prevalent (substantive) propositions. Our analysis corroborates Gough and Therborn’s conclusion that “the developmental paths of European welfare states are not likely to be repeated” (2010, 711), which has implications for the politics of social policy in the present: Policy legacies condition reform possibilities while their absence may open policy space for accelerated adoption in other fields of social policy or competing institutional solutions. Thus, while all four social policy areas we incorporate seem to be of high relevance for most countries globally, the variety of consolidation trajectories hints towards the existence of different functional equivalents to classical social policy programs.
In the next section, we lay out theoretical strands implicitly or explicitly dealing with the timing and ordering of welfare state development systematizing them into three types. details the data on social policy used, develops a conceptualization of social policy areas and introduces the methods of sequence and cluster analysis. Subsequently, presents global patterns and pathways of social policy introduction, before differentiates clusters of countries’ sequence profiles and discusses the variance of national trajectories. We conclude by highlighting how the temporal theorizations are reflected in our empirical findings and outlining the article’s contributions, limitations, and questions for further research.
Stages, timed orders, and periods in research on social policy
Notions of stages, ordering, or periods are ubiquitous in the historical and comparative research on welfare states. They are implicit in the titles of seminal works (“From the Poorhouse to the Welfare-State,” “From Proles to Industrial Workers,” “From Relief to Income Maintenance” etc.Footnote 1 ) and structure introductory chapters of handbooks (Béland et al. Reference Béland, Leibfried, Morgan, Obinger and Pierson2021; Castles et al. Reference Castles, Stephan Leibfried, Obinger and Pierson2010). While often intended to serve largely descriptive purposes, presenting developments, detailing the order and timing of key legislation, and situating them in both historical and developmental time,Footnote 2 they necessarily go beyond it. The conceptualization of stages, the interpretation of order and timing, and the delineation of periods involve imposing narrative structure and imply underlying explanatory logics (Nullmeier and Kaufmann Reference Nullmeier, Kaufmann, Francis, Leibfried, Lewis, Obinger and Pierson2010). These logics represent a distinct class of explanations of social policy dynamics, which posit that order and timing are causally important in their own right, limiting the range of possible trajectories once paths are taken, facilitating further development when thresholds are crossed or generating stable institutional equilibria, which last until they are disturbed by shocks. To investigate how and why order and timing matter, we differentiate three types of relating time and welfare state developments: Stage models, timed orders, and periodizations. Whereas stage models focus on the successive accumulation of institutions, which are implicitly directed towards some end-state, timed orders uncover patterned variation in the order and timing of adoption. In turn, periodizations emphasize self-contained episodes ideally marked by clear ruptures situated in historical time. We present their focus, uncover their implicit or explicit explanatory claims, and finally review how they have been applied in comparative welfare state research, including in studies about social policy in the global South.
Stage models
Stage modelsFootnote 3 of institutional development dominated the early social sciences and understand historical development as a sequence of successive steps. Their explanatory logic aims at identifying necessary prerequisites and presupposes a succession of steps, which correspond to a hierarchy of “tasks” to be solved. They also imply a goal towards which the steps are taken. Classical modernization theories (Rostow Reference Rostow1959), Marxist historical materialism in its orthodox form (Cohen Reference Cohen1978; Marx Reference Marx1859 [Reference Marx1977]), but also accounts of late-industrialization (Gerschenkron Reference Gerschenkron1962) are canonical examples.
More sophisticated versions do away with the strict teleology and insert elements of historical contingency, such as a distinction between “easy” and “hard” stages (Hirschman Reference Hirschman1968), or allow for steps to be taken simultaneously (Rokkan Reference Rokkan, Tilly, Pye and Ardant1975). However, a notion of linkages between stages remains essential. Empirically, stage models are often presented in narrative accounts or in terms of a deductive ideal-typical “logic of development.” New methods adapted from evolutionary biology, known as “sequential requisites analysis” (Lindenfors et al. Reference Lindenfors, Krusell and Lindberg2019), allow for systematic testing of these claims (e.g. in the case of women’s rights, Wang et al. Reference Wang, Patrik Lindenfors, FREDRIK JANSSON and Lindberg2017).
In social policy research, stage models come in two variants: One of them places social policy in a broader context of institutional development, and the second delineates stages of welfare state development itself in a process of “growth to limits” (Flora Reference Flora1986). Rokkan’s model of stages of political development (Rokkan Reference Rokkan, Tilly, Pye and Ardant1975),Footnote 4 which distinguishes “state formation,” “nation building,” “participation,” and “redistribution” is notable, not only because it places the welfare state at the apex of political development, but also because it includes education in the “nation building” phase, giving a distinct and early place to this often neglected part of the welfare state. Mass education contributes to the cultural consolidation of the national polity, which succeeds the military-economic subjugation of a territory, i.e. “state formation” in the narrow sense. It precedes mass democracies, which then redefine the substantive content of citizenship of the national population. More recent research with a global reach generally confirms the sequencing of mass education and democratization (Ansell and Lindvall Reference Ansell and Lindvall2013; Paglayan Reference Paglayan2021; Reference Paglayan2022). This logic also aligns with Marshall’s (Reference Marshall1950) notion of three successive elements of citizenship rights – civil, political, and social. Both accounts posit stages, which are linked by an underlying problem or contradiction, which is both exacerbated and counteracted by each successive stage: In Rokkan’s case, the stages are tied together by the challenges of nation-state consolidation; in Marshall’s, by the need to substantiate the promised equality of status inherent in citizenship. In both accounts, the welfare state concludes this development. As an “agenc[y] of redistribution,” it consolidates the nation-state through the “equalization of economic conditions” (Rokkan Reference Rokkan, Tilly, Pye and Ardant1975, 572); as the institutionalization of social rights it completes the status of citizenship, thereby enabling the exercise of civil and political rights.
Following the demise of modernization theory, little research on social policy outside Europe has adhered to stage models in the strict sense. Investigations of larger processes of state-building, democratization, and citizenship (Møller and Skaaning Reference Møller and Skaaning2010; Therborn Reference Therborn and Torstendahl1992) have emphasized the divergence from the European experience and the general inadequacy of stage models. However, the literature on social policy in the context of development models (Haggard and Kaufman Reference Haggard and Kaufman2009; Mares and Carnes Reference Mares and Carnes2009) makes implicit claims about stage-wise expansion, which co-vary with development models: In the case of Import Substitution Industrialization the expansion of social insurance for core workers is said to take precedence over universal services, whereas the converse is assumed to hold for export-oriented development models, which expand education and health care first (Haggard and Kaufman Reference Haggard and Kaufman2009, 69; Mares and Carnes Reference Mares and Carnes2009, 102). The explanatory logic in this case is one of successive steps, which need to align with the corresponding steps of a development model. Transitions between stages are understood to originate in the new needs brought about by successive steps of the socio-economic development model, which require also embarking towards the next stage of social policy expansion.
Stage models have succumbed to massive criticism for two interrelated reasons: First, they posit uniformity of trajectories in a world filled with various “routes to/through modernity” (Therborn Reference Therborn and Torstendahl1992) and, second, their explanatory logic implies close conditional or causal linkages between institutions. But even when successively introduced institutions, such as democracy and the welfare state, are linked functionally and hence reinforcing each other, they do not need to appear in a specific order historically. While social policy neither has democracy as a necessary condition nor follows a specific order of institutional innovation internally, reaching some stages may lower the threshold of taking the next “step.” Stripped of their determinism, stage models may provide analytical leverage to explain temporal patterns of institutional development, especially if they are able to account for variation by identifying functionally equivalent stages or phenomena of “leapfrogging” by late-comers. Yet, research has turned from the largely speculative elaboration of uniform phases to the inductive discovery of varied orders of adoption.
Timed ordersFootnote 5
In many ways, timed orders of legislation are the simplest form of providing a stylized account of institutional developments. When did countries introduce policies a, b, and c? Problems of identifying the proper dates both conceptually (e.g. What exactly is compulsory schooling?) and historically (e.g. When was it achieved?) notwithstanding, it is also one of the most economic ways to compare national developments and thus especially popular when in-depth comparative research is largely in its infancy. Time enters to inform about the order and relative timing of legislation. Hence, primary analysis tools are comparisons of ordering and years of legislation in the form of tabulation (Abbott Reference Abbott1995). Cases can be situated relative to each other (e.g. pioneers vs. late-comers) and averages are used to construct typical sequences. These can in turn be interpreted to identify outliers, which warrant further investigation. Timed orders have primarily been used to chart the consolidation within a broader policy field, e.g. of social insurance or civil rights.
These analyses are largely inductive and light on theoretical assumptions: They neither imply a necessary order (their primary interest lies in uncovering variation of order) nor anticipate eventual consolidation (the lack of specific legislation or significant deviation from the typical timing produces further research). Yet two implications are present in most analyses of this kind: First, the selection of policies tacitly assumes that they are part of the same general development. Second, the identification of typical orders assumes that single policies have distinct thresholds of adoption. To illustrate these claims, let us turn to the use of timed orders in social policy research.
Pioneers of this research tradition are Flora and Alber (Reference Flora, Alber, Flora and Arnold1981) with their analysis of welfare states in Western EuropeFootnote 6 generally and unemployment insurance specifically (Alber Reference Alber, Flora and Arnold1981). The welfare state in their conception was characterized by its “historical core”: Social insurance against the (work-related) risks of injury, sickness, old age, and unemployment; and thus by the progressive adoption of policies covering a) a set of typical risks experienced by the working class, which are b) expanded to ever larger groups within a population. Both could be analyzed using the timing and order of adoption to trace the peculiar logic inherent in social insurance. Overall adoption is driven by the needs of socio-economic modernization on the one hand, as in other functionalist accounts, but also facilitated by an institutional, i.e. endogenous, mechanism of “spillover” (Elster Reference Elster2009, 11) on the other.
Having established the general drive towards consolidation, Flora and Alber interpret the typical sequence tables as providing evidence for patterned diversity, that follows “no uniform sequence,” but still shows remarkable consistency: “[S]ocial insurance for industrial accidents came first, unemployment insurance last, with the other two systems [pensions and sickness insurance] in between” (51). This result is explained by the degree to which each of the insurances embodies a “break with liberalism” (ibid.): The degree to which social risks deviate from the principle of individual responsibility explains the relative timing of introduction.
Later analyses going beyond Western Europe have largely confirmed the empirical findings of Flora and Alber, but refrained from evaluating whether this also vindicates their twofold assumption of an inherent drive towards expansion inhibited by the degree to which socializing risks would indicate a “break with liberalism.” For 18 developed countries Abbott and DeViney (Reference Abbott and DeViney1992) show that “in all but two cases worker’s compensation, unemployment insurance, and family allowances fall in that order” (249). For East Asia, Hort and Kuhnle (Reference Hort and Kuhnle2000) find that “the sequence of establishing social security for various purposes in East and South-East Asian countries is remarkably similar to the historical pattern of European countries” (166). In a global analysis, Schmitt et al. (Reference Schmitt, Lierse, Obinger and Seelkopf2015) similarly state that “this pattern holds for the first and the last scheme up to the present day” (509). Concerning the personal scope, or inclusiveness, of social insurance, both Alber (Reference Alber, Flora and Arnold1981) and Son and Böger (Reference Son and Böger2021) point out, that agricultural and atypical/domestic workers are typical late-comers in unemployment and maternity insurance coverage respectively.
Given their descriptive simplicity timed orders remain widely used in social policy research. But identifying patterns is analytically worthwhile only when it is accompanied by theorizing about potential sources of diverging thresholds of adoption.
Periodizations
Periodizations are characterized by separate episodes marked by events in historical time. These episodes are characterized as a set of (primarily: socio-economic) conditions and a set of institutions, e.g. of social policies, which provide economic stabilization as well as political legitimacy and thus tend towards expansion. Over time this constellation becomes unstable, triggers a crisis, and gives way to new conditions, which pose a new set of challenges and require a new set of institutions. Crucially, these processes are thought to operate across cases, which live through the same “age” of development concurrently in historical time. This also explains why the set of conditions includes features of the international political economy, such as monetary regimes, global norms, or human rights, and why ruptures coincide with global events such as wars and economic crises. At the same time, concurrent developments such as sectoral economic shifts, demographic change, and the reorganization of the gendered division of labor also undermine the fit between the structures of social risk and institutions. Periodizations seldom spell out whether these transformations are brought about through endogenous or exogenous processes (Blyth and Matthijs Reference Blyth and Matthijs2017), i.e. whether the crisis of a period is of its own making.
Narratives of the “golden age” of the welfare state (Nullmeier and Kaufmann Reference Nullmeier, Kaufmann, Francis, Leibfried, Lewis, Obinger and Pierson2010) or even “capitalism” more generally (Marglin and Schor Reference Marglin and Schor1990) are examples of this type of periodization. They identify a period spanning from the end of the second World War until the final demise of the Bretton Woods system following the first oil price shock in 1973, which was marked by the expansion of social policy, in terms of programs, coverage and the level of benefits. However, as many have pointed out (Nullmeier Reference Nullmeier, Béland, Stephan Leibfried, Morgan and Pierson2021; Wincott Reference Wincott2013), analysts have overemphasized the rupture with the interwar years as well as with the alleged period of “retrenchment” and “permanent austerity” following thereafter (Pierson Reference Pierson2011). Key social policies were consolidated before the beginning of the “golden age” and grew by virtue of demographic shifts rather than institutional innovation. Whatever expansionary dynamic one might diagnose took place largely within the already established branches of social security. Similarly, many social policies proved quite resistant to large-scale retrenchment even as “Keynesian welfare-states” transformed into “Schumpeterian workfare states” (Jessop Reference Jessop1993) and their mode of governance was increasingly “neoliberalized,” as subsequent crises were not ascribed to failures of neoliberal policy prescriptions but rather reaffirmed the need to pursue them with vigor (Peck and Tickell Reference Peck and Tickell2002). Overlapping with the distinction between the “golden age” and “neoliberalism,” the literature on “new social risks” (Bonoli Reference Bonoli2005; Reference Bonoli2007) focuses on the transformation of the type of risks salient for welfare state politics rather than the distinction between expansion or retrenchment of the welfare state as a whole. It also emphasizes variations in timing between cases, which are explained in terms of the interaction between institutional path dependency and sectoral dynamics.
Periodizations of social policy development in the global South are not as common, but where they occur they follow a structure similar to the two-period distinction of the “golden age” narrative (Harris and Scully Reference Harris and Scully2015; Seekings Reference Seekings, Shapiro, Swenson and Donno2008), with an additional period sometimes added to capture recent developments of large-scale social assistance programs (Leisering Reference Leisering2018). The first period is characterized by “workerist” social policy in a context of (post-colonial) state-led industrialization (Harris and Scully Reference Harris and Scully2015; Seekings Reference Seekings, Shapiro, Swenson and Donno2008, 35) leading to a pattern of “inegalitarian incorporation” of formal sector workers and ends with the debt crisis after the Volcker shock of 1979. The period afterward, often termed “neoliberal,” is associated with the privatization of pensions and other public services (Harris and Scully Reference Harris and Scully2015, 435). These periods clearly overlap with the stage models presented in the last section, but express a stronger emphasis on international factors and hence historical time. Useful periodizations are thus stylized accounts of the interplay between a set of national as well as international conditions, which in turn translate into a set of social risks and hence require the consolidation of new institutions.
Implications of stage models, timed orders, and periodizations
Table 1 summarizes the three ways of conceptualizing institutional change in time. Each of these ways focuses on a different aspect of institutionalization as well as timing: Stage models posit a specific successive accumulation of institutions, which unfolds over time. Timed orders emphasize the order of single legislative events, including the question of relative timing. Finally, periodizations are characterized by the salience (rather than the introduction) of institutions, measured through their relative weight (e.g. in terms of expenditures) and driven by an internal logic of expansion in terms of both generosity and coverage, situated in historical time. Each type also posits that (relative) timing matters causally and conversely suggests that distinct temporal patterns are brought about by distinct mechanisms: Necessary (or facilitating) prerequisites, varying thresholds of adoption, and self-reinforcing institutional configurations embedded in global contexts.
Table 1. Three ways of relating time and welfare state developments
Source: own presentation
While each of the three types yields more or less uniform expectations with regard to the patterns of social policy development by itself, their interplay in each country over historical time yields varied trajectories of change. Even if some policies are typically adopted in a specific order, historical periods may accelerate the adoption process, resulting in different trajectories depending on the profile of adoptions at the start of that period. Similarly, while some policies serve as a prerequisite to others, the time between their adoption and successive steps may depend on the historical timing of the former.
This interplay between endogenous logics of change, as posited by stage models and timed orders, and exogenous conditions of stability, as assumed by periodizations, is implicitly present in many historical country case studies: In his work on German social protection Stolleis (Reference Stolleis2011) frames his presentation in terms of a “model of stages” (152) corresponding to “developmental stages of social formations” (27), while situating these in geographically bounded periodizations such as “industrialized states of the twentieth century” (151). Conversely, Malloy (Reference Malloy1979) in his seminal work on Brazil highlights the historical starting point of state-formation in the context of what he terms “dependent patrimonialism” marked by an “emerging international capitalist economic order” (10), while conceiving subsequent development to proceed in “phases of delayed capitalist development.”
Even though the welfare regime literature largely deals with the cross-sectional distribution of differences and similarities at a specific point in time (for exceptions see: Ferragina and Filetti Reference Ferragina and Filetti2022), the underlying assumption that “countries cluster on policy because they cluster on politics” (Shalev Reference Shalev and Shalev2009) has a historical corollary: Since cross-sectional similarities in terms of policy characteristics are outcomes of politics, which not only play out in time but are themselves conditioned by critical junctures and shaped by path dependency, they should be reflected in longitudinal similarities in terms of policy adoption.
Methods and data
In our analysis we proceed in two steps: We first identify patterns of social policy adoptions across eight policies, focusing on timing and order, as well as common pathways of social policy consolidation across four areas, focusing on order. The latter already utilizes a transformation of our data on legal adoptions of policies into sequences of states, consisting of profiles of social policy area consolidations. We, secondly, apply sequence analysis methods, treating complete sequences as our unit of analysis, to probe for the variation of trajectories across all three types of temporal regularities: First of all, the order of distinct states, which is paramount to stage models as well as timed orders. Secondly, the duration spent in each state, which corresponds to issues of relative timing as well as endogenous logics of exhaustion or crisis. And finally, the timing, which speaks to an interest in how states align with historical time (Liao et al. Reference Liao, Danilo Bolano, Cornwell, Fasang, Satu Helske, Marcel Raab, Struffolino and Studer2022).Footnote 7
In the following, we first describe our choice of policies and data as well as the rationale for constructing four social policy areas. Then we briefly introduce how the consolidation profiles of the latter are used to identify global pathways and distinct trajectories using sequence analysis and cluster techniques.
Social policy adoptions
Our study focuses on the formal de jure adoption of laws addressing different social policies. The legal introduction date does not provide information on the de facto implementation or generosity of the policy. Yet, de jure adoption gives us a fairly accurate and internationally comparable measure of the time point when political actors decide to take action to address a social issue, indicating the public recognition of a social risk (Jensen Reference Jensen2011, 407). Therefore, while the measurement only provides limited information about the extent of citizens’ actual welfare, it arguably tells us a lot about the role of the state and if and in which areas political actors deem public intervention necessary and/or justified.
Previous comparative studies on social policy adoption have commonly included four to five social policy fields, namely old-age (and partially survivors and disability) pensions, sickness-related (and partially maternity) benefits, work injury benefits, unemployment benefits, and, sometimes, family allowances (e.g. Abbott and DeViney Reference Abbott and DeViney1992; Alber 1987 [Reference Alber1982]; Collier and Messick Reference Collier and Messick1975; Hicks Reference Hicks1999; Schmitt et al. Reference Schmitt, Lierse, Obinger and Seelkopf2015).Footnote 8 While these policies constitute important aspects of public welfare provisions, they are in several interrelated ways one-sided and narrow. Firstly, they are heavily focused on cash transfers, neglecting welfare services (see e.g., Alber Reference Alber1995; Jensen Reference Jensen2011). As such, these policies are, secondly, predominantly linked to paid work, covering risks associated with commodification rather than risks associated with the availability, quality, or affordability of care or educational services. Thirdly, these branches are not only focused on workers but on male workers in particular, neglecting, in a “genderblind” fashion (Orloff Reference Orloff1993), social risks predominantly faced by women. This is not only evident in that maternity is overwhelmingly not recognized as a distinct social risk for women in paid employment (Bock and Thane Reference Bock, Thane, Bock and Pat1991), but even more so in the absence of provisions protecting against the risk of needing and providing (informal, unpaid) care work (Knijn and Kremer Reference Knijn and Kremer1997; Lewis Reference Lewis1992).
To avoid this focus on policies addressing the income security of typically male wage workers and the geographic, historical as well as substantive biases that go along with it, we employ a broader conception of social policy. We analyze the introduction of eight major distinct social policies each linked to a specific risk/situation: old-age pensions, work injury benefits, paid maternity leave, unemployment benefits, compulsory education, health care systems, early childhood education and care services (ECEC), and long-term care systems for older persons (LTC). While this set of policies reaches beyond previous international comparative macro-analyses of welfare stateness, some limitations remain. In particular, although the underlying social needs and risks may arise in many societies (with market economies), the selection is mainly based on earlier discussion of social policies in capitalist Western/European welfare states. While functionally equivalent regulations and benefits as well as provision by non-state actors can be tremendously important to fully understand welfare arrangements and levels (Seelkopf and Starke Reference Seelkopf and Starke2019), we focus on policies that reflect the public recognition of social risks through the development of functionally specific institutions. As opposed to “social policy by other means” (Seelkopf and Starke Reference Seelkopf and Starke2019), which amalgamates socio-political with other goals, the included social policies impose collective norms onto spheres governed by distinct logics and power-relations.Footnote 9 Our analysis thus traces the emergence of social policy as a potential “countervailing power based on institutional differentiation” (Gough Reference Gough, Gough and Geoffrey2004, 32) operating to attenuate market dependency and familial hierarchies. Conversely, long periods of stasis of social policy adoptions may indicate limits to this transformation and hence the presence of social policy by other means.
Previous studies on social policy introduction (e.g. Abbott and DeViney Reference Abbott and DeViney1992; Collier and Messick Reference Collier and Messick1975; Schmitt et al. Reference Schmitt, Lierse, Obinger and Seelkopf2015) mostly extract adoption dates from the heading “first law” of the Social Security Programs Throughout the World (SSPTW) report series published by the United States’ Social Security Administration and the International Social Security Association. However, the SSPTW data is problematic for several reasons (see also Grünewald Reference Grünewald2021; Knutsen and Rasmussen Reference Knutsen and Rasmussen2018). Firstly, there is no possibility to validate the data because details and sources on the first law are missing. Secondly, there is no definition of which criteria a first law needs to fulfill, which might lead to inconsistencies between cases. Thirdly, for using the data, one needs to follow the delineation of the five social security branches as outlined in the SSPTW reports, which are very broad. For instance, for the area of “sickness and maternity” it is unclear if the first law refers to cash sickness benefits, health care services and/or maternity benefits, which are all addressed in the section (SSA and ISSA 2018, 8). To circumvent these issues and to integrate a broader spectrum of clear-cut social policies not covered in the SSPTW, this study employs novel data on de jure social policy adoption stemming from the Global Welfare State Information System (WeSIS), which is the outcome of a large multiyear project dealing with the “Global Dynamics of Social Policy.”Footnote 10 In particular, we rely on the following WeSIS datasets for the policy adoptions, which can also be consulted for details of respective definitions and operationalization: Grünewald (Reference Grünewald2021) for old-age pensions, Breznau and Lanver (Reference Breznau and Lanver2020) for work injury benefits, Son (Reference Son2020) for maternity leave, Obinger and Schmitt (Reference Obinger and Schmitt2022) for unemployment benefits, Windzio et al. (Reference Windzio, Besche-Truthe and Seitzer2021) for compulsory education, de Carvalho and Schmid (Reference de Carvalho and Schmid2023) for health care systems, Böger and Seitzer (Reference Böger and Seitzerforthcoming) for ECEC, and Fischer (Reference Fischerforthcoming; see also Fischer, Sternkopf, and Rothgang Reference Fischer, Sternkopf, Rothgang, Mossig and Obinger2023) for LTC. States that have been part of the USSR or Yugoslavia and were in the datasets coded as not adopting or only adopting after independence are – in consultation with the authors of the data – recorded to take the value of a “representative” country of the larger unit (i.e. Russia or Serbia) if the adoption in these territories happened before the dissolution of the larger geographical unit. If the adoption date for Russia is before the entry of the state into the larger union, the year of absorption into the USSR is coded as an adoption year. We do the same in the case of South Africa and Namibia. For other colonized areas, we follow the operationalization employed in each dataset, which, in some cases, codes adoptions based on geographic area, not having independence as a necessary requirement. Our final dataset covers 164 countries for a time span of 1880 until 2019.
Social policy areas
To analyze sequences of social policy adoption, the order in which social policies were adopted, the timing of adoption plus the time spent in each state need to be combined. Were we to do that with the eight policy fields under investigation we would need to describe a total of 256 possible combinations, rendering the resulting sequences too complex to interpret. Hence, we combine the eight social policies into four areas. To terminologically differentiate adoptions of a single policy from the time a policy area is “introduced,” we speak of consolidation of social policy areas regarding the latter (see also Hicks Reference Hicks1999). The reasoning behind this term is that the cumulative adoptions of single policies of an area indicate the institutionalization or consolidation of the policy area. The following section outlines and justifies the differentiation of the four areas whose sequences we subsequently analyze.
Classical Western welfare state research has largely focused on work-related cash benefits such as pensions or unemployment. Even though the specific rationales behind those programs vary, they commonly address the risk of incapacity to (momentarily or permanently) earn an income on the labor market (Esping-Andersen Reference Esping-Andersen1990, 21–22). Therefore, we group the four policies old-age pensions, work injury compensation, maternity leave – addressing the women-specific risk of childbirth inhibiting labor market participation (Htun and Weldon Reference Htun and Weldon2018, 169) – as well as unemployment benefits under the area of work-related protection, setting the consolidation threshold to three out of these four programs (see also Hicks Reference Hicks1999, 25).Footnote 11 While this procedure necessarily obscures certain differences, it also allows a certain range of diversity in how different countries choose to cover work-related risks. Combining four out of eight social policies this way gives less weight to the policies traditionally considered central vis-à-vis other, previously neglected, social services, allowing us to insert them into the sequence analysis on a conceptually equal footing.
The following two areas consist of only one policy each, covering education, on the one hand, and health care, on the other. Both areas can be regarded as “‘boundary issues’ of the welfare state” (for education, see similarly Busemeyer Reference Busemeyer2015, 1–2; Immergut Reference Immergut1992, 1 regarding health). On the one hand, they are part of the welfare state because of their importance for both individual social welfare and life chances as well as societal and economic development and (re)production (Busemeyer and Nikolai Reference Busemeyer, Nikolai, Béland, Stephan Leibfried, Morgan and Pierson2021, 680–81; Jensen Reference Jensen2011, 407–8). Furthermore, education and health can easily be regarded as societal (sub)systems with their own, distinct logics, professions, and infrastructures such as doctors and hospitals, teachers and schools (see, e.g. Field Reference Field1973). The distinction of work-related social policies, education, and health care is in many instances also evident in separate institutional responsibilities for each field, for instance as regards countries’ ministries and specialized UN agencies focusing on a single area.Footnote 12 Owing to this distinctiveness, education mandates (operationalized by the adoption of compulsory education) and health care systems constitute separate areas of social policy.
In contrast to education and health care which necessitate (to large extents) professional services, the production of child and long-term care services are, in Jensen’s (Reference Jensen2011, 408) words, “very plastic, because they may be produced fairly easily by professionals or by family members without any training.” Consequently, both types of care addressing different life-cycle phases are often studied jointly (e.g. Bettio and Plantenga Reference Bettio and Plantenga2004; Daly Reference Daly2002; Esquivel Reference Esquivel2017; Ochiai Reference Ochiai2009), as they come with similar tensions regarding unpaid, informal versus (semi)professional, formal care and, relatedly, reconciliation issues of (mostly) women related to care provision and paid work. Therefore, as a fourth area, we group ECEC and LTC benefits as “care-related services” and assume consolidation of this area once at least one of these two policies is adopted, indicating a public recognition of care work/needs as a social risk (Table 2).Footnote 13
We end up with a total of 16 possible profiles, each representing a distinct combination of consolidation across the four areas of social policy. Table 3 gives an overview of these profiles which are labeled according to the different areas that are consolidated and thus in combination constitute a state. The letters are in alphabetical order and thus make no indication of the order of consolidation. Lastly, to ease the interpretation and comparison of the social policy consolidation profiles, we limit our observations to every five years. This approach allows for a good balance of nuanced understanding of the sequences of social policy consolidation profiles and the necessary reduction of complexity.
Table 2. Areas of social policy
Source: Own presentation.
Table 3. Social policy consolidation profiles
Source: own presentation.
Sequence analysis and clustering
We apply sequence analysis methods to the de jure adoption data described using these consolidation profiles. Each profile represents a distinct state and, consequently, a distinct value in the sequence, which can be understood as a “time series of categorical data” (Blanchard Reference Blanchard2011). The range of possible values, the alphabet, is constructed from combinations of cumulative consolidations, following the approach pioneered by Abbott and DeViney (Reference Abbott and DeViney1992), rather than using qualitatively distinct states as is common in life-course research.
To systematically compare and cluster sequences, we use the optimal matching (OM) algorithm. OM calculates the dissimilarity between two sequences based on the least costly set of insertions/deletions (indels) and substitutions required to transform one sequence into the other. The costs for these operations are determined based on substantive knowledge. Following Abbott and DeViney (Reference Abbott and DeViney1992), we define substitution costs using a “matching” measure between the two consolidation profiles, weighing the absence of consolidation as heavily as its presence. For indels, we set the cost at one-half of the maximum substitution cost, reflecting our assumption that the order of events is slightly more important than their duration (see also Abbott and DeViney Reference Abbott and DeViney1992; Studer and Ritschard Reference Studer and Ritschard2016).
The OM algorithm generates a distance matrix that captures the dissimilarity between countries’ social policy consolidation profile sequences from 1880 to 2019. Using this matrix, we calculate clusters via the Partitioning Around Medoids (PAM) method, as described by Kaufman and Rousseeuw (Reference Kaufman and Rousseeuw2005). After evaluating both an elbow test and a silhouette test, we selected a six-cluster solution. This choice balances the objective quality of the partitioning with the need for a manageable number of clusters to facilitate meaningful interpretation and description.
Global patterns and pathways of social policy development
Patterns of social policy adoption
To give an overview of the question of timing, Figure 1 shows the absolute numbers of adoptions of single programs per decade. We can identify a multi-peak pattern similar to what Abbott and DeViney (Reference Abbott and DeViney1992) found for five programs in 18 Western countries. Globally, social policy starts with the introduction of compulsory education, which experienced peaks in adoption numbers between the First and Second World War (WWI, WWII), the 1960s and 1990s. The introduction of health care systems follows a similar pattern peaking in the 1920s and again in the 1950s to 1970s. This bimodal pattern is also repeated for maternity leave and pensions. Work-injury and unemployment policies are skewed more towards the first half of the 20th century, while the converse holds for ECEC and LTC. In addition to these large peaks, one can identify an earlier peak of LTC introduction in the 1960s as well as the continued salience of unemployment policy adoption in the 21st century. In sum, three historical periods stand out: The interwar years, which saw widespread adoption of education, health, pensions, maternity leave, and unemployment policies. The phase of post-WWII decolonization repeated the interwar experience elsewhere and with a slight shift in focus. And finally, the late 20th century, which saw the emergence of ECEC and LTC policies. These rough periodizations underline insights into the role of war (Obinger and Schmitt Reference Obinger and Schmitt2022), the founding of the ILO (1919), decolonization (Schmitt Reference Schmitt2015), and neoliberalism (Ferragina Reference Ferragina2019) as well as the emergence of new social risks.
Figure 1. Absolute number of adoptions of social policy programs by decade.
From this first overview of adoptions in historical time, we now turn to the order in which countries introduced the different programs, regardless of when they adopted them. For Figure 2 we ranked the place of adoption for all programs and counted the number of times these programs were the first, second, third, etc. to be introduced in a country.
Figure 2. Count of position social policy programs have in countries’ introduction history.
Three patterns stand out: First, the development of social policy starts with compulsory education and work injury compensation. This supports traditional interpretations of these policies: Mass education complements the emergence of nation-states through fostering a unified national culture; work injury compensation secures economic liberalism by socializing the dangers of industrialization. Second, few countries adopt all eight policies. Unemployment insurance, LTC, and ECEC are not only laggards, they are also lacking in large parts of the world. The risks of wage labor remain more legitimate targets of public intervention than the risks originating in familial obligations to provide care. Finally, pensions, health systems, and maternity leave seem to follow no specific order.
So far, our analysis was concerned with aggregate patterns in terms of timing and order. To investigate the diversity of pathways “through” welfare state development, we turn from the analysis of adoptions to that of consolidations.
Pathways of social policy consolidation
Turning to the sequence analysis of our four social policy areas, Figure 3 gives a first overview of countries’ pathways of social policy consolidation. This figure can be interpreted analogously to those representing shifts in votes from one to another party. It is intended to give us a first overview of the order of consolidations, albeit with decontextualized pathways and a sole focus on transitions rather than duration and timing.
Figure 3. Sankey plot of social policy consolidation profile development.
Even though a small set of exceptions remains, Figure 3 broadly supports the view that social policy development follows successive stages: It starts with either of two areas, education mandates (E), or health care systems (H). From these two, three pathways branch off: The first path begins with consolidating the respective other area – health or education (E_H) – and then adding work-related protection (E_H_W). The second reverses this by first adding work-related protection to education/health care (E_W or H_W) and then proceeding to consolidate the missing area (yielding E_H_W). Finally, 22 countries (13%) “leapfrog” towards having consolidated education, health care system, and work-related protection (E_H_W) simultaneously. Only afterward do 92 countries (56%) start to socialize risks associated with care-giving (C_E_H_W). However, the equifinality of education mandates and health systems begs the question of whether these actually serve as a necessary prerequisite in the strict sense or whether this suggestive pattern is only an artifact of coinciding developments driven by (common or distinct) other factors.
To shed some light on this puzzle, in the next step, we bring in historical time of transitions and the duration of states. Figure 4 plots all sequences of social policy consolidation profiles over our whole time period from 1880–2019, with every row depicting the trajectory of a single country by distinguishing social policy consolidation profiles by color.
Figure 4. Sequences of social policy consolidation profiles for single countries.
The figure suggests five things: 1) globally, social policy is a quite recent phenomenon; 2) there is a general tendency to expand social policy over the time period; 3) countries starting their social policy consolidation pathway later tend to have a quicker succession of consolidation; 4) globally, there is no time-point in which one social policy consolidation profile clearly dominates; 5) yet, looking only at those countries that had consolidated any area of social policy, the period from the 1930s to the 1980s was a rather stagnant one having consolidated education mandate, health care system, and work-related protection. This is the “golden age of the welfare state” from a bird’s eye view: A widespread but by no means global episode in the mid-20th century.
Diverse trajectories of social policy consolidation
Clusters of social policy consolidation profiles
Every country has a unique trajectory of social policy consolidation profiles. Yet, to discover (dis)similarity of countries’ sequences, in the last step, we cluster these sequences of social policy consolidation profiles.
Figure 5 displays for every cluster one sequence that best represents all sequences in the respective clusters. To describe commonalities amongst different trajectories, we focus on these representative sequences. We start with a description of the clusters before we relate them to earlier research and classifications.
Figure 5. Representative sequences for clusters of sequences of social policy consolidation profiles.
Cluster 1: Mid-century Welfare States
Until the middle of the 20th century (around 1940) countries in cluster 1 had not consolidated any social policy area. This changed by consolidating education mandate, adding health care systems in the 1950s, and work-related social policy in the 1960s. The development ends with consolidating care services in the 2000s.
Cluster 2: Latecomer Basic Service States
Countries in cluster 2 have consolidated only the basic services education mandate and health care system as of 2015. However, their social policy consolidation sequences begin with the area of health care system, not education, and they consolidate the first area only in 1975, making them by far the cluster with the latest start point.
Cluster 3: Catch-up welfare states
Cluster 3 brings together countries that started to consolidate social policy areas around 1910. They did so in quick succession and from the 1920s onwards experienced a long time of stability with the combination of education mandate, health care system, and work-related protection. Care-related services were consolidated almost in lockstep with countries in cluster 4 in 1995. All in all, the timing of transitions suggests a catch-up dynamic in which the consolidation of areas can be interpreted as a reaction to the transitions of the pioneering welfare states (cluster 5).
Cluster 4: Latecomer welfare states
Similar to cluster 5, countries in cluster 4 start into the observation period having had already consolidated education. These countries witness almost half a century of stagnation until 1940, which prompts us to assign these states the label of “latecomers.” However, once started, they leapfrog to additionally consolidate both health care system and work-related protection during one 5-year observation period. After a period of stasis, they join countries in cluster 1 in consolidating care-related services in 2000, making them fully consolidated welfare states.
Cluster 5: Pioneer welfare states
In cluster 5 we find the earliest consolidators in every area, warranting the label of “pioneers.” Countries here start with education already being consolidated in the 1880s and health care system following suit in 1885. The next consolidations happen comparably early with work-related protection being consolidated in 1915 and care-related services in 1965.
Cluster 6: Decolonizing (proto) welfare states
Countries in cluster 6 share a quite late first consolidation involving both work-related protection and health care systems in 1960, a combination that is largely absent in other states. They consolidate education only in the 1990s; again, a quite distinct trajectory. Having reached this social policy consolidation profile, they have until 2015 not consolidated care-related services. The presence of work-related protection alongside the consolidation of basic services leads us to call them decolonized (proto) welfare states.
Geography, sequences, and routes to/through modernity
Focusing on geographical distribution, Figure 6 depicts the membership of countries in the clusters described above. All in all, some clusters are well dispersed throughout the globe while we might draw quite clear borders around others. This is best seen in the clear European cluster 5 and to a lesser degree cluster 2 and 6. The latter ones are geographically limited to large parts of South-East Africa and West Africa, respectively, and countries in the southern parts of Asia. Cluster 4 is geographically more dispersed, covering countries from most continents, albeit American countries make up the majority. Cluster 3 is found in Eastern Europe and Central Asia plus some single countries in other regions such as Brazil, Saudi Arabia and Belgium. Finally, cluster 1 is difficult to pinpoint to a region combining countries all over Africa, middle America, as well as West and East Asia.
Figure 6. World map of clusters of sequences of social policy consolidation profiles.
All in all, for most parts of the world, we are unable to clearly distinguish geographically distinct sequences. There are no typical Latin-American or East-Asian pathways, despite them being featured prominently in the literature. Rather, sequences seem to cluster based on other determinants. One possible candidate to elucidate some of the geographical (mis-)alignment are Therborn’s four routes to/through political modernity (1992). Originally devised to grasp the “patterns of the emergence, and blockages of universal suffrage in the modern world” Therborn identifies four routes to/through political modernity, which differ not only according to geography and timing but emphasize the interactive nature of political development.
The only geographically distinguished cluster presents itself due to a distinctly European development of social policy corresponding to the “pioneering European route” in Therborn’s terminology, in which “modern” forms of statehood including the welfare state develop “out of, and in struggles against, indigenous traditions” (p. 126). In contrast to these classic welfare states, we observe two distinct but converging routes, which themselves are set apart by their pre-WWII consolidation behavior. The first group, crucially the Eastern European and Central Asian countries, consolidates a relatively comprehensive profile of social policy areas already in the 1920s, the second group, including Japan, India, and China as well as large parts of the Americas and Oceania, consolidates later but in rapid succession. Both complete their sequences of social policy consolidation with the introduction of care services in the early 2000s. Taken together these two clusters overlap with two of Therborn’s routes: The “New Worlds of settler states” and the “externally induced reactive modernization.” However, due to the historical differences in take-off, these routes do not each align perfectly with a single cluster: The Eastern European and Central Asian countries, Japan, China, and Turkey align clearly with the route of “reactive modernization”; “polities which managed to survive into the modern world by adopting modern state institutions from abroad and grafting them on to indigenous rule and indigenous society.” The Americas and the Antipodean offshoots fall squarely into the category of “settler states,” even though settlement occurred at varying historical points in time. India and the Philippines are outliers with long histories of colonization in the 20th century.
The remaining states inhabit large parts of the “colonial zone,” in which “appropriation of modernity by the colonized was […] a complicated appropriation of institutions and thought of the conquerors.” Given that they are distributed across three clusters, the impact of colonialism on social policy is variegated: They share the absence of care services but differ in terms of the completeness, pace, and timing of consolidation with regard to the three other areas. Whatever colonial and pre-colonial legacies these countries may have had, they consolidated whole areas only in the period of decolonization but did so relatively rapidly. Yet, these consolidations do not seem to mark the beginning of a catch-up process but rather lead to stagnation. While, of course, speculative due to the limited timeframe of observation, these trajectories may end in a “basic services”-trap, in which limited provision diminishes public demand for expansion and vice versa, especially in contexts lacking political stability or democratic accountability. However, the restriction of social policy to basic services may also indicate the presence of functional equivalents and provision through non-state actors.
Conclusion
In this article, we have examined global social policy developments in a broad perspective and discussed them against the backdrop of existing types of social policy narratives. Systematizing the explicit and implicit notions of time and timing in comparative welfare research, we distinguish three types of temporal narratives – stage models, timed orders, and periodizations. Our analysis has given support for the existence of the sequential patterns suggested by each of these temporal theorizations. Firstly, we find that the global patterns of policy consolidation follow an – albeit not universal – typical stage model, which proceeds from education mandates and health care systems over work-related protections to care services. Secondly, analyzing social policy adoptions as a timed order has corroborated and widened classical accounts. Our extended analysis with novel data adds the break with nation-building to the break with liberalism, which has been emphasized in earlier work on European welfare state development. Thirdly, periodizations are tempting when looking at individual adoptions as well as clusters of sequences. Two stand out, which have so far not featured prominently in comparative welfare state research: The interwar years (1919–1929) as the first global age of social policy, legitimized by the newly founded ILO and interacting with projects of state-led industrialization in the non-colonized world. The period of decolonization (1949–1969) was a second truly global age of social policy, legitimized in terms of national development and dominated by the newly independent countries. Finally, the variation we find in all dimensions partly overlaps with Therborn’s “four routes”-heuristic, supporting the idea that the dynamics of social policy are contingent on how and when countries embark on their pathway of political and socio-economic modernization.
Yet, even if our analysis could identify striking patterns this does by itself not confirm that these are produced by the proposed mechanisms: Each of our findings poses corresponding questions, which could be examined in further research: Why are education mandates and health care systems interchangeable prerequisites for the consolidation of work-related protection? Or is a functionalist explanation hiding behind a sequential appearance? Do policies differ systematically in terms of their legitimacy or their beneficiaries in terms of their deservingness and hence entail diverging thresholds of adoption? Or are countries just moving along a uniform socio-economic development path producing different needs? Finally, why are some historical periods so conducive to consolidation and why do some countries fall into stagnation afterwards?
To answer these questions, one has to move beyond the descriptive ambitions of this paper and explain the individual sequences in terms of global as well as national factors. Such an explanatory account could also move beyond functionalist accounts and incorporate different theoretical strands including actor-centered factors. Especially the variation within the “colonial zone,” i.e. the three clusters that do not consolidate all areas of social policy, needs not only causal but also conceptual elaboration. While our paper is pioneering in the breadth of social policies included, our selection of social policies is still very much indebted to Europe’s 20th century experience. Not only are some policies that mark the historical origins of European social policy, labor legislation, and poor relief (later: social assistance), lacking; we also do not consider some functional equivalents which play a large role in societies dominated by agriculture, such as agrarian price controls and forbearance. At the same time, our choice to analyze only de jure adoptions and consolidations rules out tracing any shifts in salience and generosity between and within consolidated areas of social policy, biasing our description towards a diagnosis of (a lack of) convergence. We can show a persistent divergence of social policy in the post-colonial world, especially in Sub-Saharan Africa, which reflects the divergence across socioeconomic and political dimensions. Our findings, despite opening up meaningful new avenues for research, thus point towards a sobering reality (and future) for post-colonial states and their citizens.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0143814X25100676.
Data availability statement
Replication materials are available in the Harvard Dataverse at https://doi.org/10.7910/DVN/ICP1IS (Besche-Truthe Reference Besche-Truthe2025)
Acknowledgements
The authors would like to thank the participants of the Political Economy Workshop at the University of Bremen for their valuable feedback, in particular Frank Nullmeier.
Funding statement
This article is a product of the research conducted in the Collaborative Research Center 1342 “Global Dynamics of Social Policy” at the University of Bremen. The center is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – project number 374666841 – SFB 1342.
Competing interests
The authors declare none.
Open access
