Gosta Esping-Andersen's (Reference Esping-Andersen1990) The Three Worlds of Welfare Capitalism has become one of the most cited works in social policy (over 20,600 Google Scholar citations; 20 October 2014). This path-breaking work, with its identification of three distinct forms of welfare capitalism in high income countries, has become the basis for a whole academic industry described as the Welfare Modelling Business (Abrahamson Reference Abrahamson1999; Powell and Barrientos Reference Powell and Barrientos2011). According to Headey et al. (Reference Headey, Goodin, Muffels and Dirven1997: 332), it has become a canon in comparative social policy against which any subsequent work must situate itself. Abrahamson (Reference Abrahamson1999) notes that, since the publication of the book, every welfare state scholar has referred to Esping-Andersen's tripolar scheme. Scruggs and Allen (Reference Scruggs and Allan2006: 55, 69) remark that it ‘is difficult to find an article comparing welfare states in advanced democratic countries (or a syllabus on social policy) that does not refer to this seminal work’, and ‘it is hard to overstate the significance of the impact of The Three Worlds of Welfare Capitalism (TWWC) on comparative studies of the welfare state’. Its seminal status is evidenced by the extent to which it continues to be cited in articles on comparative welfare states. It also remains required reading for most (graduate) students of comparative political economy and social policy (Scruggs and Allen, Reference Scruggs and Allan2008). Kröger (Reference Kröger2011) claims that, with few exceptions, comparative social policy research is shaped by welfare regime analysis. Arts and Gelissen conclude that TWWC is a defining influence upon the whole field of comparative welfare state research (Reference Arts, Gelissen, Castles, Liebfried and Lewis2010: 569). Danforth (Reference Danforth2014) writes that the ‘three worlds’ typology has become one of the principal heuristics for examining modern welfare states. In short, TWWC is a ‘modern classic’ (Arts and Gelissen, Reference Arts and Gelissen2002).

However, TWWC has been subject to extensive criticism and debate (for example, Abrahamson, Reference Abrahamson1999; Arts and Gelissen, Reference Arts and Gelissen2002, Reference Arts, Gelissen, Castles, Liebfried and Lewis2010; Bambra, Reference Bambra2006a; Powell and Barrientos, Reference Powell and Barrientos2011). Based on eighteen OECD countries in 1980, Esping-Andersen identified ‘three worlds’ of conservative, liberal and social democratic welfare regimes. Much of the subsequent discussion has focused on the number and composition of the worlds (for example, Arts and Gelissen, Reference Arts and Gelissen2002, Reference Arts, Gelissen, Castles, Liebfried and Lewis2010; Ferragina and Seeleib-Kaiser, Reference Ferragina and Seeleib-Kaiser2011; Powell and Barrientos, Reference Powell and Barrientos2011), but studies that ‘count worlds’ and ‘catalogue countries’ may pay insufficient attention to more fundamental conceptual and methodological issues (see Powell and Barrientos, Reference Powell and Barrientos2011). In fact, the more far-reaching criticism of TWWC has not been focused on the number of worlds, but instead on the limited attention paid to gender and families in the production of welfare (Orloff, Reference Orloff1993, Reference Orloff2009; Bambra, Reference Bambra2006b).

Moreover, while the ‘three worlds’ represent Esping-Andersen's most visible contribution to social policy, TWWC's conceptual and methodological innovations are also significant. They include broadening the focus of social policy beyond state forms and institutions and linking these to political institutions and actors; and developing evaluative measures to assess the outcomes of welfare institutions in terms of decommodification (and later, in response to criticisms, of defamilialism, Esping-Andersen, Reference Esping-Andersen1999). It can be argued that a quarter of a century after the publication of TWWC, a return to its main arguments and contributions would appear to be redundant, especially given the volume of literature it has provoked. However, it can be countered that a focus on ‘three worlds’ (i.e. debating three, four or five worlds and their composition) in the welfare modelling business might have diverted attention from the broader contribution of TWWC. Together, the articles in this themed section make a case for another look at the original text. Our approach to the task is largely ‘mainstream’ (Orloff, Reference Orloff1993, Reference Orloff2009; Bambra, Reference Bambra2006b) in taking our point of departure from TWWC. The articles in this themed section give full consideration to the large literature on the ‘three worlds’, but from the perspective of the conceptual and empirical project in TWWC. This approach inevitably has some gaps in coverage, especially the gender dimension which is only developed fully after TWWC, but, we would claim, it has the advantage of rescuing insights from it which might have been sidelined.

The following summarises the main contents of the themed section. In ‘A Re-Specification of the Welfare State’, Martin Powell discusses the main conceptual developments in TWWC. In particular, the conceptual framework underlying the ‘re-specification of the welfare state’ (Esping-Andersen, Reference Esping-Andersen1990: 23) was not fully defined in Esping-Andersen's (Reference Esping-Andersen1990, Reference Esping-Andersen1999) contributions or in many subsequent contributions (Powell and Barrientos, Reference Powell and Barrientos2011). The original 1990 account was said to be based on the theoretical foundations laid down by Karl Polanyi, T. H. Marshall and Richard Titmuss, and on the criteria of de-commodification, social stratification and the neglected (as admitted by Esping-Andersen, Reference Esping-Andersen1999) welfare mix. Esping Andersen's (Reference Esping-Andersen1999) revision sought to develop the conceptual framework by stressing social risks, de-familisation and the welfare mix, but most empirical studies remained focused on de-commodification, with little attention being paid to stratification, the welfare mix or social risks (Powell and Barrientos, Reference Powell and Barrientos2004, Reference Powell and Barrientos2011; Scruggs and Allan, Reference Scruggs and Allan2008; Ferragina et al., Reference Ferragina, Seeleib-Kaiser and Tomlinson2013). Powell critically discusses the theoretical underpinnings of Esping-Andersen's original 1990 account, and in particular to what extent the work of Polanyi, Marshall and Titmuss fits with Esping-Andersen's approach. He then explores the conceptual critique of Esping-Andersen which led to the 1999 revision, based on rather different theoretical underpinnings. He concludes that many studies tend to focus on issues of statistical reliability while neglecting conceptual validity and the underlying theoretical base.

In ‘A Veritable Mountain of Data and Years of Endless Statistical Manipulation’, Armando Barrientos turns to methodological issues. He begins by reviewing the methods deployed in TWWC to analyse the main features of welfare institutions, before assessing their influence on the practice of subsequent welfare regime researchers. Considering the influence of the methodological contribution of TWWC on the literature which followed, he finds it has been limited relative to other areas. Arguably, the analytical tools of index construction and cross-section regression analysis favoured by Esping-Andersen have been ‘crowded out’ by a variety of data reduction techniques, while qualitative comparative analysis has shifted the focus of welfare regime analysis away from latent variable oriented research and towards case oriented research. The paper concludes that developing a general theory of welfare institutions will require renewed attention to be paid to appropriate analytical methods.

Beyond these overarching conceptual and methodological issues, it can be claimed that the dominant focus on ‘counting worlds’ and ‘cataloguing countries’ has been too narrow in three important dimensions: sectors, time and countries.

First, TWWC focused too narrowly on cash benefits and excluded services (for example, Jensen, Reference Jensen2008; Stoy, Reference Stoy2014). While Esping-Andersen (Reference Esping-Andersen1990, Reference Esping-Andersen1999) is correct to stress holism and criticise regimes based on individual benefits or services, focusing on certain parts of welfare regimes while ignoring others is less defensible. In ‘More than just welfare transfers’, Volquart Stoy and Daniel Buhr critically explore the possibilities open to a broader approach to welfare regimes which includes the economy. In this context, the links between welfare regimes analysis and varieties of capitalism (VoC) become important. Liberal market economies (LMEs) with a focus on coordination by the market, and coordinated market economies (CMEs) with network coordination (Hall and Soskice, Reference Hall and Soskice2001), can be mapped together with welfare regimes. This approach offers an opportunity to bring TWWC nearer to its initial ambition of providing a holistic typology of welfare capitalisms. They distinguish between a functionalist and a political explanation, which both provide plausible arguments for the existence of holistic welfare regimes, and conclude by suggesting that linking the different spheres of the welfare regime and the economy appears to be a promising research avenue.

Second, welfare regime typology analyses are often rather static (for example, Ferragina and Seeleib-Kaiser, Reference Ferragina and Seeleib-Kaiser2011; Ferragina et al., Reference Ferragina, Seeleib-Kaiser and Tomlinson2013; Danforth, Reference Danforth2014). Esping-Andersen's (Reference Esping-Andersen1990) original analysis related to data from around 1980, and most subsequent studies are also limited to a snapshot of a certain point in time. Esping-Andersen (Reference Esping-Andersen1999: 73) himself recognised that any typology of welfare regimes remains valid only as long as history stands still. In ‘The Four Worlds of Welfare reality’, Emanuele Ferragina, Martin Seeleib-Kaiser and Thees Spreckelsen offer two main advances. First, they move away from the old institutionalist base of the legal entitlements for the average production worker (APW) (see Ferragina and Seeleib-Kaiser, Reference Ferragina and Seeleib-Kaiser2011) towards an analysis of welfare regimes ordered on ‘social outcomes’ for different societal groups. Second, and largely inspired by Goodin's (Reference Goodin1988) reasons for welfare – autonomy, social stability and poverty alleviation – they select eleven outcome indicators capturing both old risks and new social risks. They test the stability of a four-fold typology across fourteen EU countries during the period 2005 to 2012. They find that, on the whole, clustering based on outcome indicators does provide a similar picture to previous welfare typologies, with the important exceptions of the classification of Germany and Ireland. In particular, the most striking finding for new social risks is that the clustering replicates very closely the theoretical four-cluster typology, but with a divide into two groups: the social-democratic and liberal countries (characterised by higher youth education and female employment), and the conservative and Mediterranean countries, with opposite characteristics.

Third, the original three worlds consisted largely of eighteen ‘advanced Western’ OECD nations. Ebbinghaus (Reference Ebbinghaus2012) reports a meta-analysis of eleven follow-up studies which range from eleven to twenty-five countries, although most focus on Esping-Andersen's sixteen to eighteen OECD countries. Critics point to the conceptual problem of a Western, Eurocentric or ‘Swedocentric’ approach, and the methodological problem of a selective and biased sample. In ‘From Worlds to Cases’, Ki-tae Kim points out that the case selection issue has long been neglected in the cross-national welfare regime studies despite its importance in securing validity and reliability. He reviews the case selection practices of thirty-three studies that produced their own welfare regime typology. He classifies them into five groups: Esping-Andersen's (Reference Esping-Andersen1990) original eighteen nations; more than eighteen nations selected from within the OECD; non-OECD nations based on regions; non-OECD nations not constrained by regions; and all nations, including both OECD and non-OECD) nations. He finds that seventeen of the thirty-three studies focus on Esping-Andersen's (Reference Esping-Andersen1990) original group of eighteen nations. The addition of Southern European nations (an ‘18+3’ group) accounts for over two thirds of the studies, leaving only seven wider studies, He finds that few studies have discussed and justified their case selection, identifying a tension between ‘isolationists’ and ‘expansionists’. While the first group (approaches 1 and 2) are sceptical or cautious of broader application, viewing welfare regime concepts as basically rooted in the Western tradition, the second group (approaches 3 to 5) are more open or willing to apply the West-based theoretical frames away from their traditional home turf. This divide requires significantly more attention to the matter of exporting the welfare modelling business to ‘other worlds’. Put another way, is theory based on Polanyi, Marshall and Titmuss an appropriate foundation for the building of welfare regimes in areas such as East Asia; is de-commodification a universal or Western construct? We do not offer a ‘Review’ article here, as all of the articles review the relevant literature, but Monica Pinilla-Roncancio provides ‘Some Useful Sources’ which includes some lesser known sources, and relevant databases.

Esping-Andersen (Reference Esping-Andersen1999: 94) acknowledged that his 1999 verdict is ‘certainly not the last word on the subject’ and ‘the question of how to identify and classify welfare regimes will remain open because . . . researchers differ in terms of what attributes they consider vital and of how to measure them’. The articles in this themed section have explored ways in which TWWC remains open. Whilst much of the focus of the literature remains fixed on ‘counting worlds’ and ‘cataloguing countries’, our list raises a different set of issues. First, insufficient attention has been paid to engaging and developing the conceptual basis of the welfare regime approach. This involves thinking through the extent to which de-commodification remains the primary criteria for assessing welfare institutions, and reconsidering dominant forms of stratification in the current setting. Second, there remains an absence of consensus over appropriate methodologies to study welfare regimes, an issue which is intrinsically connected to the previous point. It relates to the dilemma of whether welfare regimes are primarily variable-oriented latent ideal type configurations or whether they are primarily case-oriented country configurations. Beyond these overarching conceptual and methodological issues, it can be claimed that the welfare regime framework remains unduly narrow in three important dimensions: sectors, time and countries. Third, as regards sectors, welfare regime analysis remains largely focused on cash benefits, with limited development of conceptual and measurement issues relating to services. Fourth, it is also unclear how regimes evolve and change over time, and the relative strength of path-dependent and adaptation to exogenous factors. Finally, it is unclear how the regime concept or particular regimes can be extended from the original core nations to the other worlds. In fact, TWWC developed in response to the failure of linear explanations of welfare institutions. A general theory of welfare institutions can only emerge from global comparative research. This is the biggest challenge facing welfare regime analysis, especially as regards appropriate methodologies. In short, while the welfare modelling business has been very successful over the last twenty-five years, there are still significant gaps in the market that remain to be filled.