Received: 20 December 2018 Revised: 3 April 2019 Accepted: 9 April 2019

DOI: 10.1111/spol.12510

ORIGINAL ARTICLE

Thirty years of the Three Worlds of Welfare

Capitalism: A review of reviews

Martin Powell1 | Erdem Yörük2,3 | Ali Bargu3

1
Health Services Management Centre,
University of Birmingham, Birmingham, UK Abstract
2
Department of Sociology, Koç University, In the 30 or so years since the publication of Gosta
Istanbul, Turkey
Esping‐Andersen's Three Worlds of Welfare Capitalism a
3
Department of Social Policy, University of
Oxford, Oxford, UK
number of rival welfare state typologies have emerged.
This article has two broad aims. First, we review the
Correspondence
Martin Powell, Health Services Management
reviews of welfare state typologies, pointing to issues of
Centre, University of Birmingham, Birmingham often unclear case selection and a wide range of concepts,
B15 2RT, UK.
Email: m.powell@bham.ac.uk
variables, and methods, resulting in a variety of worlds of
welfare and their constituent nations. We show that there
is a great variety in the welfare modelling business at two
different levels. Reviews vary significantly in terms of the
number and composition of included studies, which has
made it difficult to sum up the “state of the art.” Individual
studies included in the reviews also vary significantly in
terms of issues such as aims, concepts, variables, and
methods. Second, we produce a new review, which adds
value as it is based on a clearer search strategy, and
includes more recent material that was not available in
earlier reviews. This finds that there is a great variety in
terms of process (concepts, variables, methods, and
number of countries) and findings (the number and com-
position of “worlds”). We argue that the country classifica-
tion seems to show less consensus that previous reviews,
with fewer “pure” nations (i.e., agreement between
studies). We suggest that in order to provide a clear point
of engagement, future reviews need to pay more attention
to a clear and explicit search strategy, including issues
such as inclusion criteria.

Soc Policy Admin. 2019;1–28. wileyonlinelibrary.com/journal/spol © 2019 John Wiley & Sons Ltd 1
2 POWELL ET AL.

KEYWORDS

Esping‐Andersen, review, review of reviews, typologies, welfare

regimes

1 | I N T R O D U CT I O N

Gosta Esping‐Andersen's Three Worlds of Welfare Capitalism (Esping‐Andersen, 1990) typology of welfare states has
been termed a “modern classic” (Emmenegger, Kvist, Marx, & Petersen, 2015). This path‐breaking work has become
the basis for a whole academic industry described as the “Welfare modelling business” (Abrahamson, 1999; Powell &
Barrientos, 2011, 2015). Abrahamson (1999) noted that, since the publication of the book, every welfare state
scholar has referred to Esping‐Andersen's tripolar scheme. According to Headey et al. (1997, p. 332), it has become
a canon in comparative social policy against which any subsequent work must situate itself. Scruggs and Allan (2006,
p. 55) stated that it is difficult to find an article comparing welfare states in advanced democratic countries (or a syl-
labus on social policy) that does not refer to this seminal work. Danforth (2014) claimed that the “three worlds” typol-
ogy has become one of the principal heuristics for examining modern welfare states.
It still has great influence on the discipline nearly 30 years after its publication, as shown by two 25 year anniver-
sary special editions of two journals. Powell and Barrientos (2015) noted that it has become one of the most cited
works in social policy (over 20,600 Google Scholar citations in 2014; now 33,805 in 2019). According to
Emmenegger et al. (2015), TWWC had an immediate impact on comparative welfare state research, and its status
has only grown since. For example, it reached a “breath‐taking” 1,600 citations in Google Scholar in 2013 alone. They
continued that it was a standard reference in virtually all social science disciplines, whereas its influence on compar-
ative welfare state research can hardly be overstated. For example, in recent years, almost 50% of all articles pub-
lished in Journal of European Social Policy have referred to the book.
However, in the nearly 30 years since publication, it has been subject to extensive debate and critique (eg Arts &
Gelissen, 2002, 2010; Arcanjo, 2006; Bambra, 2007b; Powell & Barrientos, 2011). Bambra (2007b) argued that as a
result of these criticisms, a number of rival welfare state typologies have emerged, each based on different classifi-
cation criteria, and each trying to capture in its own way what a welfare state actually does.
This article has two broad aims. First, we review the reviews of welfare state typologies based on Esping‐
Andersen (1990). A review of reviews is not common in social policy, but it points to issues of often unclear case
selection and a wide range of concepts, variables, and methods, resulting in a variety of worlds of welfare and their
constituent nations. Second, we produce a new review, which adds value as it is based on a clearer search strategy,
and includes more recent material that was not available in earlier reviews. Turning to the structure of the article,
after briefly outline the main issues of the welfare modelling business, we review the existing reviews. Then we move
to producing an updated review and discuss the results and implications.

1.1 | Typologies

More typologies have emerged since Bambra (2007b) stated that rival typologies have resulted in the welfare state
modelling literature being in a state of confusion. A number of scholars have reviewed these typologies (Arcanjo,
2006; Arts & Gelissen, 2002, 2010; Bambra, 2007b; Ferragina and Seeleib‐Kaiser (2011); Ebbinghaus, 2012; Kim,
2015), with two broad aims of examining the content of the classification (number and composition of worlds) and
discussing the overall quality of typologies.
Arts and Gelissen (2002) reconstructed several typologies of welfare states in order to establish, first, whether
real welfare states are quite similar to others or whether they are rather unique specimens; and, second, whether
there are three ideal‐typical worlds of welfare capitalism or more. They selected six classifications, which they “think
POWELL ET AL. 3

draw attention to interesting characteristics of welfare states not directly included in Esping‐Andersen's classifica-
tion” from the “vast array of welfare state typologies”: Esping‐Andersen (1990), Leibfried (1992), Castles and Mitchell
(1993), Siaroff (1994), Ferrera (1996), Bonoli (1997), and Korpi and Palme (1998) their Table 1; We assume that this
means six classifications in addition to Esping‐Andersen, 1990). Later, they included a further five studies in a table
titled “Empirical robustness of the three‐worlds typology”: Kangas (1994), Ragin (1994), Shalev (1996), Obinger and
Wagschal (1998), and Wildeboer Schut et al. (2001). They argued that even when different indicators are used to
classify welfare states, some countries emerge as standard examples, approximating certain ideal‐types: Liberal
(USA), Continental/Conservative (Germany), and Social‐Democratic (Sweden). However, consensus seemed to end
here. For example, according to some, Italy can best be assigned to the second, corporatist/continental/conservative
type, but belongs, according to others, along with Greece, Spain, and Portugal to a distinctive Mediterranean type.
The same holds for Australia, which may either be classified as liberal or is the prototype of a separate, radical welfare
state. Hybrid cases, such as the Netherlands and Switzerland are a bigger problem. They concluded that Esping‐
Andersen's typology “neither passes the empirical tests with flying colours, not dismally fails them” (p. 153).
Arcanjo (2006) reviewed 13 empirical studies carried out between 1994 and 2005. Her selection is based on two
criteria: the attempt to develop a constructive response to the three critiques of the “three worlds”; the diversity of
conceptual and methodological options. This diversity (her table 10) required great prudence when comparing the
results obtained, due in particular to the use of data relating to a time‐span of more than 20 years, during which
period, the European systems of social protection underwent many changes. She noted that the typologies analysed
are based on different conceptual and methodological options. This justified the a priori comparison of the analytical
focus, indicators, countries selected, and the time period of observation (her table 8). She stated that for the 12 coun-
tries selected by all of the authors, only six countries are given the same classification: Ireland and the United King-
dom, France and Germany, and Norway and Sweden. She continued that when a wider analytical criterion (i.e., all 22
countries) is adopted, 14 countries have the same classification (Ireland, the United Kingdom, Canada, New Zealand,
and the United States; France, Germany, Luxembourg, and Japan; Norway and Sweden; and Greece, Portugal, and
Spain). She made a number of concluding points. First, all of the studies confirmed the existence of at least three
regimes, but in total, six types were identified (the three worlds of Esping‐Andersen, together with Radical, Southern,
and Eastern European). Second, the existence of a specific regime for the four countries of Southern Europe was sup-
ported in three studies. Third, the classification of Eastern European countries was somewhat inconclusive. Lastly, the
United States, Germany, and Sweden can be identified, respectively, as prototypes of the liberal, conservative, and
social‐democratic regimes.
Bambra (2007b) aimed to compare classifications and determine which are currently of the most utility: to “sift
the wheat from the chaff” in terms of welfare state regime theory. She selected 12 studies based on income main-
tenance, not services such as health or education: six welfare state typologies (her table 1), and six welfare state tax-
onomies (her table 2). Bambra drew on Bonoli's (1997) argument that the welfare state typology literature can be
divided into two halves: one of these examines “how much” (i.e., the quantity of welfare provision), whereas the other
examines “how” (the Bismarck–Beveridge funding dichotomy). She argued that it is possible to determine which, if
any, of the competing theories of welfare state regimes are the most useful in terms of accounting for welfare state
variation in one‐dimensional (how much or how) and/or a two‐dimensional manner (how much and how). She used
analysis of variance (ANOVA), multivariate analysis of variance (MANOVA), and discriminant analysis (DA) to test the
different welfare state typologies. She concluded that the four typologies of Leibfried (1992), Castles and Mitchell
(1992), Ferrera (1996), and Bonoli (1997), and the two taxonomies of Kangas (1994) and Obinger and Wagschal
(1998) emerge as the “wheat” with Ferrera's (1996) typology accounts for the most variance and is therefore the
most useful of the competing welfare state classifications.
Arts and Gelissen (2010) presented 11 studies examining the “empirical robustness of welfare state models.”
These studies tested the goodness‐of‐fit of the welfare state models using statistically more sophisticated and robust
methodologies, with some also generating new data sets to inquire into the robustness of the models over time.
However, they did not present any inclusion criteria. This is important, not least as some subsequent reviews such
 A comparison of studies included in review articles

4
TABLE 1

Arts and Arts and Ferragina and

Gelissen Arcanjo Bambra Gelissen Seeleib‐Kaiser Ebbinghaus
(2002) (2006) (2007) (2010) (2011) (2012) Kim (2015) N

Bambra (2004) Bambra (2004) 2

Bambra (2005a) 1
Bambra (2005b) Bambra (2005b) Bambra (2005b) 3
Bambra (2006) Bambra (2006) Bambra (2006) Bambra (2006) 4
Bonoli (1997) Bonoli (1997) Bonoli (1997) Bonoli (1997) 4
Castles and Castles and Castles and Castles and 4
Mitchell (1992) Mitchell (1992) Mitchell (1992) Mitchell (1992)
Castles and Castles and Castles and Castles and 4
Obinger (2008) Obinger (2008) Obinger (2008) Obinger, (2008)
Esping‐Andersen Esping‐Andersen (1990) Esping‐Andersen Esping‐Andersen Esping‐Andersen 5
(1990) (1990) (1990) (1990)
Esping‐Andersen 1
(1999)
Ferragina, 1
Seeleib‐Kaiser,
and Tomlinson
(2012)
Ferreira and 1
Figueiredo (2005)
Ferrera (1996) Ferrera (1996) Ferrera (1996) 3
Gal (2004) Gal (2004) 2
Gallie and Gallie and 2
Paugam, (2000) Paugam, (2000)
Goodin (2001) Goodin (2001) 1

POWELL
Hudson and 1
Kuhner, (2009)

ET AL.
(Continues)
 POWELL
TABLE 1 (Continued)

Arts and Arts and Ferragina and

ET AL.
Gelissen Arcanjo Bambra Gelissen Seeleib‐Kaiser Ebbinghaus
(2002) (2006) (2007) (2010) (2011) (2012) Kim (2015) N
Kangas (1994) Kangas (1994) Kangas (1994) Kangas (1994) Kangas (1994) Kangas (1994) 6
Kautto (2002) 1
Korpi and Korpi and Korpi and Korpi and 4
Palme (1998) Palme (1998) Palme (1998) Palme (1998)
Leibfried (1992) Leibfried (1992) Leibfried (1992) 3
MacMenamim 1
(2003)
Obinger and Obinger and Obinger and Obinger and Obinger and Obinger and Obinger and 7
Wagschal (1998) Wagschal (1998) Wagschal Wagschal Wagschal (1998) Wagschal Wagschal
(1998) (1998) (1998) (1998)
Obinger and Obinger and 2
Wagschal Wagschal (2001)
(2001)
Pitruzzello (1999) 1
Powell and Powell and Powell and Powell and Powell and 5
Barrientos (2004) Barrientos Barrientos Barrientos Barrientos
(2004) (2004) (2004) (2004)
Ragin (1994) Ragin (1994) Ragin (1994) Ragin (1994) Ragin (1994) Ragin (1994) Ragin (1994) 7
Saint‐Arnaud Saint‐Arnaud Saint‐Arnaud Saint‐Arnaud 4
and Bernard and Bernard and Bernard and Bernard
(2003) (2003) (2003) (2003)
Schroder (2009) Schroder (2008) Schroder (2009) Schroder (2009) 4
Scruggs and Allan Scruggs and Scruggs and Scruggs and 4
(2006) Allan (2006) Allan (2006) Allan (2006)
Scruggs and 1
Allan (2008)

(Continues)

5
 6
TABLE 1 (Continued)

Arts and Arts and Ferragina and

Gelissen Arcanjo Bambra Gelissen Seeleib‐Kaiser Ebbinghaus
(2002) (2006) (2007) (2010) (2011) (2012) Kim (2015) N
Siaroff (1994) Siaroff (1994) 2
Shalev (1996) Shalev (1996) Shalev (1996) Shalev (1996) Shalev (1996) Shalev (1996) 6
Soede Soede (2004) 2
(2004)
Talme (2013) 1
Wildeboer Schut, Wildeboer Wildeboer Wildeboer 4
Vrooman, Schut et al. Schut et al. Schut
and de Beer (2001) (2001) (2001) et al. (2001)
Vrooman (2009) Vrooman (2009) Vrooman 3
(2009)
Vrooman (2012) 1
12 13 12 11 23 13 24

POWELL
ET AL.
POWELL ET AL. 7

as Ebbinghaus (2012) and Kim (2015) take this selection to be the starting points for their reviews. Arts and Gelissen
(2010) concluded that the empirical evidence pertaining to Esping‐Andersen's original three worlds typology is mixed.
First, they argued that his typology has at least some heuristic and descriptive value, but a case can be made for
extending the number of welfare state regimes perhaps to four or five, with a Mediterranean grouping being the most
consistently suggested extension. Second, some cases came close to Esping‐Andersen's three ideal type regimes,
there are no pure cases, with some clear hybrids. Third, typologies based on social programmes beyond the original
case of cash benefits did not conform so easily to Esping‐Andersen's typology.
Ferragina and Seeleib‐Kaiser (2011) reviewed 23 quantitative studies analysing welfare regimes. They identified
“pure countries” (classified more than 80% of the time in the same regime type) of Sweden, Norway, and Denmark
(social‐democratic); France, Germany, and Austria (Christian‐democratic), and the United States (liberal). They
found that only Switzerland and the Netherlands were not classified in the same regime type in more than
50% of the studies, with both countries being hybrids, classified in the literature in four distinct “worlds.” Their
aim was to bring all these studies together in order to verify whether Esping‐Andersen's typologies constitute a
good guidance for comparative social policy analysis. However, comparing these studies is fairly complicated for
two reasons: first, typologies were based on different policy domains/indicators; and second, typologies were
based on data collected at different points in time. Their final classification was based on all 23 studies because
the purity of Esping‐ Andersen's ideal types should be verified by considering all sorts of indicators and policy
domains. However, when discussing the “divergent typologies” from Esping‐ Andersen's measurement, they
analysed only studies based on cash transfers and those that included a mix of indicators (including indicators
for social transfers) because they relate closely to the concepts of “decommodification,” “social stratification,”
and “defamilialisation” (used by Esping‐Andersen's to classify countries). They claimed that many deviations from
Esping‐Andersen's typology are simply explained by the inclusion of indicators related to the healthcare and edu-
cation system.
Ebbinghaus (2012) reported a “meta‐analysis” of 13 studies (Esping‐Andersen, 1990, 1999) and the 11 studies
reported in Arts and Gelissen (2010). He pointed out that the number and coverage of cases varied considerably from
a low of 11 countries to as many as 25 countries, although most range around Esping‐Andersen's (1990, 1999) 16 to
18 countries. Moreover, there were clear differences in the methods and indicators of studies.
Kim (2015) reviewed studies that produced their own welfare regime typology by synthesising the lists compiled
by Arts and Gelissen (2010) and Ferragina and Seeleib‐Kaiser (2011), followed by “backward and forward reference
list checking.” This resulted in a total of 33 studies, which were divided into five groups: Esping‐Andersen's (1990)
original 18 nations, OECD members, non‐ OECD regions, non‐ OECD nations, and all nations. For the purpose of
his review, he focused on his first two groups of 24 studies.
The broad conclusions from these reviews present a confusing picture (cf Bambra, 2007b). A comparison of
the reviews is given in Table 1. It shows that a total of 37 studies have been included in at least one review.
The number of studies included in reviews varies from 11 to 24, and this is not explained simply by the pool
of eligible studies increasing over time. Some studies were chosen by only one of the seven reviews (Bambra,
2005a; Esping‐Andersen, 1999; Ferragina et al., 2012; Ferreira & Figueiredo, 2005; Hudson & Kühner, 2009;
Kautto, 2002; MacMenamim, 2003; Pitruzzello, 1999; Scruggs & Allan, 2008; Talme, 2013; Vrooman, 2012),
although clearly some of the later studies could not have been included in earlier reviews. On the other hand, only
Ragin (2004) and Obinger and Wagschal (1998) (in German) were included in all reviews, whereas Kangas (1994)
and Shalev (1996) were included in six of the seven reviews. Even Esping‐Andersen (1990) was included only in
five reviews.
Scholars such as Ebbinghaus (2012) and Kim (2015) pointed to a case selection problem in the study of welfare
regimes, namely, that few studies justify their choice of countries. As Kim (2015) put it, taking their lead from
Esping‐Andersen (1990), few studies have discussed and justified their case selection. It is possible that their conclu-
sions are linked to the particular (and often unclearly justified) samples of studies. For example, some included studies
appear to be more concerned with defamilization rather than decommodification. Our review of reviews suggests a
8 POWELL ET AL.

different case selection issue in that few reviews provided a clear search strategy with inclusion/exclusion criteria
(but see Ferragina & Seeleib‐Kaiser, 2011), resulting in a large variety in the choice of studies. In our view, the large
variety of studies covered in the reviews, the unclear search strategy of some of them, and the fact that some are
now rather dated justifies a new review.

1.2 | Search strategy

Our starting point was to draw on the studies in the existing (above) reviews, resulting in 37 studies. We then carried
out a search to capture any other earlier studies and to update the list. In order to include books and chapters, we
used a Scopus search with the following search terms:
Social OR Welfare OR Esping OR World OR Regime AND Typology OR Taxonomy OR Cluster OR Class*.
The Search resulted in 1,044 hits. We then carried out forward and backward tracking on the studies. Our inclu-
sion criteria (cf Ferragina & Seeleib‐Kaiser, 2011) were that the study must include a classification; based on
social/welfare criteria (rather than economic criteria such as Varieties of Capitalism); cite Esping‐Andersen, and refer
to his at least one of his concepts; based on income maintenance or welfare as a whole (rather than individual ser-
vices such as health or education); be “mainstream” (i.e., not based on gender or de‐familization alone); and focuses
on “advanced capitalist democracies” (i.e., with the majority of cases being Esping‐Andersen's (1990) 18 nations of
Esping‐Andersen).
Two of these points may require more explanation. First, the term mainstream follows Orloff (1993, 1996,
2009) in focusing on research such as Esping‐Andersen (1990) that does not thematise gender. In other words,
it focuses on decommodification rather than defamilization. The large literature on the gendered analysis of wel-
fare states, which arose partly as a reaction to Esping‐Andersen's neglect of gender, focuses on defamilization, but
this term has produced different spellings, definitions, and operationalizations, as well as a number of cognate
terms such as “dedomestication” and “degenderization” (e.g., Kröger, 2011; Kurowska, 2016; Lohmann & Zagel,
2016; Saxonberg, 2013).
Moreover, although Esping‐Andersen (1999) discussed defamilization, he defined it in a very different way to
feminist scholarship in terms of the freedom of the family rather than women's freedom from the family (Bambra,
2007a). As Orloff (2009) noted, Esping‐Andersen (1999) blunted the radical edge of the concept. She added that
“every time I read an article in a nonfeminist journal citing Esping‐Andersen and no feminists on questions to do with
mothers employment and gender, I'm beyond vexed.” This review examines the decommodification (Esping‐Ander-
sen, 1990) rather than mainstream (Esping‐Andersen, 1999) or feminist concepts of defamilization. This is not
because defamilization is unimportant, but because it is important, and requires separate and more extended analysis
(e.g., Kröger, 2011; Kurowska, 2016; Lohmann & Zagel, 2016; Saxonberg, 2013).
Second, the selection of nations is restricted to studies that focus on Esping‐Andersen's (1990) original group of
18 “advanced capitalist democracies.” Similarly to gender, recent years have seen a significant literature that extends
regime analysis into a much wider group of nations (e.g., Kim, 2015). Esping‐Andersen's regime analysis has been
termed “Eurocentric” and “Swedocentric,” associated with “ethnocentric western social research,” a “Western lens”
and a “social democratic bias” (Powell & Kim, 2014). There are a number of arguments that suggest that the worlds
of welfare may be a historically and geographically bound typology based on particular notions of the welfare state
(e.g., Estevez‐Abe, 2008; Ferragina & Seeleib‐Kaiser, 2011; Rice, 2013; Powell & Kim, 2014). It is still an open ques-
tion whether Esping‐Andersen's (1990) approach extends to more recent OECD nations such as Korea and Turkey
(Powell & Yörük, 2017), let alone very different non‐OECD nations (e.g., Gough, Wood, Barrientos, & Bevan, 2004).
This means that we excluded some studies included in the earlier reviews. Indeed, we excluded the most included
studies such as Ragin (1994) and Kangas (2004) as they were based on one element of income maintenance, and
Obinger and Wagschal (1998) as this is based on the wider “Family of Nations” approach rather than Esping‐
Andersen (1990). This resulted in a final list of 21 studies.
POWELL ET AL. 9

2 | RESULTS

Table 2 focuses on the processes of the studies. The underlying concepts exhibit great variation. Indeed, it is notice-
able how few studies clearly discussed the three concepts of Esping‐Andersen (1990) of decommodification, strati-
fication, and the welfare mix. Perhaps—unsurprisingly—the closest is the updated replication study of Talme
(2013). Some studies focused on decommodification (e.g., Bambra, 2004; Scruggs & Allan, 2006). Others aimed to
stress original concepts perhaps under‐played by Esping‐Andersen (1990) such as the welfare mix (Powell &
Barrientos, 2004) and stratification (Scruggs & Allan, 2008). As suggested by the different terminology of their typol-
ogies, some studies such as Korpi and Palme (1998) and Hudson and Kühner (2009) had rather different aims. The
variety of the underlying concepts was related to a variety in variables. The number of variables showed significant
differences from a minimum of two (Bonoli, 1997) to a maximum of 30 (Danforth, 2014).
Some studies aimed to widen the original stress on transfer payments to include services (e.g., Bambra, 2005b;
Jensen, 2008). We have included studies such as Bambra (2005b) and Jensen (2008), which examine both
transfers and services, but exclude (e.g.) Stoy (2014), which focuses solely on services. However, we have not
included the most “included” studies of Ragin (1994) and Kangas (2004). As Vrooman (2012) noted, Ragin
(1994) analysed pension benefits, whereas Kangas (1994) focused on health insurance. He continued that “such
partial analyses can neither corroborate nor rebut Esping‐Andersen's classification, which refers to the entire
configuration of social security and labour market institutions.” Similarly, Ferragina and Seeleib‐Kaiser (2011)
argued that many deviations from the typology reported in the literature are simply the result of including traits
from other policy fields (education and health care) that should be discarded on theoretical grounds. Esping‐
Andersen (1999, pp. 73–74) stressed some criticisms of the TWWC are irrelevant because they are not
addressing welfare regimes but individual programmes. However, he admitted that his original typology focused
rather one‐sidedly on income maintenance, but income‐transfer programmes capture but one side of the welfare
state (pp. 87–88).
Finally, there is a significant variety of methods. Some studies aimed to replicate the original Esping‐Andersen
methodology of creating groups based on how nations cluster around the mean of decommodification scores, which
has been criticized (see e.g., Barrientos, 2015; Pitruzzello, 1999; Shalev, 1996). Some studies used some type of clus-
ter analysis, whereas others used Fuzzy set ideal type analysis, principal component analysis (PCA), multiple corre-
spondence analysis (MCA), and structural equation modelling (SEM). However, some appeared to use “eyeballing”
of a 2 × 2 matrix.
Table 3 shows that the variety in findings is related to the variety in process. The number of nations varies
from a minimum of 11 (Vrooman, 2012) to 23 (Hudson & Kühner, 2009), although 11 of the studies focused
on 18 nations. However, the description of those nations varied, including “European nations,” “capitalist
democracies,” “advanced capitalist democracies,” “advanced OECD democracies,” “OECD countries,” “postindustrial
countries,” and “original Esping‐Andersen countries.” The number of clusters varied from three to five (or nine
“overall types”) of Hudson and Kühner (2009), although Scruggs and Allan (2008) found no “mutually exclusive
groups.” The cluster names closely followed Esping‐Andersen's terminology, with the addition of labels such as
“Southern,” “Latin,” or “Peripheral” regimes. However, Pöder and Kerem (2011) found a “post‐Communist”
regime. A few studies used rather different terminologies such as Bismarckian/high‐spending (e.g., Bonoli, 1997);
productive/protective (Hudson & Kühner, 2009), and basic security, encompassing, and targeted (Korpi &
Palme, 1998).
The difference in terminology made classifying nations across studies difficult. We have assumed that terms such
as “socialist” and “social democratic” on the one hand and “corporatist,” “continental,” and “conservative” were similar
enough to be part of the same grouping across the studies. The country classification seemed to show less consensus
than the review of Ferragina and Seeleib‐Kaiser (2011; see above). However, our study suggests that most of the
nations are placed in the same broad group by only around 50% of the studies (i.e., their “medium internal consis-
tency countries”). We found that no nation reached their pure threshold, with the greatest level of agreement for
 A comparison of studies included in review articles

10
TABLE 2

Cluster
Study Aim Concepts Data Method N nations N clusters composition

Esping‐ To classify Decommodification; De‐commodification: Additive 18 OECD countries 3 Liberal (AUS,

Andersen Western stratification; SSIB data files (1980) indices CAN, IRE,
(1990) welfare welfare mix Stratification: NZ, UK, USA)
states as Social Security Conservative
belonging to programs (FIN, FRA,
one of three throughout GER, JAP,
“worlds of the world (1981) IT, SWI)
welfare ILO the cost of social Social
capitalism.” security (1981) Democratic
OECD Measuring (AUT, BEL,
Health Care NL, DEN,
1960–1983 NOR, SWE)
Castles and Regime and Redistribution; Trade Union density Nations placed in 18 OECD 4 Liberal (CAN,
Mitchell welfare state types of welfare Non‐Right Incumbency four quadrants countries FRA, IRE,
(1992) typologies that capitalism GINI coefficients of pre of 2 × 2 matrix SWI, USA)
examine and post transfer Conservative
redistribution Net redistribution (GER, ITA,
process NET)
Radical (AUS,
NZ, UK)
Non‐right
hegemony
(AUT, BEL,
DEN, FIN,
NOR, SWE)
Bonoli Two‐dimensional “quantity” of Social expenditure Nations placed in 16 European 4 Bismarckian/
(1997) classification of welfare as a percentage four quadrants nations high‐spending
welfare states provision (the of GDP, of 2 × 2 matrix (BEL, FRA,
“how much” average 1989–1992 GER,LUX,NED)
dimension) Percentage of social Bismarckian/

POWELL
different expenditure financed low‐spending
models of

ET AL.
(Continues)
 POWELL
TABLE 2 (Continued)

Cluster

ET AL.
Study Aim Concepts Data Method N nations N clusters composition
welfare through (GRE,ITA,
provision contributions POR,ESP,SWI)
(the “how” Beveridgean/
dimension) low‐spending
(IRE,UK)
Beveridgean/
high‐spending
(DEN,FIN,
NOR,SWE)
Korpi and To present social Causal factors Bases of Create ideal types 18 OECD Four types of Basic Security
Palme insurance affecting entitlement; based on three countries institutional (CAN, DEN,
(1998) institutions and the institutional benefit level aspects of social structure IRE, NET, NZ,
the strategies aspects the principle; insurance (five ideal SWZ, UK, US)
of equality welfare state; employer‐ programmes types) Corporatist
which they the effects of employee and place (AUT, BEL,
embody institutions cooperation in countries in FRA, GER,
program one of ideal types ITA, JAP)
governance accordingly Encompassing
development (FIN, NOR,
of social SWE)
insurance Targeted
programmes: (Australia)
Social Citizenship
Indicator
Program (SCIP)
Other data:
Luxembourg
Income
Study (LIS)
To re‐examine and Mostly OECD (1991; Dominant mode 18 OECD 3 Universalist
update 1990s EA 1994) and ILO data of solidarity; countries DEN, NOR,

(Continues)

11
 12
TABLE 2 (Continued)

Cluster
Study Aim Concepts Data Method N nations N clusters composition
Esping‐ welfare state State; Market, for Labour market Dominant SWE, FIN,
Andersen classification Family; Risk; institution data locus of NED (and
(1999) defamilization (average production solidarity; partly UK)
workers approach) Degree of Residual (AUS,
Defamilialization data decommodification CAN, NZ,
mostly from USA (and
OECD (1993; 1996) partly UK)
Social
Insurance:
(AUT,
BEL, FRA,
GER, ITA, JAP
Not classified:
IRE, SWZ
Goodin Classify welfare Post‐productivist Work ‐welfare Nations placed in 18 advanced 4 (5 corporatist Low spending
(2001) regimes with their autonomy: income dimension: 2 × 2 four quadrants OECD divided by corporatist
demands on work adequacy, temporal matrix (high/low): of 2 × 2 matrix countries spending) (GER, IRE, ITA)
and welfare adequacy, and Labour force High spending
(two indicators) minimimal participation corporatist
conditionality rate and Social (AUT, BEL,
Security FRA)
expenditure as Post‐productivist
% of GDP (NET)
Social
Democratic
(DEN, FIN,
SWE)
Liberal (AUS,
CAN, JAP,
NZ, NOR,

POWELL
SWI, UK, US)

(Continues)

ET AL.
 POWELL
TABLE 2 (Continued)

Cluster

ET AL.
Study Aim Concepts Data Method N nations N clusters composition
Saint‐ To assess the Indicators that (1) Characteristics Hierarchical 20 OECD 4 clusters Liberal (CAN,
Arnaud applicability capture specific of governmental Cluster countries (new: Latin) US, UK, NZ,
and of EA clusters arrangements programmes Analysis AUS,IRE, ICE)
Bernard 1empirically between e.g. spending on Conservative
(2003) markets, the (education, (BEL, FRA,
state, and health etc.); % of GER, AUT,
families in government outlays NED, ITA, ESP)
the production and (2) Social situation Social democratic
distribution of the variables (SWE, FIN,
resources required e.g. Unemployment NOR,
for the well‐being rate; GDP DEN)Latin
of people annual growth; (GRE, POR)
Life expectancy;
Fertility Rate
(3) Political
participation
variables
e.g. Voter
turnout; Level
of trust; Union
membership
Powell and To provide a Moving from passive (1)Public Hierarchical 21 OECD 3 clusters Social
Barrientos classification to active politics spending on and k‐means countries Democratic
(2004) based on of the welfare state social security cluster (FIN, DEN,
welfare mix Including ALMPs into (2) Education analysis SWE, NOR,
the typology (3) ALMP: index FRA, NET)
of strictness Conservative
of employment (ITA, POR,
protection, and FRA, ESP,
ALMP spending GER, NZ,
BEL, AUT,
GRE)

(Continues)

13
 14
TABLE 2 (Continued)

Cluster
Study Aim Concepts Data Method N nations N clusters composition
Liberal (AUS,
IRE, US,
UK,
SWZ,
CAN, JAP)
Bambra To include health Reevaluation of Original Esping‐ Similar methods 18 advanced 5 Based on Cash
(2005) care services decommodification Andersen indices, and weights as capitalist and Health
into the welfare based EA Typology plus health Esping‐Andersen democracies Services
state typology (cash‐benefits) care index comparison:
Inclusion of social with three factors: Scandinavian/
services (health (1) Private health High (FIN,
services) expenditure as a NOR, SWE)
percentage of GDP Conservative1/
(2) Private hospital focus
beds on cash (GER,
as a percentage of SWZ, NET)
total Conservative2
bed stock (3) The (AUT, BEL,
percentage of the CAN, DEN,
population covered FRA, ITA)
by the health Liberal1/low
care system health (AUS,
CAN, IRE,
Liberal 2/high
health (IRE,
UK, NZ)
Ferreira and To examine Different Patterns of Hierarchical 14 and 22 countries 2 before Before
Figueiredo welfare welfare Welfare cluster analysis enlargement Enlargement:
(2005) regime regimes Provision; And PCA 3 after 1)EL, ESP, IT,

POWELL
structures before and welfare POR
before and Outcomes

(Continues)

ET AL.
 POWELL
TABLE 2 (Continued)

Cluster

ET AL.
Study Aim Concepts Data Method N nations N clusters composition
after EU after EU and stratification 2)BEL, DEN,
enlargement enlargement effects GER, GRE,
IRE, NET,
AUT, FI,
SWE, UK)

After
Enlargement:
1 (BEL, DEN,
GER, FRA,
NET, AUT,
SI, FI,
SWE, UK)
2 (IRE, CZ, EST,
LV, LT, HU,
PL, SK)
3 (ESP, ITA, POR)
Bambra To assess New Pensions Following Esping‐ 18 OECD 3 Finds EA roughly
(2006) the robustness decommodification (five indicators); Andersen countries typology to
and validity of measure Unemployment calculations be robust
Esping‐ (five indicators); with new Low
Andersen's Sickness (five decommodification decommodification
typology indicators) measure (AUS, IRE;
over time JAP; NZ, UK;
New USA)
decommodification Medium
index decommodification
(AUT, BEL,
CAN, DEN,
FRA, GER,
ITA, NET,
SWZ)

(Continues)

15
 16
TABLE 2 (Continued)

Cluster
Study Aim Concepts Data Method N nations N clusters composition
High
decommodification
(FIN,
NOW, SWE)
Scruggs To replicate Results Benefit Generosity Recalculations of 18 postindustrial 3, but Low generosity
&Allan welfare state suggest that Index composed benefit generosity/ countries question if (USA, JAP,
(2006) typology with there is little of unemployment, decommodification the idea of AUS, ITA,
updated/new evidence sickness, and with same regimes IRE, UK)
Esping‐Andersen's of national pension benefits methodology can be Medium
decommodification regime of creating uphold generosity
index coherence dimensions if there (NZ, CAN,
across as Esping is not a high AUT, FRA,
these three Andersen/ correlation FIN, GER)
programmes but updated across social High generosity
(insurance, measurements insurance (NET, SWZ,
pension, programmes BEL, DEN,
and sickness) generosity NOR, SWE)
Scruggs To examine Reanalysis calls Dimensions Same methodology of 18 postindustrial No mutually ‐
&Allen welfare state into question of welfare creating dimensions countries exclusive
(2008) typology with previous state as Esping groups
updated/new empirical stratification: Andersen/ found and
Esping‐Andersen's results that 1)Corporatism but updated many
social stratification imply effects number measurements countries
index of welfare 2) Etatism score
stratification, 3) Means‐tested either on
and raise poor relief more
questions 4) Private pensions than one of
about 5) Private health the
the degree of spending three original

POWELL
empirical 6)Average EA
support universalism dimensions

(Continues)

ET AL.
 POWELL
TABLE 2 (Continued)

Cluster

ET AL.
Study Aim Concepts Data Method N nations N clusters composition
for “welfare 7) Average benefit or
regimes,” as equality on none
opposed
to “welfare
programmes.”
Jensen To emphasize transfer 1)Transfers as a Hierarchical 18 OECD 3 (or 4) Social Services
(2008) the role of component % of social cluster countries Cluster Results
welfare services and welfare expenditure analysis (others available
in typologies of service 2)Welfare services for health)
welfare regimes components as a % of social Social democratic
expenditure (DEN, FIN,
Health care NOR, SWE)
spending as a Continental
% of social (AUT, BEL,
expenditure FRA, GER,
Social care ITA)
spending as Liberal (AUS,
a % of social CAN, IRE,
expenditure JAP, NET, NZ,
ESP, UK, USA)
Hudson and To examine Social investment 1) Education Fuzzy set 23 OECD 4 ideal types Pure/ideal types:
Kühner welfare and protection investment ideal type countries at (9 overall Productive
(2009) state clusters measures Training analysis 3 time points types) protective (FIN)
with the investment Protective (BEL,
inclusion 2) Employment GER)
of concept of protection Productive (US,
productive Income protection NZ)
welfare EPL index by Weak (AUS, UK
vs. existing OECD for at threshold)
protective employment Hybrid:
protection.

(Continues)

17
 18
TABLE 2 (Continued)

Cluster
Study Aim Concepts Data Method N nations N clusters composition
welfare Weak protective:
operationalization (ESP; FRA; CZ;
JAP; POR)
Weak productive
(CAN)
Weak productive
protective
(GRE, IRE,
SWZ, ITA,
KOR)
Protective plus
(SWE, NET,
AUT)
Productive plus
(DEN, NOR)
Pöder and To construct Social Model 19 Variables in Principal 22 countries (original 5 Mediterranean
Kerem geographic approach total: Component EP + Eastern Europe) (ESP, GRE,
(2011) divisions of (more (1) Total Govt Analysis (PCA) HUN, ITA)
European social encompassing: expenditure Continentals
models health, education, (2) Health (CZE, BEL,
LM, Family, Indicators AUT, FRA)
Social protection) (3) Family policy Post‐communist
indicators (GER, SLO,
(4) Education POL, LIT,
indicators LAT, EST)
(5) Labour Market Anglo‐Americans
Indicators (POR, IRE,
Social Protection UK USA)
indicators Nordics (DEN,

POWELL
SWE; NOR; FIN)

(Continues)

ET AL.
 POWELL
TABLE 2 (Continued)

Cluster

ET AL.
Study Aim Concepts Data Method N nations N clusters composition
Ferragina To analyse Family policies; Decommodificationand Multiple 18 OECD countries 4 Radicals (NZL,
et al. family policy dynamic temporal defamilialization correspondence AUS)
(2012) and approach to analysis (MCA) Liberal (UK,
unemployment welfare regime (and cluster analysis) IRE, USA,
protection to analysis; SWZ, ITA,
show a dual integration of JAP)
transformation of the work‐ Social
welfare states welfare and Democratic (FRA,
care‐welfare SWE, NOR, DEN)
nexi Christian
democratic
(GER, FIN,
NLD,BEL,
AUT)

very difficult to
clearly classify
(Austria, Italy,
Japan,
Belgium, Canada
and Switzerland)
Van der To test Esping EA rejected on Restricted to the Test construct 18 advanced capitalist 3 + 2 hybrids Liberal (AUS,
Veen and Andersen the five social validity using democracies JPN, USA)
van der models and grounds of insurance structural Conservative
Brug classifications construct characteristics of equation (BEL, FRA,
(2013) validity welfare states model (SEM) GER, ITA)
Universalist
Propose: 1) Original data (DEN, NET,
parsimonious from EA NOR, SWE)
measurement Scruggs and Allan Hybrid 1(AUS,
model limited replication data CAN, FIN,
to institutional SWZ)

(Continues)

19
 20
TABLE 2 (Continued)

Cluster
Study Aim Concepts Data Method N nations N clusters composition
characteristics Hybrid 2(IRE,
of social NZ, UK)
insurance
Vrooman To reassess Connection Similar to EA with Nonlinear principal 11 OECD countries 3 (for regimes,
(2012) the link between added variables component not culture)
between the regime and on informality analysis (PCA) Liberal (USA,
structural and culture and culture AUS, UK,
cultural aspects CAN)
of social security Corporatist
(DE, FR, BE)
Social
democratic
(DK, SE, NO)
Danforth To analyse Holistic 30 measures: Repeated cluster 18 capitalist democracies 3 1(AUT, BEL,
(2014) regimes understanding Decommodificaition; analysis over in 5‐year intervals from DEN, FIN,
over time and Public provision of time in 5 year 1950 to 2000 FRA, GER,
operationalization social increments ITA, NET,
of welfare state; services; Population NOR, SWE,
path dependence coverage; Income SWZ)
redistribution; Post‐ 2(CAN, IRE,
tax/transfer JPN, UK,
poverty; USA)
Defamilialization; 3(AUS NZL)
Activation In year 2000
for model
with all 30
measures on
7 dimensions

POWELL
(Continues)

ET AL.
 POWELL
ET AL.
TABLE 2 (Continued)

Cluster
Study Aim Concepts Data Method N nations N clusters composition
Talme To examine if the Same Same variables Replication 18 original EA countries 3 (for 13 Countries
(2013) three‐folded concepts as EA, but from of EA countries) that fall
typology is as EA CWED. under same
still valid regime type
or
decommodification
and stratification:
Liberal (AUS,
CAN, NZ,
USA)
Conservative
(AUT, FIN,
FRA,
Socialist (BEL,
DEN, IRE,
NET, NOR,
SWZ

21
 22
TABLE 3 Country classification

Modal
Regime
EA CM BO EA GO SAD PB BA FF BA SA SA JE HK PK FST VB VR DF TA # (MR) # in
Country (90) (92) (97) (99) (01) (03) (04) (05) (05) (06) (06) (08) (08) (08) (11) (12) (12) (12) (14) (14) Included Classification MR

AUS L R RE L L L LL LD LG ‐ L W R L L 3 L 17 Liberal 11

AUT SD NRM SI CH C C C 1 MD MG ‐ C PT+ C CD H1 1 C 18 Cons 10

Corp

BEL SD NRM BSH SI CH C C C 1 MD HG ‐ C PT C CD C C 1 S 20 Cons 10

Corp

CAN L L RE L L L LL MD MG ‐ L W/ ‐ H1 L 2 L 17 Liberal 9
PD

CZ 2 W/PT C 3 ‐ ‐

DEN SD NRM BVH U SD SD SD C 1 MD HG ‐ SD PD+ N SD U SD 1 S 20 SD/S 11

ESP BSL C C 3 L W/PT M 7 Cons 2

Corp

EST 2 PC 2 ‐ ‐

FIN C NRM BVH U SD SD SD SC 1 HD MG ‐ SD PP (+) N CD H1 1 C 19 SD/U 8

FRA C L BSH SI CH C C C 1 MD MG ‐ C W/PT C SD C C 1 C 20 Cons 13

Corp

GER C C BSH SI CL C C CC 1 MD MG ‐ C PT PC CD C C 1 19 Cons 13

Corp

GRE BSL LA C W/PP M 5 Cons/ 2

Bis

HU 2 M 2 ‐ ‐

ICE L 1 Liberal 1

IRE L L BVL N/C CL L L LL 2 LD LG ‐ L W/PP AA L H2 2 S 19 Liberal 10

ITA C C BSL SI CL C C C 3 MD LG ‐ C W/PP M L C 1 18 Cons/Bis 10

POWELL
(Continues)

ET AL.
 POWELL
TABLE 3 (Continued)

Modal

ET AL.
Regime
EA CM BO EA GO SAD PB BA FF BA SA SA JE HK PK FST VB VR DF TA # (MR) # in
Country (90) (92) (97) (99) (01) (03) (04) (05) (05) (06) (06) (08) (08) (08) (11) (12) (12) (12) (14) (14) Included Classification MR

JAP C SI L L LD LG ‐ L W/PT L L 2 12 Liberal/ 7

Low
Gen

KOR W/PP 1 W/PD 1

LUX BSH 1 BSH 1

LVA 2 PC 2 ‐ ‐

LTU 2 PC 2 ‐ ‐

NET SD C BSH U PP C SD CC 1 MD HG ‐ L PT+ CD U ‐ 1 S 19 SD/S/U 6

NOR SD NRM BVH U L SD SD SC HD HG ‐ SD PD+ N SD U SD 1 S 19 SD/S/U 13

NZ L R RE L L C LH LD MG ‐ L PD R H2 3 15 Liberal 7

POL 2 PC 2 ‐ ‐

POR BSL LA C 3 W/PT AA 6 Bis/Cons 2

SVK 2 PC 2 ‐ ‐

SVN 1 1 ‐ ‐

SWE SD NRM BVH U SD SD SD SC 1 HD HG ‐ SD PT+ N SD U SD 1 S 20 SD/S/U 14

SWZ L L BSL N/C L L CC MD HG ‐ W/PP L H1 1 S 15 Liberal 5

UK L R BVL RE/U L L L LH 1 LD LG ‐ L W AA L H2 L 2 19 Liberal 10

USA L L RE L L L LD LG ‐ L PD AA L L L 2 L 17 Liberal 12

Number 18 17 16 18 18 20 21 18 21 18 18 18 18 23 22 18 18 11 18 13 (18)
of
Countries

(Continues)

23
 24
TABLE 3 (Continued)

Modal
Regime
EA CM BO EA GO SAD PB BA FF BA SA SA JE HK PK FST VB VR DF TA # (MR) # in
Country (90) (92) (97) (99) (01) (03) (04) (05) (05) (06) (06) (08) (08) (08) (11) (12) (12) (12) (14) (14) Included Classification MR

N 3 4 4 3 5(4) 4 3 5 3 3 3 0 3 3 5 4 3+2 3 3 3
Regimes

Note. EA (90), Esping‐Andersen (1990); L, liberal; C, conservative; SD, social democratic; CM (92) Castles and Mitchell (1992); R, radical; NRM, non‐right Hegemony; BO (97), Bonoli
(1997); BSH, Bismarckian/high‐spending; BSL, Bismarckian/low‐spending; BVH, Beveridgean/high‐spending; BVL, Beveridgean/low‐spending; EA (99), Esping‐Andersen (1999); U, uni-
versalist; RE, Residual; SI, Social Insurance; N/C, not classified; GO(01), Goodin (2001); CL, low spending corporatist; CH, high spending corporatist; P, post‐productivist; SAD(03),
Saint‐Arnaud and Bernard (2003); LA, latin; PB(04), Powell and Barrientos (2004), BA(05), Bambra (2005); SC, Scandinavian; CC, conservative focused on cash; C, conservative; LL, liberal
low health; LH, liberal high health; FF(05), Ferreira and Figueiredo (2005): 1; 2; 3 (after enlargement); BA(06), Bambra (2006); HD, high decommodification; MD, medium
decommodification; LD, low decommodification; SA(06), Scruggs and Allan (2006); LG, low generosity; MG, medium generosity; HG, high generosity; SA(08) Scruggs and Allan (2008):
no mutually exclusive groups found; JE(08), Jensen (2008); HK(09), Hudson and Kühner (2009) Ideal types: P, productive protective; PT, protective; PD, productive; W, weak and combi-
nations for hybrids and PD+, productive plus; PT+, protective plus; PK(11), Pöder & Kerem (2011); M, Mediterranean; C, Continentals; PC, Post‐communist; Anglo‐Americans; N, Nordics;
FST(12), Ferragina et al. (2012); CD, Christian democratic; VB(12), Van der Veen and van der Brug (2013); H1 = Hybrid 1; Hybrid 2; VR (12), Vrooman (2012); DF(14), Danforth (2014)
(2000 model chosen) 1, first cluster; 2, second cluster; 3, third cluster; TA(14), Talme (2013): L, Liberal; C, Conservative; S, Socialist.

POWELL
ET AL.
POWELL ET AL. 25

Sweden (social democratic: 14/20 = 70%), Germany (conservative: 13/19 = 68%), France (13/20 = 65%), and Austra-
lia (liberal: 11/17 = 65%). The lowest level of agreement is for Switzerland (5/15 = 33%), Portugal (2/6 = 33%), Neth-
erlands (6/19 = 32%), and Spain (2/7 = 29%). Our review includes a number of post 2011 studies, which they could
not have included, but it seems that different inclusion criteria seems to produce different results. It is difficult to
state that one set of inclusion criteria is “better” than another, but this does underline the point about the need to
be clear about inclusion criteria.

3 | D I S C U S S I O N A N D CO N C L U SI O N S

This article has reviewed the reviews of welfare state typologies, pointing to issues of often unclear case selection
and a wide range of concepts, variables, and methods, resulting in a variety of worlds of welfare and their constit-
uent nations. Second, we produced a new review, which adds value as it is based on a clearer search strategy and
includes more recent material that was not available in earlier reviews. This finds that there is a great variety in
terms of process (concepts, variables, methods, and number of countries) and findings (the number and composi-
tion of worlds). We argue that the country classification seems to show less consensus that previous reviews, with
fewer pure nations (i.e., agreement between studies). We suggest that, in order to provide a clear point of engage-
ment, future reviews need to pay more attention to a clear and explicit search strategy, including issues such as
inclusion criteria.
In more detail, we have pointed to the great variety in the welfare modelling business at two different levels. It
has been shown that reviews vary significantly in terms of the number and composition of included studies, which
has made it difficult to sum up the “state‐of‐the‐art.” This problem is compounded by often unclear search strategies,
resulting in trying to draw conclusions from some studies with very contrasting aims, and based on different con-
cepts, variables, and methods. In our view, more attention must be paid to the search strategy, including issues such
as inclusion criteria. This means that subsequent critics have at least a clear point of engagement, allowing points of
disagreement to be made explicit. Moreover, it has been shown that the individual studies included in the reviews
also vary significantly in terms of issues such as aims, concepts, variables, and methods. In our view, the production
of further typologies may be near the point of diminishing returns. We suggest that further work in the field should
consider a number of issues.
First, theoretical or conceptual issues should be clearly discussed (Powell, 2015). Arts and Gelissen (2010) argue
that it is the theory that creates the typology, not the typology creates the theory. Esping‐Andersen drew on power
resources theory, or as Arts and Gelissen (2010: 582) put it, the “power resources” cum “path dependency” paradigm.
However, they continued that most of the alternative typologies lacked a firm theoretical foundation. The original
and modified conceptual position of Esping‐Andersen (1990) was based on decommodification, stratification, and
the welfare mix, with social risk and defamiliziation added later (Esping‐Andersen, 1999). If authors wish to signifi-
cantly depart from these concepts, then they need to justify their reasons for this decision. Second, the variables
must be clearly linked with the concepts (see Yörük, Öker, Yıldırım, & Yakut‐Çakar, 2019 for variable selection prob-
lem that undermines validity and reliability in the welfare modelling business). Third, the range of nations should be
carefully considered. Fourth, the range of sectors covered should include benefits and services rather than single
areas. The aim should be towards holism, covering as much of the welfare state or regime as possible, allowing a view
of the wood rather than the individual trees. Fifth, methods should be “formal” and based on some notion of cluster-
ing rather than on informal or arbitrary methods. Finally, data should be recent (e.g., Talme, 2013) or longitudinal (e.g.,
Danforth, 2014). The original data of Esping‐Andersen (1990) is now nearly 40 years old. Much has changed over this
period, and examining a static or frozen landscape of almost ancient history can tell us little of today's social policy.
Clearly, some elements of Esping‐Andersen's (1990) pioneering study would not match these suggestions, but
he has later pointed to some of these issues (Esping‐Andersen, 1999). It is fitting to leave (almost) the last word to
Esping‐Andersen (1999, p. 94): his re‐consideration of some points were “certainly not the last word on the
26 POWELL ET AL.

subject. The question of how to identify and classify welfare regimes will remain open because … researchers dif-
fer in terms of what attributes they consider vital and of how to measure them.” However, future researchers
need to be equally clear of their approach.

ORCID
Martin Powell https://orcid.org/0000-0002-9148-5087
Erdem Yörük https://orcid.org/0000-0002-4882-0812
Ali Bargu https://orcid.org/0000-0002-8729-3832

RE FE R ENC ES
Abrahamson, P. (1999). The welfare modelling business. Social Policy and Administration, 33(4), 394–415. https://doi.org/
10.1111/1467‐9515.00160
Arcanjo, M. (2006). Ideal (and real) types of welfare state, working paper 2006/06, Lisbon: Department of Economics, Tech-
nical University of Lisbon.
Arts, W., & Gelissen, J. (2002). Three worlds of welfare or more. Journal of European Social Policy, 12(2), 137–158. https://doi.
org/10.1177/0952872002012002114
Arts, W. A., & Gelissen, J. (2010). Models of the welfare state. In S. Leibfried, J. Lewis, H. Obinger, & C. Pierson (Eds.), The
Oxford handbook of the welfare state (pp. 569–583). Oxford: Oxford University Press.
Bambra, C. (2004). The worlds of welfare: Illusory and gender blind? Social Policy and Society, 3(3), 201–211. https://doi.org/
10.1017/S147474640400171X
Bambra, C. (2005a). Worlds of welfare and the health care discrepancy. Social Policy and Society, 4(1), 31–41. https://doi.org/
10.1017/S1474746404002143
Bambra, C. (2005b). Cash versus services: “Worlds of welfare” and the decommodification of cash benefits and health care
services. Journal of Social Policy, 34(2), 195–213. https://doi.org/10.1017/S0047279404008542
Bambra, C. (2006). De‐commodification and the worlds of welfare revisited. Journal of European Social Policy, 16(1), 73–80.
https://doi.org/10.1177/0958928706059835
Bambra, C. (2007a). Defamilisation and welfare state regimes: A cluster analysis. International Journal of Social Welfare, 16(4),
326–338. https://doi.org/10.1111/j.1468‐2397.2007.00486.x
Bambra, C. (2007b). “Sifting the wheat from the chaff”: A two‐dimensional discriminant analysis of welfare state regime the-
ory. Social Policy and Administration, 41(1), 1–28. https://doi.org/10.1111/j.1467‐9515.2007.00536.x
Barrientos, A. (2015). “A veritable mountain of data and years of endless statistical manipulation”: Methods in the Three
Worlds and after. Social Policy and Society, 14(2), 259–270. https://doi.org/10.1017/S1474746414000578
Bonoli, G. (1997). Classifying welfare states: A two‐dimension approach. Journal of Social Policy, 26(3), 351–372. https://doi.
org/10.1017/S0047279497005059
Castles, F., & Mitchell, D. (1992). Identifying welfare state regimes: The links between politics, instruments and outcomes.
Governance, 5(1), 1–26. https://doi.org/10.1111/j.1468‐0491.1992.tb00026.x
Castles, F., & Mitchell, D. (1993). Families of Nations. Aldershot: Dartmouth.
Castles, F., & Obinger, H. (2008). Worlds, families, regimes: Country clusters in European and OECD area public policy. West
European Politics, 31(1/2), 321–344. https://doi.org/10.1080/01402380701835140
Danforth, B. (2014). Worlds of welfare in time: A historical reassessment of the three‐world typology. Journal of European
Social Policy, 24, 164–182. https://doi.org/10.1177/0958928713517919
Ebbinghaus, B. (2012) Comparing welfare state regimes: Are typologies an ideal or realistic strategy? Draft Paper presented
at ESPAnet Conference, Edinburgh, September 6–8.
Emmenegger, P., Kvist, J., Marx, P., & Petersen, K. (2015). Three worlds of welfare capitalism: The making of a classic. Journal
of European Social Policy, 25(1), 3–13. https://doi.org/10.1177/0958928714556966
Esping‐Andersen, G. (1990). The three worlds of welfare capitalism. Cambridge: Polity Press.
Esping‐Andersen, G. (1999). Social foundations of post‐industrial economies. Oxford: Oxford University Press. https://doi.org/
10.1093/0198742002.001.0001
Estevez‐Abe, M. (2008). Welfare and Capitalism in Postwar Japan. Cambridge: Cambridge University Press.
Ferragina, E., & Seeleib‐Kaiser, M. (2011). Welfare regime debate: Past, present, futures? Policy & Politics, 39(4), 583–611.
https://doi.org/10.1332/030557311X603592
POWELL ET AL. 27

Ferragina, E., Seeleib‐Kaiser, M., & Tomlinson, M. (2012). Unemployment protection and family policy at the turn of the 21st
century. Social Policy and Administration, 47(7), 783–805.
Ferreira, L. V., & Figueiredo, A. (2005), Welfare regimes in the EU 15 and in the enlarged Europe: An exploratory analysis,
working paper 176, Faculdade de Economia do Portugal.
Ferrera, M. (1996). The “Southern Model” of welfare in social Europe. Journal of European Social Policy, 6(1), 17–37. https://
doi.org/10.1177/095892879600600102
Gal, J. (2004). Decommodification and beyond: A comparative analysis of work‐injury programmes. Journal of European Social
Policy, 14(1), 55–69. https://doi.org/10.1177/0958928704039795
Gallie, D., & Paugam, S. (2000). Welfare regimes and the experience of unemployment in Europe. Oxford: Oxford University
Press.
Goodin, R. (2001). Work and welfare: Towards a post‐productivist welfare regime. British Journal of Political Science, 31(1),
13–39. https://doi.org/10.1017/S0007123401000023
Gough, I., Wood, G., Barrientos, A., & Bevan, P. (2004). Insecurity and welfare regimes in Asia, Africa and Latin America: Social
policy in development contexts. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511720239
Headey, B., Goodin, R., Muffels, R., & Dirven, H.‐J. (1997). Welfare over time: Three worlds of welfare capitalism in panel
perspective. Journal of Public Policy, 17(3), 329–359. https://doi.org/10.1017/S0143814X00008576
Hudson, J., & Kühner, S. (2009). Towards productive welfare? A comparative analysis of 23 OECD countries. Journal of Euro-
pean Social Policy, 19(1), 34–46. https://doi.org/10.1177/0958928708098522
Jensen, C. (2008). Worlds of welfare services and transfers. Journal of European Social Policy, 18(2), 151–162. https://doi.org/
10.1177/0958928707087591
Kangas, O. (1994). The politics of social security. In T. Janoski, & A. Hicks (Eds.), The comparative political economy of the wel-
fare state (pp. 346–364). Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9781139174053.015
Kautto, M. (2002). Investing in services in West European welfare states. Journal of European Social Policy, 12(1), 53–65.
https://doi.org/10.1177/0952872002012001636
Kim, K.‐t. (2015). From worlds to cases: Case selection and “other worlds” in the welfare modelling business. Social Policy and
Society, 14(2), 309–321. https://doi.org/10.1017/S1474746414000554
Korpi, W., & Palme, J. (1998). The paradox of redistribution and strategies of equality. American Sociological Review, 63(5),
661–687. https://doi.org/10.2307/2657333
Kröger, T. (2011). Defamilisation, dedomestication and care policy: Comparing childcare service provisions of welfare states.
International Journal of Sociology and Social Policy, 31(7/8), 424–440. https://doi.org/10.1108/01443331111149860
Kurowska, A. (2016). (De)Familialization and (de)genderization–Competing or complementary perspectives in comparative
policy analysis? Social Policy and Administration, 52(1), 29–49.
Leibfried, S. (1992). Towards a European welfare state. In Z. Ferge, & J. E. Kolberg (Eds.), Social policy in a changing Europe
(pp. 245–279). Frankfurt: Campus‐Verlag.
Lohmann, H., & Zagel, H. (2016). Family policy in comparative perspective: The concepts and measurement of familialization
and defamilialization. Journal of European Social Policy, 26(1), 48–65. https://doi.org/10.1177/0958928715621712
MacMenamim, I.(2003), Is there an East‐Central European variety of democratic capitalism? A twenty‐two country cluster
analysis, Working Paper 5, Centre for international studies
Obinger, H., & Wagschal, U. (1998). Drei Wohlfahrtsstaates? In S. Lessenich, & I. Ostner (Eds.), Welten des
Wohlfahrtskapitalismus (pp. 109–135). Frankfurt, Germany: Campus Verlag.
Obinger, H., & Wagschal, U. (2001). Families of nations and public policy. West European Politics, 24, 99–114. https://doi.
org/10.1080/01402380108425419
Orloff, A. S. (1993). Gender and the social rights of citizenship: The comparative analysis of gender relations and the welfare
states. American Sociological Review, 58, 303–328. https://doi.org/10.2307/2095903
Orloff, A. S. (1996). Gender in the welfare state. Annual Review of Sociology, 22(1), 51–78. https://doi.org/10.1146/annurev.
soc.22.1.51
Orloff, A. S. (2009). Gendering the comparative analysis of welfare states: An unfinished agenda. Sociological Theory, 27(3),
317–344. https://doi.org/10.1111/j.1467‐9558.2009.01350.x
Pitruzzello, S. (1999). De‐commodification and the worlds of welfare capitalism. A cluster analysis. Florence: European Univer-
sity Institute.
Põder, K., & Kerem, K. (2011). "Social Models" in a European Comparison. Eastern European Economics, 49(5), 55–74. https://
doi.org/10.2753/EEE0012‐8775490503
28 POWELL ET AL.

Powell, M. (2015). “A re‐specification of the welfare state”: Conceptual Issues in “The Three Worlds of Welfare Capitalism”.
Social Policy and Society, 14(2), 247–258. https://doi.org/10.1017/S1474746414000529
Powell, M., & Barrientos, A. (2004). Welfare regimes and the welfare mix. European Journal of Political Research, 43, 83–105.
https://doi.org/10.1111/j.1475‐6765.2004.00146.x
Powell, M., & Barrientos, A. (2011). An audit of the welfare modelling business. Social Policy and Administration, 45, 69–84.
https://doi.org/10.1111/j.1467‐9515.2010.00754.x
Powell, M., & Barrientos, A. (2015). Introduction: Twenty five years of the welfare modelling business. Social Policy and Soci-
ety, 14(2), 241–245. https://doi.org/10.1017/S147474641400058X
Powell, M., & Kim, K.‐T. (2014). The ‘chameleon’ Korean welfare regime. Social Policy and Administration, 48(6), 626–646.
Powell, M., & Yörük, E. (2017). Straddling two continents and beyond three worlds? The case of Turkey's welfare regime.
New Perspectives on Turkey, 57, 85–114.
Ragin, C. (1994). A qualitative comparative analysis of pension systems. In T. Janoksi, & A. Hicks (Eds.), The comparative polit-
ical economy of the welfare state (pp. 320–345). Cambridge: Cambridge University Press. https://doi.org/10.1017/
CBO9781139174053.014
Rice, D. (2013). Beyond welfare regimes: From empirical typology to conceptual ideal types. Social Policy and Administration,
47(1), 93–110. https://doi.org/10.1111/spol.12001
Saint‐Arnaud, S., & Bernard, P. (2003). Convergence or resilience? A hierarchical cluster analysis of the welfare regimes in
advanced countries. Current Sociology, 51(5), 499–527. https://doi.org/10.1177/00113921030515004
Saxonberg, S. (2013). From defamilialization to degenderization: Toward a new welfare typology. Social Policy and Adminis-
tration, 47(1), 26–49. https://doi.org/10.1111/j.1467‐9515.2012.00836.x
Scruggs, A., & Allan, J. (2006). Welfare‐state decommodification in 18 OECD countries: A replication and revision. Journal of
European Social Policy, 16(1), 55–72. https://doi.org/10.1177/0958928706059833
Scruggs, L., & Allan, J. (2008). Social stratification and welfare regimes for the twenty‐first century: Revisiting The Three
Worlds of Welfare Capitalism. World Politics, 60(4), 642–664. https://doi.org/10.1353/wp.0.0020
Shalev, M. (1996). The privatization of social policy? Occupational welfare and the welfare state in America, Scandinavia, and
Japan. Basingstoke: Palgrave Macmillan.
Siaroff, A. (1994). Work, welfare and gender equality: a new typology. In D. Sainsbury (Ed.), Gendering Welfare States (pp.
83–100). London: Sage. https://doi.org/10.4135/9781446250518.n6
Soede, A. J. (2004). Unequal welfare states: Distributive consequences of population ageing in six European countries. Den Haag:
SCP, Sociaal en Cultureel Planbureau.
Stoy, V. (2014). Worlds of welfare services: From discovery to exploration. Social Policy and Administration, 48(3), 343–360.
https://doi.org/10.1111/spol.12006
Talme, L. (2013). Do capitalist welfare states still consist of the good, the bad and the ugly? Unpublished Thesis, Lund Uni-
versity, Sweden.
Van der Veen, R., & Van der Brug, W. (2013). Three worlds of social insurance: On the validity of Esping‐Andersen's welfare
regime dimensions. British Journal of Political Science, 43(2), 323–343. https://doi.org/10.1017/S0007123412000257
Vrooman, J.C. (2009). Rules of relief. The Hague, the Netherlands: Netherlands Institute for Social Research.
Vrooman, J. C. (2012). Regimes and cultures of social security: comparing institutional models through nonlinear PCA. Inter-
national Journal of Comparative Sociology, 53, 444–477. https://doi.org/10.1177/0020715212469512
Wildeboer Schut, J. M., Vrooman, J. C., & de Beer, P. (2001). On worlds of welfare. Institutions and their effects in eleven welfare
states. The Hague: Social and cultural planning Office of the Netherlands.
Yörük, E., Öker, I., Yıldırım, K., & Yakut‐Çakar, B. (2019). The variable selection problem in the three worlds of welfare liter-
ature. Social Indicators Research, 1–22. https://doi.org/10.1007/s11205‐019‐02070‐7 online first

How to cite this article: Powell M, Yörük E, Bargu A. Thirty years of the Three Worlds of Welfare Capitalism: A
review of reviews. Soc Policy Admin. 2019;1–28. https://doi.org/10.1111/spol.12510