Worlds of Welfare Services: From Discovery to Exploration



A recent thread of debate in social policy research has been the ‘discovery’ of welfare services. Previous comparative studies in this field have been largely erratic and have led to different results. This ambiguity is mainly due to flaws inherent in the data sets. In order to overcome these problems, this article uses an alternative approach of operationalizing welfare services. Employment patterns in the welfare sector provide a holistic picture of welfare services regarding quantity, kind, and organization. Cluster analysis leads to a four-cluster structure that bears high resemblance to the conventional welfare regime typology by Esping-Andersen and its subsequent advancements. These findings are set in the context of the welfare regimes literature in order to enhance our understanding of the functioning of welfare regimes. The study suggests that the ideological orientation of the welfare state is a good starting point for a holistic framework of welfare regimes combining the transfer and the service component.


A recent thread of debate in social policy research has been the ‘discovery’ of welfare services. After a long phase of neglect, welfare services have been identified as a distinct component of the welfare state (Jensen 2008). However, comparative research in this field has been erratic and has led to different results. Whereas some authors claim that the new worlds of welfare services are similar to the old worlds of transfer (e.g. Anttonen and Sipilä 1996; Esping-Andersen 1999; Jensen 2008), others find only partial consistency (Kautto 2002) or deny any congruence (Bambra 2007; Rauch 2007). As a consequence, worlds of welfare services resemble the dark side of the moon; we know that they exist but we do not really know what they look like.

It is argued that the ambiguity of existing research in the field of welfare services is mainly due to four major flaws:

  1. It does not take the welfare regime approach seriously as it ignores the welfare mix, i.e. the way in which welfare services are provided.
  2. It is sector-specific focusing either on social care or on healthcare. These approaches, however, do not offer a holistic picture of welfare services.
  3. It only covers a limited number of countries.
  4. It uses outdated data.

The latter is problematic as ‘[a]ny typology of welfare regimes […] remains valid only as long as history stands still’ (Esping-Andersen 1999: 73).

Any new study that seeks to shed more light on welfare services has to overcome these problems. Accordingly, this article addresses these aspects and aims at providing a holistic and up-to-date typology of welfare services regimes. In contrast to other studies looking at expenditure or provision levels, this study is based on welfare sector employment patterns as a measure for welfare services. It is argued that this measure is better suited to this purpose as it not only provides information on the quantity of welfare services but also on their kind and how they are organized.

This article follows an inductive approach. Engaging in the discussion on ideal-typical welfare regime analysis, I argue why it is important to analyze welfare services regimes distinctively. After presenting a literature review and outlining the distinctiveness of this study from previous work, I employ quantitative methods in order to derive empirical results. A hierarchical cluster analysis is run that results in a four cluster structure of welfare services regimes. These findings are set in the context of the welfare regime literature lest only to add another cluster structure but to enhance our understanding of the functioning of welfare regimes.

Ideal-typical Welfare Regime Approach

The publication of Esping-Andersen's seminal work on The Three Worlds of Welfare Capitalism (1990) has sparked a vast array of academic endeavour that is neatly summarized under the label ‘welfare modelling business’ (Abrahamson 1999). Most of the subsequent studies were real-typical comparisons, largely preoccupied with testing the suggested ideal-types (for an overview, cf. Arts and Gelissen 2010). This led to an academic debate ‘heavily slanted towards countries rather than worlds as the unit of analysis, and to whether specific countries actually represent a particular world’ (Powell and Barrientos 2011: 70). On the one hand, the frequent cross-checking of the ideal-types with actual developments is necessary as ideal-types only provide benefit if they are a valid representation of reality (Arts and Gelissen 2002: 138–40). On the other hand, they are no means in themselves and classifying single countries within the typology is of little use when the research stops here as this discussion is unlikely to generate further theoretical insights. In contrast, Esping-Andersen (1990) demonstrated with respect to the power resource theory that ideal-types can be gainfully employed in theory development – either as a dependent variable or as an independent variable (e.g. Johnson 2005 with respect to EU social policy; Jaeger 2006 with respect to public attitudes).

Despite the merits of an ideal-typical welfare regime approach, the debate on ideal-types has not really gotten off the ground (for notable exceptions, cf. Aspalter 2006, 2011). The original formulations of Esping-Andersen are still largely taken for granted and replicated. This is especially problematic with respect to social services. Although Esping-Andersen's original clustering was based on the decommodification score that does not take welfare services into account, the typology has been readily imposed on the worlds of welfare services as well. However, as Jensen (2008) showed with respect to welfare state expenditure, the transfer and the service component of a welfare state do not always correspond. Accordingly, simply applying the assumptions of conventional welfare regime research to the analysis of welfare services can be misleading. What is required instead is a better understanding of the functioning of the welfare services regimes in their own logic. In a later step, this knowledge can be integrated into a holistic framework of welfare regimes (Jensen 2008: 160, 2011a,2011a: 404).

The ‘Discovery’ of Welfare Services

The study of welfare services has made some headway in the last years (Kautto 2002; Jensen 2008, 2011a). A crucial question in this field of research – as in all social scientific issues – is the aspect of operationalization. How do we measure welfare services in order to arrive at meaningful conclusions? There is no across-the-board answer to this question as the appropriateness of an indicator depends on the theoretical perspective underlying the analysis (Green-Pedersen 2004). Thus, what is required for the purpose of this article is a measure that is able to capture the central theoretical assumptions of the welfare regime approach regarding welfare services. In contrast to functionalist explanations of welfare state development, the welfare regime approach argues that ideological (as claimed by power resource and party difference theory) and path-dependent factors (as stated by new institutionalism) have led to and maintained distinct institutional set-ups based on different conceptions of social security. The measure of welfare services should therefore be able to account for these institutional differences. Furthermore, as the welfare regime approach is based on holism (Powell and Barrientos 2011: 75), the indicator should be able to look at welfare services as a whole and not merely at specific programmes. The variable selection has to be based on these criteria.

Alber (1995: 134) suggested three variables for measuring welfare services: expenditure, staffing and take-up ratios. So far, only two of these approaches have been employed: using expenditure levels or provision levels.

A conventional approach in the literature has been the operationalization of welfare services via disaggregated expenditure data (Kohl 1981; Kautto 2002; Jensen 2008; Starke et al. 2008). The use of social expenditure data as a measure of cross-national variation in welfare state effort has been widely criticized (Esping-Andersen 1990; Allan and Scruggs 2004: 497–8; for a discussion, cf. Siegel 2007). From the perspective of the welfare regime approach the major flaw of this proxy is that it does not measure institutional characteristics (Green-Pedersen 2004: 9). Jensen (2011b) argues that cross-national institutional differences are less pronounced in the field of welfare services than in the field of transfers as healthcare and education are characterized by free access and equal treatment in most Western countries. Although this argument certainly has its merits with respect to eligibility criteria and benefit type, it ignores the fact that the organization of welfare services is characterized by a further institutional component that differs markedly between different countries: the welfare mix. The welfare mix is one of the defining characteristics of welfare regimes. Esping-Andersen distinguishes between state, market and family as ‘three radically different principles of risk management’ (Esping-Andersen 1999: 35; cf. Esping-Andersen 1990). The relative importance of each of these actors reflects the division of tasks and labour in a welfare regime that has historically evolved based on the ideological orientation within a country. The concept of the welfare mix has gained particular importance in the field of welfare services as it is characterized by the interplay of all actors (Evers 2011: 265). An analysis of cross-national variation in welfare services that does not include the welfare mix is inevitably incomplete as it ignores a major source of difference.

Esping-Andersen (1990) did not base his regime classification on expenditure but on provision levels that were coded into the decommodification index. It was suggested by Jensen (2008: 158) that decommodification corresponds to the transfer component, whereas defamilialization relates to social care services of a welfare state. Accordingly, relying on provision levels for the operationalization of welfare services, the defamilialization index could be used for regime differentiation. Empirical tests of the consistency of the defamilialization concept with the existing typology by Esping-Andersen led to ambiguous and sometimes contradictory results (Esping-Andersen 1999; Bambra 2004, 2007). A major flaw of this approach is its programme-specific orientation as the defamilialization concept does not capture welfare services in general but social care services in particular. Healthcare services that outnumber social care services in most countries are completely ignored. Thus, the defamilialization index would not shed light on the ‘forgotten half’ (Jensen 2011a) but on the ‘forgotten fraction’ of the welfare regime. Furthermore, this approach also fails to take the welfare mix into account and, hence, is strictly speaking also only about numbers. Whereas the first approach asks ‘how much’ money is spent on welfare services, the second considers ‘how many’ services are provided. Both fail to account for the way these services are provided. However, it is exactly this question that would contribute most to a better understanding of welfare regimes.

As an alternative approach of operationalizing welfare services, I use employment patterns in the welfare sector.1 Broadly speaking, welfare sector employment encompasses employees and self-employed in the fields of healthcare, childcare and long-term care. This method holds the promise to avoid the problems of the conventional two approaches as, first, it allows for a comprehensive analysis of welfare services beyond policy-borders and, second, it takes into account the different dimensions of welfare services including the welfare mix. In other words, it could possibly provide a holistic picture of welfare services regarding quantity, kind and organization.


As noted above, a good deal of work within the welfare regime debate has focused on the empirical validation of the advocated typologies. Thereby, a variety of methods has been used of which ‘cluster analysis has proved the most effective and widely used technique to identify welfare regimes’ (Powell and Barrientos 2004: 91).

Cluster analysis is a rather simple, explanatory method aimed at ‘finding groups in data’ (Kaufman and Rousseeuw 1990, cf. Everitt et al. 2011). The basic idea is that similar cases are identified and merged into groups. The resulting clusters are characterized by high internal homogeneity and high external heterogeneity. In general, different cluster methods can be discerned. Due to the inductive approach of this study, I use the Ward's linkage cluster analysis which belongs to the hierarchical agglomerative methods. Hereby, in the beginning all cases are treated as distinct clusters. Stepwise, the two clusters with the highest similarity are merged. This is measured as the lowest possible increase in the total-within-cluster residual sum of squares (Everitt et al. 2011: 77). This process is repeated until all cases are merged into one cluster. Based on the resulting fusion values, it is then the task of the researcher to decide upon a meaningful number of clusters. The ultimate criterion for the number of clusters is the theoretical fit of the result.

However, by running a multivariate analysis of variance (MANOVA) it is possible to assess the statistical distinctiveness of the clusters. The resulting F-values report how well the dimensions discriminate between the clusters. MANOVA only reveals if the cluster means differ significantly but do not tell which clusters are responsible for this variation. More detailed analysis can be achieved by post-hoc tests that look for differences between all possible pairs of group means. Whereas MANOVA reveals in general that the cluster means are statistically distinct with respect to one dependent variable, post-hoc tests compare the clusters pairwise in order to ascertain which of the clusters actually differ with respect to the dependent variable. Thus, whereas the F-test can been seen as a method of testing the statistical robustness of the cluster structure, the post-hoc test provides more insights on the cluster differences and is thus a valuable tool for the interpretation of the clusters. There are different post hoc methods. In cases where all pairwise combinations are of potential interest, Maxwell and Delaney (2004: 199) recommend the Tukey-HSD test.

Although a MANOVA allows some sort of statistical test, cluster analysis is not subject to formal quality criteria. Therefore, it runs the risk of being used arbitrarily. The findings are very sensitive to the inclusion or omission of variables. Therefore, the selection of variables must be subject to careful scrutiny and theoretical explanation; a demand that is often only insufficiently paid attention to.

To summarize the methodical procedure: as the first step, I explain my variable selection; this variable selection is used for cluster analysis; the cluster structure is subsequently tested by a mulitivariate analysis of variance and a post-hoc test.

Variable Selection and Data

The selection of variables must be derived from the research question. It is the aim of this article to portray a holistic picture of the worlds of welfare services. The approach of using welfare sector employment data allows for capturing different dimensions of welfare services.

Quantity of services

Welfare services stand out against welfare transfers due to their labour intensity. They are provided from people to people and are thus person-related in a double sense (Naegele 2011: 404). As no welfare service can be provided without personnel, the level of employment in the welfare sector is a promising indicator for the quantity of welfare services. Thus, the employment level in the welfare sector as a percentage of total civilian employment is included.

It appears reasonable to distinguish between employment levels in the fields of healthcare and social care. Looking at expenditure levels, Jensen (2008: 158–60) argues that it is primarily social care services that account for welfare regime differentiation (cf. Esping-Andersen 1999: 105). As their provision is ideologically contested, there are marked differences in social care expenditure between the welfare states. In contrast, public health services are hardly subject to any ideological debate, which leads to similar expenditure levels and makes them a bad proxy for the welfare services component of welfare regimes. This assessment has to be qualified as Jensen creates an artificial homogeneity in healthcare spending by his country selection as he ignores the Southern European and the Central and Eastern European welfare states. Furthermore, this problem only arises when the focus is solely on quantity aspects such as expenditure data. Healthcare provision is still likely to account for regime differences when its organization is taken into consideration. Therefore, healthcare and social care services are included in this analysis.

Kind of services

It is not only the quantity of welfare services that matters but also what kind of services are provided. Of course, the kind of welfare services could be broken down into very detailed activities. However, as it is the aim of this article to identify the general regime logic, I distinguish the kind of services along one central dimension that I label the level of care intensity. Broadly speaking, it is possible to distinguish between skill-intensive services with a low level of personal care orientation and caring services with lower skill requirements. Differences in the welfare regimes' care intensity do not merely result from the varying importance of social care services. Instead, the care intensity is a general orientation that is reflected in the fields of healthcare and social care services alike. Thus, measuring the care intensity of welfare services reveals central aspects of a regime's overall orientation and its functioning logic. The variable is operationalized as the share of personal service workers (ISCO-88, Group 5.12) employed in the welfare sector.

Provider of services

Despite the conceptual primacy that the welfare mix has gained with Esping-Andersen's Social Foundations of Postindustrial Economies (1999), it has not been adequately reflected in the welfare modelling business (Powell and Barrientos 2011: 81, 2004: 86). This analysis includes the share of public sector employment as a proxy for the welfare mix.

Although public employment is probably the most prominent component of the welfare mix, internationally comparative research has been impeded by a lack of adequate data. As Tepe (2009: 6) states ‘[a] primary obstacle in comparative research is the availability of consistent public employment data’. This is especially true when sector-specific data are needed. Early work used the Nordic WEEP Data Bank (Kolberg and Esping-Andersen 1991) but the data end with the year 1985. The ILO Database on Public Sector Employment provides sector-specific data for only 15 countries of this study. Moreover, it is based on a broad concept of public employment encompassing also social security funds and non-profit institutions (Hammouya 1999: 3–4). Thus, a consistent source for internationally comparable data does not exist. Subsequent studies have tried to solve this problem differently. Whereas some researchers employed micro-data from the Luxembourg Income Study (LIS) (Gornick and Jacobs 1998; Mandel and Shalev 2009), others used overall civilian government employment as a proxy given its high correlation with welfare state employment (Huber and Stephens 2000: 328). I suggest a different approach. The National Accounts of OECD Countries, General Government Accounts (OECD 2011b) disclose general government spending on salaries to employees by sector. Dividing the compensation paid to public employees in the welfare sector by the adjusted in-kind expenditure (see below) provides us with an approximation of the share of public welfare sector employment. Despite all legitimate criticism with this approach, the indicator provides a meaningful measure which can be shown by cross-checking with some of the national employment data available (see table 1).

Table 1. Selected cross-check for public employment data in the welfare sector
CountryCalculated valueNational statistical instituteDifference
Source: public welfare employment: national statistical institutes; total welfare employment: OECD Labour Force Statistics.
Portugal (2005)43.7%43.3%+0.4%

Of course, the share of public employment is only an approximation of the welfare mix. Esping-Andersen (1999: 35) distinguished between three different providers of welfare services: state, market and family. The category of the market can be further divided into for-profit and non-profit organizations suggesting the picture of a ‘welfare diamond’ (Evers et al. 1994). Given this differentiated concept, the reliance on the share of public employment as a proxy for the welfare mix suffers from two limitations. First, the informal service provision within the family cannot be captured with this approach. Second, by using public sector employment data, the market is treated as a residual category encompassing profit and non-profit providers. Given the differences between both sectors, this simplification is problematic. Although the third sector plays a prominent role in many areas of the welfare regime, it is seldom treated as an independent type of service provider. Ranci (2002: 32) observes that the existing welfare regime typologies suffer from two limitations: either they do not take into account the third sector or they are not limited to the welfare sector. This neglect of the non-profit sector in the welfare regime literature is largely due to the lack of internationally comparable data in this field. Although the Johns Hopkins University Center for Civil Society Studies and the Economic Statistics Branch of the United Nations Statistics Division launched an initiative for standardized data collection almost two decades ago, this has not yet resulted in a comprehensive data base. Therefore, a comprehensive international comparison is precluded. Despite these limitations, the approach of using public sector employment data has its particular merits as it captures a good deal of cross-national variation in the organization of welfare services leading to distinct welfare regimes.

Payer of services

The question of who bears the costs has always been a crucial one for the development of welfare services. As social services are very labour-intensive and hence in most cases too expensive for private purchase, it required the subsidization by the state for the formalization of service provision. This can be exemplified by the example of the development of long-term care (LTC) services. Historically, the family was the traditional provider of care services and still constitutes the first line of care provision in most of today's societies. It was only with the increasing takeover of responsibilities by the state that formal care provision soared. Scandinavian countries were the first to experience a service expansion in the wake of the introduction of public LTC schemes in the late 1960s. In contrast, it was only with the adoption of comprehensive LTC policies in the mid-1990s that a market for formal LTC provision developed in countries such as Germany and France. In other countries that still lack comprehensive LTC policies such as Italy, Greece or the Central and Eastern European countries, formal care services are still underdeveloped. Consequently, an analysis of welfare services must take the payment structure into account. Of course, it would be interesting to examine the welfare mix with regard to spending. However, with the exception of the healthcare sector no reliable data exist in this field. Given this lack of adequate data, this study relies on the level of public expenditure as an indicator for state intervention. This restriction can be justified on the grounds of the state's dominance in this field that has been described above.

This dimension is expressed in terms of adjusted social expenditure. The focus on benefits in kind is still too broad as not only personnel costs are contained but also spending on pharmaceuticals, housing allowances and technical infrastructure (cf. ESSPROS 2008). Thus, by subtracting these expenses an adjusted spending level is calculated.3 The data is expressed as a percentage of gross domestic product.

Of course, the data set used in this study is not beyond reproach. However, it will enable us to spot more than just the peak of the iceberg (see table 2).

Table 2. Data sources
VariableNameYearData Source
  1. Note: all data were standardized by z-transformation.
Employment healthcareemp_health2008Eurostat LFS data by occupation and economic activity (from 2008, NACE Rev. 2 at the two digit level) (1 000) (lfsa_eisn2) requested from Eurostat (NACE group 86). Data for Australia, Canada, New Zealand and the USA estimated via linear regression with NACE group 86 as the dependent variable and group Q as the independent variable (source: OECD Labour Force Statistics; for the USA: US Bureau of Labor Statistics)
Employment social careemp_care2008Eurostat LFS data by occupation and economic activity (from 2008, NACE Rev. 2 at the two digit level) (1 000) (lfsa_eisn2) requested from Eurostat (NACE group 87 and 88). Data for Australia, Canada, New Zealand and the USA estimated via linear regression with NACE group 87 and 88 as the dependent variable and group Q as the independent variable (source: OECD Labour Force Statistics; for the USA: US Bureau of Labor Statistics)
Care intensityemp_int2008Eurostat LFS data by occupation and economic activity (from 2008, ISCO-88 at the two digit level) (1 000) (lfsa_eisn2) requested from Eurostat. Data for Australia, Canada, New Zealand and the USA estimated via linear regression with ISCO-88 group 5.1 as the dependent variable and group 5 as the independent variable (source: for Australia, Canada and New Zealand: ILO LABORSTA; for the USA: US Bureau of Labor Statistics – SOC2000 converted into ISCO-88 classification on the basis of the conversion table provided by the National Crosswalk Service Center)
Public welfare employmentemp_public2007National Accounts of OECD Countries – Vol. IV: General Government Accounts and OECD Labour Force Statistics; for Australia: national statistical institute
Adjusted public social spending on benefits in kindexp_inkind2007OECD Social Expenditure Database; data on pharmaceutical expenditure from OECD Health Data 2010 and World Health Organization Health Database


The cluster analysis results in a four cluster structure that has proved to be statistically significant as revealed by the F-values.4 It bears high resemblance to the conventional welfare regime typology by Esping-Andersen and its subsequent advancements. A liberal, a conservative and a social democratic cluster can be clearly distinguished. A fourth cluster comprises the Southern as well as the Central and Eastern European countries (see tables 3 and 4, and figure 1).

Figure 1.

Difference in characteristics of welfare service regimes (cluster means, z-transformed)

Table 3. Worlds of welfare services
LiberalConservativeSocial democraticRudimentary
  • Australia
  • Canada
  • Finland
  • France
  • Iceland
  • Ireland
  • New Zealand
  • UK
  • Austria
  • Belgium
  • Germany
  • Netherlands
  • Switzerland
  • USA
  • Denmark
  • Norway
  • Sweden
  • Czech Republic
  • Greece
  • Hungary
  • Italy
  • Poland
  • Portugal
  • Slovak Republic
  • Spain
Table 4. Differences in characteristics of welfare services regimes
  1. Note: number denotes the cluster average. An asterisk denotes a significant difference between the means of two clusters at the 0.05-level based on the Tukey-HSD test.



emp_health6.3%  *
emp_care5.1% **
emp_int31.3% * 
exp_inkind7.4% **



emp_health 6.2% *
emp_care 5.3%**
emp_int 23.6%* 
exp_inkind 7.0%**



emp_health  7.1%*



emp_int  *22.9%
emp_public **43.1%

The classification of certain countries appears to be surprising at first glance and would require a more detailed analysis if the article was preoccupied with real types. However, as the focus of this article is on ideal-types, the four clusters are interpreted in the context of the welfare regime literature. Although the academic debate has moved from welfare state regimes to welfare regimes, the role of the welfare state appears to be a good starting point for an analysis of welfare services regimes. As welfare services are generally too expensive for private purchase, the development of a welfare service sector is dependent on state intervention.

In general, the welfare state has three channels of influence on welfare services: first, public funding, second, public employment and, third, public regulation. This study focuses on the first two channels for two reasons. First, regulatory policies have been under-researched in the welfare state literature (Powell and Barrientos 2011: 75) and have only recently been considered in policy-oriented research (Wendt 2009). Including them in the analysis would necessitate a study on its own. Second, it is difficult to translate qualitative aspects such as regulatory policies into numbers as this study would require.

Regression analysis reveals that the welfare state through the two channels of influence accounts for at least half of the variation in the level of healthcare and social care services as well as in the degree of care intensity. This suggests that in an international comparative perspective the ideological orientation and the capacity of the welfare state are shaping factors of welfare services regimes (see table 5).

Table 5. Influence of state intervention on welfare service development (regression analysis)
  1. Note: */**/*** denote significance at the 5%, 1% and 0.1% levels respectively.
emp_public  0.19*


In its original formulation by Esping-Andersen (1990), the liberal welfare state was characterized as residual in the sense that it only adopts a narrow definition of social risks. The label ‘liberal’ refers to the ideology of non-interference by the state into its citizens' private lives. Consequently, liberal welfare regimes are based on a distinction between ‘public’ and ‘private’ spheres (O'Connor et al. 1999: 3). This divide runs through the field of welfare services that is marked by a dual structure. On one side, as the provision of healthcare services is beyond the capacity of the individual, it is in many countries financed and carried out by the state. Universal healthcare schemes lead to a high level of welfare services in this field. On the other side, social care services are perceived as an individual responsibility that receives little state support. Social care policies were adopted relatively late and apply means-tested eligibility criteria in order to limit state support to the needy. Social care services are mainly provided by independent, for-profit companies. Whereas well-off households can afford to purchase care services at the market and poor households receive state support, for all other households the family is the primary care provider by default. Recent reforms have rather strengthened this approach. As a consequence, the liberal welfare regime exhibits high levels of healthcare services and modest levels of social care services.


Based on the subsidiarity principle the welfare states in continental Europe were reluctant to intervene in the field of social service provision. Historically, these welfare states have always been considered ‘service-lean yet very “transfer-heavy”’ (Esping-Andersen 1997: 67). But as a result of economic, social and demographic change, the high reliance on the family for social service provision has become more and more dysfunctional. As a consequence, Bismarckian welfare regimes have been forced by necessity to become more employment-friendly and gender-neutral (Hemerijck and Eichhorst 2010: 323). This is not to say that they have embarked on a wholehearted shift towards defamilialization. On the contrary, they have tried to limit the welfare service extension by introducing cash for care schemes that give especially low-skilled women incentives to stay at home (Morel 2007; Palier 2010a: 376). As a result, the scope of welfare services and the care intensity has risen but falls way short of the high level of service provision in the social democratic welfare regimes.

The conservative welfare states have a very low share of public employment in the welfare sector. Welfare services are predominantly provided by non-profit and for-profit providers whereas the state exerts only a low level of influence. This trait is founded in the logic of the welfare regime. As Palier (2010b: 35) notes with respect to the transfer component ‘[u]sing the notion of “welfare state” to designate the social protection systems of Continental Europe is misleading’. The same applies for social services as well. Neither is the state involved in the administration of funds (which is done by social insurance bodies) nor in the actual service delivery. In these countries, the state is largely confined to legislation and exerts only limited influence on the service provision.

Social democratic

In the social democratic welfare regime, the welfare state is highly interventionist in order to ensure social equality. As female labour market participation has long been a primary goal, social democratic welfare states put a strong focus on defamilialization and used public sector employment expansion as a labour market policy. In Sweden, the state took over responsibility for elderly care in 1956 when the law that held children responsible for their parents' welfare was abolished. As a consequence, public expenditure for welfare services is high in order to finance the well developed childcare, long-term care and healthcare services. This results in a high level of welfare service provision. The share of public sector employment is comparably high ensuring control over service quality. In other words, the social democratic service state offers and provides qualitatively high welfare services for a high share of its citizens.

Recently, the uniqueness of the social democratic welfare regime with regard to welfare services has been questioned. Rauch (2007) for example argues that ‘it does not seem to make much sense to speak of a particular Scandinavian social service model’. Looking from the perspective of welfare sector employment, this assessment is clearly wrong. As table 3 shows, the social democratic welfare regime differs significantly from the other regimes in almost all respects. Instead, it can be claimed that the high level of welfare service provision is the most distinguishing feature of the social democratic welfare regime (Kautto 2010: 592).


The label ‘rudimentary welfare state’ was coined by Leibfried (1992) with regard to the Southern European countries. It alluded to the fact that social security was only partially institutionalized in these countries which has been especially true in the field of welfare services. Its main characteristics in this field are the limited role played by the state and the high degree of familialism (Naldini 2003). The low degree of state penetration of the sphere of welfare services is reflected in comparatively low levels of social expenditure, healthcare and social care services. From an ideal-typical perspective, these countries share the logic of subsidiarity and the main features of the conservative welfare regime to an exaggerated degree. Thus, in this regard they can be characterized as a ‘“discount edition of the Continental model” rather than a regime in its own right’ (Abrahamson 1999: 405). However, these countries have introduced national healthcare services which give them a distinct trait (Ferrera 1996) that is reflected in a medium level of public sector employment.

Despite some differences from a real-type perspective, the Central and Eastern European countries share these basic characteristics of the Southern European countries. They have undergone a massive transformation and financial problems have stalled an expansion of the healthcare systems. Having inherited a defamilialized welfare regime from the era of communism, the countries embarked on a trend of re-familialization (Saxonberg and Szelewa 2007: 353). Accordingly, they exhibit low levels of social expenditure and welfare services. From their general outline they fit well the characteristics of the conservative welfare regime (Aspalter et al. 2009) but the welfare state takes only a rudimentary role.


The neglect of welfare services in the construction of The Three Worlds of Welfare Capitalism (Esping-Andersen 1990) has been a major criticism. Subsequent work in the wake of the recent ‘discovery’ of welfare services has tried to elaborate on the congruence of the new worlds of services with the old worlds of transfer. However, research in this field has been erratic and has led to inconclusive results. This article aimed at overcoming existing problems in welfare services research by using welfare sector employment patterns as a proxy. This approach has proven to be very promising as it provided a holistic picture of welfare services comprising their quantity, kind and organization. By giving insights based on new variables, more countries and the most recent data, this analysis was able to contribute to the debate on the role of welfare services in different welfare regimes.

Having shed light on the dark side of the moon, this article was able to reveal that the dark side looks very similar to the bright side. The typology by Esping-Andersen was confirmed not only to apply to the worlds of transfers but to the worlds of welfare services as well. By setting the cluster results in context of the welfare regime literature, a liberal, a conservative and a social democratic welfare regime as well as a group of countries with a rudimentary welfare state could be distinguished. From an ideal-typical perspective, this study was able to show how the different ideological orientations of the welfare states translate into distinctive welfare services regimes. This suggests that the role of the welfare state is a good starting point for a holistic framework of welfare regimes combining the transfer and the service component. The liberal welfare state with its distinction between private and social risks, the conservative welfare state with its focus on the preservation of the social structure and the limited role of the state, the social democratic welfare state with its high level of intervention in order to ensure social equality, and the rudimentary welfare state that lacks the resources for a comprehensive social security system – these are all well-known notions from the social policy literature. However, it has never been explicitly spelt out what this means to welfare services. The findings of this article reveal that the new worlds of welfare services resemble the old worlds of transfers. Adjusting our knowledge on the welfare state in the light of the insights from the worlds of welfare services is likely to generate an improved understanding of welfare regimes in general.


  1. 1

    The understanding of welfare sector employment is based on the standards laid down in the International Standard Industrial Classification (ISIC) Rev. 3 developed by Eurostat, the International Labour Organization (ILO) and the Organisation for Economic Development and Co-operation. ‘Employment in the health and social sectors includes people working in the following groups of the International Standard Industrial Classification (ISIC) Rev. 3: 851 (Human health activities), 852 (Veterinary activities) and 853 (Social work activities). The data are based on head counts, not taking into account whether people are working full-time or part-time. Data for all countries come from labour force surveys, so as to achieve greater comparability’ (OECD 2011a: 60).

  2. 2

    ISCO-88 is a classification of occupational information designed to achieve international comparability. The International Standard Classification of Occupations (ISCO-88) differentiates between ten major groups and 27 subgroups of jobs that are related to four skill levels.

  3. 3

    In concrete terms, the adjusted spending on in-kind benefits has been calculated by subtracting in-kind expenditure for pharmaceuticals, housing, accommodation, survivors and free schemes from the overall level of in-kind benefits.

  4. 4

    F-values by dependent variable: emp_health 23.762, emp_care 29.222, emp_int 9.954, emp_public 19.425, exp_inkind 15.842.