How useful are typologies of welfare states?
Given the complexity and depth of questions faced by undergraduate students of Social Policy, let alone academic researchers, the classification of different welfare states into typologies is only natural. As an aid to academic study, the formulation of typologies represents a logical progression in the field of welfare state research.
When Titmuss gave his lecture on ‘Social Administration in a changing society’ in 1951, he noted that at the start of the study of Social Policy (or administration) in 1913, most issues discussed were entirely contemporary, designed to prepare those who wanted to work in the charitable of voluntary field. After World War Two, Titmuss and other academics such as Marshall, started laying greater theoretical foundations – examining the welfare state that was growing around them and its motivations.
The intensive study of ‘typologies’ of welfare states, widely recognised as being prompted by Esping-Anderson’s ‘The three worlds of Welfare capitalism’ (1990), is a natural progression – an attempt, in search of a fundamental theory or law, to bring together earlier theoretical work on the welfare state and its foundations. The use of typologies is useful because it provides a tool that can be used to simplify complex comparative work, thus making it easier to work on a universal theory of welfare state formation and development.
Similarly, typologies can also provide ‘an approximate picture of the range of options available to policy-makers for policy reform’ (Bonoli, 1995, 352). However, despite these apparent attributes, the concept of using typologies as a methodological tool has not been entirely uncontested. Those such as Baldwin have disputed that typologies have any explanatory power whatsoever, claiming that they most certainly cannot be useful in the formulation of theories about current and future development of welfare states (Arts & Gelissen, 2002).
Most criticism along these lines deals with the apparent impossibility that, given the massively varied complexity of welfare states across the world, no typological theory is capable of summarising these differences except in a deeply misleading way. Esping-Anderson (1990, 2) fully accepts that a ‘trade-off’ must be made, which this means that it will be impossible to make ‘detailed treatments’ of differences between regimes, but asserts that this is the ‘price to be paid for making grand comparisons’. This seems commonsensical to me.
We simplify the levels of detail about economic activity by creating an assumption of ‘homo-economicus’ and even though we know this model not to be completely accurate, no-one would deny that the results the model produces are not instructive. Similarly, in creating a model of welfare-state typologies, we look to gain insights into the nature of welfare states, which we can then use in further theoretical work. The ideal types used in typologies are, as Klant (1984, cited in Arts and Gelissen, 2002) notes, a ‘representation of a reality, which cannot (yet) be described using laws’.
As long as we recognise that the typologies themselves are not ever claimed to be, even by their proponents, complete explanatory theories, then they can usefully be used to gain insights into the past, present and future of welfare states. Therefore, I will move to look quickly at the different varieties of typological survey that have been conducted, before then spending some time comparing the different sorts of ‘welfare state regimes’ that emerge. Different typological classifications of welfare states vary in both the amount, and kind, of variables used for analysis.
The most basic attempt to categorise welfare states has been to compare the levels of public expenditure between countries. For example, it is posited that the Scandinavian countries, all spending in excess of 30% of GDP on social expenditure in the period 1989-92, constitute a type of welfare state regime significantly different to the kinds experienced in the United Kingdom or in Italy, countries which both experience much lower levels of social spending.
However, such crude classification seems to go against any conception of varied mixed economies of welfare – while two countries may appear identical in terms of their % GDP social spend, they may be completely different on the level of overall welfare provided by the other welfare ‘pillars’ – the voluntary and family sector. Similarly, other, more recent models have also tried to classify welfare states using a one-dimensional approach.
Ferrera (1993) uses the ‘coverage model’, which, rather than looking at the quantity of welfare provided, instead looks at the method of provision – specifically whether welfare states provide services on a universal basis, or on an occupational basis. However, as Bonoli (1995) points out, the problem arises in distinguishing between the generosity of benefits offered by welfare states of the same ‘method’ typology – for example, the Netherlands and Switzerland are both included in the Ferrera’s same set of welfare states regimes, despite massive differences between the extent of welfare provision between those two states.
Esping-Anderson’s (1990) approach was two-dimensional in the sense that it considered two factors – the degree of decommodification and social stratification. These variables are explicitly outcome, rather than means, based – decommodification, for example, is defined as ‘the degree to which individuals or families can uphold a socially acceptable standard of living independently of market provision’. Thus, independence is deemed identical, whether achieved through state or voluntary provision.
This practical approach implies a value-judgement that it does not matter if states actually provide the welfare or not, that it is just outcomes that matter, which may be controversial to some socialists. However, this must be considered justified as a modelling assumption. Despite praise for the two-dimensional nature of Esping-Anderson’s approach, some questions have been raised about whether the correct two variables were chosen.
Bonoli (1995), raises concern that what he perceives as being the two fundamental distinctions between different kinds of welfare states – ‘how’ and ‘how much’ – are merged in both of Esping-Anderson’s variables, and that as a result, they are both ‘taken into account only so far as they affect the decommodifying [or stratifying] impact of social policy’. He therefore suggests two different variables – level of social spending and the method (measured in terms of the % of benefits that are contribution-based) of welfare delivery.
Similarly, most other criticisms of Esping-Anderson have not disputed the methodological use of typologies, but instead have sought to question the variables used to measure the relative congruence of different welfare states in order to form sets of welfare regimes. While requirements for brevity make it impossible for me to outline all here, a useful summary can be found in the tables of Arts & Gelissen’s 2002 article. While the indicators/factors used in typology classification vary widely, the extent of similarity in the output of empirical testing is striking.
All models find three or four different typologies, which, rather than being completely distinctive, are similar in terms of characteristics and the country assignments. To start, let us examine Esping-Andersen’s typology, which distinguishes between three clusters of welfare regimes, each of which represents a different ‘world of welfare capitalism’. The three ‘ideal-typical’ regime types are the liberal, the conservative (or corporatist), and the social-democratic regimes.
The main characteristic of the liberal welfare regime is the important role assigned to the market and the strong emphasis placed upon individual responsibility. In general, only if the market fails will the state interfere. Benefits are means-tested and targeted for low-income dependents, who accordingly make the intensive use the welfare state. Financed by taxation, this leads to significant income redistribution. In contrast, the state is generally passive with regards to gender issues – the market determines the position of women, who are neither encouraged nor discouraged by the government to work for pay.
In the conservative/corporatist welfare regime less stress is placed on the role of the market. The regime’s main goal is to preserve existing status/class differentials, a task which is often left to other non-state actors such as the church, classes and the family. They play a crucial role in society, and the state supports them in this role. Social benefits are more elaborate than in liberal welfare regimes and are usually organized along occupational lines, with negligible amounts of redistribution.
With regard to the position of women, the breadwinner model (husband as breadwinner, wife as caregiver) is implied and perpetuated through the design of the welfare system, which thereby systematically discourages women from participating in the labor market. Finally, the social-democratic welfare regime distinguishes itself from the other two models in that much more emphasis is laid on the interventionist role of the state. The idea of equal rights for all citizens is guaranteed by the social democratic state, which delivers extensive, universal welfare services not usually subject to means testing.
The welfare regime is designed around active-labour policies – both men and women are expected to participate fully in the labor market, and therefore the government makes arrangements to overcome any gender-obstacles to doing this, for example, by providing extensive free childcare. Leibfried (1992) and Ferrera (1996) initially criticized Esping-Anderson’s group of typologies for the omission of what they called a ‘Latin-Rim’ model of the welfare state, characterized by strong family-centric properties and an immature and selective social security system that offered poor benefits and lacked a guaranteed minimum benefit system.
Esping-Anderson (1999) accepted that more emphasis should be put on the family-effects of welfare regimes, and made a distinction between familalistic regimes on the one extreme (e. g. , Italy), and defamilialising regimes on the other extreme (e. g. , Sweden). A familialistic regime was defined to be where “public policy assumes that households must carry the principal responsibility for their members’ welfare”, as opposed to a de-familialising regime “which seeks to unburden the household and diminish individuals’ welfare dependence on kinship” (1999, 51).
Given the late inclusion of a ‘Latin-Rim’ model in Esping-Anderson’s work, we have four examples of welfare-state regimes that we can compare with other models. Bonoli’s (1997) classification, based on the expenditure and method factors as described earlier, produced four ideal-types – Continental, British, Nordic and Southern – to which he assigns countries through the empirical work. These clearly correspond respectively with Esping-Anderson’s Conservative, Liberal, Social-Democratic and Latin-Rim typologies.
The British regime is characterized by a low level of social expenditure which is mainly financed through general taxation rather than contributions – this fits in with Esping-Anderson’s description of the Liberal regime as being limited in scope and mildly redistributive in nature. Similarly, Bonoli’s Nordic regime, with low levels of contributions but high levels of overall spending, brings to mind Esping-Anderson’s social democratic regime, characterised by universal benefits for all on the basis of citizenship and not contribution.
There are minimal differences in state-placement between the two writers’ typological systems. The Netherlands and Belgium are perhaps the only significant variation, placed in the ‘continental’ typology by Bonoli but in the ‘Social Democratic’ by Esping-Anderson. Such disagreements about the placement of the Benelux states are repeated in various other schemes of classification – such as in work by Korpi and Palme (1998, cited in Arts & Glissen, 2002) and by Visser and Hemerijck (1997, cited in Arts & Glissen, 2002).
Such variability of outcome on ‘border cases’ between typologies returns us to the original debate about the methodological justification and accuracy of such explanatory systems, and thus it seems to be a sensible place to conclude this essay. It would not be right, though Esping-Anderson (1999) has tried, to try and plead the ‘unique’ nature of these ‘rogue’ states – the initial simplification and assumptions made by any typological system are done so on in full knowledge that detailed individual characteristics, the ‘uniqueness’ of each regime will be obscured.
Rather than attempting to remove previously imposed assumptions, we must accept the limits of typological classification. Ideal-type classifications are exactly that – ideal, and there will undoubtedly be some states that are ‘hybrid’. The continued existence of such apparent ‘anomalies’ gives purpose to further theoretical work – to examine the interactions and movements between the different ‘ideal types’ of states, and to identify whether there is any trend that will make the occurrence of such a hybrid regimes more common and explainable in the future.