Three (Two? Four? Five?) Worlds Of Welfare Capitalism (US Public Spending In Context-Interlude)

by: Paul Rosenberg

Sat Apr 11, 2009 at 18:30

Although attempts to analyze and categorize welfare state began as far back as the 1950s, the commonly-recognized watershed work was the 1990 book, The Three Worlds of Welfare Capitalism by Gosta Esping-Andersen. While others have criticized his work, and more different types have been suggested, there seems to be no discarding of his basic accomplishment, only additions and revisions, whereas earlier efforts have primarily been mined for spare parts.  In this diary, I'm going to focus primarily on a 2001 paper that contains a much more sophisticated re-analysis that ends up largely confirming the Three Worlds hypothesis in one sense, before exploring another sense in which two of the worlds appear to meld, leaving only two distinct worlds.  I will also make some reference to other approaches.  My basic perspective is that there is no one "right" model of how many "worlds" there are, but that the "Three Worlds" model is particularly useful for relating the US to how mature Western European systems work.

The paper I'm drawing on is Luxembourg Income Study Working Paper No. 276, "Measuring Welfare State Performance: Three Or Two Worlds Of Welfare Capitalism?" (PDF) by Paul de Beer, Cok Vrooman and Jean Marie Wildeboer Schut.

As I mentioned in the previous diary, the Luxembourg Income Study is a  cross-national data archive and research institute, which just so happens to be located in Luxembourg. The first LIS working paper was published in 1985. The  LIS Database includes income microdata from a large number of countries at multiple points in time. A second, more recently-started database LWS, contains wealth microdata.  It is one of the major economic datasources in the world, and a crucial focus of international comparative studies. I'm going to be drawing on several other LIS working papers in this diary and the next.

Section two of "Measuring Welfare State Performance" starts off with a description of the detailed analysis underlying its categorization of welfare states, which ends up broadly confirming the original "The Worlds" analysis.  I'm going to quote extensively, first from the introductory passage, then from its descriptions of the original "Three Worlds" welfare system types, and then from the much more carefully calibrated descriptions used in the paper itself.

A Detailed Analysis

The paper relies on principle component analysis (Wikipedia description here), which allows a large number of measurements over a large number of subjects to be reduced to a small number of categories into which all the subjects can be placed.  (In this case, the authors surveyed 11 countries with 58 characteristics to come up with their 3-fold grouping.) It's the analytical tool used by Pew, for example, in coming up with its typology of political groupings.  PCA does not predetermine in advance how many groupings there will be.  However, it's sometimes as much an art as a science, coming down to a matter of judgment which possible grouping should be chosen.  On the other hand, often there is only one viable choice, though it's often possible that a small number of subjects won't be accounted for.  (Here, the Netherlands doesn't fit, sharing characterstics of two different categories--although in a somewhat structured way.)  In this case, it simply means that the researchers were not assured in advance of finding some three-fold categorization.  The fact that they did so came from the data, rather than any initial supposition.

Broadly, the authors explain:

The application of the non-linear principal component analysis (Princals or 'principal components analysis by alternating least squares') results in an optimal scaling of both the countries and the institutional characteristics, in which countries that share a lot of characteristics get similar scores on the dimensions, while countries that have not much in common are placed far apart. Figure 1 shows the scaling of both the countries and th institutional characteristics.

I reproduce a somewhat altered form of Figure 1  below, after the descriptions of the original "Three Worlds" typology.  For clarity's sake, I have taken the lists of characteristics which were part of the figure, and placed them into three text blocks immediately following the figure.

More specifically, the authors say:

This section examines the extent to which countries may indeed be clustered into three types on the basis of the institutional features of their welfare state.

This seeks to provide empirical support for the three Esping-Andersen welfare state types. To this end a principal components analysis has been conducted on 58 institutional characteristics of eleven western welfare states, namely Australia, Belgium, Canada, Denmark, France, Germany, the Netherlands, Norway, Sweden, the United Kingdom and the United States. The choice of these countries is based on their central position in Esping-Andersen's classification, and on the availability of comparative empirical data on income distribution needed for the further analysis in this paper (see below).

The institutional characteristics in terms of which the countries have been compared with one another concern the funding of the welfare state (including tax rates and social security contributions), the regulation and structure of the labour market (including the minimum wage, collective agreement coverage rate, labour force participation of women and older people and the share of the service sector in employment), and the level and coverage of the social security provision for old age, surviving dependants, children, disability, unemployment and social assistance.

Here, then are their decriptions of the welfare state clusters:

Social-Democratic (Socialist) Welfare States

Sweden, Denmark and Norway form a cluster of social-democratic welfare states. Within this cluster Sweden turns out to share its position as a country which most closely approximates the ideal-type of a social-democratic welfare state with Denmark. Norway has a less pronounced social-democratic regime. These social-democratic welfare states are characterized by a comprehensive system of social protection. The high level of social services is revealed by, among others, the high replacement rate for unemployed families with children and the relatively high level of the old age and survivors pension schemes. These generous social security benefits result in high average and marginal tax rates. The social-democratic countries also spend large amounts on active labour market policies. The labour force participation of women is high, which is facilitated by generous parental leave schemes.

Conservative (Corporatist) Welfare States

Germany, Belgium and France form a cluster of corporatist welfare states, which it is not possible to rank in terms of 'purity'. Characteristic for the corporatist welfare states are the generous unemployment and disability insurance schemes for (former) employees. The position of the traditional male breadwinner is also safeguarded by the protective services for children and parenthood, e.g. relatively high child benefits and long-term pregnancy, childbirth and parental leave. In the field of labour relations the collective agreement plays a central role. Both the labour participation of women and of older men is much lower than in the social-democratic countries.

Liberal Welfare States

The United Kingdom, Canada, Australia and the United States form a liberal Anglo-Saxon cluster. As expected, the United States turns out to be the predominant liberal welfare state, followed at a short distance by Australia. The UK and Canada are at a significantly greater distance from the liberal ideal-type, and also share certain institutional characteristics with the social-democratic welfare states. The social security systems of the liberal countries can in many respects be characterized as 'residual'. The replacement rate of most benefits is relatively low and they have a limited duration. Much emphasis is put on means-tested social assistance schemes. A distinguishing feature of the collectively financed social security for the old age is the relatively low fl at rate benefit. There are few public facilities for children or parentage. The labour market is hardly regulated at all. Although not many public resources are spend on active labour market policies, the participation rate of women is higher than in the corporatist countries.

And here is the chart, showing how the countries clustered as a result of their analysis.  In the original, the top left dotted line pointed to the list of "Corporatist Features", the dotted line at the bottom left pointed to the "Social Democratic Features," and the dotted line at the middle right pointed to the list of "Liberal Features."

CORPORATIST FEATURES:
Labour market:
- moderate active labour market policy (0.5 - 0.8% GDP)*
- low labour participation rate of women aged 25-54 (<65%)*
- low labour participation rate of men aged 55-64 <60%)
- low labour participation rate of men aged 65+ (<5%)
- high coverage of collective labour agreements (>90%)
- low share of employment in services sector (<65%)
Disability:
- risque social: coverage restricted to (certain) employees
- risque social: moderate threshold (65-79%)*
- risque social: wage related benefits
- risque professionel: full wage compensation
Funding:
- largely financed by contributions of employers/employees (>50%)
- no tax allowances for non-working spouses
Surviving dependants pensions:
- no collective widow insurance for all inhabitants*
Occupationalism:
- frequent arrangements for particular occupational groups
- high level of social protection for civil servants (>$500 per capita)
Costs of children:
- relatively high family allowances for non-poor couples
- relatively low family allowances for poor single parent families
- high tax allowance for children
Leave facilities:
- long duration of parental, pregnancy, and childbirth leave (incl. non-paid period>1,5 y)

SOCIAL-DEMOCRATIC FEATURES:
Labour market:
- intensive active labour market policy (>1.1 % BNP)*
- high labour participation rate of women aged 25-54 (>75%)*
- large share of employment in health/social work (26-30%)
- no statutory minimum wage
Disability:
- risque social: low threshold (<65%)*
Unemployment and social assistance:
- high net replacement rate after 5 years unempl., families with
children (>71%)
- high net replacement rate after 5 years unempl., single parent families (>69%)
Funding:
- high marginal tax rate breadwinners (62-64%)
- high average rate of income tax and social security contributions (>26%)
Old age pensions:
- high minimum level of public benefits non-ex-employees
- high coverage of earnings-related occupational pension schemes (100%)
Surviving dependents pensions:
- public insurance widows: high flat rate benefit*
- no separate public insurance for widows of former employees
Costs of children:
- relatively low family allowance poor couples with children
Leave facilities:
- long duration of earnings-related parental, pregnancy, and childbirth leave (>26 weeks)

LIBERAL FEATURES:
Labour market:
- little active labour market policy (<0,8% GDP)*
- moderate labour participation rate women (65-74%)*
- large share of employment in wholesale and retail business (>22%)
- large share of employment in financial services (>11%)
Disability:
- risque social: high threshold (>80%)*
Funding:
- no tax exemptions on labour (at the wage level of an average production worker)
- low yield of taxes and social security contributions (<38%)
- no or low employers contributions for social security (max. 8%)
- low marginal tax rate single persons (<43%)
- low marginal tax rate double earners (<42%)
Unemployment and social assistance:
- short duration of earnings-related unemployment insurance (non means-tested) (max. 1 year)
- low net replacement rate at the start of unemployment for wage of 2/3 average production worker (60-69%)
- low net replacement rate at the start of unemployment for wage of average production worker (60-69%)
- low net replacement rate after 5 years unemployment, single persons (<39%)
- low net replacement rate after 5 years unemployment, couples without children (<50%)
- high share of means-tested social assistance in total social security spending (>15%)
Old age pensions:
- public benefits are partially or completely means-tested
Surviving dependants pensions:
- public insurance widows: low flat rate benefit*
- public benefits are partially or completely means-tested
Costs of children:
- low level of public provisions
- low average family allowances (< €140)
- family allowance income-related or means-tested
- relatively low family allowance for non-poor single parent families
Leave facilities:
- low level of pregnancy benefit (<70% of earnings)
- no pregnancy leave for all employed women

Two Worlds?

Significantly, the largest portion of the paper was devoted to an analysis of income redistribution, income inequality, social welfare, and poverty, each in a separate section.  When that analysis was completed and combined, the authors found that there was no discernible difference between the social democratic and the conservative welfare state models--or the Netherlands, which shared characteristics of both.  This result lead to the notion of "Two Worlds"--the liberal welfare state, and all the rest.

Although no clear line can be drawn between the social-democratic and corporatist welfare states in terms of their performance, there is a clear line between these two types and the liberal welfare states. With the exception of social welfare and 'absolute' poverty on the basis of the Dutch statutory minimum, the liberal welfare states are nearly always at the bottom of the table....

Since no systematic differences between the social-democratic and corporatist welfare states emerge from the analysis, the Netherlands as a 'hybrid' welfare state is not so much in between these two types as in the midst of a joint group of social-democratic and corporatist welfare states....

This conclusion is certainly no reason for rejecting Esping-Andersen's typology. As noted earlier Esping-Andersen did not draw up his typology so much with a view to the traditional income-protection and redistributive functions of the welfare state as to the correlation between the welfare state, social stratifi cation and the labour market. This paper, it should be stressed, was not designed to investigate the relevance of Esping-Andersen's typology for these aspects of a modern capitalist society. If one is particularly interested in the extent to which welfare states fulfi l the traditional functions of income protection, combatting poverty and reducing income disparities, however, it would be more accurate to speak of two instead of three 'worlds of welfare state capitalism'.

There are multiple ironies here for American conservatives.  First of all, conservatives everywhere like to speak both of universal truths and principles, and of respect for "organic" local and national traditions.  But these two principles are clearly in conflict, since European conservative welfare states do a great deal of the sort of things that American conservatives just love to rail against.  Indeed, this analysis shows that they are just as enthusiastic as their more openly socialist (aka "social democratic") brethren.

A deeper irony is that since the social democratic and conservative clusters have virtually indistinguishable outcomes overall, the differences between them are primarily cultural/institutional and normative/purposive.  Unfortunately for American conservatives, this is precisely where US and European conservatives diverge the most sharply, since European conservatives are quite the statists, although they do share with American conservatives the desire to sustain the position of elites, and the distinctiveness of group indentities.    It is, by contrast, the social democratic welfare states that have not, traditionally, been motivated by a desire to strengthen the national government, a feature which Esping-Gosta noted most clearly, for example, in the high levels of benefits provided for government employees in conservative welfare states such as Germany and Austria.

But, of course, it's the liberal welfare states, including the US, which are most to the Anerican conservatives' liking, as they do the least amount of providing for people's welfare outside the scope of the marketplace.

Four Worlds?

Following the publication of The Three Worlds of Welfare Capitalism in 1990, there was a vigorous debate in which other divisions were suggested, particularly focused on some regions that it was argued ought to be considered separately.  One of the strongest arguments was for a "Mediterranean" model (Spain, Italy, Greece) in which families played a much larger role, and spending levels remained lower.

LIS Working Paper No. 498, "The Struggle between Equity and Efficiency: Evidence from the Luxembourg Income Studies" is an example that draws on the four-worlds model, using slightly different nomenclature, and it produces graphs such as the following: (which will be discussed in the following diary):

Five Worlds... And More?

A fifth "world" that was proposed early on was the separate "world" of newly-developing Asian welfare states, which also seem to share similar characteristics, in terms of smaller size and greater dependence on family, though they differ in other regards.  However, as time has gone on, and more attention has been paid, differences within the Asian "world" have increasingly been noted.  A recent paper taking this approach to a high level--"Clusters Representing Process: Welfare Capitalism in East and Southeast Asia" by Myoung-Shik Kim--uses a similar analytical technique to identify three different "worlds" within the Asian model.  

It seems clear to me that this is an evolving approach that will provide much better insight over time into how welfare states can be compared, contrasted and evaluated, taking us fr beyond the primative state of "debate" commonly seen here in America.  How many "worlds" there are may well be a question with no one answer.  Much more useful is the framing of questions in terms of a relevant range of parameters of interest, and an analysis that helps provide deeper insight into how those parameters interact by studying who different states form into natural groupings with distinctly different sorts of outcomes.