Different types of welfare states? A methodological deconstruction of comparative research.

Research on modern welfare states has been strongly influenced by the theory that they develop according to patterns, which form distinct regimes--liberal, corporatist, and social democratic. These regimes are characterized by several key variables, among which the decommodification of labor is heavily weighted. This article examines the operational assumptions, measures, and calculations used in the most widely cited empirical study around which distinct regime theory has developed over the last decade. The findings reveal critical methodological weaknesses in the conceptualization and quantification of decommodification measures, which form the empirical cornerstone of distinct regime theory.

**********

Theory and research on the development of the modern welfare states tend to be concentrated around two lines of analysis which emphasize either impersonal forces of structural-functional change or the socio-political forces of contending group interests. The structural-functional approach is associated with convergence theory, which posits that over time the programs and policies of welfare states in the advanced industrial countries develop a considerable resemblance to one another. The socio-political approach is associated with distinct regime theory, which holds that there are systematic variations in programs and policies reflected in particular types of welfare states that emerge from different socio-political adaptations (Van Voorhis, 1998).

Among the numerous studies that have advanced distinct regime theory, the most influential contribution to date is Esping-Andersen's (1990) Three Worlds of Welfare Capitalism. The significance of this work is three-fold. First, this impressive quantitative comparative study of 18 countries greatly extended the boundaries of empirical socio-political welfare state analysis--providing a strong quantitative socio-political response to Wilensky's (1975) landmark functionalist study of convergence. Second, the work builds upon the theoretical contributions of Titmuss (1974) and other earlier theorists, and refined the differences between welfare states by identifying three distinct regimes. Third, in focusing on separate welfare state regimes, this study offers an explanation for the causes of welfare state differences.

The distinct regime model identifies a liberal, corporatist, and social democratic paradigm by formulating a systematic comparison of how policies and programs reflect: a) the degree to which labor is decommodified; b) the relationship of entitlements to need, contributions, or citizenship; and c) the type of the public-private mix in social provisions, particularly pensions. The three regimes and their distinguishing characteristics are represented in Table 1 below. In addition to the characteristics in Table 1, the three regimes are seen as creating different systems of social stratification that help to determine and maintain class and status differentiations. In this schema, the Liberal regime is associated with poor relief that maintains class distinctions based on income; the Corporatist regime is identified with contributory social insurance that sustains differentiation based on occupational status; and the Social Democratic regime is linked to middle-class universalism and social equality. Over the last decade this classification of distinct regimes has had a substantial impact on the conceptualization of comparative welfare state research. One of the most recent examples is Goodin, Headey, Muffels and Dirven's 1999 analysis of The Real Worlds of Welfare Capitalism--a rigorous study of three countries in the Esping-Andersen sample identified as liberal (U.S.), social democratic (the Netherlands) and corporatist (Germany) regimes.

The empirical claim to the identification of distinct regimes is supported by quantitative measurement of several key characteristics, among which the most elaborate and systematic analysis is devoted to the decommodification of labor--a widely-cited analysis that provides the clearest case for the clustering of three regimes. An entire chapter is addressed to the discussion and quantification of decommodification, which is operationally defined by an index that measures the relative degrees of decommodification offered by three separate social insurance programs--old age pensions, sickness and unemployment benefits. The concept of decommodification represents the idea that social policies of modern welfare state provide a level of income maintenance, which allows individuals to "opt-out of work", thereby reducing the necessity to sell their labor at any price in order to survive. Hence, social welfare benefits create a buffer against human labor being treated merely as a commodity that can be bought for the lowest price. The distinct regime thesis suggests that different welfare states foster varying degrees of decommodification, which can be measured by examining rules for eligibility, disincentives and benefit levels.

The division of welfare capitalism into three worlds of Liberal, Corporatist and Social Democratic regimes is an ambitious formulation, and as such has drawn various criticisms. The critiques leveled against this formulation usually involve conceptual and theoretical assessments of the broad conclusions from this study, rather than an in-depth review of the empirical analysis (Orloff, 1993; Sainsbury, et. al, 1994; Baldwin, 1996; Sorenson, 1996, Overbye, 1996). It has also been argued that the division of three worlds is based on a relatively small number of subjects (N= 18) and that the data which are utilized may not accurately portray the unique ideological and class compositions of the countries involved (Kvist & Torfing, 1996). This article provides a detailed analysis of the operational assumptions, measures, and calculations used to create the decommodification index, which represents not only the cornerstone concept in the theoretical foundation of distinct regimes, but offers the empirical glue for clustering of these regimes. This analysis is conducted on three levels: 1) the selection and definition of the decommodification indicators within programs 2) the selection and definition of programs scored on these indicators and 3) the calculation of the overall decommodification score--assumptions and alternatives.

Operational Definition: What's Included and What's Left Out?

In constructing the decommodification index several different indicators were employed to operationally define the degree to which old age pensions, sickness, and unemployment programs permitted individuals to maintain a livelihood without reliance on the market. For example in the case of pensions, an index was constructed using the following variables: 1) the replacement rate of the minimum pension benefit for an average production worker 2) the standard replacement rate of a pension benefit for a normal worker 3) the required period of contribution to become fully vested (inversely counted) and 4) the individual contribution rate. The two measures of replacement rates (variables 1 and 2) were then given a double weight in the final pension score. Presumably, the reason for including both the minimum and average pension benefit is to determine the minimum income regardless of what one contributed to the system. A higher level minimum benefit (higher replacement rate) acts as a greater decommodifier of labor than a lower level benefit, or put another way, the less the benefit is tied to employment more it decommodifies labor.

All 18 countries in Esping-Andersen's sample were ranked on each of these four indicators according to a three point scale--1--low decommodification 2--for medium or 3--for high decommodification--in which the designations of high and low are based on scores of plus or minus one standard deviation from the mean. The cumulative value of these scores was then weighted by the percent of the relevant population covered by the pension program, which produced the pension decommodification score. A similar procedure was performed for both sickness and unemployment benefits, using replacement rates and other slightly different indicators appropriate to each benefit.

A critical examination the individual variables included in this operational definition reveals serious flaws. Two of the variables in the pension decommodification index--1) the minimum pension replacement rate ("minimum pension as a percent of normal worker earnings net of taxes") and 2) "the standard pension replacement rate (net)for a single person" (Esping-Andersen, 1990, p. 49)--are such similar measures that the replacement rate variable is essentially counted twice, with each count multiplied by a factor of two. As a result of this procedure, replacement rates effectively account for 66% of the pension score in the decommodification in index and 40% of the two other scores (sickness and employment) in the index. But to what extent does a comparison of replacement rates specify how well social welfare benefits permit workers "to uphold a socially acceptable standard of living independently of market participation?" (Esping-Andersen, p. 37) There are two basic problems with the way this index is operationalized.

First, international comparisons are plagued by problems related to standardization and differences in the real value of the benefit relative to the cost-of-living. The best way to assess the comparative value of pension benefits relative to the cost of living in each country is to operationalize these benefits on the basis of parity adjusted replacement rates (Amzallag, 1995). Offering greater standardization for comparative purposes, these rates are usually calculated using a simple formula of Average Production Worker-wage/Average Pension benefit * Purchasing Power Parities (Whiteford, 1996). By factoring in the purchasing power parities, this measure of replacement rates can be used as a proxy for comparing levels of benefits in different countries and the degree of social protection afforded by different welfare states. In using pension replacement rates that are not adjusted for purchasing power parities, the decommodification index provides a crude measure for comparative purposes.

If the use of soft measures is a critical observation on what is included in the index, the second problem concerns what is not included, which raises a more fundamental issue that challenges the basic validity of this index. The measurement of pensions, for example, does not account for policy variations in the normal retirement age (NRA). All countries base eligibility for pensions on criteria that include the attainment of a minimum age. If country A has a 10% higher wage replacement rate than country B, but requires a worker to be 3 or 4 years older in order to retire--3 or 4 years is more than 10% of the life expectancy when the normal age of retirement is 65--which country's pension policies offer a higher degree of decommodification?

Even if variables such as the normal retirement age were included, a larger issue remains. That is, the replacement incomes furnished through the three public programs--pensions, sickness, and unemployment--ignore a wide range of financial supports related to pensions, sickness, and unemployment benefits that welfare states are increasingly providing through social policies which do not show up as direct public expenditures. As Adema (1999) points out, international comparisons of social welfare that measure only direct cash transfers convey an incomplete view of public social efforts. He suggests it is hazardous to draw serious conclusions about comparative benefits on the basis of gross spending indicators. A full understanding of the scope and value of social benefits provided by modern welfare states requires calculations based on a comprehensive ledger that includes both publicly mandated private benefits as well as voluntary private social benefits.

Publicly mandated private benefits, which usually involve employer payments for absence from work due to sickness and pension contributions to employer-based pension plans, have been legislated in many countries including Denmark, Belgium, Germany, the Netherlands, Sweden, the United States, Norway, and the United Kingdom. Replacement rates that are calculated on public old age pensions create a misleading picture of the real standard of living that retired workers can sustain when publicly mandated private pension benefits are also counted.

In the same vein an accurate accounting of the extent to which social welfare provisions promote the decommodification of labor must include the cash value of voluntary private benefits, such as employer-provided pensions, health insurance, unemployment compensation, severance pay, and sickness compensation. Although these voluntary provisions are not publicly mandated, they are publicly subsidized to varying degrees through tax concessions. Indeed, as shown by Adema's (1999) rigorous calculations, when international comparisons of social welfare transfers involve a comprehensive accounting of direct cash benefits as well as the value of other public measures and subsidizes, there is a remarkable leveling of differences between the traditionally high public Social Democratic countries and Liberal regimes such as the United Kingdom and the United States. Based on the conventional measures, for example, Denmark ranks 1st and the U.S. ranks 12th in gross public social expenditure. But when social spending is adjusted for taxation of benefits, tax expenditures, publicly mandated private benefits and voluntary private benefits, the U.S. moves to 6th place and Denmark falls to 7th place.

Looking closely at how programs in the United States were scored on the decommodification index reveals some of the difficulties of comparative research that attempts to cover a wide and complex range of policies. According to the operational definitions employed in this index, the United States does not provide a cash benefit for Sickness--receiving a score of zero for that category. Esping-Andersen does not elaborate on the definition of "Sickness Benefits". If this program refers to "sick leave" with benefits provided when short-term illness prevents work, arguably many U.S. employers offer the functional equivalent as a part of an employees' tax subsidized fringe benefits. However, the decommodification power of sickness programs is measured empirically by ranking the 1) replacement rates for the first 26 weeks of illness 2) number of weeks of employment required prior to illness to qualify 3) number of days before benefit is paid and 4) length of receipt. Although the exact definition is unclear, these criteria suggest that sickness benefits under consideration involve something more than fringe benefit sick days provided by an employer in the U.S., or payments from European employers for short-term illness (Prins, 1990). In many countries when short-term sickness benefits are exhausted, the employee is then paid a disability benefit (SSA, 1996). Since the duration and level of benefits vary between countries, an argument can be made that three separate U.S. programs, 1) the disability portion of Supplemental Security Income (SSI) 2), the Disability Insurance part of Old Age Survivors and Disability Insurance (OASDI) and 3) Worker's Compensation would accommodate the definition of sickness benefits.

Supplemental Security Income is a social assistance program which provides a cash grant to persons unable to engage in any gainful activity by reason of a disability expected to last for a year or more. There is no stipulation that a program must be social insurance-based in order to qualify for inclusion in the decommodification index. (Australia's means-tested pension program is included, but received a reduced score.) If classification as social insurance were a prerequisite, a second program could have been counted in the index, namely, Disability Insurance, part of Old Age Survivor's and Disability Insurance (OASDI), the United State's largest social welfare program. While the American program of Disability Insurance may be less generous that its European counterparts, the system is similar enough to merit inclusion. Disability benefits are payable to persons who are unable to engage in substantial gainful activity due to a physical or mental impairment, expected to last for 12 months or more (SSA, 1996).

A third program "Workers' Compensation" (WC) also provides sickness benefits for workers who are hurt or injured on the job, or who develop job related diseases. While all but three states in the U.S. have instituted a form of Workers' Compensation, it is not a federal program and the lack of standardization between state programs could be a significant obstacle to developing a measure for an index. Moreover, WC may be more narrow than many sickness benefits furnished by other countries since it is linked to job related injury. However, most states provide relatively high replacement rates--usually two-thirds of the worker's salary, plus coverage for incurred medical expenses. While it may have been difficult to incorporate any or all of these programs into the decommodification index, failure to acknowledge the widespread benefits of short-term sick leave or existence of three separate programs which provide long-term support for people with disabling illnesses imprecisely categorizes the US as a country without a sickness benefit.

Quantification: Inadvertent Weighting

In an effort to quantify the levels of decommodification, Esping-Andersen adds the specific decommodification scores for each of the three programs--pensions, sickness, and unemployment--to arrive at an overall decommodification index score, shown in column three, Table 2. The 18 countries are ranked from low (18) to high (1) levels of decommodification, based on their cumulative index scores. As illustrated in Table 2, these countries divide evenly into three clusters characterized by high (Social Democratic), moderate (Corporatist), and low (Liberal) levels of decommodification, lending empirical support to the distinct regime classification.

An analysis of variance was conducted to assess the distinctiveness of the three groups from a statistical perspective. As noted in Table 3, the Tukey's B procedure supports the distinct regime interpretation in the sense that the mean decommodification scores for each group are significantly different from one another.

However, before drawing too firm a conclusion about how well the results of the decommodification index support the distinct regime thesis, it is worth taking a closer look at the additive method by which the overall index score was quantified. The data in Table 4 show that pension benefits had the highest decommodification scores, averaging 10.7, followed by sickness benefits, which averaged 9.2 and unemployment benefits, which averaged 7.1. If the objective is to compare countries relative to each other on all three program areas, then adding the decommodification scores inadvertently weights pensions (with a mean of 10.7) almost 50% more than unemployment (with a mean of 7.10). Since, as noted earlier, the way pension scores were originally calculated lent decisive value to wage replacement rates, the disproportionate weight allotted to pension scores in the overall index makes the "decommodification index" almost a proxy for wage replacement rates of old age pensions. The problem here is not simply that the pension scores are weighted 50% more than unemployment scores, but that country rankings differ considerably on these two program areas. As illustrated in Table 4, Sweden, for example, ranks 1st on the degree of decommodification in the area of pensions and 10th on unemployment programs.

The logic of using different programs to operationally define decommodification suggests that if the countries being studied actually fell into three distinct clusters (high, medium, and low levels of decommodification) the pattern would be internally consistent across different program areas. According to this logic, the development of an overall index should give equal weight to how the countries cluster on each program. Rather than adding the raw decommodification scores for each program, a more precise approach to weighting the programs equally would be to standardize the raw scores by calculating z-scores for each program (shown in Table 4). Then, by adding the z-scores to compute the overall decommodification index, each of the three component programs (sickness, pensions, and unemployment) contributes equally to each country's position on the final ranking. When the overall index is computed based on standardized scores for each component program, the findings show that the 18 countries no longer divide so neatly into the three groupings of distinct regimes, which was produced by the index based on the raw scores for each program. Indeed, as illustrated in Table 5, while the rank-order changes are not large, these changes are such that exactly one-third of the countries change their group position, shifting either up or down between the Social Democratic the Corporatist regime categories and between the Corporatist and the Liberal regime categories.

Distinct Regimes: Empirical Validation or Heuristic Value?

In conclusion, while Esping-Andersen's ground breaking study offered the first vigorous effort to verify distinct regime theory, there are serious methodological weaknesses in both the conceptualization and quantification of the decommodicifation index, which represented the empirical cornerstone of this study. In examining how this variable was constructed we find that the decommodification index appears to be more a proxy for replacement rates of old age pensions than a broad measure of several dimensions of unemployment, old age and sickness benefits. When the index is adjusted so that unemployment and sickness benefits are weighted the same as old age pensions, a large proportion of the sample countries shift their rank-order position from one regime to another. An appraisal of what is included and what is left out of the operational definition of decommodification, suggests that this measure fails to standardize the real value of benefits (via purchasing power parities) and that the selection of programs from different countries does not represent a comprehensive account of the benefits provided in these areas, which diminishes the ability to make a valid systematic comparison. Thus, despite an elaborate effort to quantify the distinct liberal, corporatist, and social democratic regimes, the empirical analysis falls far short of robust validation.

The inadequate empirical validation of distinct regime theory, however, does not negate its heuristic value as way of thinking about and trying to categorize modern welfare state. Indeed, the analysis of the Three Worlds of Welfare Capitalism has fueled considerable research into welfare state typologies--northern versus southern Europe (Leibfried, 1993, Ferrera, 1996), protestant versus catholic (Castles, 1993) and differences between the Anglo-American welfare states. Responding to Esping-Andersen's categories Leibfried (1992) distingishes among Germanic, Scandinavian, Anglo-Saxon and Levantine (Latin rim countries) regimes. In Three Worlds of Welfare Capitalism--or Four (1991) Castles and Mitchell suggest a fourth type of welfare state--dubbed radical liberal--can be identified which is more redistributive in nature than the liberal welfare state identified in the three worlds typology.

It is important to recognize that the distinct regimes rest on an empirically fuzzy foundation upon which to build other empirical analysis. As noted earlier, Goodin et. al. (1999) apply Esping-Andersen's typology in their study of three countries representing Social Democratic, Liberal, and Corporatist regimes. Selecting the Netherlands as the country representative of Social Democratic regimes, they conclude that the "social democratic regime is the best of all possible worlds," based on a highly detailed (over 40 pages of tables) comparative analysis of outcome criteria for the three countries (Goodin, et. al. 1999, p. 260.). At the same time, however, after reexamining the original typology of Three Worlds Esping Andersen (1999, p 87) concluded that based on a more comprehensive analysis the Netherlands shifts from the Social Democratic regime to "squarely a member of the conservative Continental European fold"--or a Corporatist regime.

Table 1

The Three Worlds of Welfare Capitalism

                                           Degree to Which
Type of      Proto-typical  Philosophical     Labor is
Regime         Countries        Basis       Decommodified

Liberal      Australia      Classical      Low
             Canada         Liberalism
             United States

Corporatist  Austria        Conservative   Moderate
             France         Social Policy
             Germany

Social       Denmark        Socialism/     High
Democratic   Sweden         Marxism
             Holland

Type of       Entitlement      Type of
Regime         Based On     Public/Private Mix

Liberal      Need           Market dominated /
                            residualist

Corporatist  Contribution   State Dominated/
                            Occupational Related

Social       Citizenship    State Dominated/
Democratic                  Universal

Table 2
Overall Decommodification Index and Regime Clusters

                       Decommodification
Country          Rank     Index Score

Australia        (18)         13.0
United States *  (17)         13.8
New Zealand      (16)         17.1
Canada           (15)         22.0
Ireland          (14)         23.3
United Kingdom   (13)         23.4

Italy            (12)         24.1
Japan            (11)         27.1
France           (10)         27.5
Germany          (9)          27.7
Finland          (8)          29.2
Switzerland      (7)          29.7

Austria          (6)          31.1
Belgium          (4/5)        32.4
Netherlands      (4/5)        32.4
Denmark          (3)          38.1
Norway           (2)          38.3
Sweden           (1)          39.1
                          Mean = 27.2

Source: Esping-Andersen, G. (1990).
The three worlds of welfare capitalism.
Princeton, NJ: Princeton University
Press, pp. 52.

* According to the additive approach
described by Esping-Andersen the U.S.
score should be 14.2

Table 3
Analysis of Variance: Decommodification Index Groupings

                                    Mean
           (I) Group  (J) Group  Difference  Std. Error  Significance

Tukey HSD    1.00       2.00      -8.8000 *     2.102        .002
                        3.00     -16.4667 *     2.102        .000

             2.00       3.00      -7.6667 *     2.102        .006

* The mean difference is significant at the .05 level

Table 4
Raw and Standardized Decommodification Scores for Old-Age
Pensions, Sickness Benefits and Unemployment Insurance

                   Pensions          Sickness          Unemployment

                     Raw                Raw                Raw
               Rank  DS *    Z    Rank  DS *    Z    Rank  DS *    Z

Australia      (16)   5.0   -6.7  (15)   4.0  -1.29  (16)   4.0  -1.62
Austria        (6)   11.9   0.36  (3)   12.5   0.84  (11)   6.7  -0.21
Belgium        (2)   15.0   1.27  (10)   8.8  -0.09  (4)    8.6   0.78
Canada         (13)   7.7  -0.88  (14)   6.3  -0.79  (7)    8.0   0.46
Denmark        (2)   15.0   1.27  (1)   15.0   1.46  (6)    8.1   0.52
Finland        (4)   14.0   0.98  (7)   10.0   0.21  (13)   5.2  -1.00
France         (5)   12.0   0.39  (9)    9.2   0.01  (12)   6.3  -0.42
Germany        (12)   8.5  -0.64  (5)   11.3   0.54  (8)    7.9   0.41
Ireland        (15)   6.7  -1.17  (11)   8.3  -0.21  (5)    8.3   0.62
Italy          (9)    9.6  -0.32  (8)    9.4   0.06  (14)   5.1  -1.05
Japan          (8)   10.5  -0.05  (13)   6.8  -0.59  (15)   5.0  -1.10
Netherlands    (7)   10.8   0.04  (6)   10.5   0.34  (1)   11.1   2.08
New Zealand    (10)   9.1  -0.46  (15)   4.0  -1.29  (6)    4.0  -1.62
Norway         (3)   14.9   1.24  (2)   14.0   1.21  (2)    9.4   1.19
Sweden         (1)   17.0   1.86  (1)   15.0   1.46  (10)   7.1  -0.01
Switzerland    (11)   9.0  -0.49  (4)   12.0   0.71  (3)    8.8   0.88
UK             (12)   8.5  -0.64  (12)   7.7  -0.36  (9)    7.2   0.05
United States  (14)   7.0  -1.08  (16)   0.0  -2.29  (9)    7.2   0.05

Mean                 10.7                9.2                7.1
S.D.                  3.4                4.0                1.9

* Decommodification Score

Calculations based on data from: Esping-Andersen, G. (1990). The
three worlds of welfare capitalism. Princeton, NJ: Princeton
University Press, pp. 50. For scoring procedure see Three worlds,
pp. 77-78.

Table 5
Rank-Order of Welfare States Based on Alternative Computation of
The Decommodification Index

                       Decommodification     Rank       Z
                Rank         Score        Based on Z  Score

Australia       (18)         13.0           (18)      -4.59
United States   (17)         13.8           (16)      -3.33
New Zealand     (16)         17.1           (17)      -3.38
Canada          (15)         22.0           (13)      -1.13
Ireland         (14)         23.3           (11) *    -0.77
United Kingdom  (13)         23.4           (12) *    -0.96

Italy           (12)         24.1           (14) *    -1.31
Japan           (11)         27.1           (15) *    -1.74
France          (10)         27.5           (10)      -0.02
Germany         (9)          27.7           (8)        0.31
Finland         (8)          29.2           (9)        0.19
Switzerland     (7)          29.7           (6) *      1.10

Austria         (6)          31.1           (7) *      0.98
Belgium         (4/5)        32.4           (5)        1.96
Netherlands     (4/5)        32.4           (4)        2.45
Denmark         (3)          38.1           (3)        3.25
Norway          (2)          38.3           (1)        3.65
Sweden          (1)          39.1           (2)        3.32

                          Mean = 27.2
                          S.D. = 7.7

* Countries which change grouping when rank based on Z score

Source: Esping-Andersen, G. (1990). The three worlds of welfare
capitalism. Princeton, NJ: Princeton University Press, pp. 52.

References

Adema, W. (1999) Net Social Expenditure. Labour Market and Social Policy Occasional Papers No. 39. Paris: Organization for Economic Cooperation and Development, Table 3.

Amzallag, J. (1995). Replacement ratios: Comparability and trends. Twenty-fifth General Assembly, Nusa Dua: ISSA.

Baldwin, P. (1996). Can we define a European welfare state model? In Greve (Ed.) The Scandanavian Model in a Period of Change. Basingstoke: Macmillan.

Castles, F. G. (1993). On religion and public policy: Does Catholicism make a difference? Working Paper, Public Policy Program, Australian National University, Canberra.

Castles, F. G. & Mitchell, D. (1991). Three worlds of welfare capitalism or four? Luxembourg Income Study, Working Paper 63. Luxembourg: CEPS/INSTEAD

Esping-Andersen, G. (1990). The three worlds of welfare capitalism. Princeton, NJ: Princeton University Press.

Esping-Andersen, G. (1999) Social foundations of postindustrial economies. New York: Oxford University Press.

Ferrera, M. (1996). The "southern model@ of the welfare state in social Europe. Journal of European Social Policy, 6, 1, 17-37.

Goodin, R,. Bruce Headey, Ruud Muffels, and Henk-Jan Dirven, (1999) The Real Worlds of Welfare Capitalism. Cambridge, England: Cambridge University Press.

Kvist, J. & Torfing, J. (1996). Changing welfare state models. CWR Working Paper 5, Centre for Welfare State Research.

Leibfried, S.(1992) "Towards a European Welfare State? On Integrating Poverty Regimes into the European Community" in Zsuzsa. Ferge and Jon Kolberg, eds. Social Policy in a Changing Europe. Boulder: Westview Press.

Leibfried, S. (1993). Towards a European welfare state? In P. Jones (Ed.) New perspectives on the welfare state in Europe (pp. 135-56). London: Routledge.

Orloff, A. S. (1993). Gender and the social rights of citizenship: The comparative analysis of gender relations and welfare states. American Sociological Review, 58, 304-328.

Overbye, E. (1996). Pension politics in the nordic countries: A case study. International Political Science Review, 17, 1, 67-90.

Prins, R. (1990). Sickness absence in Belgium, Germany (FR) and the Netherlands. Masstricht: NIA.

Sainsbury, D. (1994). Gendering the welfare state. Security Bulletin, 50, (1), 5-11. London: Sage

Sorensen, A. M. (1996). Gender equality, family decline and social exclusion in modern society. Paper presented at the European Science Foundation Conference on Social Exclusion and Social Intergration in Europe, Blarney, March, 1996.

Titmuss, R. M. (1974). Social policy: An introduction. London: George Allen & Unwin.

Social Security Administration (1996). Social security programs throughout the world--1995. Washington, D.C.: Social Security Administration.

Van Voorhis, R. A. (1998). Three generations of comparative welfare theory: From convergence to convergence. European Journal of Social Work 1 (2) 189-202.

Whiteford, P. (1996). The use of replacement rates in international comparisons of benefit systems. International Social Security Review, 48, 3-30.

Wilensky, H. L. (1975). The welfare state and equality. Berkeley: University of California Press.

REBECCA A. VAN VOORHIS
California State University
Department of Sociology & Social Services

Rebecca A. Van Voorhis Department of Sociology & Social Services California State University Hayward, CA 94542-3048 [e-mail]