271wep02.qxd

17/12/03

11:52

Page 20

Swiss Worlds of Welfare K L A U S A R M I N G E O N , FA B I O B E RTO Z Z I A N D GIULIANO BONOLI

Are there different worlds of welfare in Swiss cantons, analogous to different national welfare state regimes? Are the welfare regimes made up of functionally related sub-regimes in the field of employment, education, taxation and social security, as argued in recent analyses of ‘varieties of capitalism’? And can the variations between such cantonal welfare regimes be explained by the same politico-institutional variables that account for the development of national welfare states? These are the guiding questions of this article. We find strong empirical evidence for large inter-cantonal variation in welfare state policies. The sub-regimes are not functionally connected, however. Their variation is related to different sets of variables, indicating a weak functional link between them. Developments and causal structures are specific to the various subregimes. In addition, it is socio-economic and not politicoinstitutional variables that have had the strongest imprint on the overall structure of cantonal welfare regimes. Like most federal countries, Switzerland has developed over the years a multi-tiered system of social protection, where financing, decision making powers and implementation are shared between the different levels of government. The present arrangement is the result of one and a half centuries of social policy making, which has been characterised by an ever increasing role for the federal government. Core policy areas like old age pensions, unemployment insurance or health insurance have been transferred from the cantonal to the federal government, so that in 1998 approximately 58 per cent of national social expenditure was distributed by the federal government. Cantons, however, retain substantial room for manoeuvre in several important areas of social policy: for instance, social assistance is entirely decided upon at the cantonal level. Income tax is collected separately (and at different rates) by the three different levels of government: federal, cantonal and communal. Family policy, social services and education are also areas of cantonal sovereignty (Tables 1 and 2). West European Politics, Vol.27, No.1 (January 2004), pp.20–44 PUBLISHED BY FRANK CASS, LONDON

271wep02.qxd

17/12/03

11:52

Page 21

21

SWIS S WO RLDS OF W ELFARE

TABL E 1 R E S P O N S IBI L I T I E S OF CANTONS AND F E DER ATIO N IN TH E FIELD O F SO C IA L S E CURI T Y I N 2000

Programme

Type of Programme

Legislation

Implementation*

Old age and survivors insurance Disability insurance Supplementary benefits

Federation Federation Federation

Cantons Cantons Cantons

Unemployment insurance Accident insurance Health care insurance Family allowances

Universal insurance Universal insurance Universal insurance/ means-tested Social insurance Social insurance Universal insurance Social insurance

Cantons Federation Private health funds Cantons

Unemployment assistance Social assistance

Means-tested Means-tested

Federation Federation Federation Cantons (Federation)** Cantons Cantons

Cantons Cantons

Notes: * In the implementation column we only mention the administrative level that holds the main responsibility for the concrete implementation of the scheme, regardless of the financing level. ** Family allowances for farmers are under the responsibility of the federal level, for the other categories of workers family allowances are under the responsibility of the cantons. Source: Obinger 1998, modified. TABL E 2 S O C I A L EXP E NDI T URE P E R ADMI NI S T RATIV E LEV EL, 1998, A S % O F TO TA L

Total Social Expenditure Old age insurance (a) Disability insurance (a) Health care and health care insurance Other social insurances (b) Social assistance 1 (wide definition) (c) Social assistance 2 (narrow definition) (d)

Federation

Cantons

Municipalities

All

58.4 84.5 75.0 64.4 42.3 31.0

27.3 12.4 19.8 29.2 40.8 35.5

14.3 3.1 5.3 6.4 16.9 33.5

100.0 100.0 100.0 100.0 100.0 100.0

0

41.8

58.2

Notes: (a) Public contributions, without employers’ contributions. (b) Includes old-age and disability complementary benefits, unemployment insurance, accident insurance and family benefits funds. (c) Includes assistance to old people, help to needy people, labour offices, help to Swiss in foreign countries and other assistance tasks. (d) Includes only help to needy people, 1995. Source: BSV, Schweizerische Sozialversicherungsstatistik, 1999 and 2001.

On the one hand, the increasing role played by the federal state, the small size of the country and the extent of economic and cultural exchange within the country may lead us to expect an overall low level of variation in

271wep02.qxd

22

17/12/03

11:52

Page 22

WE S T E U R O P E A N P O L I T I C S

the extent and shape of social protection at the cantonal level. On the other hand, within the policy areas that have remained under their sovereignty, the room for manoeuvre of cantonal governments is subject to few constraints. Cantonal social laws must conform to the federal constitution and, if a federal law exists on the subject matter, must not go against it. We can thus expect a high degree of cantonal variation in these policy areas. As a matter of fact, empirical analysis of the cantonal fiscal and social policies in the 26 cantons that make up Switzerland reveals profound differences in the instruments that are used and in the size of cantonal welfare states. Tax loads and tax scales vary considerably. The lowest taxload is found in the canton of Zug, the highest (more than twice as high) in the canton of Neuchâtel (BFS 2001: 801).1 Tax progression (taxation index for highest/lowest income class) in the canton of Vaud is almost six times as high as in the canton with lowest progression, the canton of Obwalden (Vatter et al. 2002).2 Cantons like Argovia or Obwalden know of no cantonal welfare scheme, while a canton like Geneva has six schemes in place (Wyss 1999).3 Considering these variations, it is arguable that some cantonal welfare states follow a liberal minimalist approach, other are more generous and egalitarian in their orientation, and yet others have a clearly conservative orientation, based on a traditional vision of the family. Hence some of the variation of Western European welfare states could be reproduced within Switzerland itself. Demands for some form of fiscal harmonisation are regularly put forward in the political arena (essentially by left-wing political actors) but have so far failed to be transformed into policy measures. Against this background, this article has two main objectives. First, it provides an account of the variation that exists among cantonal welfare states within Switzerland. It does this by applying to Swiss Cantons the framework developed by Esping-Andersen (1990) for the comparison of national welfare states. Second, it looks for explanations for the observed pattern. Theories that have been developed by students of social policy in relation to the development of national welfare state are found most useful in explaining the pattern of cantonal variation. By looking at social policy in the member states of a federal country, this article takes forward the debate on the relationship between federalism and social policy. In fact, even though a substantial amount of literature exists today on this theme, most of it is actually concerned with social protection at the federal level. Systematic comparisons of policy at the level of the member states are to a large extent missing. Yet the room for manoeuvre that some federal constitutions allow in social policy means that the living conditions and life experiences of some individuals can be very different in different part of the same country.

271wep02.qxd

17/12/03

11:52

Page 23

SWIS S WO RLDS OF W ELFARE

23

In the case of Switzerland, this analysis is of crucial importance. In international comparisons of welfare states, Switzerland has long been classified as one of the least developed, or a ‘welfare laggard’ to use the jargon of social policy students. It is only in recent years that Switzerland has caught up to the median Western European welfare state (Armingeon 2001; Bonoli and Mach 2000; Lane 1999). Yet this view of a liberal or conservative welfare state with strong liberal traits is based essentially on the observation of national social programmes only. In fact, these programmes, in some cantons more than in others, are supplemented by others, and together they impact upon the living conditions and life experiences of Swiss residents. Analysing cantonal welfare states in the same manner as national welfare states raises the question of whether theories and techniques are not stretched in an inappropriate way. We believe that the welfare regime approach can be helpfully applied to the study of cantonal social protection systems. Cantons have different political systems, social structures and political legacies. There is no reason why these independent variables should not have similar effects on cantonal social policies as they have on the national level on the welfare state of a country. The article starts with some theoretical issues (section 2) and moves on to present the research design and databases used in this study (section 3). Our results are then presented, first in relation to the variation that exists among cantonal welfare states (section 4), and second as an attempt to explain such variation (section 5). T E R M S , T H E O RY A N D H Y P O T H E S E S

In using the term ‘worlds of welfare’ we refer to the typology developed by Esping-Andersen (1990). Although we apply the distinctions between a liberal, a social democratic and a conservative-Catholic welfare regime, we do not suggest that empirical reality is close to these ideal types,4 in particular in differentiating three blocs of cantonal welfare systems of which each is rather homogeneous internally. Rather, we argue that, for example, the ‘social democratic’ elements in a given canton are stronger than in another canton. In other words, we assume three ideal types of welfare states: in a liberal welfare state, size of transfer payments are limited and public schemes are geared to provide the last resort to people in need (safety net), giving not much more than minimum of existence. Private insurance against risks of modern society is given preference over public schemes, and government tends to favour subsidies to private schemes instead of running its own public scheme. In a Catholic-conservative welfare state there are massive transfers with the aim of maintaining the status which the

271wep02.qxd

24

17/12/03

11:52

Page 24

WE S T E U R O P E A N P O L I T I C S

beneficiary has acquired while being in employment. The Catholicconservative welfare state is oriented towards the maintenance of the family with one (typically male) breadwinner, in particular if he or she is a Swiss national and not foreign. The social democratic welfare state also has massive transfers. However, it does not aim at status preservation, but at equality among all members of the population, irrespective of gender, former employment status or nationality. Worlds of welfare may also be linked to the goals of party families. The social democratic left has been in favour of the socialist welfare state; the prime example being the Scandinavian countries (Castles 1982, 1999; Korpi 1983; Esping-Andersen 1990; Schmidt 1998). Catholic-centrist parties have been the major forces bringing about the continental welfare state with its emphasis on the traditional family structure and status preservation (Kersbergen 1995). Liberal parties support a marginal (or lean) welfare state, prime examples for this correlation being found in the Anglo-Saxon countries, and in particular the USA. Our analysis of cantonal variations in social polices is based one these theoretical approaches, which were developed as a result of empirical research comparing nations. We start by identifying four distinct dimensions that differentiate between different worlds of welfare, and then move on to review likely explanations for the observed pattern. Dimensions and Explanations of Worlds of Welfare The typology developed by Esping-Andersen focuses on the core schemes of the welfare state, pensions, unemployment insurance, sickness insurance, and so forth. The level and the kind of benefits delivered by these schemes constitute our first dimension, which we call social security. Our second dimension is also based on the institutional structure of redistributive mechanisms, but instead of focusing on the benefits provided, it deals with the financing of public expenditure. It is labelled ‘taxation’. These two dimensions are used in most comparative research on welfare states. However, recent political discourse in most Western societies and academic debates, in particular with regard to the programmatic development of social democratic parties, have pointed to two additional dimensions of social policy, closely linked and interacting with social security schemes: education and employment regimes, which we also wish to include in our analysis. Education is important since it can make people largely independent from welfare state provisions like unemployment benefits. Hence investment in human capital reduces the need for social security expenditure. In addition, increasing human capital fosters economic productivity and hence competitiveness. Recent research has shown that left-wing parties have chosen the strategy of building up human capital by educational and public

271wep02.qxd

17/12/03

11:52

Page 25

SWIS S WO RLDS OF W ELFARE

25

sector expansion, while liberal and conservative parties tried to increase economic growth by low taxes and low public investment (Boix 1997, 1998). A similar finding has been presented by Frank Castles (1999: 174–85). In addition to the impact of social democratic incumbency, he finds a difference between liberal – mostly Anglo-Saxon – countries and Catholic-conservative countries, to which nearly all continental European countries belong. In the 1970s and even 1980s, the Anglo-Saxon countries have spent more on education than the continental European nations. However, by the 1990s this difference has dwindled. Looking at the historical development, one could argue that the Catholic-conservative countries have been the most resistant to expansion of public education, while the left has been most favourable to increasing budgets for education. Finally, the employment regime is the fourth dimension we use to distinguish among the cantonal worlds of welfare. The term denotes the patterns of employment and unemployment and the patterns of wage regulation. We argue that it may substitute public social policies or create challenges to these schemes. Cases in point are the former Australasian systems of collective arbitration, guaranteeing a ‘decent standard of living’ to everyone (Castles 1985). On the other hand, conflictual labour relations produce industrial unrest, whose outcomes (e.g. unemployment) have to be covered by public policy. Hence it comes as no surprise that employment patterns are linked to types of welfare. This is clearly shown in analyses by Iversen and Wren (1998) or Jochem and Siegel (2000), relying on ideas developed by Scharpf and Schmidt (2000) and Huber and Stephens (2001). Recently Hall, Soskice and others have argued that labour relations, employment patterns, skills formation and welfare state policy are closely linked together (Hall and Soskice 2001; Estevez-Abe et al. 2001). Empirically, this link is by no means clear-cut, as demonstrated by Alber (2000) for employment patterns or Daly (2000) for the case of gender differences in employment. However, there are good theoretical and empirical arguments for assuming that social democratic welfare states foster a combination of low unemployment, high employment for both sexes through adequate labour market policies and generous child care provisions. In contrast, conservative-Catholic welfare states support employment regimes with high employment of males, low employment of females and generous schemes which allow families with children to rely on only one breadwinner. Liberal welfare states are characterised by high rates of labour force participation for both sexes without providing much public childcare or similar means. The functional equivalent is low taxes and low social security contributions making the employment of child carers affordable to parents with good professional/vocational training. While the absence of high unemployment is considered crucial for the social

271wep02.qxd

26

17/12/03

11:52

Page 26

WE S T E U R O P E A N P O L I T I C S

democratic welfare state, it is more acceptable in Catholic-conservative states – in particular if it does not hit male breadwinners – and in liberal states. In the latter, unemployment is considered to be the result of malfunctioning labour markets in which it does not make sense to intervene and in which government is not allowed to intervene. Analysing cantonal welfare states along these four dimensions, we not only take into account the most recent state of research in comparative welfare policy analysis (Hall and Soskice 2001; Scharpf and Schmidt 2000; Huber and Stephens 2001), but we are also able to contribute to the debate on the interconnectedness of these four dimensions, using a most similar systems comparative design. The extension of Esping-Andersen’s (1990) typology to these other dimensions is straightforward, since according to the argument policies in these dimensions are causally if not functionally related to each other (see Hall and Soskice 2001; Estevez-Abe et al. 2001). Western societies have not only developed different systems of social security, but have followed substantively different trajectories in the fields of employment and education as well. As we shall see below, this divergence can be observed among Swiss cantons. In order to explain it we make reference to the recent research on the development of the national (see Esping-Andersen 1990; Schmidt 1998; Castles 1999; Pierson 2000) and cantonal welfare states (Vatter 2002; Freitag 2000a, 2000b). On this basis, we can identify the following explanatory variables: (a) The strength of left (social democratic), centre (Christian democratic/Catholic) and liberal/right political parties. Each of these parties supports the expansion of ‘its’ welfare state. However, with one exception,5 cantonal governments are broad coalitions where the major means of decision making are negotiation and mutual accommodation. In 1998, three cantons were governed by a five party coalition. In nine states, there was a government made up of four parties and in 13 cantons three parties shared power (Vatter 1998; Germann 1999: 406–8). Hence party competition in the sense of policy seeking is constrained within cantonal governments. However, the relative strength of the three major political forces – social democrats, Christian democrats and liberals – in the cantonal electorate might be important for the willingness of the other political parties to make concessions. In other words, due to the broad coalition in government, we cannot expect a ‘pure’ type of welfare state reflecting the preferences of one political party. On the other hand, the ability to lean towards a particular type may be conditioned by the relative strength of the particular political camp that supports that type.

271wep02.qxd

17/12/03

11:52

Page 27

SWIS S WO RLDS OF W ELFARE

27

(b) Institutions that foster welfare state expansion. The size of government coalitions is of importance in this respect. The broader the coalition, the more likely are bargains between the various partners including compensatory payments for the clientele of the parties. Hence the chances for a lean welfare state to diminish with the increasing size of government coalitions. (c) Institutions that constrain welfare state expansion. Direct democracy is such an institution (Immergut 1992; Obinger 1998; Bonoli 1999). There are various arguments for this effect. One observation is that in principle it is easier to have a majority against compared to a majority for a new policy that might be redistributive, and that future advantages are less visible than the costs incurred doing the reform. A more elaborate argument points to different logics of population and politicians. Politicians tend to increase spending in order to be reelected; citizens tend to restrict spending since they know that they have to pay it via taxes (Kirchgässner et al. 1999). The more direct democracy is used in a canton, the better the chances for a liberal welfare state. (d) Deep-seated cultural values of a given society. Comparative policy analysis has pointed to Catholicism as a major force impacting on welfare states (Castles 1994, 1999). Hence we expect that in Protestant cantons (i.e. cantons with a large share of Protestants) the liberal and socialist welfare state has the highest likelihood to flourish, while in Catholic cantons the development of a conservative welfare state is most likely. (e) Needs and opportunities. These are general conditions of public policy: the more affluent and economically advanced a region, the more it can afford to nurture a costly welfare state. On the other hand, the more pronounced the needs for a welfare state the larger welfare expenditures. In the context of Switzerland, four such needs are prominent: the share of elderly, who need health care and – sometimes – social assistance (Ergänzungsleistungen); the number of unemployed who need support by active and passive labour market policy; the number of foreigners, who usually are at the bottom of the income ladder and hence often in need of welfare state support; and finally the needs produced by urban living conditions. The more urbanised a canton, the less likely it is that family networks can function as substitutes of public social security and health care.

271wep02.qxd

17/12/03

11:52

Page 28

28

WE S T E U R O P E A N P O L I T I C S

D ATA A N D R E S E A R C H D E S I G N

Our dependent variable is the type of cantonal welfare state in the late 1990s, which we determine with reference to our four policy-based dimensions: social security, taxation, education and employment. For each policy area, we assign every canton to one of Esping-Andersen’s regimetypes. In order to assign a canton to a given welfare regime, we use indicators or set of indicators discriminating between one ideal type of welfare and the two other types. We derive these indicators from theoretical reasoning with regard to the underlying goals of liberal, conservative and socialist political forces and from the findings of empirical comparative research, being able to show that in international comparisons these political goals in fact have resulted in different types of welfare states. Our indicators are unidimensional (e.g. higher or lower social expenditure), so how do we discriminate between the three worlds of welfare? The basic idea is to combine two indicators or sets of indicators denoting two different sub-dimensions. These sub-dimensions have been chosen so that they distinguish one type of welfare regime from the remaining two. By intersecting the two sub-dimensions, we obtain a fourfold classification where three cells correspond to the three worlds of welfare. The fourth cell is a residual type for which on theoretical reasons there should not be empirical instances. However, if we relax the assumption of just three worlds of welfare, there might be very different mixes of different policies constituting cases that cannot be neatly attributed to one of Esping-Andersen’s types of welfare state. Technically, we use as the cut-off point the arithmetic mean of the respective variable. In most cases we have more than one indicator for one sub-dimension. For example, in the case of social security the number of schemes might be as important as the expenditures. Provided these indicators were significantly correlated (p.05. Left parties: Social democratic, green, democratic-social parties Centre parties: Christian-Democratic parties and Union of Independents (Landesring der Unabhängigen) Right parties: Radicals, Swiss People’s Party

(.03)**

Employment

Protestantism

Explanatory Variables

TA B LE 6 E XP LA NATO RY P OWE R O F C U LTU R A L, IN STITU TIO N A L A N D SO C IO -EC O N O MIC VA RI A BLES O F TH E SWISS WO R LD S O F WELFA R E (N A G ELK ERK E R -SQ U A RES)

11:52

Institutional

17/12/03

Socioeconomic

271wep02.qxd Page 37

271wep02.qxd

38

17/12/03

11:52

Page 38

WE S T E U R O P E A N P O L I T I C S

Table 6 offers an overview of the general results of the analyses. Hardly any systematic correlation can be found for the employment regime. It is not dependent on cultural, institutional or political variables. Similar results are obtained for taxation. However, there we have a clear institutional effect: direct democracy exerts an important influence on the type of taxation regime.9 Even when we control for the impact of unemployment – the other significant explanatory variable – in a multivariate model, the overall effects remain significant. The more popular votes in a canton, the lower the likelihood of a conservative or social democratic regime of taxation. This finding supports an institutional interpretation of worlds of welfare as far as the taxation dimension is concerned. In the field of public revenue the impact of party families is very weak, while the extent to which parliamentary decision is superseded by popular vote is of prime importance. This fits nicely into theories stating that all parties tend to expand public revenues in order to finance public services, and so to improve their chances of re-election. In contrast, citizens are primarily interested in low tax loads (Steinmo 1993; Kirchgässner et al. 1999). In cantons where voters have a strong (direct) say, tax loads will therefore tend to be smaller than in cantons where parliaments operate under fewer plebiscitary restrictions – irrespective of the political make-up of the governing coalition. On the dimension of education, the strength of the left in government and the size of coalition have an impact on the type of education: the stronger the left, the greater the likelihood of high educational expenditures, and similarly, the larger the size of the governing coalition, the greater the likelihood of higher spending on education. Adding successively the other significant variables as controls in a multivariate model, the effects remain significant in the case of the strength of left-wing government (except if urbanisation is the second variable), while the size of the coalition becomes insignificant (except if the strength of the left is used as a control). Although our findings are not very robust, they do fit in with theories according to which left-wing parties try to steer on the supply side, in particular with regard to human capital and public infrastructure. This supply-side effect of welfare state policies has frequently been overlooked in recent comparative research (though see Boix 1997, 1998; Castles 1999), where the focus has been on demand side policies such as fiscal measures and social security schemes. As in the analysis of the dimension of social security, the sole political effect emanates from the distinction between left and non-left governments. In does not matter whether government is composed of conservative-Christian (i.e. mostly Catholic parties) or right-liberal parties (the Radicals and the Swiss People’s Party). An obvious explanation points to the distinction between centrist parties favouring state intervention and

271wep02.qxd

17/12/03

11:52

Page 39

SWIS S WO RLDS OF W ELFARE

39

right-liberal parties fighting for the free play of market force. By its very origin, the Swiss People’s Party has been a classic protectionist party (Junker 1968), and it has remained so until today with regard to selected protectionist measures for farmers. Programmatically, the Swiss Radicals is not a coherent liberal party, but is composed of cantonal sections with different ideological orientations (Ladner 1999). In historical perspective, the radicals never consistently supported the free play of market forces; in addition they have been in strong support of conservative-statist and protectionist policies (Gruner 1964). For the explanation of dimensions of systems of social security, leftwing government, direct democracy and the size of coalition produced significant results in the expected direction. Once we controlled for the robustness of these findings by including all other significant independent variables in a multivariate model, these political and institutional variables continued to produce significant results only in two or three of the possible six combinations: size of coalition became insignificant when direct democracy, urbanisation, unemployment or numbers of foreigners were added. Direct democracy lost significance in the combination with left-wing government, urbanisation and size of coalition. Finally, left-wing government was insignificant in tandem with direct democracy, urbanisation, unemployment and numbers of foreigners. Urbanisation is obviously a major socio-economic explanatory variable for cantonal social security systems. Left-wing power is strongly and positively correlated with urbanisation (.64): the more urban a canton, the higher the share of left-wing parties in government. On the other hand, the more rural a region, the better the odds for centrist parties (correlation with urbanisation: -.54). In contrast, the power of right-liberal government is not significantly related to urbanisation. Hence, one could argue that urbanisation is the major background variable explaining both worlds of welfare and the political strength of the left in Swiss cantons. CONCLUSIONS

Are there different worlds of welfare in Swiss cantons, analogous to different national welfare state regimes? Are the welfare regimes made up of functionally related sub-regimes in the field of employment, education, taxation and social security, as argued in recent analyses of ‘varieties of capitalism’? And can the variations between such cantonal welfare regimes be explained by the same politico-institutional variables that account for the development of national welfare states? These have been the guiding questions of this article. We find strong empirical evidence for large intercantonal variation in welfare state policies. The sub-regimes are not

271wep02.qxd

40

17/12/03

11:52

Page 40

WE S T E U R O P E A N P O L I T I C S

functionally connected, however. Their variation is related to different sets of variables, indicating a weak functional link between them. Developments and causal structures are specific to the various sub-regimes. And, in addition, it is socio-economic and not politico-institutional variables that have had the strongest imprint on the overall structure of cantonal welfare regimes. Although socio-economic variables – in particular urbanisation – turn out to be the major force shaping cantonal ‘worlds of welfare’, we also have indications of significant influence of institutional and political variables. In the field of taxation direct democracy is an important force shaping the cantonal taxation regime. In education, left-wing political parties support the expansion of education system. In the field of social security, it is once again the left and direct democracy that have an impact. But we have to admit that the robustness of these findings is limited. It is noteworthy that if there is a major political effect on the worlds of welfare in Swiss cantons it is between the left and all other parties. The differences between the Christian-democratic and right-liberal forces have no significant impact on the welfare state policies of these parties on the cantonal level. To what extent can the findings of this paper be generalised to other democratic nations. We see three contributions: first, there are huge differences in cantonal welfare state policies that have an impact on the people’s living conditions. We would not expect such strong differences to exist inside Germany or Austria, simply due to constitutional rules which would not allow for such variation. Hence, in analysing social policies in federal states, we need to specify which type of federalism pertains in a particular country.10 Probably a useful distinction is that between federalism (as in the US, Switzerland or Spain), unitary federalism (as in Germany and Austria) and unitary systems (as in France). Considering the wide difference of state welfare policies in Switzerland and Germany, we hypothesise that the major dividing line does not run between unitary and both types of federalist system, but between federalism on the one hand and unitary systems and unitary federalism on the other. Second, this study completes existing analyses of the impact of federalism on social policy, which usually focus exclusively on the effects on the federal level of social policy making. For instance, in the case of Switzerland, most studies (e.g. Cattacin 1995; Kriesi 1995; Obinger 1998) identify the delaying effect of federalism on the development of a federal welfare state. We add another perspective: in federal states like Switzerland, the welfare state is the outcome of the interaction between federal and state policies, whereby state policies are not only of marginal importance. Third, we find that there is a missing correlation between the various dimensions of the welfare state. Our study does not support the idea that a particular system of social security corresponds to a particular system of

271wep02.qxd

17/12/03

11:52

Page 41

41

SWIS S WO RLDS OF W ELFARE

taxation, employment or education. This corresponds to the finding by Alber (2000) and suggests the need for a more differentiated analysis of dimensions of the welfare state. Probably in the field of employment, markets and cultural variables are of utmost importance, whereas in taxation it is institutions that play a particularly strong role. The role of politics might be restricted to education and social security. ACKNOW LEDG E ME N T An earlier version of this paper was presented at the workshop ‘Federalism and the Welfare State – Comparative Perspectives on the Old and the New Politics of the Welfare State’, Hanse Institute for Advanced Study, Delmenhorst, 24–27 May, 2002, and benefited from comments made by workshop participants. To a large extent this paper is based on a data set gathered by Adrian Vatter, Markus Freitag and Christoph Müller (University of Berne, 2002). We are grateful to these colleagues for generously sharing their database with us. This paper was written in the context of the research project ‘Convergence towards a unique “continental” model of welfare state in Europe? The importance of regional integration for Swiss welfare state reform’. The project started in February 2001 at the Institute of Political Science at the University of Berne and is supported by the Swiss National Science Foundation in the framework of the Swiss Priority Program Switzerland: Towards the Future. NOTES 1. The chosen indicator is the global index of fiscal charge and refers to the year 1999. Whereas in Zug this index represents 58.7% of the inter-cantonal average, in Neuchâtel it reaches 125.3% of the average. 2. The tax progression index is provided by Vatter et al. (2002). Data refer to year 2000. The tax progression index is obtained by dividing the taxation index (cantonal, communal and church tax on incomes) for the highest income class (900,000–1,000,000 Swiss francs) by the taxation index for the lowest income class (7,500–30,000 Swiss francs) as a percentage of the Swiss average (=100). Whereas in the canton of Vaud this index is 2.73, in the canton of Obwalden it is 0.46. 3. Social assistance excluded. 4. Esping-Andersen (1990) himself admits the non-existence of pure cases in empirical reality and the general lack of purity of the empirical models, since they often combine elements from different ideal-typical regimes. 5. Only one political party, the Christian Democrats, governs Appenzell Inner-Rhodes. However, the party system in this canton is at an embryonic stage; there are hardly differentiated and organisationally consolidated party organisations. Hence the notion of one-party government is not appropriate (Steiner 1969). 6. The variable has been calculated as follows: The z-values of child benefits for the fourth child and of birth grants have been added and then z-standardised. From this z-scores the zscores of the variable ‘expenditures on pre-school facilities’ have been subtracted. This variable indicates the weight given to the goal of sustaining traditional family forms compared to the weight given to liberal and social democratic goals of giving working women the right to have a family. 7. The standard measure for the heterogeneity, the standard deviation (sd), has one crucial shortcoming: the absolute magnitude of the standard deviation increases linearly with the mean of the variables. In order to compare several groups with respect to their relative homogeneity this has to be taken into account. A suitable technique is the calculation of the coefficient of variation (CV), being the standard deviation divided by the mean. 8. Correlation matrix (Cramers’-V *: p