THE VARIETY OF SOCIAL CAPITAL IN WELFARE STATE REGIMES – A COMPARATIVE STUDY OF 21 COUNTRIES Open Access

Only a little over a decade ago Gösta Esping-Andersen (1990) laid the foundation for a comparative research tradition of welfare state regimes, in all of which the spheres of responsibility of the state, the local community, family and relatives in producing social security vary remarkably. The relationships of formal institutions and informal communities both of which produce social security are notably different. Particularly from the point of view of informal social security, it is reasonable to ask whether the formal social security system generates and feeds the informal social security systems or, on the contrary, avoids the development of the informal social security produced by families, friends and local communities. In this context the concept of social capital is important: it refers to the social resources that are important in the formation of well-being of individuals (Field 2003: 44–70; Halpern 2005: 73–112).

As a concept social capital is made of many elements and is therefore incoherent. Nevertheless, its most essential features seem to be those emphasising social networks, social trust and norms of interaction. On the basis of their findings in the United States, both Coleman (1988) and Putnam (2001) have been concerned about the erosion of social capital, which is manifested in the dissolution of social networks and as a whole as a declining significance of civil society. In recent years discussions have arisen in which the connections of the welfare state and social capital have been addressed. In these discussions the relation between the welfare state and social capital seems unclear: on the one hand, it can justifiably be argued that the institutions of the welfare state erode social capital at the level of social networks (Habermas 1973; Offe 1984; Wolfe 1989; Etzioni 1995; Fukujama 2000). This is sometimes called the crowding out hypothesis (Oorschot and Arts 2005). On the other hand, it is at least as justifiable to assume that the institutions of welfare states evoke feelings of trust and security within their citizens (Hall 1999; De Hart and Dekker 1999; Siisiäinen 1999; Freitag 2001; Rothstein 2001; Sheepers et al. 2002; Rothstein and Stolle 2003; Oorschot and Arts 2005).

One thing seems be clear, however. There is still only very little empirical research conducted on how social networking, informal social support or trust are connected to the way in which states participate in producing social security. There are exceptions to this naturally. Scheepers et al. (2002), for example, have studied the extensiveness of family ties and bonds of friendship among elderly people in various welfare states. Oorschot and Arts (2005) have tested the crowding out hypothesis on the European welfare states in a way which comes close to our approach.

This article is organized as follows. First we clarify what we mean by social capital and introduce the variables we used to measure social capital. After that we place our observations in the discussion on welfare state regimes and introduce our regime classification, that is, the countries included in our research by regime. We want to find out how the amount of social capital varies by welfare state regime and discuss what kinds of regime characteristics this variation could be connected to. At the final stage of our analysis, we build multi-level models to differentiate the individual-level and country-level variance of social capital indicators.

Robert Putnam's definition crystallizes the central elements of social capital that are repeatedly present in the literature of the field: ‘[–], social capital refers to connections among individuals – social networks and the norms of reciprocity and trustworthiness that arise from them’ (Putnam 2001: 19).

Putnam's definition is very close to the definition developed by the OECD that has been taken as a point of departure in extensive research projects measuring social capital by British and Australian statistical authorities, for example. According to the OECD definition, social capital is a ‘network together with shared norms, values and understandings that facilitate co-operation within or among groups’ (Healy and Cote 2001). Social capital is often used to refer to social networks that are ‘personal relationships which are accumulated when people interact with each other in families, workplaces, neighbourhoods, local associations and a range of informal and formal meeting places’ (ABS 2000; Harper 2003). On the other hand, social capital is also used to refer to trust and the norms of reciprocity.

Social networks lay the foundation for social interaction, which in turn brings about trust, expectations of norms emphasising reciprocity, and co-operation. There are many kinds of social networks, however. In the discussion it seems an established tradition to divide social capital in terms of networks into three groups (see, e.g., Putnam 2001; Woolcock 2001; Harper 2003):

• -

  bonding (or exclusive) social capital

• -

  bridging (or inclusive) social capital

• -

  linking social capital

The concept of bonding social capital refers to strong ties among family members, friends or tightly structured ethnic or religious groups. These kinds of networks bind their members tightly together and they can be efficient in producing social support and control but they can also be exclusive towards other people outside the network. Strong ties can also be suffocating from the viewpoint of the network members. Bridging social capital refers to weaker social networks; they are not as tight and dense as bonding networks but they enable broader interaction and provide opportunities to develop social interaction that benefits individuals. For example, social networks that emerge in working life, non-governmental organizations or free time hobbies are often weak but nevertheless more useful because they offer a channel for free social interaction. By the concept of linking social capital Woolcock (2001) refers to the kinds of social relationships that emerge between those with various positions of power and citizens. Examples of this are the various public and private service organizations in which the service providers and customers meet.

In addition to social networks, the concept of social capital is closely connected to the concepts of trust, reciprocity, values and norms. These concepts are usually used to refer to the subjective consequences of social networks: functioning networks enable the emergence of reciprocal norm expectations and social values. They also further the flow of information. Coleman (1988) in particular has given emphasis to the primary importance of social networks in the emergence of reciprocity and trust. In dense networks, people make observations on each other's activities and maintain the rules of behaviour through sanctions and rewards. The normative order creates predictability and trust in the activities. Communities oriented towards intensive social interaction evoke trust in fellow-men and, on the other hand, increase co-operation among people. However, if bonding social capital is in too dominating position in the community, it can cause conflicts between too inwardly oriented groups. It is therefore evident that there has to be certain equilibrium between different types of networks. In comparing welfare state regimes it might be interesting to deliberate the question how bonding, bridging and linking capital shape into social capital and how the different sub-areas of social capital are interconnected.

In the discussion on measuring social capital, one common observation comes across: social capital is a multidimensional phenomenon (see, e.g., Grootaert et al. 2003). Unfortunately this is where the unity ends. In practice, however, all empirical designs repeatedly mention at least different social networks (both bonding and bridging) and generalized trust, and often also different forms of civic engagement. In this article, we had in our use the ISSP 2001 (International Social Survey Programme), the theme of which was social networks in particular. This data covers variables which are relevant as measures of four dimensions of social capital: bonding social capital, bridging social capital, generalized trust and social support. The formation and contents of the variables we used will be described in greater detail in connection with the presentation of the results.

According to Esping-Andersen (1990: 26, 32; and 1999: 35, 74), welfare state regimes are systematic complexes of legal and organizational functions regulating the relations between the state, the economy and the households, which have emerged and differentiated as a result of historical development. Esping-Andersen names three ideal-types: a conservative, liberal and social-democratic welfare state regime. The states of the real world correspond to these ideals to a variable extent: they are hybrid or mixed forms (Arts and Gelissen 2002: 139).

According to Esping-Andersen, welfare state regimes differ in two basic dimensions: the degree of de-commodification and the type of stratification/solidarity. De-commodification refers to the degree of freedom from market dependency, to the extent to which social-political benefits are social rights independent of markets (and family relations). Stratification refers to how extensive social political systems are, to the universality of benefits. It is typical of the social-democratic regime to have a high degree of de-commodification and extensive universalism. The de-commodification of the liberal regime is low and its universalism limited. By its de-commodification and stratification the conservative regime can be placed between these two. Stratification of the liberal regime leads to a growing difference between the well-off and the not-well-off. In the conservative regime solidarity is corporatistically differentiated and covers only part of the population. The system protects the benefits of the well-off. The aim of the social-democratic regime's stratification is to decrease the differences between citizens and to maintain individual rights.

Later on it has often been suggested that Esping-Andersen's division of regimes into three is not sufficient and that there are more than three regimes. One addition to the regime types is the welfare state regime of the Mediterranian (South European, Latin Rim) countries (Liebfried 1992; Ferrera 1996; Bonoli 1997; Arts and Gelissen 2002: 142–6). Its degree of de-commodification is typically low and universalism limited, the influence of Catholicism is great (apart from Greece), and it is characteristically familist (cf. Gough 2000: 131–42). Including countries outside Western Europe and North America has made it necessary to consider how these new countries fit in the earlier regimes. One wide discussion deals with the Antipodean countries, Australia and New Zealand. They have been characterized as countries of the liberal regime type (Castles 1996), typical of which are also a low degree of de-commodification and limited universalism focusing on minimum security and means testing (Castles and Mitchell 1993; Castles 1996, 1998; Esping-Andersen 1996,1999: 75; Arts and Gelissen 2002: 146–7). Problems of the same kinds surfaced when the study expanded to concern Asia and other parts of the world.

From the point of view of the study, one essential problem concerning categorization is into which division of regimes the earlier socialist countries fall. Do they form one or more than one regime or do they fall into the West European division of regimes into three? We are dealing with a large geographical area where countries are very different from each other (Heidenreich 2003, Ferge and Juhàsz 2004; Manning 2004; Potucek 2004). In the socialist period they were yet surprisingly uniform in terms of their social policy and social security (Deacon 2000: 147). Their staple foodstuffs were subsidized, rents low, differences between the minimum wage and the average wage were small and the social benefits were therefore close to the wage level (ibid.: 147; Oorschot and Arts 2005: 12). After the collapse of socialism, unemployment and inflation have eroded the real value of social benefits, and the World Bank recommends as one means of solving financial problems the dissolving of the old social security system (Deacon 2000: 159). Differences in development between the countries are great. Despite this, we can say that after the transition period in most former socialist countries, universalism is still widespread although hanging in the balance, and because of inflation the degree of de-commomodification is low in these countries. They thus differ from liberal welfare state regimes so distinctively that it feels appropriate to group them into a separate post-socialist welfare state regime. The same solution has been arrived at in another, earlier study (Oorschot and Arts 2005: 12).

We will thus be dealing here with five welfare state regimes: liberal, conservative, social democratic or Nordic, Mediterranean and post-socialist welfare state regimes. Our data cover 21 countries; in fact, there are 22 cases because we divided Germany into its western and eastern parts and placed the parts in different regimes. The liberal welfare state regime is represented by Great Britain, the United States, Canada, Australia and New Zealand (Castles 1996; Esping-Andersen 1999: 90); the conservative welfare state regime by Austria, France, Switzerland and western Germany. The Nordic welfare state regime includes Norway, Denmark and Finland. Unfortunately Sweden, known as the flagship of the social-democratic regime, is not included in the data. The Mediterranean welfare state regime is represented by Spain, Italy and Cyprus. The East European countries, eastern Germany, Hungary, the Czech Republic, Slovenia, Poland, Russia and Latvia are grouped as the post-socialist welfare state regime.

Research on the social capital of the welfare states seems to be in an interesting phase. Two research lines deviating from each other seem to have emerged. Both are based on Robert Putnam's (2001) ideas but arrive at nearly opposite hypotheses. The first interpretation, which we call liberalist, suggests that the relationship between social capital and public social security is reverse: the more extensive and developed the social security of the welfare state regime, the smaller amount of social capital there is within the regime (see, e.g., Scheepers et al. 2002). The other interpretation, which could be called social-democratic, suggests that an extensive and developed social security can be an indication of a great amount of social capital. For example, Kumlin and Rothstein (2003) draw attention to the fact that social-democratic welfare state regimes do not behave according to the liberalist hypothesis. In them, the relationship between social security and social capital is positive. A developed social security can be an indication of solidarity and the kind of social capital that springs from citizens’ trust in the political system.

The first-mentioned hypothesis seems to lean on a liberal, Anglo-American glamorised concept of community, in which the public and the private are set against each other and in which genuine privacy is based on private interdependencies (cf. McLean et al. 2002: 8). The fact that these hypotheses are opposite to each other may be caused by the differences in the conceptions of state. In liberal thinking, the state and the individual are set against each other. The state limits the individual and the functioning of civil society. In the social-democratic line of thinking, the state is seen to provide individuals with rights and to increase individuals’ possibilities to act. In addition, it may be that the idea of liberal social capital is relatively instrumental and functional in assuming that interaction first and foremost aims at creating social security. It may nevertheless be the case that interaction and the social capital it produces emerge in a situation in which social security has been guaranteed by public administrative measures. Interaction is also always of intrinsic value or expressive. It does not have to aim at reaping benefits, but it can still produce beneficial effects. A symbolic sense of community creates feelings of belonging, trust and security.

Where does social capital actually come from, what makes it to be? Examining this question particularly from the viewpoint of trust, Hooghe and Stolle (2003: 3) differentiate from each other two very general main explanations. On the one hand, trust and civic attitudes can be explained mainly at the level of social interaction in one's local community. This is what the authors call the society-centred approach. On the other hand, the institution-centred approaches start from the idea that citizens’ trust also emerges as a result of the functioning of public administration and politics. We will later apply Hooghe's and Stolle's conceptual differentiation to concern social capital in general and the relationship between the welfare state and social capital in particular. The line of hypothesis that can be called liberalist clearly follows the society-centred approach and the social democratic line of hypothesis follows institution-centred approach. In the next pages we will simply refer to two hypotheses: the society-centred hypothesis and the institution-centred hypothesis.

Oorschot and Arts's study (2005) of testing the crowding out hypothesis of the welfare state, found no evidence in favour of it. Their results mostly support the institution-centred hypothesis. In this study, we will continue and hopefully complement the work of Oorschot and Arts by testing not only the crowding out hypothesis, but also two opposite hypothesis on the existence of the different kinds of social capital in different kinds of welfare states: the society-centred hypothesis and the institution-centred hypothesis.

In our research, the empirical data consist of the ISSP 2001. Of the data in question we selected for analysis the variables that describe relations among family and friends, civic engagement, social support and personal trust. At the first stage, our method of analysis is simple: we calculated for variables the means by country and by regime. For each group of variables we also calculated sum variables which have been returned to the scale of original variables. For these sum variables we also gave country- and regime-specific confidence intervals for the reader to be able to independently make comparisons between the countries and country groups. The tables comprise so much data that we had to limit our interpretations to include only the main observations. At the second stage of analysis we will build multi-level models to differentiate the individual-level and country-level variance of social capital indicators.

Let us begin by looking at bonding social capital. The ISSP questionnaire used has a quite extensive set of variables which concern relations among family and friends. Here we limited our observation to the closest family members and friends. Table 1,² shows how often respondents see their brother or sister, grown-up child, mother or father or close friend. Also the variable describing the number of friends is included in the table.

By studying the sum variable we can see that the Mediterranean regime clearly differs from others as a regime with a high degree of bonding social capital. There are just a few single countries in other regimes with equally high scores of bonding social capital: the Czech Republic and Slovenia in the post-socialist regime and Australia as the only exception in the liberal regime. The sum score of the post-socialist regime is the second highest, but much lower than in the Mediterranean regime. There is a considerable variation between countries of the post-socialist regime. In eastern Germany, Hungary, Poland and Latvia, bonding capital is at about the same level as in the liberal and Nordic welfare state regimes.

Do the other welfare regime states then differ in the relations between family and friends? There seem to be differences but they are relatively small. However, between the liberal countries and the Nordic countries there is a statistically significant difference. In line with the society-centred hypothesis, there is little more bonding capital in the liberal regime than in the Nordic countries (see Oorschot and Arts 2005: 19), with the exception of Norway. The conservative regime is placed according to the assumptions of the society-centred hypothesis. Its citizens seem to have a larger amount of bonding social capital than those living in the Nordic countries. The society-centred hypothesis is thus supported by the result. However, it is slightly confusing that countries that have the least amount of bonding social capital (Finland, Latvia, France, Canada and New Zealand) represent different regimes.

It can thus be said that in Mediterranean and in some East European societies family ties are stronger than elsewhere. This observation supports the conception that the areas in question have a strong family-oriented tradition. Although the birth rate has in recent years been declining in Eastern and Southern Europe and families are therefore becoming smaller (Castles 2003), this development seems to appear as a slight decline in the significance of social relationships of families.

Civil participation in different kinds of voluntary activities of various associations or groups, such as political, occupational, religious or free time activities, probably represents bridging social capital as its most typical. Engagement in this kind of activity was mapped out in the ISSP 2001 by a seven-item question. By looking at the sum variable formed out of these, we can see that the respondents were categorized by welfare state regime fairly logically, although now mainly in line with the institution-centred hypothesis and contrary to the society-centred hypothesis. The order is the following: the largest amount of bridging capital can be found in the Nordic countries, after them come the liberal regime countries, the conservative regime countries, then the Mediterranean regime, and finally, the post-socialist societies. Differences between all these regimes are statistically significant with a risk level of 1 per cent. All the countries of the Nordic regime are among the five highest scores. Countries of the liberal welfare state regime also have high scores, bridging capital in New Zealand being at the top level. In the conservative regime, most bridging capital can be found in France and western Germany. Here the difference between the former western and eastern Germany is quite clear.

Our findings are largely comparable with earlier studies that have investigated voluntary association membership in a comparative design. For example, Curtis et al. (1992) as well as Curtis et al. (2001) have made the observation that in the North American and Nordic countries participation in voluntary civic engagement has been highest compared to most other western countries. In a similar vein, our findings correspond well to the results gained from recent studies conducted on civic engagement in East European post-socialist societies (see, for example, Uslaner 2003; Wall et al. 2003).

Now an interesting question arises: how do the other elements of social capital, such as informal social support and generalized trust, vary within these two types of regimes?

The relationship of informal social support with social capital and welfare state arrangements may be complex in the sense that social capital in the form of social networks, on the one hand, creates a foundation for the emergence of social support; on the other hand, society's official support and service systems lessen the need for informal support. It would therefore make sense to assume that there would be an especially large amount of social support in the countries where informal social networks are exceptionally well functioning but that at the same time have a lot of social inequality and social problems. It feels natural to assume that in the society-centred countries the amount of social support is greater than in the institution-centred countries.

Social support was measured in the ISSP 2001 with four questions asking if the respondent has helped someone outside the household with housework or shopping, lent quite a bit of money to another person, spent time talking with someone who was a bit down and depressed or helped somebody to find a job. Table 3 shows the means concerning these questions by country and by regime. The sum variable we calculated on the basis of the variables shows that support does not significantly vary between the Nordic, liberal and conservative welfare state regimes. The most surprising result, contrary to the assumptions of society-centred hypothesis, is that in the Mediterranean and post-socialist regimes the level of support is lower – not higher – than elsewhere. In the society-centred countries informal social support is not exceptionally common. How is it to be understood that a substantial amount of bonding capital is not producing a substantial amount of social support? Bonding capital does not seem to be as instrumental as the society-centred hypothesis assumes, but rather the opposite (cf. Wall et al. 2001). It has to be especially noted that in the Nordic countries the level of social support is as high as in most countries despite a high social security level provided by the state.

If we look at the individual countries, social support is most often given, deviating from the regime assumption, in Russia and in the United States. Although these countries are in different leagues in the level of living comparison, they seem to share a strong tradition in giving informal social support. If this observation supports the argument that where there is informal social support the need for it is greatest, the United States would to a greater extent be a poor relief state like many European post-socialist countries. Despite the long period of neo-liberalism, this notion does not seem believable, and the conclusion is not supported by findings from Mediterranean countries where social inequality is common. It is interesting that although it is characteristic of Mediterranean countries to have strong bonding social capital, it does not seem to refer to strong social support. Maybe a strong sense of family community really is so exclusive in nature that it rather prevents the emergence of social support networks than makes them possible. And maybe it really is so that social networks are limited to close relationships in these countries (Flap and Völker 2003), which decreases the emergence of social support. Among the countries with highest social support are France, Norway and western Germany, which means that there are countries with high social support in every regime, except for the Mediterranean one.

How about generalized trust? How does it divide on the one hand among welfare-state regimes and on the other hand among society-centred and institution-centred countries? Generalized trust was measured in the ISSP 2001 with three items (see Table 4). Measured by the sum variable of these items, generalized trust can be found most often in the Nordic welfare state regime (cf. Oorschot and Arts 2005: 14, 15 and 19). All the countries within it are in the company of five highest scores. They are followed by the countries of the liberal regime, whose score increases by the presence of Australia, New Zealand and Canada. The level of generalized trust in Great Britain and the United States is at the level of the conservative welfare state regime. In the third place comes the conservative welfare state regime with an exceptionally high amount of generalized trust in Switzerland. In the fourth place is the Mediterranean welfare state regime. The least amount of trust can be found in East European post-socialist societies (cf. Oorschot and Arts 2005: 15, 16 and 19). The order of the regimes is exactly same as in the case of bridging social capital, and differences between all these regimes are statistically significant with a risk level of 1 per cent.

The greater the generalized trust, the more developed the regime in question. According to Uslaner (2003), the level of trust in society can be explained by the distribution of welfare resources: when inequality increases, trust decreases. There are two reasons for that. First, an even distribution of welfare resources creates within a majority of the population an optimistic attitude towards life: it is worthwhile for the individual to participate in interaction with others, because interaction creates possibilities to be successful in the future. Second, an even distribution of resources consolidates shared values and experiences of the community. Correspondingly, if welfare is distributed very unevenly, negative stereotypes of other groups begin to form, and distrust and tension increase. In this respect, differences in trust between the Nordic countries and other western countries become understandable.

It follows from what has already been said that generalized trust is greater in the institution-centred countries than in the society-centred countries. Our results point to the fact that bonding capital produces a remarkably smaller amount of generalized trust than bridging capital. The Nordic welfare state as a regime of high bridging capital and generalized trust will also endure in international comparison on social capital. This also means that in developing social capital the Nordic countries can present good examples. The fact that bonding capital is scarce there can point to the modernization development of social capital. Although the amount of bonding capital can in this development decline, informal social support prevails, bridging capital increases and generalized trust grows.

As a conclusion of this section, let us look at the relations between our social capital dimensions. Table 5 shows the correlations between the various sum scores of social capital in country level.

As already shown above, the correlation between bonding and bridging capital is clearly reverse (–0.45, P<0.05). In other words, in countries with a lot of bonding capital, there is little bridging capital and vice versa. Also the relationship between bonding capital and social support is reverse, but even more strongly so (–0.62, P<0.01). The connection between bridging capital and social support instead is positive (0.47, P<0.01). An extremely significant observation is also that trust correlates strongly (0.65, P<0.01) with bridging capital, but the connection of trust with bonding capital is not statistically significant.

Above we have calculated from the individual-level data the country-specific means. By examining these means we have made conclusions about the differences between single countries and welfare state regimes in the occurrence of social capital. What we have done can however be criticized from some methodological perspectives. First of all, by using this method, we also at the same time lost a lot of information connected to the individual-level variation in social capital. In other words, we should be able to assess how large a part of the variation in social capital can be explained by differences among individuals and how big a part of the variation stems from differences among the countries. Second of all, to make conclusions about differences in the amount of social capital between countries or groups of countries, we would have to control the central factors that at the individual level explain social capital, because some of the differences detected at the country level in social capital may be explained by a compositional effect of individuals (see, e.g., Leyland and Groenevegen 2003: 270). Third of all, if differences among the countries prove significant even if we controlled the individual-level explanators, we should still be able to explain to what extent the differences in the type of welfare states explain the country-level variation.

Because our research setting and data can be considered to be hierarchically organized, with individuals (level 1) living in certain countries (level 2), we will in the following carry out an analysis referred to as multi-level analysis by employing the MLwiN programme. We calculated models for all dependent variables in three phases. We first calculated the so-called basic models with an intercept only. By using these basic models we can divide the variance of variables describing social capital into the individual and country levels. In the second phase, we calculated the models in which the possible individual-level independent variables we chose are controlled and examine how the residual variance behaves at both levels. In the third phase, we brought along one country-level variable, regimes, and after that examined again changes in the residual variance. Finally, we tabulated the parameter estimates with their tests.

As individual-level independent variables we used all those common background variables that are included in the ISSP 2001 data and out of which the comparable information on all countries we studied has been collected. The individual-level variables we used in the models are shown in Table 7.

In the following we look at how the variance is divided between two different levels for every dimension of social capital separately (see Table 6). The baseline model of bonding ties shows that of the total variance of the variable, about seven per cent is at the country level. This variance does not drop significantly even if individual-level factors were taken into account. In the third model bringing in the variable describing welfare state regimes drops significantly the country-level residual variance (from 0.105 to 0.060): the typology of welfare regimes can thus be considered to be a significant explanator of social capital. In addition, it is worth noticing that individual-level factors explain the total variance notably effectively.

Of the variance of the variable describing participation, 20 per cent are found at the country level and controlling the individual-level factors even slightly increases the variance at the country level. Adding a variable that describes welfare state regimes into the model drops the country-level residual variance (from 0.123 to 0.032) even more than expected. We can thus conclude that variation in social capital of the bridging type is significant even if the individual-level factors were controlled and that variation is strongly explained by the type of welfare regime. The individual-level variables in our use instead did not very strongly explain the variation in bridging social capital.

The country-level variance of social support looks lowest when compared with other dimensions of social capital: it is only 4.4 per cent of the total variance. Yet it also seems to remain at the same level despite controlling the individual level factors. The variable describing welfare state regimes also drops the country-level residual variance of this variable clearly (from 0.038 to 0.026). We can thus conclude that even if the country-level variance in social support is not very significant, it is still clearly explained by the welfare state type.

Correspondingly the country-level variance of our variable describing generalized trust is 11.5 per cent of the total variance of the variable. In this case, also the country-level variance remains nearly unchanged even if we controlled the individual-level factors. The variable describing welfare state regimes drops in this case also the country-level residual variance significantly (from 0.067 to 0.027). The welfare state type is thus clearly significant in explaining the country-level variation in trust.

Table 7 shows the parameter estimates for the final models. Here it is not important to comment on the significance of individual-level variables in the models. However, Table 7 presents the individual-level parameters so that the readers can, if they so desire, compare them with observations in other studies. We instead focus on the dummy variable describing the welfare type in which the reference category used is the Scandinavian welfare regime.

As can be observed above, the factors that most effectively explain bonding social capital can be found at the individual level; in fact, excluding gender, all our background variables seemed to be connected to bonding social capital. Our most important observation is that bonding social capital can indeed be found most in the Mediterranean countries. Between other welfare state types instead there does not seem to be significant differences when individual-level factors are controlled. In the case of bridging social capital, differences between welfare state types are more diverse. There seems to be no difference between the Nordic and the liberal models. Instead, in countries of conservative type there seems to occur less bridging social capital than in the Nordic countries. The clearest difference lies in that both Mediterranean and post-socialist countries have clearly less social capital of the bridging type than the countries of the Nordic (and liberal) type. In the case of generalized trust the result is parallel but even clearer. When the individual-level factors are controlled, trust seems to occur most in countries of the Scandinavian type and it seems to grow scarcer when shifting to the liberal, conservative, Mediterranean and post-socialist types. The differences seem to be smallest in the case of social support. There seems to be only one slightly negative deviation from other countries standing out: the Mediterranean countries.

Our research is guided by two competing hypotheses on the relationship between social capital and the development of the welfare state. The first one, the society-centred hypothesis, assumes that social capital emerges first and foremost in close communities such as families, local communities, or in civic engagement. Thinking of welfare state regimes, this means that strong welfare states rather weaken than strengthen the preconditions for the growth of social capital. According to this assumption, social capital should be found particularly in the undeveloped welfare state regimes. The competing institution-centred hypothesis in turn starts from the fact that social capital develops above all in societies with a strong political system and a very well-functioning public administration. Along these lines, social capital should occur in the strong welfare state regimes in particular.

Our research findings point to the fact that both hypotheses are partly correct. The society-centred hypothesis is correct in the sense that social networking based on family ties in particular (bonding social capital) is stronger in the Mediterranean and in post-socialist regimes than in other regimes. Differences between the liberal welfare state regime and the Nordic regime instead are small in this respect.

However, the main part of our research results supports the institution-centred hypothesis. First, voluntary civic engagement (bridging social capital) is most common in the social-democratic and liberal welfare state regimes and least common in the post-socialist and Mediterranean welfare state regimes. Second, generalized trust is strongest in the social-democratic Northern Europe, after which come the liberal and conservative regimes. The least amount of trust can be found in Eastern and Southern Europe where social security is weak. As far as the level of informal social support is concerned, we found out that it is surprisingly low in the Mediterranean welfare state regime but does not seem to vary much between the other welfare state regimes. Strong bonding social capital does not seem to turn into informal social security in the Mediterranean countries, where there are many social challenges and where formal social security systems are limited. On the other hand, the score of social support of the Nordic countries shows that developed social services do not seem to indicate a decline in informal social support.

Observations in our study in part complement Oorschot and Arts's (2005) observations that the welfare state does not seem to prevent the forming of social capital. In addition, as far as we understand, we have been able to diversify the picture of the relationship between social capital and the welfare state: the welfare state may have a negative impact on some forms of social capital, on some the impact is non-existent and on some the impact is apparently positive. It is however easy to agree with the notion of Oorschot and Arts that in the future a more detailed study should be carried out to find out what kinds of mechanisms actually give birth to those in part multi-direction connections that we observed between the different manifestations of welfare state regimes and social capital.

From the viewpoint of further research, an extremely interesting question in our opinion is how are the best preconditions for the emergence of informal social support brought about and how important a role the different elements of social capital actually play here. What kind of social capital is needed to advance the arrangements of informal relations of caring? According to our findings, we could assume that concentrating on bonding capital is after all not the decisive factor in the favourable development of modern societies. A decline in bonding capital is naturally a negative thing when it makes everyday life's interaction networks thin and poor. It is nevertheless the more important, the scarcer the possibilities to act in larger interaction networks. The scarcity of civic engagement can be an indication of a society plagued by political and social problems. Bonding capital does not seem to essentially advance informal support. It may be that bonding capital networks are too exclusive for that. The emergence of informal support seems to be linked more clearly with functioning bridging capital, the kind of extensive civic engagement that produces trust in people one does not know personally. It may very well be that the development of bridging capital is the road in the direction of both functioning informal support and functioning political democracy. It can also be a mistake to see the informal and formal care systems as rivals or mutually exclusive.

Juha Kääriäinen PhD (born 1957) is a Senior Researcher at The Police College of Finland. He has been working at the University of Tampere, Finland, as a researcher in 1985–1993, and as a lecturer of social policy at the University of Tampere in 1993–2005. He has also been a visiting researcher at the London School of Economics in 1997. His doctoral thesis was accepted in 1994 and it handled criminal careers of Finnish recidivists. His main research areas cover criminality and other social problems and the questions of local communities as a social recourse.

Heikki Lehtonen is a Professor of Social Policy at the University of Tampere, Department of Social Policy and Social Work. His main research interests include studies on university and adult education, on social theory and social policy, and on the recession of 1990s and social policy in Finland (‘Did the Crisis Change the Welfare State in Finland?’ In Kalela, J., Kiander, J., Kivikuru, U., Loikkanen, H. A. and Simpura, J. (eds.) ‘Down from the heavens, up from the ashes. The Finnish economic crisis of the 1990s in the light of economic and social research’. Government Institute for Economic Research. VATT Publications 27:6. Helsinki 2001. (Together with Simo Aho, Jarmo Peltola and Mika Renvall). ‘The State of the Welfare State in Europe Anno 1992 and Five Years Later: Finland’. In Pacolet, J. (ed.), 'The State of the Welfare State in Europe Anno 1992 and Five Years Later and Beyond. Studies in economic transformation and public policy’. APF Press, Toronto 2002. (With Simo Aho).