Wendy Stone & Jody Hughes¹

Australian Institute of Family Studies

Paper to ‘Competing Visions’ National Social Policy Conference,

hosted by the UNSW, 4 — 6 July 2001, Sydney.

There is much policy and academic interest in the capacity of the concept ‘social capital’ to generate both micro level outcomes such as family wellbeing as well as macro level outcomes such as efficient economies, democratic polities and active communities (see for example the work of Bourdieu 1993; Coleman 1988, 1990; and Putnam 1993, 1995; as well as Winter 2000 for a discussion of social capital in Australian public policy). However, while much has been hypothesised about the potential of social capital to achieve a range of outcomes, little is known about how social capital operates empirically, how it is generated, nor how it is distributed across Australian families, localities and communities of interest.

Conversely, the social capital concept has attracted much criticism — both for its underlying assumptions and theoretical basis as well as for the rapidity of its uptake and usage (see for example Fine 2000). Much of this criticism is made on the basis of political position, limited theoretical critique and/or poor quality empirical scrutiny (ie that typically fails to link empirical work to social capital theory — see Stone and Hughes 2000 for discussion).

Indeed, much of both the popularity and criticism of the concept lie on under-explored assumptions about the way social capital is manifest, generated and distributed empirically — and, ultimately, how it might lead to a host of other outcomes.

This paper draws on the Families, Social Capital and Citizenship (FSAC) survey (2001) to address some of these gaps. Specifically, the paper sets out to answer the following:

It is critical that in order to properly test the ‘social capital thesis’, the first of these questions is addressed first. Only once we achieve an empirical understanding of the nature of social capital can we begin to consider questions of how social capital is distributed among different parts of the population, and ultimately how social capital relates to a host of outcomes attributed to it. Given this, the initial analysis of the FSAC survey presented in this paper focuses on the first of these questions. Some initial findings are presented for questions two and three.

The findings presented in this paper are preliminary only. Detailed analysis of what is underlying many of the results presented in this paper will be undertaken in future work. As well, future work of the FSAC project will link social capital empirically to a series of family, community, economic and political outcomes, to further explore the extent to which the social capital concept adds to our understanding of these phenomenon.

1.1 A theoretically informed measurement framework

There are inherent dangers in researching any social concept without regard to a rigorous theoretical framework. In social capital research, these dangers are rife and have led to a host of problems. Most notably, failure to undertake empirical social capital research without regard for the link to social capital theory has led to the following four problems: first, that social capital measurement and ‘practice’ is rarely theoretically informed; second, that empirical work rarely understands social capital as a resource to action; third, that there is a general failure in empirical work to recognise social capital as a multidimensional concept; and finally, that there is a general failure to recognise that social capital will vary by network type and social scale (Stone and Hughes 2000; see also Portes 1998 and Paxton 1999).

In order to empirically investigate social capital in a theoretically informed way, a conceptual framework for researching social capital in family and community life has been developed within the work of the Families, Social Capital & Citizenship project (see Stone 2001).

As set out in the FSAC framework, in order to achieve theoretical rigour in social capital measurement, a clear understanding of the concept, upon which to base an empirical framework, is essential. In line with the work of Bourdieu (1993), Putnam (1993) and Coleman (1988) — the key theorists of social capital ² — the concept social capital can be understood as networks of social relations which are characterised by norms of trust and reciprocity and which lead to outcomes of mutual benefit. Understood in this way, social capital can be thought of as a resource to action, which may lead to a broad range of outcomes.

Table 1 presents the measurable components of social capital. It identifies networks, trust and reciprocity as the key dimensions. The types of social networks in which social capital can inhere are infinite. The FSAC study is comprehensive, and is concerned with family based networks as well as links with kin/in-laws, friends and acquaintances, neighbours, workmates, people known through civic groups, as well as ties to institutions and feelings about other people more generally. The study is also concerned with the quality of social relations in each of these types of networks — indicated in Table 1 by the different types of trust and reciprocity governing these different type of network relations.

The table also presents a range of network characteristics which previous studies indicate may influence the nature and extent of social capital within a given network: types of network (whether they are made of familiars or others); network size; the proximity of networks; network density (the extent to which networks are overlapping ie whether members of one type of group know members of another); as well as network homogeneity (the ‘mix’ or diversity of connections in social networks).

Source: Stone 2001 (modified).

In sum, social capital is a concept concerned with both the quality and structure of social relations and is said to be influenced by characteristics of the social structure in which it inheres (see particularly the work of Coleman 1988, 1990).

Social capital is thought to have different capabilities, in relation to the outcomes it can achieve, depending upon the configuration of network types and network characteristics, as well as the norms governing respective networks.

In particular, recent theoretical work (from the World Bank in particular, through the work of Narayan 1999 and Woolcock 2000) provides a framework for classifying the different capabilities of social capital, and introduces the notions of bonding, bridging and linking social capital. ‘Bonding’ refers to ‘close’ ties people typically rely on to get by on a day to day basis; ‘bridging’ refers to having a diversity of relations across different network types; and ‘bridging’ refers to links with power and institutions of authority. Future work of the FSAC project will more fully explore the relationship between these capabilities and a host of economic, civic and political outcomes.

1.2 Data and method

This paper draws on survey data collected in 2000/2001 as part of the Families, Social Capital and Citizenship project being undertaken by the Australian Institute of Family Studies. A total of 1506 respondents participated in the survey. These were drawn from a national random sample of households (with at least one person aged 18 years or over). Interviews were conducted via the telephone using the Institute’s Computer Assisted Telephone Interviewing system (CATI), and were 32 minutes duration on average ³.

The FSAC survey is based on the conceptual and empirical framework presented above. The survey collected information about respondents’ social networks and the quality of relationships in those networks, as well as information about a range of outcome measures plus detailed demographic information.

The data contain three main biases. The first of these relates to the use of the telephone as a vehicle for data collection. Hence, those who are highly mobile, those who do not have a telephone or have a silent number, as well as those who ‘vet’ calls via answering machine, are under-represented in this study.

The second bias also relates to the survey methodology employed to collect the FSAC survey data. Given limited resources, the survey was only conducted in English language. Where householders were unable to undertake the survey in English, these households were excluded from the study.

The third main bias is also an effect of both the telephone methodology as well as limited resources, and concerns the gender balance within the survey data. The data contain a significant over-representation of women (69.2 per cent) at the expense of men (who make up 30.8 per cent of respondents). Where gender is not controlled for in data collection, this gender pattern is typical.

These biases should be kept in mind when interpreting results. However, given the exploratory nature of the survey and study, and the lack of theoretically grounded empirical social capital research, these results still provide one of the most detailed means of testing the ‘social capital thesis’.

2. The Nature of Social Capital

As described above, in order to understand how social capital operates empirically, it is first necessary to recognise social capital as a multidimensional concept and gain an understanding of the nature of each dimension before exploring how these operate together empirically.

The following section of the paper introduces and documents each of networks, norms of trust and norms of reciprocity, according to different network types.

2.1 The dimensions of social capital: networks, trust & reciprocity

The FSAC survey asked respondents about both their informal social relations and connections, and their more formal ties to civic and community groups, organisations and institutions

While everyone has at least one informal tie, there is considerable variation in the size of informal networks, and in family and friendship networks in particular.

There is also great variation in the extent to which informal networks are family based, with the proportion ranging from less than one per cent to 100 per cent. On average, around a third of respondents’ total network of informal connections is made up of family ties.

Note 1. Includes partner (married or defacto) and relatives other than children (eg. parents, siblings etc).

* = max number allowed.

One of the key concerns within the social capital literature is the notion of civic ‘decline’. To explore this issue, the FSAC survey asked respondents about whether they belonged to any of a series of types of civic groups/organisations and, if so, how many of each type they belonged to. Membership of groups of this kind has been used as a key measure in the social capital empirical literature to date, and often as a single measure (eg. see Putnam 1993, 1995).

The table below indicates that 86 per cent of respondents belonged to at least one group or organisation, and the average number of groups respondents belonged to was three (median, mean = 4.3) ⁵. Sporting groups were the most common, then community/welfare groups, arts/educational groups, child related groups and professional groups/trade unions.

The table below indicates that, of those who belonged to at least one group or organisation, 81 per cent had made personal ties through this involvement (about 70 per cent of all respondents).

One way of measuring the extent to which respondents have what is known as ‘linking’ social capital (ties to those in authority and power) is to examine the extent to which respondents have connections to people in powerful institutions. The survey listed a series of organisations and institutions and asked respondents, for each one, whether they personally knew someone in the organisation or institution they could contact if they needed information or advice about something (they could include family, friends and work-associates).

* Assume respondents have personal tie/s to church if they attend church at least a few times a year (71 per cent of respondents attended church at least once a year).

The table above indicates that 93 per cent of respondents had personal ties with at least one type of institution, whether it be the church, the media, the legal system, the government, or big business; and the table below indicates that people had personal ties with three different types of institutions on average.

Social capital theory dictates that the quality and structure of social networks are inextricably linked. Thus, the FSAC survey not only asked respondents about the extent and structure of their connections, but also about the quality of those connections, as measured through norms of trust and reciprocity, respectively. Importantly, norms of trust and reciprocity were asked about for each respective network type separately.

The following table presents the mean levels of trust respondents reported across a range of networks and social scales.

The following figures indicate that trust of family within the household is the highest reported type of trust, and differs from trust of other types of family beyond the household. The average level of trust reported for family members in the household is high (9.62), and this level shows little variation across households (standard deviation for this item = 1.01).

The figures also indicate that, overall, people tend to report higher levels of trust in those they are close to and are familiar with, than in other people or institutions. There is a clear clustering of high levels of trust reported among familiars, with people reporting overall high levels of trust (greater than 8 out of a scale of 0 to 10) for kinship networks, in-laws and friends.

Trust levels reported for workmates, people in the local area, and Australians more generally show a slightly different pattern, although each remains quite high.

Notably, the levels of trust, or ‘confidence’, reported in the various institutions listed above, are notably lower than for other types of networks. These results show that trust of the media, of union, big business and all levels of government (ranging from federal government at 4.50 to local government at 5.03) are relatively low — with each attracting an average score of between 4.31 and 5.03 only. Confidence in the public service, the legal system, churches rate slightly higher (up to 5.63), with the police force attracting the highest average level of confidence (6.96).

Differences in levels of trust within different types of networks (above) are confirmed by factor analysis. Results of Principal Components factor analysis using a standard Varimax rotation procedure indicate that although there is a relatively high overall level of communality between the items (ie that we can think of ‘trust’ as one dimension), there are four different elements, or ‘factors’, which comprise ‘trust’ (and which jointly explain 52.6 per cent of variance).

The first of these is trust of family within the household (Eigenvalue = 1.06, explains 6.2 percent of variance). This single item reflects respondents’ reports of the extent to which the family they live with can be trusted to act in their best interests (note that respondents not living with either a partner, child or blood relative were excluded from analysis for the purposes of this paper).

Interestingly, trust of family/kin beyond the household as well as family-in-law appear more closely related to an overall sense of trust of other close network ties, than to trust of family within the household. These two items form part of the second factor, which can be thought of as ‘trust of familiars’ (Eigenvalue = 1.96, explains 11.5 per cent of variance). As well as trust among kin and in-law networks, this factor represents the extent to which respondents trust friends and workmates.

The third factor reflects the extent of respondents’ trust in people more generally, beyond those known to the respondent. This factor can be thought of as representing ‘generalised trust’, or trust in others (Eigenvalue = 1.15, explains 6.7 per cent of variance). This factor represents responses to the two items: ‘Generally speaking, most people in my neighbourhood can be trusted’ and ‘Generally speaking, most people can be trusted’.

The fourth factor represents respondents’ sense of trust in institutions. This factor incorporates 10 items relating to the extent of confidence in a range of institutions (Eigenvalue = 4.78, explains 28.2 per cent of variance), and is most distinct from each of the other factors.

The distinctions found in the data between different types of trust is largely supported in the social capital literature (see for example Dasgupta 1988; Uslaner 1999; Cox and Caldwell 2000; Black and Hughes 2000).

Levels of reported reciprocity across different types of social networks show a pattern similar to those for norms of trust.

The pattern of mean levels of reciprocity reported by respondents across different network types largely reflects the pattern found for trust (above). As in the case of trust, the mean level of reciprocity among family members within the household is higher than that for other family members and in-laws living outside of the household and all other ties. Again, there is little variation in levels of reciprocity reported for family households (standard deviation = 1.33) — less than for any other type of network.

Results indicate high levels of reciprocity in operation across other types of informal networks. Non-household based family networks and friendship networks each score highly on this measure (all scoring an average over 8 on the 0 to 10 scale). Mean levels of reciprocity reported among workmates, as well as people in the local area, and people in general are similar — each approaching an average of around seven.

Factor analysis procedures confirmed that norms of reciprocity also operate differently in different types of social networks, or social scales. Results of Principal Components analysis (again using Varimax rotation) showed that norms of reciprocity group together in similar ways to norms of trust.

The factor analysis procedure indicated that ‘reciprocity’ can be thought of as comprising reciprocity among family households (Eigenvalue = 1.00, explaining 14.3 per cent of variance), reciprocity among other familiars (Eigenvalue = 2.23, explaining 31.8 per cent of variance), as well as a more generalised norm of reciprocity reflecting the way respondents feel about people in the local area, and people in general (Eigenvalue = 1.19, explaining 17.1 per cent of variance).

*** P < 0.005 , ** P < 0.01 ,* P < 0.05

2.2 How does social capital in one network relate to social capital in
another?

One of the key hypotheses posed in the social capital literature is that social capital within one type of social relationship, or at one ‘social scale’, will foster the generation of social capital at another. That is, social capital within one network will translate into social capital in other aspects of social life.

In particular, it is most typically assumed that social capital within the family — good quality familial relationships — will generate social capital in other spheres of life (for example increased democracy, trust in institutions, and neighbourhood cohesion) (see Winter 2000a for discussion of the assumptions made about family life in the social capital literature).

Whether social capital generation does occur in this way — leaving aside for the moment the question of how this might occur — is an open empirical question. The question addressed by this section then, is how does social capital within one set of social relationships, at one social scale, relate to social capital within another?

In order to begin to untangle the complex ways in which social capital (as a multidimensional concept) operates across different social scale, the analysis presented below reports on how each of the core dimensions of social capital — networks, trust and reciprocity — relate to one another across different network types.

First we examined the relationship between the size of family and non-family networks and found a fairly strong positive correlation between the two ⁶ (r(1488)= .3108, P = <.001). That is, people with large family networks also have large non-familial informal networks (friends, neighbours and/or work contacts).

The correlations between the separate components of family and non-family networks suggest that this may be primarily because those with large family networks (kin and/or in-laws) also tend to have large friendship networks (r(1499)=.2550, p<.001) ⁷.

We then examined the relationship between the total size of these informal networks and the number of group memberships and institutional ties and in both cases found a fairly strong positive correlation - people with large informal networks are also more likely to belong to lots of groups and/or organisations and to have broad institutional ties (that is, to know people in many different types of institutions).

These findings indicate mixed support for the ‘crowding out thesis’ that asserts that an abundance of one type of network can undermine the existence of another. Most interestingly, these early findings suggest little support for the notion that ‘too much family’ is related to fewer ties in civic and organisational groups — rather, no relationship was found between the extent of family networks and civic ties. On the other hand, networks of non-family familiars (friends, neighbours and workmates) appear complementary to civic/organisational ties. This is perhaps not surprising as friendships and other ties are often formed through or facilitate engagement in civic groups.

It is important to note, however, the findings presented here only report the size of networks, without regard to the quality of those relationships — other core dimensions of social capital ⁸.

Correlation procedures were also used to explore the question of how trust and reciprocity within one type of network, at one social scale, relate to trust and reciprocity at other social scales.

The following tables present results about the relationships between trust within family household, trust of familiars, generalised trust and institutional confidence, as well as between norms of reciprocity within the family household, among familiars and a generalised norm of reciprocity, respectively.

*** P < 0.005 , ** P < 0.01 ,* P < 0.05

*** P < 0.005 , ** P < 0.01 ,* P < 0.05

These figures suggest a ‘flow on’ effect may be operating between the norms within one set of social relations, and norms within another set of social relations, where that next set of relations is ‘near by’ in terms of social scale. The figures indicate, for example, that norms of trust and reciprocity operating within family households are positively associated with quality relationships among non-household based familiars (.3015 for trust and .3118 for reciprocity). Norms or trust and reciprocity within networks of familiars outside of the household show a quite strong positive correlation with generalised norms of trust and reciprocity (.3071 for trust and .3403 for reciprocity). However, there is only a weak positive correlation between the quality of relations within family households and more generalised norms of trust and reciprocity (.1240 for trust and .0995 for reciprocity).

The first of these sets of results also indicates this trend is sustained for patterns of institutional confidence. Trust in institutions is most highly associated with generalised trust and trust in familiars (.2979 and .2280 respectively), and most weakly associated with norms of trust among family households (.1240).

2.3 The influence of social structure: network size, density and heterogeneity

As set out in the introduction at 1.1 above, a key part of the social capital thesis is that the structure of social relations influences the quality of those relations. In particular, it is hypothesised that size of networks, the density of relations (how well relationships from different types of networks are integrated), as well as the diversity of social networks, affect the quality of social relations within those networks (as well, ultimately, as the extent to which social capital is able to affect outcomes of different types).

In respect to informal networks, the correlations below indicate that the size of each type of network is not typically related to levels of trust and reciprocity within that network, although there is one apparent exception to this rule — there is a strong positive correlation between the number of family members in the household and levels of trust and reciprocity among family members in the household, ie. the larger the number of family members in the household, the higher the levels of trust and reciprocity reported.

While the size of each other type of informal network is not related to trust within that network, the table indicates that the number of neighbours respondents know is significantly correlated with their trust of people in the local area and with the degree of reciprocity in the local area; that belonging to groups and organisations is significantly but very weakly correlated with generalised trust; and that having broad institutional ties is significantly but very weakly correlated with confidence in institutions.

***=P < 0.005, **=P < 0.05 and * =P < 0.01.

1. Correlation between number of people from previous jobs that respondent keeps in contact with and levels of trust and reciprocity among current work mates.

2. Composite measures of generalised trust and reciprocity which combine people in local area with "most people/Australians".

The social capital literature suggests that the degree to which networks are ‘dense’ - the extent to which network memberships overlap — will effect norms of trust and reciprocity within networks. For example, network density will affect the ability of persons in one context to call on assistance to solve a problem in another.

Network density is explored in the FSAC survey primarily via measures of the extent to which friends of respondents know one another, as well as the extent to which respondents report that family members generally know one another’s close friends.

There is a great degree of variation in the density of friendship networks but little variation in the density of family-friendship networks (most respondents reported that family members know each other’s close friends very well, and only 6 percent disagreed with this statement). However, there is also a strong positive correlation between the two measures (r(753)=.6979, p<.001), so both were combined to create a single indicator of the density of informal networks.

There appears to be a positive but very weak relationship between this composite measure of informal network density and levels of trust and reciprocity among familiars, offering some support for Coleman’s (1988) thesis. Respondents with dense informal networks report significantly higher levels of trust (r(1466)=.0795, p<.002) and reciprocity (r(1460)=.0643, p<.05) among familiars, although these correlation coefficients are very small.

When we examined the two measures of density separately, it is clear that it is dense or overlapping family and friendship networks ("family members know each other's close friends") that is the most potent predictor of levels of trust and reciprocity, rather than dense friendship networks (friends know each other). Respondents with dense family-friendship networks reported higher levels of trust in family in the household (r(748)=.2997, p<.001) and reciprocity in the household (r(752)=.2533, p<.001), higher levels of trust in friends (r(747)=.1959, p<.001), and greater reciprocity among friends (r(744)=.2046). Thus, having dense or overlapping family and friendship networks is associated with higher levels of trust and reciprocity in both network types. Respondents with dense family-friendship networks also tend to report higher levels of trust and reciprocity among family beyond the household, family beyond the household (kin and in-laws), and work-mates; although these relationships are weaker ⁹.

There appears to be little relationship between the density of informal networks and generalised trust (not significant), generalised reciprocity (r(1416)=.0733, p<.006), or trust in institutions. However, when we examined the separate measures of density we again found that having overlapping family and friendship networks was related, albeit weakly, with higher levels of generalised trust (r(722)=.0864, p<.05) and reciprocity (r(728)=.1538, p<.001), and confidence in institutions (r(1135)=.0747, p<.05).

The literature also suggests that the heterogeneity of group or network membership may influence the levels of trust within networks, the extent to which trust of familiars translates into generalised trust of strangers, and the extent to which norms within networks are shared. However, some argue that heterogeneity increases social capital and other outcomes (Grootaert 1998) while others argues that homogenous networks are most conducive to social capital (Portney and Berry 1997; Sampson et al 1997). Others (eg Stolle 1998) suggest that the effect of network heterogeneity will vary according to the homogeneity of the community context being investigated.

We examined two indicators of the heterogeneity of networks: ethnic diversity and educational diversity. We examined both the ethnicity and education levels of the members of respondents’ informal networks, and the extent to which the groups and organisations they belong to are made up of people from mixed ethnic and educational backgrounds.

In roughly forty per cent of cases in the FSAC survey, respondents’ informal networks ¹⁰ included at least one person who spoke a language other than English at home as well at least one person who spoke English only.

When these groups were compared in terms of levels of trust, some differences were found. First, those whose informal networks were homogenous in terms of ethnicity had significantly higher levels of trust in familiars (t(1467)=3.59, p<.001), and significantly higher levels of reciprocity among familiars (t(1462)=2.09, p<.05), although the levels of trust and reciprocity did not appear to be substantially higher (means=8.3 compared with 8.0 and 8.16 compared with 8, respectively).

Secondly, people with ethnically diverse informal networks also had significantly higher levels of generalised trust (t(1409)=2.30, p<.05) and reciprocity ¹¹ (t(1424)=2.75, p<.01), although again the levels were not substantially higher (means=6.9 compared with 6.7, for both trust and reciprocity). However, when people in the local area are examined separately from 'most people', it is clear this pattern applies only to people in the local area ¹² - there was not a significant difference between the two groups in terms of trust and reciprocity in 'most people'.

Ethnic diversity within informal networks does not appear to be related to levels of confidence in institutions, as no difference between the groups was found.

In terms of ethnic diversity of formal networks, 47 per cent of respondents who were in groups and/or organisations reported that the ethnic background of group members was mostly the same, but a fairly high proportion (34 per cent) also reported they were mostly mixed (19 per cent reported that the ethnic background of group members was 'somewhat mixed').

Those who belonged to groups and/or organisations that were homogenous in terms of ethnicity again had higher levels of generalised trust (mean=6.9) than did those who belonged to groups that were somewhat mixed (mean=6.8) or mostly mixed (mean=6.7); but the difference was not significant at the .05 level (p<.10). There was no difference between these groups in respect to level of confidence in institutions.

Most people in the study had informal networks that were somewhat mixed in terms of education ¹³. In about a quarter of cases (24 per cent) the level of education of all close personal ties (partner and close friends) was the same as the respondent’s, and in only 19 per cent of cases were informal networks very mixed in terms of education (ie. at least one person had tertiary education or above, and at least one person had less than year 12 education).

When those with homogenous, somewhat mixed and very mixed informal networks in terms of education were compared in terms of levels of trust and reciprocity among familiars, no differences were found.

About fifty per cent of respondents who were in groups and/or organisations reported that the education level of groups members was mostly mixed and about 50 per cent reported that education levels were mostly the same.

Again, the extent to which these groups was mixed was found not to be significant in terms of levels of trust and reciprocity among familiars, people in general and levels of confidence in institutions.

In order to further the dimensions of social capital, as well as to begin to understand how social capital is distributed, we examine the distribution of the core dimensions of social capital according to a range of individual level characteristics of respondents, as well as to their family circumstances and in relation to some key characteristics of the localities respondents live in.

The individual and family level variables explored include: sex of respondent, relationship status of respondent, presence of children, family type, ethnicity, educational attainment, labour force participation and financial wellbeing.

3.1 How do the dimensions of social capital relate to sex of respondent?

The table below indicates that there is a significant difference between men and women in terms of the mean number of family members in the household, with women reporting a larger number of family members in the household on average. There is also a significant difference between men and women in terms of the number of institutional ties they have: men have a significantly larger number of institutional ties.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

While there appears to be little difference between men and women in the levels of trust and reciprocity they report across network types, women appear more trusting overall. In particular, women report significantly higher levels of trust and reciprocity among their informal networks outside the home. They are also more likely to agree that most people — in the local community and society generally — try to be helpful.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

3.2 How do the dimensions of social capital relate to family characteristics?

3.2.1 How do the dimensions of social capital relate to relationship status?

Respondents were split into two groups based on their relationship status: married/defacto versus not married or defacto. These groups will be referred to as partnered and not partnered.

The table below indicates that overall, partnered respondents have larger networks than do those who are not partnered. Not surprisingly, the number of family members in the household is higher for those who are partnered. Those who are partnered also have larger informal networks outside the household. Partnered respondents have both larger family and non-family informal networks, but the larger family networks are simply due to the addition of '‘in-laws'’ (rather than any difference in the size of kinship networks). Partnered respondents do tend to have larger friendship networks and know more of their neighbours, but there is no difference between the two groups in terms of the tendency to keep in contact with people from previous jobs.

There is no difference between the two groups in terms of the number of groups and organisations they belong to, but partnered respondents have more institutional ties on average.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

A comparison of the two groups in terms of mean levels of trust and reciprocity reveal that, where differences exit between the two groups, levels of trust and reciprocity are typically higher among those who are partnered. Partnered respondents expressed significantly higher levels of trust in family members in the household, and higher levels of generalised trust and confidence in institutions.

While the two groups don’t differ in terms of levels of trust in familiars, when the different types of informal networks are examined separately, partnered respondents appeared to have higher levels of trust in kin (p<.05. Means=8.9 compared with 8.7) and in work mates (p<.005. Means=6.8 compared with 6.3), but there was no difference between the two groups in terms of trust in friends.

The pattern is largely the same for reciprocity. Reciprocity among family within households was higher among partnered respondents, and partnered respondents perceived higher levels of reciprocity to exist among people generally.

While levels of reciprocity among familiars do not appear to differ between the two groups, when the different types of informal networks are examined separately, partnered respondents report higher levels of reciprocity within kinship networks (p<.05. Means=8.7 compared with 8.4) and among work mates (p<.05. Means = 7.0 compared with 6.7), but there was no difference between the two groups in terms of reciprocity among friends.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

3.2.2 How do the dimensions of social capital relate to the presence of children?

Respondents were split into two groups based on whether or not they had a child under 18 years in the household.

The table below indicates first and not surprisingly that respondents with at least one child under 18 years at home live with more family members on average. There is no difference between these two groups in terms of the size of informal networks outside the household, except when the different types of informal networks are examined it appears that those with children at home have somewhat larger family based networks.

While there is no difference between the two groups in terms of the number of group memberships or institutional ties, those with children at home are significantly more likely to be involved with at least one group or organisation (89% vs 85%; c (1)=6.429, p<.05.).

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

A comparison of the two groups in terms of mean levels of trust and reciprocity reveal that there is little difference between the two groups in respect to levels of trust. Those without a child under 18 years in the household expressed higher levels of trust in familiars, but the difference between the two groups was very small. In contrast, those with a child under 18 years in the household had greater confidence in institutions.

The pattern is different for reciprocity. Reciprocity among family within households was higher where there was no child under 18 in the household, as was reciprocity within networks outside the household. When the different types of informal networks are examined separately, it appears to be higher levels of reciprocity in kinship and friendship groups that is responsible for this effect.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

3.2.3 How do the dimensions of social capital relate to family type?

Putting together information about respondents’ relationship status plus whether or not respondents have dependent children in the household allows us to comment on the overall relationship between ‘family type’ and the dimensions of social capital.

The table below indicates that there are significant differences between different types of families in terms of network size. Couple families have the largest network of familiars and sole parent families have the smallest. When the different type of informal networks are examined it is clear this pattern applies to both family and non-family networks (friends, neighbours and work mates).

While the number of group memberships does not vary by family type, respondents living in couple and children households appear most likely to belong to at least one group or organisation (90%), and single respondents living without children the least likely (83%)¹⁴.

Finally, couple families report a significantly larger number of institutional ties.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

Couple families report the highest levels of trust and reciprocity in the household, the highest levels of generalised trust and reciprocity, and the greatest confidence in institutions. It is striking that sole parent families, in contrast, report the lowest levels of trust and reciprocity across all network types. There is least difference between the groups in respect to trust of familiars.

Note: Child(ren) means a dependent of the respondent or partner, aged 18 years or under.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

3.3 How do the dimensions of social capital relate to ethnicity of respondents?

Respondents in the study were divided into three groups based on whether they spoke English only when at home or with their family, spoke mainly English when at home or with their family, or spoke another language primarily.

The table below indicates that there are significant differences between these three groups in terms of the mean number of family members in the household, with those who speak another language primarily having the largest number of family members in the household.

There is also a significant difference between the three groups in terms of group memberships and institutional ties: those who speak primarily English (but also another language) have the highest number of group memberships, and those who another language primarily have the lowest number of both group memberships and institutional ties. While there is no significant difference between the three groups in terms of the size of informal ties outside of the house, those who speak another language primarily have the fewest on average.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

The table below indicates that there is a significant difference between these three groups in terms of trust in familiars - trust of familiars was significantly higher among respondents who spoke English only at home or with their family, and lowest among those who spoke another language primarily. When levels of trust in different types of networks of familiars are examined separately however, this pattern holds only for trust of kin ¹⁵. Those who spoke English primarily but also another language (at home or with their family), report the highest levels of trust in friends¹⁶.

There is not a significant difference between these three groups in terms of levels of reciprocity among familiars. However, when reciprocity in different types of networks of familiars are examined separately, it appears that levels of reciprocity in friendship networks are higher among those who report speaking English only at home or with their family¹⁷.

Generalised trust is also highest among those who speak English only (there is no difference between the other two groups). Again when the separate measures are examined, this pattern holds for ‘trust in most Australians’¹⁸, but no difference is found between the groups in terms of trust in people in the local area.

There was no difference between the three groups in terms of generalised reciprocity or confidence in institutions.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity at .05 level.

3.4 How do the dimensions of social capital relate to education?

The table below indicates that network size differs significantly by level of education. First, the number of family members in the household is somewhat higher for those with lower levels of education. Second, there appears to be a positive linear relationship between level of education and the size of informal networks. However, when family and non-family networks are examined separately, it is the size of non-family networks that increases with education level; there is no differences between the groups in terms of the size of family networks. Furthermore, when the different types of non-family networks are examined, there is a linear relationship between respondents’ level of education and the number of friends and former work contacts they have, but education level is not related to the number of neighbours they know.

Third, there is also a clear linear relationship between respondents’ level of education and the number of group memberships and institutional ties they have (with tertiary educated respondents having particularly large numbers of group memberships and institutional ties).

In sum, the size of informal networks of friends and work mates, and the number of groups memberships and institutional ties all increase with the education level of respondents. However, family and neighbourhood networks do not vary by education.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

A comparison of means in the following table indicates that there is no clear relationship between level of education and forms of trust. This finding is strangely inconsistent with all previous research in this area!

In terms of reciprocity, while there is certainly not a linear relationship between education and measures of reciprocity, the perception that people are generally helpful (locally and/or in Australia) is strongest among those with low levels of education (less than yr 12) and weakest among those with high levels of education (tertiary or above). If we examine the two measures of generalised reciprocity separately, the pattern holds for both, but it is stronger for reciprocity in the local area (p<.001. Mean = 7.1 for respondents <yr 12, and 6.35 for tertiary).

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

3.5 How do the dimensions of social capital relate to respondents’ labour force participation?

The table below indicates that some aspects of network size differ by respondents’ labour force status. The number of family members in the household is highest for respondents who are not in the labour force (NILF) and lowest for those who are retired.

Also, the number of group memberships and institutional ties differs by labour force status of respondent - those in paid work have the highest number of group memberships and the largest number of institutional ties, while those who are unemployed have the lowest.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

A comparison of means in the following table indicates that levels of trust and reciprocity among familiars and among people generally, is consistently highest among retired persons.

When those in paid work are compared to those who are unemployed or NILF, it appears that trust in familiars is highest among those who are NILF and lowest among those who are in paid work. In contrast, those in paid work have the highest levels of generalised trust (there is no significant difference between the other two groups in terms of generalised trust).

The picture is not so clear with reciprocity. Reciprocity among familiars is again highest among those NILF but is lowest among those who are unemployed. The perception that people are generally helpful is also strongest among those NILF but equally low among those in paid work and those who are unemployed. However, if the measures of generalised reciprocity are examined separately it seems that reciprocity among people in the local area is perceived as highest by those who are unemployed (mean=6.82) and lowest by those who are in paid work (mean=6.51); while the reverse is the case in terms of reciprocity among ‘most Australians’: it is perceived as highest among those who are in paid work (mean=6.87) and lowest among those who are unemployed (mean=6.42).

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

The self reported financial stress measure used in the survey distinguished between respondents who reported to be ‘living comfortably’ (31.3 per cent of respondents), ‘doing alright’ (32.4 per cent), ‘just about getting by’ (26.2 per cent), ‘finding it quite difficult’ to get by (7.1 per cent) and a small though significant number who reported ‘finding it very difficult’ to get by (3.1 per cent, n = 46).

The table below indicates that there is a clear linear relationship between level of financial wellbeing and the size of informal networks: those who are living comfortably have the largest informal networks and those who are finding it difficult financially have the smallest. Examination of the size of each separate type of network for each group confirms that this relationship holds for all informal network types except neighbours (family, friends and previous work-mates).

There is clearly also a linear relationship between respondents’ level of financial wellbeing and the number of group memberships and institutional ties they have: those who report feeling financially well off have the highest number of group memberships and institutional ties, and those who are least well off have the fewest.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test for homogeneity of variance.

Results indicate that the extent to which respondents report that they and their family are getting by financially is related to the levels of trust and reciprocity they report across each type of network.

The following two tables show a clear pattern in which levels of reported trust and reciprocity across each type of network are higher among those who report either living comfortably or doing alright/getting by, than for those who report finding it quite or very difficult to get by financially.

It is striking that this pattern holds across networks within the family household, among other familiars, generalised norms as well as for trust in institutions. Indeed, levels of confidence respondents report in institutions demonstrate differences between those doing well and those struggling, most clearly. Those living comfortably report an average level of trust in institutions of 5.37 (on the standard scale of 0 to 10, where 0 is no confidence and 10 is complete confidence) while those who report finding it very difficult to get by report an average level of institutional confidence of only 3.89.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

4.1 Geographic proximity

One of the central tenets of the social capital thesis is that unless you use it, you lose it. Previous Australian research indicates that, as might be expected, the more close network ties are, the more contact people are likely to have with them (see for example Millward 1998). While there are exceptions to this general pattern (for example in the case of migrants, and highly mobile people who have regular contact with persons living long distances away), we might assume that geographic proximity is likely to increase the norms of trust and reciprocity that exist among network members, as favours are more frequently exchanged, and relationships are frequently negotiated and renegotiated.

The following two tables present data about the proximity of informal networks to respondents (whether members of different types of informal networks live in close proximity to respondents), as well as the proportion of each network type living in the local area.

The key thing to note in the tables below is the considerable variation in proximity of family and friends.

Note 1. Includes boyfriend, girlfriend or partner with whom don't live.

Note 1. Includes boyfriend, girlfriend or partner with whom don't live.

Examination of the correlation between geographic proximity of social networks and reported levels of norms of trust and reciprocity provides little support for the predication that the closer networks are, the more likely these are to be trusting and reciprocal. The proportion of respondents’ informal networks in the neighbourhood/ local area is positively related to trust and reciprocity in the local area. However, it bears little relationship to trust/reciprocity within informal networks of family and friends¹⁹, and no relationship to generalised trust/reciprocity, or trust in institutions.

4.2 The rural — urban divide

Seventy-eight per cent of the FSAC sample reside in capital cities or other metropolitan centres (n = 1176). Around twenty-one per cent live in rural or remote areas (n = 311).

Given inherent differences in the ‘communities’ and nature of community life in which urban and rural dwellers reside, we might expect to see differences in the norms governing the family, informal, general and institutional networks of rural compared with urban respondents.

Results indicate people in rural and remote communities have slightly higher levels of trust and reciprocity in their informal networks as well slightly higher levels of ‘generalised’ trust and reciprocity. There were no significant differences found between rural and urban respondents in the extent to which they trust institutions overall. However, when the level of institutional confidence urban and rural dwellers reported was examined for each institutional type separately, significant differences are seen between the levels of confidence respondents reported in both federal government and state government.

Urban residents reported higher levels of confidence in each of these levels of government (4.58 in federal government and 5.01 in state government) than their rural counterparts (who reported an average level of confidence of 4.19 in federal government and 4.56 in state government respectively).

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

4.3 The relative socio-economic status of areas

In order to explore the relationship between social capital and the socio-economic circumstances of local areas respondents live in, we used two measures, both based on ABS (1996) Socio-Economic Indexes for Areas (SEIFA).

The first of these distinguishes between respondents living in areas of low and high disadvantage. The second distinguishes between respondents living in relatively low and high areas of economic wealth²⁰.

There were no significant differences found between levels of trust and reciprocity reported for family networks and networks of familiars using either locational measure. Generalised norms of trust and reciprocity, as well as trust in institutions, do appear to vary, however, in relation to locality type. Respondents living in areas of less disadvantage, and with higher levels of economic resources appear on average to have slightly higher levels of trust and reciprocity (generalised and institutional).

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

***=P < 0.005 , **=P < 0.01, *=P < 0.05.

1. Fails Levene test of homogeneity of variance.

As outlined above, the concept ‘social capital’ has attracted much academic and policy attention as well as much critique. Regardless of whether one is ‘in favour’ of the concept or not, it is undeniable that at a time of much social change, social capital discourse has focused attention upon new ways of thinking about social connectedness, and has re-focused attention on the importance of ‘the social’. Thus far, however, debate has raised more questions than it has addressed.

This paper has presented initial findings of the Families, Social Capital & Citizenship project, which aims to address some of the considerable gaps in our knowledge about how social capital is manifest, generated and distributed.

To do so, the paper has presented some early findings around three key questions: what is the nature of social capital; how is social capital distributed among individuals and families; and how is social capital distributed across communities. Much of the paper focused upon the first of these questions.

5.1 Summary of initial findings according to key questions

In order to begin to understand how social capital is manifest empirically, we first examined what each of the core dimensions of social capital — networks, norms of trust and norms of reciprocity — look like empirically. We also explored how these operate together within different network types.

Beginning with an examination of the extent to which people appear ‘connected’ in different types of social networks, results suggest that almost all people have some kind of informal ties (with family and/or friends, neighbours workmates), although there is great variation in the size of informal networks and the extent to which these networks are made of family members (ranging from less than one per cent as a proportion, to 100 per cent). As well, it appears that the vast majority of people (86 per cent) belong to some sort of community group or organisation (ranging from involvement in children’s groups, sport, hobbies, education and arts, self-help groups, community organisations and charities, among others). A surprisingly high proportion of respondents (93 per cent) also reported having at least one tie with an institution (of power/authority).

Together, these findings provide early evidence for high levels of what has been termed ‘bonding’, ‘bridging’ and ‘linking’ forms of social capital — relating to ties with familiars, ties with people different to or distant from oneself, and ties to power.

Examination of the norms governing these relationships indicates that people express high levels of trust and reciprocity overall. Levels of trust and reciprocity tend to be highest among familiars and relatively high in general (the extent to which people trust others in the local area and report that local people are willing to help each other out; as well as a broader sense of trust in people in general, and belief that people are generally willing to help each other out). Mean levels of trust appear lowest in institutions.

Analysis of these data confirmed that we can think about four main ‘types’ of trust and reciprocity: among family households; among other familiars (family and non-family out of the household); generalised norms of trust and reciprocity; as well as institutional confidence. It also appears, as we might expect, that these norms of trust and reciprocity are positively and highly correlated with one another within each type of network.

As a first step to understanding how social capital at one social scale relates to social capital at another, we examined how one dimension of social capital relates to that same dimension in another type of network.

We found a fairly strong positive correlation between the size of family and non-family networks. That is, people with large family networks also have large non-familial informal networks (friends, neighbours and/or work contacts). It also appears that people with large informal networks are also more likely to belong to many civic groups and/or organisations and are more likely to have broad institutional ties (that is, to know people in many different types of institutions), particularly those with large non-familial informal networks.

With regards to the question of whether trust at one level (for example within the family) generates trust at another level (for example within the community) we found mixed results. Correlations between the different types of trust indicate that there is a significant positive correlation between trust of family in the household and trust of familiars; between trust of familiars and generalised trust; and between generalised trust and institutional confidence (these correlations were around .3 and were significant). However, the relationship between family trust and generalised or institutional trust was weaker (for example .12 between family trust and institutional confidence). These findings support the idea of a ‘ripple out’ effect for trust, rather than a transference of one type of trust into another.

Another key hypothesis this paper began to explore concerns the relationship between the structure and quality of relations. The work of Coleman (1988, 1990) in particular, suggests the size, density and diversity (heterogeneity) of networks affect the norms governing those relationships.

In exploring these three types of network characteristics, we found mixed support for the social structure thesis. We found that the size of networks is not typically related to the levels of trust and reciprocity within that network. However, there is evidence to suggest that network density (the extent to which networks among family and friends overlap) is positively related to trust and reciprocity of familiars. Similarly, the diversity of informal networks (measured through the extent that networks were mixed in terms of ethnicity) appears related, albeit weakly, to the extent of trust and reciprocity among familiars (those who were most homogenous had higher levels of trust and reciprocity). Educational diversity did not appear to be related to norms of trust and reciprocity within informal networks.

The final part of the paper briefly examined the way the core dimensions of social capital are distributed among Australian individuals, families and communities.

We examined a number of individual and family characteristics, including sex of respondent, family type, main language and language spoken at home, educational attainment, labour force status and self-reported financial wellbeing. We found evidence to suggest that a number of these individual and family level characteristics are related to the size of people’s networks, as well as to their overall reported levels of trust and reciprocity across these networks.

A number of these individual and family level factors are likely to be jointly affecting results and need to be explored in more detail before the relationships between these variables and social capital can be fully understood. In particular, it appears that the level of self-reported financial wellbeing (which is generally negatively related to the size of people’s social networks as well as the levels of trust and reciprocity they report) may be affecting the relationships observed between social capital and family type and labour force participation, respectively.

We also briefly examined how the dimensions of social capital are manifest in different types of geographic areas. In doing so, we found some differences between rural and urban based respondents, as well as some differences between respondents according to the level of socio-economic disadvantage and wealth of their local areas. Again, it appears that while some differences were found across all these variables, more detailed analysis is necessary before the relationship between locality and social capital can be fully understood.

The initial findings presented in this paper indicate some support for ‘the social capital thesis’. Levels of trust and reciprocity are highly correlated within network types and these are affected by the structure of social relationships, such as the density and diversity of social networks. These early empirical results appear to provide support for the conceptual framework developed in the Institute’s Families, Social Capital and Citizenship project for thinking about social capital and family and community life.

As well the paper suggests that the nature and distribution of social capital may vary according to some social and community characteristics, particularly financial wellbeing, labour force status, the extent to which networks are proximate, urban-rural divides, and the level of resourcing of localities in which respondents live.

Clearly, the findings presented in this paper are preliminary only and will be more fully explored in future work of the project. That work will analyse in more detail the relationships between the core dimensions of social capital, before further examining how social capital is distributed among individuals, families and communities.

As well, in order to further test the ‘social capital thesis’, future work of the FSAC project will link the empirical measures of social capital explored in this paper to a range of family, community, economic and political outcomes. We will examine whether indeed social capital is as important as the literature suggests, and in doing so will test this against alternate theories (for example that it is the ‘power’ rather than quality of networks that leads to key outcomes, or some combination of the two). Only in this way will we be able to determine how well the social capital concept explains the relationship between social relations and outcomes of interest — and how well it might meet the policy expectations of it.

Black, A. and Hughes, P. (2000) ‘Social capital in urban life’, Paper presented to Quality of Life Conference, Singapore, February 2000.

Bourdieu, P. (1993) Sociology in Question, Sage, London.

Coleman, J. (1988) ‘Social capital in the creation of human capital’, American Journal of Sociology 94:S95-120.

Coleman, J. (1990) Foundations of Social Theory, The Belknap Press of Harvard University Press, Cambridge MA.

Cox and Caldwell (2000) ‘Making Policy Social’, in Winter, I (ed) Social Capital and Public Policy in Australia, Australian Institute of Family Studies, Melbourne, pp. 43-73.

Dasgupta, P. (1988) ‘Trust as a commodity’, in D. Gambetta (ed.) Trust, Basil Blackford, Oxford, pp. 49-72.

Finch, J. (1989) Family Obligations and Social Change, Polity Press, Cambridge.

Finch, J. and Mason, J. (1993) Negotiating Family Responsibilities, Tavistock Routledge, London.

Fine, B. (2000) Social Capital versus Social Theory: Political economy and social science at the turn of the millennium, Routledge, London.

Paxton, P. (1999) ‘Is social capital declining in the United States? A multiple indicator assessment’, American Journal of Sociology, Vol. 105, No. 1, pp. 88-127.

Portes, A. (1998) ‘Social capital: its origins and applications in modern sociology’, Annual Review of Sociology, Vol. 24, pp. 1-24.

Portney, K. E. and Berry, J. M. (1997) ‘Mobilizing minority communities: social capital and participation in urban neighbourhoods’, American Behavioral Scientist, Vol. 40, No. 5, pp. 362-644.

Putnam, R. D. with Leonardi, R. and Nanetti, R. Y. (1993) Making Democracy Work: Civic traditions in modern Italy, Princeton University Press, Princeton, New Jersey.

Putnam, R. D. (1995) ‘Bowling alone: America’s declining social capital’, Journal of Democracy, Vol. 6, No. 1, pp. 65-78.

Sampson, R. J., Raudenbush, S. W. and Earls, F. (1997) ‘Neighborhoods and violent crime: a multilevel study of collective efficacy’, Science, 277, pp. 918-924.

Stolle, D. and Rochon, T. R. (1998) ‘Are all associations alike? Member diversity, associational type, and the creation of social capital’, American Behavioral Scientist, Vol. 42, No. 1, pp. 47-65.