1 Introduction: Family Structure and Social Capital

The initial conceptualization of social capital was closely linked to the family. Although Coleman’s (1988, 1990) description of social capital, including how it is created and its properties, could encompass different organizations and networks, family was prominent in his illustration of the benefits of social capital. He studied the impact of social capital on youth’s achievement, using measures such as presence of both parents in the household and employment of mothers. Since then, several studies have examined the impact of family social capital on development of children and youth using different measures of both social capital and childhood and youth outcomes (see for example, Bianchi and Robinson 1997; Boisjoly et al. 1995; Hoffert et al. 1998; McLanahan and Sandefur 1994; Modell 1994; Runyan et al. 1998; Teachman et al. 1997). Findings in these studies have not been consistent, possibly because measures used in attempts to provide empirical evidence of the effect of social capital on children’s outcomes, such as number of parents, number of siblings, and church attendance, have been, as Morrow (1999: 748) described, “crude and somewhat arbitrary”.

There are reasons why family structure, often measured by whether or not both parents are present in the household, is used as a measure of social capital. Compared to two parents, single parents would not have as much time and attention to interact with their children. Family disruption through divorce often leads to change of residence that in turn breaks established relations in previous residence. However, empirical findings show that these reasons for using family structure as a proxy for social capital do not always hold true. For example, development outcomes of children and youth in two-parent step families are not much better than those in one-parent families (Kerr and Michalski 2007; McLanahan and Sandefur 1994; Teachman et al. 1996); single-parenting has no effect on time spent with children (Bianchi and Robinson 1997); and, geographic mobility has not affected education outcome of children from high income families (Hoffert et al. 1998).

Furstenberg and Hughes (1995) suggest that, rather than using a unitary concept of social capital, it may be more useful to relate the various outcomes of children and youth to different types of social capital such as parental social network and embeddedness in the community, thus in effect, giving up on family structure (or presence of one or two parents) as indicator of social capital. But, there could be benefits to understanding how family structure relates to social capital, particularly because different family types have emerged from changes in formation and dissolution of unions over the past few decades. Types of families could differ in the ways of acquisition, management, and deployment of social capital to benefit the members, including children.

An assumption behind the research of Coleman and others is that intact family has greater social capital than lone-parent family. This assumption has not been examined in depth for reasons such as lack of conceptual clarity about social capital itself, as well as lack of data to measure it. This study therefore aims at a better understanding of social capital and its relationship to different types of families using a more focused concept of social capital and relevant empirical data.

We start with definition of social capital, focusing on one that is amenable to measurement. We then describe data and methodologies, discuss the results of different models—mainly for influence of family structure on social capital. We include also other relevant variables related to social capital, such as work status, religiosity and length of stay in the neighbourhood. We conclude with possible explanations and implications of our findings.

2 Social Capital Defined

Coleman (1990) and Bourdieu (1985) are often cited as the early proponents of the concept of social capital, both of whom have drawn upon the sociological tradition pioneered by Durkheim (1951). Coleman’s concept of social capital assumes that “individuals are embedded in a system of normative obligations created by social consensus” (Furstenberg 2005: 810). This system is drawn upon by families to benefit the members, an idea that is similar to Bourdieu’s that families’ symbolic and material resources could be mobilized for the benefit of its members (Furstenberg 2005). The introduction of the concept of social capital among political scientists is attributed to Putnam (1995, 2000) who has in turn drawn from the writings of de Tocqueville (1945) on political participation (Furstenberg 2005). In this line of thinking, social capital is seen in terms of social trust and civic participation.

Social capital is also viewed in terms of “networks”. Portes (1998: 8) defines social capital as the “ability to secure benefits through membership in networks and other social structures”. The Policy Research Initiative (PRI 2005; Franke 2005:9) adapts a similar definition: “Social capital refers to the social networks that may provide access to resources and social support.” A definition that is amenable to measurement and most useful for empirical research is the one by Stone et al. (2003:56): “networks of social relations characterized by norms of trust and reciprocity”. In an earlier paper, Stone and Hughes (2002: 2) distinguished three types of networks: (a) informal ties with kin, families, friends, neighbours, and workmates; b) generalized relationships with local people, people in civic groups, and people in general; and, c) relationships through institutions. With the purpose of measuring social capital, they also identified dimensions of networks, which include size and extensiveness (for example, number of neighbours personally known), density and closure (that is, whether network members know each other), and diversity (ethnic, education, and cultural mix of networks). An advantage of using type of networks and diversity in measuring social capital is that they help distinguish between the “bonding” and “bridging” nature of social capital (Granovetter 1973, 1995; Gittel and Vidal 1998; Woolcock 2001, 1998; Erickson 2003). Close relationships or “strong” bonds engender sense of belonging and are usually confined to a limited number of individuals. The notion of “bridging” social capital or “weak” bonds refers to a wider outreach and therefore may be more useful, say, for economic outcomes. A specific variant of bridging social capital is the “linking” social capital, which refers to a relation with people in position of power.

Thus, social capital takes different forms, has multiple dimensions, and can be measured for various units of analysis. For many proponents (for example, Bourdieu 1985; Lin 2001; and Astone et al. 1999), social capital is an attribute of individuals. For others, such as Coleman (1990) and McLanahan and Sandefur (1994), social capital is also an attribute of familiesFootnote 1 and communities. Putnam’s (1995, 2000) concept of social capital—and that of others, particularly those working at the World Bank (for example, Serageldin and Grootaert 2000; Narayan and Cassidy 2001), is for an even larger group such as regions or nations. The definition of social capital based on networks as proposed in this study implies individuals as units of analysis.

In whatever way social capital is defined, possibly specifying its multidimensional aspects, it is imperative not to confound the determinants of social capital from its outcomes. A criticism in the literature on social capital is that it is frequently confounded with its effects (see for example, Portes 1998; Fine 2001; Edwards 2004; Morrow 1999). In more recent literature (for example, Lin 2001; Narayan and Cassidy 2001; Stone and Hughes 2002; and PRI 2005), the frameworks of analysis differentiate the outcomes of social capital from its determinants. According to Lin (2001: 245–246), the determinants include “the factors in the social structure and each individual’s position in the social structure, both of which facilitate or constrain the investment of social capital.” Investment of social capital is expected to yield returns in terms of better social, economic, political, and health outcomes. At the individual level, the outcomes include better physical and mental health, life satisfaction, wealth, power and reputation (Lin 2001: 246), or the capacity to “get by” and to “get ahead” (Stone and Hughes 2002:2).

In line with the above discussion, this study focuses on family structures as a determinant of social capital that is measured by three types of networks: informal network, relationship through civic groups and people in general, and relationships through institutions. The first type of network broadly falls within the sociological stream of thinking about social capital, while the latter two fall within political science.

3 Data and Methods

3.1 The 2003 General Social Survey

The general social survey on social engagement was conducted by Statistics Canada with 24,950 respondents representing a target population of all persons in Canada 15 years and older, excluding residents of Yukon, Northwest Territories, and Nunavut, and all-time residents of institutions (Statistics Canada 2004). This study will focus on 8,250 women who, at survey date, were 30–64 years old, the ages at which variation in family structures is greatest.Footnote 2

The survey gathered information on a wide-range of topics including the respondent’s civic engagement, social networks, and participation in clubs, associations, and organizations, and voting and volunteering. The survey also collected information on the person’s background including education, work status, cultural background, health and well-being and information on his/her parents and partners.

3.2 Variables Used in the Analysis

3.2.1 Measures of Social Capital

We use information from the survey to derive measures of network dimensions—that is, the network size, norms of trust and reciprocity, and diversityFootnote 3. Appendix 1 lists the variables used to measure the types of networks and their dimensions. For informal networks, the information on the number of relatives and friends, and the number of neighbours that one knows, serves as indicators of size of informal networks. The information on the level of trust in one’s family, people at work, or in one’s neighbours is used as measures of trust and reciprocity. The respondents were also asked how similar their friends are with regards to level of education, family income, age, or ethnic group. This specific information is used for measuring the diversity dimension.

For networks through generalized relationship with people and civic groups, we use the information on whether or not the respondent was a member or participant in different organizations. Aggregating the response provides the number of organizations the individual is involved in, which is used as an indicator of network size. With the family in focus, we derive two different measures following a distinction made by Coleman (1990: Ch.22) between primordial structures that are based on or derived from family itself (such as neighbourhood and religious groups) and purposive structures that are independent of the family (such as firms, trade unions, and professional associations). The measure of diversity is based on questions as to whether the people that one met through the organizations were similar in terms of education, income, ethnic group, and age. Responses to questions of trust in strangers are used as an indicator of trust in people in general.

For the third type of network, the relationship with institutions, the level of confidence in various institutions such as the police, health care system, school system, etc., is used as an indicator of the trust dimension.

Appendix 1 lists the questions as they were stated in the survey. The answers to these questions were coded (or recoded, when necessary) so that the direction of answers would move from low to high social capital; for example, from none to several friends and relatives, or from cannot be trusted to can be trusted a lot, and so on. For the set of questions on diversity, the direction is from least to the most diverse; that is, from all are from the same group to none is from the same group.

3.3 Statistical Methods

3.3.1 Reliability Tests and Factor Analysis

We used factor analysis to obtain more parsimonious (latent) measures from the several responses used as indicators of the various dimensions of social capital. In instances where a measure is categorical or binary as in the case of whether or not a respondent is a member of an organization, we summed up responses to questions on membership in a number of organizations. Whenever the level of measurement (rank or interval) allows, reliability tests were done to find out which variables were correlated. These groups of variables were factor analyzed, and factor scores were derived for measures of the following dimensions: (a) size of informal networks of friends and relatives, (b) trust in people in the neighbourhood, (c) income-education-age diversity of friends, (d) income-education-age diversity of members of organizations, (e) confidence in government institutions, and (f) confidence in business institutions.

3.3.2 Bivariate and Multivariate Analysis

We used bivariate analysis to detect differences in the dimensions of networks by family structure categorized as follows:

  1. 1.

    Living with Children: (a) Intact-Married; (b) Intact-Cohabiting; (c) Step-Married; (d) Step-Cohabiting; (e) Lone Parent

  2. 2.

    Not Living with Children: (a) Married; (b) Cohabiting; (c) Never Married; (d) Divorced or separated; (e) All Others including the widowed and other living arrangements.

These categories are combinations of motherhood and marital statuses. Living with children (or motherhood status) could affect the acquisition of social capital. As Furstenberg (2005: 813) notes “…the presence of children requires parents to reach out to potential connections in the larger kinship system and the neighbourhood, through involvement in local community institutions”. Inclusion of marital status is meant to capture the differences in stability of relationship conducive to involvement with people outside of the family, with the community, and with institutions. Marriage is assumed to be more stable than cohabiting relationship.

To see whether the relationship between family structure and measures of social capital holds after controlling for other variables, we did appropriate multivariate analyses, progressively including in the models family structure, demographic (age), socio–economic (education, work status, income), cultural (religiosity, migration status), geographic (region of residence, urban–rural) and personal situation variables (length of stay in neighbourhood and self-perceived health status). Different scales of measurement (binary, rank, or interval) of the dependent variable call for building different models such as binary logistic, ordinal, or ordinary least squares regression models. To simplify the presentation of results, we show only the model coefficients (and their associated levels of significance) in the tables, which are to be interpreted in relation to the reference category; that is, positive coefficients indicate higher, and negative coefficients indicate lower, level of social capital compared to the reference groupFootnote 4.

As mentioned earlier, our discussion of the results will focus on the differences in social capital by types of families. However, we will also present the results for three of the control variables that have been often used as indicators of social capital: (a) work status, as women’s employment has been cited as a possible reason for decline of social capital; (b) religiosity, as the value systems held by those who frequently attend religious services are assumed to foster social capital; and, (c) length of stay in neighbourhood, as mobility breaks ties with neighbours, communities, and schools when children are present.

Fractional weights were used in all the statistical procedures to take into account the complex sampling procedure used for the survey.

4 Results

4.1 Descriptive and Bivariate Results

Table 1 presents the descriptive statistics of combined marital and parental (living or not living with children) statuses of women aged 30–64 at the time of survey. A little more than half (57%) of women aged 30–64 in Canada were living with children. Most (65%) of these mothers were married, a fifth were lone mothers, and the rest (about 5% each) were mothers cohabiting with their partners, married step-mothers, or step mothers in cohabiting relationships. Of the women not living with children, more than half (55%) were married. Many of these married and formerly married (that is, divorced, separated, widowed) women once lived with children who might have left the parental home.

Table 1 Canadian women aged 30–64 by combined motherhood and marital status 2003
Full size table

Table 2 shows the mean scores of the informal network indicators classified by marital and parental statuses of Canadian women aged 30–64 from the survey. These results from various bivariate analyses confirm that social capital does vary by family structure. Only a few of mean scores for all women by parental status are not significant at 1% level. In a multivariate analysis, however, some of these differentials by family structure either decline or disappear with the inclusion of other socioeconomic variables. Appendix Table 5, for example, presents the six (ordinal) regression models for the number of neighbours known. The bivariate model (Model 1 in Appendix Table 5) shows that step mothers have significantly smaller size of neighbourhood network compared to married mothers (with a coefficient of − 0.418), but after controlling for other variables, especially for length of stay in the neighbourhood (Model 6), the difference is no longer significant.

Table 2 Mean scores of informal network indicators by combined marital and parental status Canadian women aged 30–64, 2003
Full size table

In the final model (Model 6 in Appendix Table 5), mothers living with children either in intact-cohabiting, step-married or step-cohabiting relationship are no longer significantly different from the reference women (who are married and living with children) in their number of neighbours known. In contrast, lone mothers living with children and all women not living with children in whatever marital status are highly significantly different from the reference women. Although three sub-categories of marital/parental status are not significantly different, the overall variable “marital/parental status” or “family structure” significantly contributes to the explanation of the dependent variable “number of neighbours known” even in the presence of other socioeconomic covariatesFootnote 5.

Tracing the changes in influence of family structures on the various indicators of social capital as control variables are progressively introduced (as shown in Appendix Table 5) is not only interesting but also provides insights into the more complex relationships between family structure and other control variables. But an adequate discussion of these findings requires a much longer paper. In the interest of providing an overall view of the differences in the three types of social capital, our discussion will focus on the results of the “final” models, that is, models wherein all the independent variables have been included.

4.2 Social Capital Through Informal Networks: Differentials by Family Structure

Table 3 shows the results from the final regression model for the various dimensions of informal network—its size, norms of trust and reciprocity, and diversity. As mentioned in the methods section above, the type of regression model used for each indicator depends on its type of measurement, and this is shown in the last row of Table 3. Two variables (trust in family and ethnic diversity of friends) are binary and the models are binary logistic. Two more variables are ordinal and the models are ordinal logistic. The remaining three variables are factor scores and the models are ordinary regression models. The (pseudo) R-squares of all these models range from 6 to 26%. The ordinal logistic regression models have threshold values (or cut-off points) for respective categories, and they are provided at the bottom of the table.

Table 3 Final models of informal network indicators, Canadian women aged 30–64, 2003
Full size table

A thorough discussion of the impact of all the socioeconomic covariates in the model on various indicators of social capital is beyond the scope of this paper. In line with the aim of this paper, we will focus on the differentials by family structure and point out only some salient impacts of other covariates. The family structure variable has mothers in intact family as reference category, denoted in the table as “married with children”.

4.2.1 Size of Networks, and Trust and Reciprocity

Children do connect parents to networks beyond the family and help increase the size of neighbourhood network. As can be seen in Table 3, compared to women with spouses or partners and living with children (that is, all mothers except lone mothers), women not living with children have significantly smaller number of neighbours known (Table 3, col. 2). Children also play a role in generating greater trust in neighbours. Women who are not living with children in all marital statuses, except married women, have significantly lower levels of trust in neighbours than married women living with children (col. 5). That married women not living with children do not differ from married mothers in their level of trust in neighbours may be partly because many of them may have lived with children who have left to live on their own.

The absence of their children’s fathers has negative influence on lone mothers’ social capital. Compared to mothers in intact families, they fare considerably worse: their network size is smaller—they have fewer relatives, friends, and neighbours—and their level of trust in people in the family, at work or school, or in the neighbourhood is significantly lower. But compared to divorced or separated women not living with children (that is, neither children nor partners are present), lone mothers are in a better position. For the number of neighbours known (Table 3, column 2), for example, the coefficient for lone mothers is −0.381, whereas the coefficient for divorced or separated women is −0.893, indicating that lone mothers are more likely to know more neighbours than divorced or separated women not living with children. In between lie cohabiting women not living with children (that is, partners are present but children are absent) who are likely to know fewer neighbours (a coefficient of −0.648), while cohabiting women living with children (that is, both children and partners are present) are in a much better position (with a coefficient of −0.199).

Marital status matters as well, with marital disruption as an influential and differentiating factor. Among women living with children, the informal network of intact families differs significantly from those in step-families. Stepmothers, whether married or cohabiting, have significantly lower levels of trust, especially trust in family members, than mothers in intact families (col. 3). Married stepmothers have also significantly lower trust in neighbours (col. 5). Among women who are not living with children, similar differences exist between the married and the divorced or separated; that is, divorced women have fewer relatives and friends (col. 1), fewer neighbours that they personally know (col. 2), and have lower levels of trust particularly in people at work (col. 4) and in the neighbourhood (col. 5). All these may be an indication that when marriage breaks down, much more than the family is dissolved; the networks of friends, relatives and neighbours are disrupted as well. Furthermore, subsequent remarriage or cohabitation does not seem to mitigate the impact. An often cited factor related to disruption is physical mobility, that is, marital dissolution frequently necessitates a change of residence. However, this effect is net of the influence of length of stay in the neighbourhood, which has been controlled for in this analysis.

The type of family formation—marriage or cohabitation—does not seem to matter greatly when there are children. The informal networks of married and cohabiting women in intact families do not differ much, as indicated by coefficients (intact–cohabiting) that are not, or only weakly, significantly different from the reference category (married with children). Similarly, among mothers living in stepfamilies, there are more similarities than differences between the married stepmothers and cohabiting stepmothers. This finding indicates that when children are born within a union (or when there are children in the family), formal marriage no longer matters very much in terms of the relationship with friends, relatives, and neighbours.

4.2.2 Diversity of Friends

The last two columns of Table 3 also show the results from the final regression models on indicators of diversity in social status (measured by a factor score derived from information on education, income, and age) and ethnicity of friends. Married mothers stand out distinct from women in all other categories in that their friends are most similar to them in terms of income, education, and age. All the coefficients (in col. 6) are positive and statistically significant, indicating that the friends of women belonging to all other categories are more diverse in social status than the friends of married mothers. Furthermore, as indicated by the positive, statistically significant coefficients (col. 7), the friends of lone mothers (with 0.280 coefficient), never married (0.465) and divorced or separated women not living with children (0.286) are more ethnically diverse than friends of married mothers.

As diversity of networks is meant to capture the difference between “bonding” and “bridging” social capital, these results indicate that married mothers in intact families have stronger “bonds” and women in other categories have weaker. It would be tempting to conclude that the weak “bonding” social capital of women in the other categories is compensated for by the greater diversity of friends and thus greater “bridging” social capital that is generally regarded as useful in many ways, such as for generating economic outcomes. However, there are indications that this may not be the case. As discussed above, the size of informal networks is largest among married mothers. Furthermore, the coefficients for variables education and income (Table 3) show that women with higher education or income are more likely to have friends of similar social status and age. In other words, married women may have homogenous friends but these friends are more likely to have higher education and income like themselves. In contrast, women in other categories may have more heterogeneous friends in terms of social status and ethnicity, but many of these friends may also have low education and income, and thus their “bridging” or “linking” social capital would not be higher than those of the married mothers.

4.3 Social Capital Through Informal Networks: Differentials by Work Status, Religiosity, and Length of Stay in Neighbourhood

Early proponents of social capital assumed that employment of women decreases the social capital as they would not have time to interact with their neighbours. As shown in Table 3, this is partly supported by data. Compared to non-employed women, employed women, whether working part-time or full-time, know fewer neighbours (col. 2). However, working women have significantly greater number of friends and relatives (col. 1), and their levels of trust in people, in the family or in the neighbourhood do not differ from the unemployed (cols. 3 and 5). Furthermore, the trust in people at work or at school is significantly lower among women with part-time employment than women employed full-time (col.4). (The question of trust in people at work or in school were asked only of those who were employed or in school; at age 30–64, very few women belong to the “not-employed, or in school” category, and thus comparison is made mainly between part-time and full-time employed.) These results indicate that working outside the homes reduces interaction with others in the neighbourhood but increases access to networks through the workplace.

In the literature, one also finds religiosity used as an indicator of social capital. This seems to be warranted. As seen in Table 3, highly religious women are more likely to have more friends and relatives and know a greater number of neighbours (cols. 1 and 2). Further, women who profess no religion have significantly lower level of trust in people than women who are highly religious (cols. 3, 4, and 5).

The results for length of stay in the neighbourhood, often used as an indicator of social capital, show that it is indeed a determinant of social capital, but only as it refers to embeddedness in the community—that is, the longer the stay in the community, the greater the number of neighbours known and the greater the level of trust in people in the neighbourhood (cols. 2 and 5). Understandably, the number of friends and relatives does not depend on the length of stay in the neighbourhood as friendship and kinship are not limited by place of residence.

Diversity of friends does not differ significantly by length of stay in the neighbourhood, and only weakly by work status. Friends of full-time employed women seem to be somewhat more ethnically diverse (col. 7). Religiosity is associated with ethnic diversity of friends, which is open to two possible explanations: attendance in religious services facilitates friendship with people belonging to different ethnic groups, or the highly religious people come from diverse non-mainstream ethnic groups, which lends to a greater probability of friendship with the more plentiful mainstream ethnic groups.

4.4 Social Capital Through Membership in Organizations, and Through Institutions: Differentials by Family Structure

Table 4 presents the results from the final models of membership in civic groups, trust in people, confidence in institutions, and diversity of civic groups. The variable “membership in primordial organization” is binary and therefore follows a binary logistic regression model. Two variables, namely “membership in purposive organizations” and “ethnic diversity of members”, are ordinal and the models are ordinal logistic regression models. The remaining variables follow the ordinary regression models. As seen at the bottom of Table 4, the (pseudo) R-square values range from 4 to 31%.

Table 4 Final models of membership in civic groups, trust in people, confidence in institutions, and diversity of civic groups, Canadian women aged 30–64, 2003
Full size table

The effects of family structure on membership in primordial organizations (namely, religious-affiliated groups, school groups such as parent teacher associations, and neighbourhood or community associations such as block parents and neighbourhood watch) are similar to those of number of neighbours known (as shown in Table 3 and discussed above). As seen in Table 4 (col. 1), except for the coefficient for step-married mothers, all the coefficients are negative, implying that compared to married mothers, all other women are less likely to be members of primordial organizations. Much larger (negative) values for all groups of women not living with children imply much lower proportion of membership in primordial organizations. This is a further evidence of the active role that children play in connecting parents to communities, this time through family-related organizations.

As for trust in people (col. 3), married mothers in intact or stepfamilies and married women not living with children have higher levels of trust in people in general than lone mothers, cohabiting women with or without children, women who never married, and divorced or separated women not living with children. It may be that positive experience in marriage reinforces the trust in people while marital disruption reduces trust not only in family members but also people in general. However, as the analysis is based on cross-sectional data, a ‘selection effect’ may be operating here; that is, women who have greater trust in people may be more likely to get married and to stay marriedFootnote 6.

Membership in purposive organizations is not significantly (or only weakly) influenced by parental or marital status. This is to be expected given that these organizations or associations do not have much to do with families.

Lone mothers and the divorced or separated women have significantly lower confidence in institutions, both government (col. 4) and business (col. 5). This sets them apart from the never married and cohabiting women not living with children, who like them, have also lower trust in people in general (col. 3). It may be reasonable to surmise that experience of marital breakdown or dissolution has also eroded the confidence in all types of institutions, be they government or business, just as it weakens their trust in people—whether in the family, in the neighbourhood, or people in general.

The diversity in social status of members of organizations to which the individuals belong is significantly greater for lone mothers and divorced or separated women not living with children than for married mothers (col. 6). This is an indication of weaker bonding social capital but, whether this translates to stronger bridging social capital is open to question. Given that marital dissolution often leads to a decrease in income for women, it seems likely that their “linking” capital, that is, their bridge towards people with more power and resources, is less than that of married mothers even though membership in their organizations is more diverse.

The never married women stand out in the greater diversity of membership of organizations that they belong to, both with respect to social status (col. 6) and ethnicity (col. 7). Possibly, having neither spouses nor children makes them more free to join associations with greater diversity of members, just as they associate with people from all walks of life, as seen in their diversity of friends (shown in Table 3 in the last section). Recall however that their size of networks, whether informal networks of friends and neighbours or formal civic groups, is smaller than those of married women.

4.5 Social Capital Through Membership in Organizations, and Through Institutions: Differentials by Work Status, Religiosity, and Length of Stay in Neighbourhood

Contrary to the oft-cited assumption that employment of women leads to lower membership in civic organizations, Table 4 shows that employed women (whether full time or part-time) are more likely (than women not employed) to be members of primordial or purposive organizations (cols. 1 and 2). This implies that time (or lack thereof) is only one factor that influences membership in organizations. Another factor is exposure to and knowledge of organizations. Unemployed women may have the advantage of having more time but employed women may have a greater advantage in terms of the latter. The organizations to which employed women belong have also more diverse membership in terms of social status.

As seen in Table 4, religiosity has a significant positive influence on membership in organizations (cols. 1 and 2), whether primordial or purposive, and on confidence in institutions, government or business (cols. 4 and 5). When taken with the results on informal networks (Table 3), this indicates that religiosity is a determinant of, and could continue to be used as a proxy indicator for, social capital. In comparison to women who profess no religion or women whose attendance in religious services is infrequent, highly religious individuals have higher social capital whether seen in terms of size of informal or formal networks, in norms of trust and reciprocity, or in confidence in institutions. The organizations to which they belong are also more likely to be diverse in social status and ethnicity (cols. 6 and 7).

In contrast, length of stay in a neighbourhood has its influence mainly on membership in organizations and trust in people—those with <3 years of stay in the neighbourhood is less likely to be members and have lower trust (cols. 1, 2, and 3).

5 Conclusion

Studies that examined the impact of social capital on children’s outcome and well-being most often measured social capital crudely by whether or not children live with one or two parents. Their explanations usually focused on intra-family social capital, indicated for instance by parental time spent with children, parental expectations, monitoring of children’s activities, and parent–child communication. Many of these explanations are parenting practises, far from the original concept of social capital and its implications such as networks of relationships buttressed by norms of trust and reciprocity. In contrast, this study looks at the embeddedness of different types of families in the community through informal networks of families, relatives, friends, and neighbours, and networks through organizations and institutions.

Making use of various measures of network size and norms of trust, our study shows that social capital is greater in intact families than in lone parent families. Compared to lone mothers, mothers in intact families (especially married mothers) have larger informal networks, are members of more primordial and purposive organizations, have greater trust in people in the family, in the neighbourhood, and in people in general, and have greater confidence in government or business institutions. Most indicators that we have used also show that social capital of mothers in step families is in between that of married mothers in intact families and lone mothers. Children help embed families in the communities. This is most clearly seen in the number of neighbours known, trust in neighbours, and membership in primordial organizations, all of which are greater among mothers than among women not living with children.

The cross-sectional data used in the analysis precludes putting forward definitive causes for the differentials in social capital by family structures. In an attempt to advance the discussion, however, we offer some plausible explanations. As with intra-family social capital, lack of time to interact with the community may be one reason for social capital deficits of lone mothers. However, the similar low social capital of divorced and separated women who are not living with children indicates that lack of time is not a sufficient explanation, since women not living with children would presumably have greater amount of time to interact with friends, relatives, and neighbours or be members of organizations should they desire to do so. Another possible explanation therefore could be sought in the experience of marital disruption, common to lone mothers and to divorced and separated women. Marital dissolution, often accompanied with acrimony and severance of ties with family members, possibly brings about breaking of ties with informal and formal networks and consequently decreases trust in people.

The explanations cited above do not however, hold for the significantly lower social capital (in comparison to married mothers) of never married and cohabiting women without any children. A more plausible explanation may be sought in “selection effect”; that is, women who marry are selected for certain characteristics (such as sociability, and desire to settle or for stability) conducive to networking in the community, which are not dominant among those who have not married.

Measures of diversity of friends and organization members are meant to capture the strength of bonding social capital (implying the greater the homogeneity, the greater the bonding) and the extent of bridging social capital (implying the greater the diversity, the better the linkage and bridging). With few exceptions, the measures of social status diversity (measured in terms of education, income, and age) and of ethnic diversity indicate that married mothers have friends who are more similar to them in terms of social status and ethnicity than lone mothers, divorced and separated, and never married women. While this could be taken to mean that married women have greater bonding social capital, it is difficult to conclude that they have lower bridging or linking social capital. For one, their networks of friends and the number of organizations to which they belong are larger. Furthermore, if there is polarization in family life—that is, those with lower social status are more likely to form unions, have children at earlier age, and experience marital dissolution, which is probably the case among younger women in Canada—the networks of married mothers may consist of women with similar (higher) education and income as they have (Ravanera and Rajulton 2006). The homogeneity of networks of women in intact families may be a result of deliberate choices but the network diversity of women in other categories may have come about by chance. In sum, this means that measures of diversity may not be good indicators of differences in bridging social capital by family structures.

The results for other family-related variables deemed to affect social capital are mixed. Employed women tend to know fewer neighbours but they have greater number of relatives and friends and belong to greater number of organizations, both primordial and purposive, than unemployed women. That is, while employment reduces one form of social capital, particularly in relation to the neighbourhood, being employed connects women to additional network in and beyond the workplace. Religiosity has all the expected effects. Compared to women who profess no religion, highly religious women are more likely to have larger networks, both formal and informal, and have greater levels of trust in people and confidence in institutions. The social status and ethnic diversity of their friends and of members of organizations to which they belong is also greater. Understandably, the length of stay in the neighbourhood is positively related mainly to number of neighbours known, trust in neighbours, and membership in primordial organizations.

Our study provides further evidence of the vulnerability of lone mothers, who comprise about 20% of all mothers. Their vulnerability is accentuated by the finding in this study that lone mothers do not have as extensive a network—informal or through organizations—to fall back on as mothers in intact families have. What sets lone mothers apart from divorced and separated women (who also have lower social capital) is the presence of children who need their care. Reaching out to women who have undergone marital dissolution is made complicated by their having low confidence in institutions, public or private.