Structuring Social Capital: Social, Cultural and Institutional Dimensions
Nan Lin Yang-chih Fu Chih-jou Chen
Academia Sinica
This is a draft; do not quote without permission
5/ 24/2008
Paper presented at the International Conference on Social Capital, May 28, 2008 at the Academia Sinica, Taiwan. Data used in this paper were drawn from the thematic research project “Social Capital: Its Origins and Consequences", sponsored by Academia Sinica, Taiwan, through its Research Center for Humanities and Social Sciences, and the Institute of Sociology. The principal investigator of the project is Nan Lin. We acknowledge the assistance of Siyin Lee in the data analysis.
Lin, Fu and Chen 1 Abstract
Using the attribute data (gender, strength of ties and relationships) in the position generator from three societies (US, China and Taiwan) we examine how the societies are differentiated by how these attributes differentially describe access to social capital (embedded resources). We consider possible differential patterns relative to social (friendships), cultural (nuclear, extended and pseudo-kinships), and institutional (school, work, clan ties) dimensions. Results show that different patterns can be discerned between the US and China samples. For the US respondents, friends emerge as an important source in accessing better social capital. For the Chinese respondents, kin ties and institutional ties are more important. Taiwan seems to fall in-between with the continued but weakened significance of kin ties, the cont inued significance of institutional ties, and the increasing significance of friends. We suggest that these patterns are more consistent with an industrialization explanation than a cultural or political-economic regime explanation. The data also shed light on the nature of guanxi in Chinese societies and on the prevalence of the social principle (homophily) transcending developmental or economic processes.
Lin, Fu and Chen 2 Structuring Social Capital: Social, Cultural and Institutional Dimensions
For the past two decades, the position generator has provided a standard measurement instrument for social capital (Lin & Dumin, 1986; Lin et al., 2001; see the entire volume of Social Capital: An International Research Program, 2008).
The principal idea is to ascertain an individual’s access to various valued resources through social ties. Typically, occupations are assumed to carry significant resources
(capital) in most societies, so the instrument may consist of a list of sampled occupations. The instrument, asking an individual to indicate whether she/he knows of others having sampled occupations, may yield indications of the richness (access to high-status occupation) as well as diversity (number of sampled occupations accessed) of social capital for the person – access to capital through social ties. Thus the position generator is a theory-based instrument for social capital. As an estimation of social capital, the instrument has yielded reliable and valid empirical evidence of returns on social capital in life chances (e.g., market competition and social solidarity
(Lin, 2005)). Aggregation of such indexes and returns afford comparisons across social groups (gender, race\ethnicity, etc.) as well as across societies. The instrument can also be adapted for other types of valued resources (van der Gaag et al., 2005; van der Gaag et al., 2008) and for groups/organizations (Lin, 2006; Son & Lin, 2008).
One additional feature of the instrument is that it may also incorporate measurement of particular characteristics of the occupant of each accessed occupation
(e.g., gender and race\ethnicity) as well as the relations (e.g., strength of the tie and the role relationship) between the occupant and the focal individual. These attributes can help estimate sources of inequality in social capital. For example, one may pose the question: whether certain types of social ties (e.g., men or whites) are more likely to accessing rich or more diverse resources? This question is posed based on two
Lin, Fu and Chen 3 considerations: (1) the homophily principle favors individuals of certain characteristics in accessing social ties with similar characteristics and (2) individuals and groups of certain characteristics tend to occupy higher-status or accessing more diverse occupations in most societies. With information about attributes of tie characteristics (e.g., whether the occupant accessed is man or woman, white, black or
Latino), it becomes possible to assess whether access through men rather than women and white ties rather than minority members, for e xample, tends to reach rich and/or diverse positions – better social capital.
Multiple attributes incorporated in the position generator afford simultaneous examination of multiple characteristics of occupants and social relations as they are associated with rich or more diverse social capital. It then becomes possible to estimate the relative significance of attributes (e.g., is having a male tie more important than having a white tie to gain better social capital?).
The objective of this paper is to ascertain the relative significance of attributes for accessing richer and/or more diverse occupations as sampled in the position generator instruments in a study of three societies – US, Taiwan and China. Since the data come from three national surveys, we can further seek to understand if different structures of social capital exist across societies.
In the remainder of the paper, we will introduce the position generator instrument that secures attributes in regard to certain characteristics of the accessed occupants.
We then introduce some conceptual considerations about how the associations of these attributes and the accessed social capital may be expected to vary across the three societies. A brief introduction to the surveys and the da ta will be followed by analyses. The paper will conclude with some reflections of the conceptual issues.
THE INSTRUMENT - THE POSITION GENERATOR AND THE TIE ATTRIBUTES –
Lin, Fu and Chen 4 A typical position generator poses a question, “I am going to ask some general questions about jobs some people you know may have. These people include your relatives, friends and acquaintances (acquaintances are people who people who know each other by face or name). If there are several people you know who have that kind of job, please tell me the one that occurs to you first.” Then, a list of sampled occupations is provided and the respondent is asked to indicate if she/he knows someone in that occupation (“Is there anyone you know who is a nurse?”). The sampling of the occupations should be from a list of occupations listed in a rank order of “values” – for example, status or prestige scale scores. They could be either sampled from equal-intervals along the scale, or some random sampling process.
Not all initially sampled occupations should be used, as some of them may be obscured and unfamiliar to most people. Preferences should be given to occupations at or close to sampled occupations in rank which have substantive prevalence in the labor force (e.g., the size of occupants) thus are more familiar to the population.
Typically, the sample should consist of at least 10 or 15 occupations to provide reasonable variance in the data. A further refinement of the sample may be to add some occupations oriented toward certain social groups in the population (e.g., female-dominant occupations). In our surveys, we sampled 22 occupations.
An important feature of the position generator, as in contrast to the name position, is that it does not evoke any particular topics or issues on which the ties are based, thus avoiding the problem of a lack of knowledge of the population of topics or issues from which a probability sampling may be drawn. Also, the position generator differs from the name generator in that it focuses on positions in a social hierarchy rather than individual actors in the hierarchy. Thus the position generator estimates the extent to which an actor can access different positions in the hierarchy. In so far as such positions are reasonable estimates of valued resources in the hierarchy, the
Lin, Fu and Chen 5 position generator is a reasonable instrument to estimate accessed valued resources – social capital. Thus the primary aim of the instrument is to estimate access to capital in the hierarchy and its validity depends on (1) the reasonable representation of capital captured by the positions (Do the positions reflect differential valued resources?) , and
(2) a reasonable representation of the positions sampled (Do the sample of positions reflect the range of such positions?). In contrast, the name generator elicits other actors, regardless of their positions in the social hierarchy. It is well suited to study the ego-networks linking to other actors. Estimation of their relative vertical position and range is helpful to describe the clustering of hierarchical positions among stronger ties for ego, but it does not inform us as to whether ego can indeed access other positions in the hierarchy.
Still another feature of the position generator is the less likelihood of eliciting strong ties, as compared to the name generator, with its limited number of names elicited. However, it does not completely overcome the tendency to elicit stronger ties. An actor may know multiple occupants at a sampled position. Typically, as in our surveys, we elicit the first occupant that comes to mind. This strategy has the advantage of efficiency (quick response) in a survey, but it also tends to draw stronger rather than weaker ties , as we assume stronger ties come first to mind. 1 Thus, the position generator does not eliminate the tendency to draw out occupants of stronger ties for a sampled position. Inclusion of a reasonable size of sampled occupations may increase the likelihood of the inclusion of some weaker ties, but it does not eliminate the general tendency of stronger-tie inclusions. Nevertheless, we should remind ourselves that attributes examined for the occupants also may be biased toward descriptions of stronger ties.
When a respondent indicates knowing someone who occupies a sampled position
(occupation) , a series of questions follow to identify attributes – characteristics of the
Lin, Fu and Chen 6 occupant and relations between ego and the occupant. In our survey, we identified
(1) relationships (ranging from spouse, parent, to co-worker, boss, close friends, to ordinary friends, people from the same groups or associations), (2) gender of the occupant, (3) whether ego knows the person through the spouse, (4) how long ego has known the person, (5) how close ego perceives to be with the person (very close, close, so, so, not close or not close at all), and (6) in the US sample, the race\ethnicity of the person. For the present paper, we skipped spouse-mediated ties (it deserves a more intensive analysis) and years of knowing each other (as it is highly correlated with the closeness measure).
SOME CONCEPTUAL CONSIDERATIONS
We borrow some previous work on social relations and societal characterizations to formulate some tentative conceptual considerations as to what to expect in such a comparative study. First, we assume that there are three major components as bases of social relations: social, cultural and institutional. Social components consist principally of friends. For the time being, we identify two types of friends: close friends and ordinary friends and assume that this differentiation reflects strength of tie: ranging from very strong ties to very weak ties on such characteristics as intimacy, intensity of relations, frequency of interactions and reciprocity of services
(Granovetter, 1973). The cultural component consists of mostly ascribed relationships including kin ties. It is possible to differentiate kin ties into several groups: (1) the nuclear family consisting of spouse and children, (2) the vertical family ties: parents and parents-in-law, and (3) lateral and extended family ties, including other relatives not in the nuclear or vertical relationships. Institutional ties include both ascribed and constructed institutions such as clan (compatriot) ties, work ties, school ties, ties built on services and indirect relationships , including neighbors.
Lin, Fu and Chen 7 The three societies – US, Taiwan and China – may be aligned at different points along the three components. The cultural component may reflect the strength of kin ties, ranging from nuclear family ties to extended family ties. The institutional component may reflect saliency of various types of institutions ranging from ascribed such as clan or compatriots to constructed such as schools and work. The social component includes both close and ordinary friends. If we assume that social relations are layered – from the core to the peripheral as in binding relations, bonding relations, and belonging (Lin, 1986; Lin & Ensel, 1989; Son et al., 2008) , it is conceivable that social, cultural and institutional ties pla y different roles in the layered relations in different societies. What, then, explains the juxtaposition of the various components in the layered relations?
Three explanations may differentiate how these societies align layers of social relations along the three components. One explanatory scheme is industrialization and its associated process such as division of labor, urbanization and mobility. One may argument, for example, as a society becomes more industrialized, the kin ties become more nuclear and decrease its contributions to the core of social relations.
At the same time, institutional ties have moved from the ascribed to the constructed, and also become more peripheral in relations. For friends, two things may happen.
One is that it becomes generalized and detached from kin and institutional ties. In other words, friendship can be developed independent of cultural and institutional relations. Second, the strength of friendship becomes meaningful. Unlike the less industrialized society, where friendship is attached to kin ties ( “pseudo-family ”) and institutional ties ( a member of a clan is a friend), friends can be differentiated meaningful as close or ordinary friends. Thus some close friends may move into the core of relations. If we assume that US, Taiwan and China occupy three points in this industrialization process, then we should expect to see differentiations of social
Lin, Fu and Chen 8 relations in such configurations.
A second explanatory scheme is culture. Principally it argues that the demarcation of nuclear versus extended families is a critical determinant in social relations. Further other non-kin ties assume significance only in so far as they are seen as part of the extended family – members of a pseudo-family. Institutional ties
(e.g., clan and work ties) are also significant as they may become part of this pseudo-family. For this argument, we would expect that kin ties are salient for
China and Taiwan but not for US. Other relations assume salience if and when they are aligned with kin ties. In general, then, social relations in Taiwan would be expected to look more like those in China than in the US.
A third explanatory scheme reflects the economic -political regimes. The three societies may be differentiated by being either capitalis tic – US and Taiwan – and transitional from planned and command regime to the capitalistic regime - China. In this scheme, we expect Taiwan to be closer to the US than to China. We should see that extended and institutional ties losing their grips in Taiwan. On the other hand, friends gain significance and become detached from kin ties and institutional ties (e.g., work, school, and neighbor ties) in Taiwan.
We will examine the data and see how they would fit into these explanatory schemes. We suspect the fit will not be clean as there are only three observations.
However, it would be informative to see how the data sustains each of the explanations.
THE SAMPLING PLANS
USA:
A random-digit dialing telephone survey was conducted, November 2004 –
March 2005, among adults (between twenty-one and sixty-four years of age) currently
Lin, Fu and Chen 9 or previously employed in the United States. A CATI system was utilized.2 As the survey unfolded it became clear that the response rates from minorities (especially
African Americans and Latinos) were lower than that from whites. To ensure a reasonable representation of minorities in the sample to reflect the census distribution, we imposed an additional sampling criterion – to seek out qualified African
Americans and Latinos to approximate the census distribution. To estimate and control potential bias that such a sampling modification might incur in the data, we created a dummy variable, quota, to distinguish all sampled respondents after the imposition (value=1) and took it into account in all analyses. In all studies conducted so far, the bias was minimal and insignificant.
The final sample consisted of 3000 respondents , for a response rate of 45 percent. For the present study, only those who were currently working were used, for a study sample size of 2,317.3 The sample splits evenly between men and women. Slightly more than two-thirds (70 percent) of the sample are non-Latino white, and about 12 percent are Latinos and African Americans, respectively. About
12 percent of the respondents are foreign born. Most of the respondents are either married (63 percent) or have never been married (22 percent), while 12 percent are divorced.
Taiwan
The survey was conducted in 2004 in interpersonal interviews with respondents in an island-wide three-state stratified probability sample of adults between the ages of 21-64. At the first stage, the island was divided into ten clusters in accordance to urbanization and economic development criteria and proportions of cities and towns were sampled from each cluster. At the second stage, systematic sampling of neighborhoods was drawn from each sampled city\township. At the
Lin, Fu and Chen 10 final stage, systematic sampling of households was drawn from each sampled district.
The response rate is 48 percent. The total sample size was 3,281, of which 13 percent were housewives , 8 percent were either unemployed or students, and 4 percent were retired. The study sample consisted of 2,406 respondents, including those who were currently employed with complete answers on all measurements in the analysis, excluding housewives, those retired, or those not currently participating in the labor force. The study sample consisted of more men than women (59 percent to 41 percent). The respondents’ average age was 40, with an average of ten years tenure with current job. As to the education, 44 percent of the respondents had advanced degrees (associate college or higher), and only 23 percent received secondary education or lower. The average monthly income was in the range of NT$
30,000 to NT$ 40,000 (about US$ 900 to US$ 1200).
China
The targets of the survey were adults, ranging in age between twenty-one and sixty-four, currently or previously employed, registered and residing in urban cities. The sampling plan was a multistage systematic probability sample. At the initial stage, households in all urban cities were sequenced. Clustering of nineteen consecutive households w as the basic unit of sampling at this stage. The desired sample size was determined to be 3,535 in order to provide adequate sub-samples for possible subgroup (gender, region, and so on) analyses. An interval was thus established and a random start number was drawn. Starting from the top of the list of all households-by-cities, clusters of households were sampled, resulting in 184 clusters from 167 cities.4
At the second stage, all qualified respondents in each sampled household were identified, and the one whose birth date was closest to June 30 was designated as the sample respondent. A personal interview was conducted by professional
Lin, Fu and Chen 11 interviewers between November 2004 and March 2005. After up to three attempts with each designated respondent and further sampling from adjacent households if previous households failed to yield respondents and the response rate was 35.7 percent. The relatively low response rate was due to several factors. We held the sampled respondents rigidly and without replacement. Respondents who could not be contacted and interviewed after the initial attempt and follow -ups were counted as failure. We also found that more and more urban residents in China are becoming less inclined to be interviewed, and we made no effort to force their participation.
The effects of the low response rates are difficult to estima te. It is hard to find the appropriate census data from the sampled cities for currently or previous employed adults aged twenty-one to sixty-four. However, as will be shown below, to whatever extent possible comparison of sample characteristics with Chinese census data was made. Strong correspondence in key parameters assures us that the sample provides credible estimates for the population under study. To further ensure against possible sample biases, the study will incorporate critical control variables in all analyses.
The total sample size was 3,529. Approximately, the sample was evenly distributed between men (49.4 percent) and women (50.6 percent). T his is consistent with the Chinese census (51.5 percent of 2005 urban residents are male).
The a verage age was 39. The average year of education was 11.4. This is a higher number than the 2004 1 percent national survey data for the urban population, where the average year of education was 9.4 for urban residents aged 6 and over. However, since we are focusing on employed adults only, an elevation of education should be expected.
The median income was in the range of RMB$10,000 – 14,000 per annum. 5 The
2005 national survey showed a median of RMB$10,493 for urban residents. For sectors of work unit, our survey showed that 56.4 percent of the respondents were in
Lin, Fu and Chen 12 the state sector, and 8.8 percent were in the collectives. The 2005 1 percent national survey showed that the corresponding percentages were 57.9 and 7.1 respectively.
We therefore conclude that the study sample of employed urban residents shows a high degree of representation of the national parameters. For the present study, only those currently employed are included (N = 2706).
MEASUREMENTS
The sampled positions in the position generator for the three surveys are listed in
Appendix.
Male and female ties. We counted the number of male ties and female ties to the accessed positions for each respondent.
Strong and weak ties. From the closeness question, we grouped responses (a dummy varia ble for each respondent) the “very close” and “close” categories as strong tie and the “not close” and “not close at all” categories as weak tie. Then, we summed the total strong and weak ties used by each respondent to access positions .
Race\ethnicity. For US data, we created categorical variables of white, black,
Latino, Asian, Native American and other for each tie and added up the total number of such ties used for each respondent.
Relationships. There were 29 categories of relationships used in the instrument.
We used factor analysis to find clusters of relationships. Details will be discussed later in the paper.
Social capital indexes. We constructed the typical indexes: upper reachability
(the highest status score of accessed occupations), extensity (the number of accessed occupations) and the range (the difference between the highest and lowest statuses of accessed occupations). For this paper, we only used reachability and extensity as the range is substantively correlated with the other two indexes.
Lin, Fu and Chen 13 Social capital position clusters. We also conducted factor analysis for the 22 occupations for access responses. Details of this analysis will be discussed later in the paper.
DISTRIBUTION OF TIE CHARACTERISTICS
Table 1 presents the distribution of tie characteristics in the three surveys. The average number of accessed occupations (extensity) is 7.43 for the US and 7.52 for
China, with similar variances. However, the extensity average for Taiwan is 9.15, significantly higher than both the US and China. One possible source of the variation is the larger of percent of men in the labor force, as reflected in the Taiwan survey (59 percent being male). We examined the correlations between being a male and the extensity score and found it to be -.8 for US, .12 for China and .00 for Taiwan.
So, the presence of more men in the Taiwan sample does not explain the higher extensity. For the time being, we have no explanation for the higher extensity score
(indicating greater diversity of social capital) for Taiwan.
(Table 1 about here)
The analysis of the gender of ties shows that there are about equal number of male and female ties used in the US (3.86 and 3.55 respectively), but more male ties than female ties in Taiwan and China. We then checke d the correlations between being male and using male ties. For the US, the correlation is .09; for China it is .26, and for Taiwan it is .28. Therefore the principle of homophily (being a male and naming male ties) is relatively weak in the US, but relatively strong in Taiwan and in
China. The effect of homophily may be the greatest in China as the lopsided choices for male ties rather than female ties (4.91 versus 2.61) cannot be accounted for by the gender distribution in the sample (almost equal number of male and female respondents in the China sample). We conclude that gender homophily operates in
Lin, Fu and Chen 14 different degrees across societies.
For the US sample, we find that white ties account for about 70 percent of ties used, black ties for 13 percent, and Latino ties 10 percent. Again, checking against the sample race/ethnicity distribution (white accounting for 70 percent of respondents, blacks 12 percent and Latinos 13 percent), they seem to reflect the effect of homophily. When we verified this possib le linkage with correlations between race\ethnicity of the respondents and the use of race\ethnic ties we find that the association is .56 for whites naming white ties, .76 for blacks naming black ties and .70 for Latinos naming Latino ties. Thus, the hom ophily holds again - but it is stronger for black and Latino respondents than for white respondents. That is, homophily varies across racial\ethnic lines in the US. It is striking that blacks and
Latinos access positions through co-racial and co-ethnic ties, despite the presence of significantly more whites in the population.
These findings confirm the general finding that homophily – a principle of networking choices – overpowers the structural effect – a principle of demographic determination. However, it differs on attributes in different societies. The effect on gender homophily is strong in Taiwan and China, but not so strong in US. For US, instead, racial\ethnic homophily is very strong. We suspect segmentation of society is along the gender line in Taiwan and China and along the racial\ethnic lines in US.
We then examined the patterns of tie strengths across the three societies. As can be seen in Table 1, naming strong ties (e.g., very close or close) accounts for 50 percent of ties in the US sample, 41 percent in the Taiwan sample and 45 percent in the China sample. Naming weak ties accounts for 20 percent in the US sample, and
12 percent each in the Taiwan and China samples. Also, about 30 percent of the ties are in the middle (“so, so”) category in the US, 47 percent in the Taiwan sample and
43 percent in China. That is, there is a greater tendency for US respondents to name
Lin, Fu and Chen 15 ties as either strong or weak rather than “so, so” as compared to the Taiwanese and
Chinese respondents. In particular, naming weak ties is much more infrequent for the Taiwanese and Chinese respondents than American respondents (i.e., the ratio is about one to two). We conclude, tentatively, that the use of strong ties is pervasive in all three societies, much more so than the use of weak ties. It may reflect the real strength of strong ties – the tendency for associating with others closer in the social and spatial space. We will further discuss this issue later.
The greater use of weaker ties in US as compared to that in Taiwan and China may be due to two explanations: industrialization and cultural influence. As mentioned earlier, industrialization brings about loosening of family ties and stronger ties. Yet, culture may also dictate the more reliance on stronger ties. We need further data to sort out between the two alternative explanations.
DISTRIBUTION OF RELATIONSHIPS
In the position generator, when a respondent affirmed access to an occupation, we asked about the occupant: “what is his/her relationship to you?” Listed were 28 relationships for the respondent to check. If multiple relationships applied, we asked the respondent to indicate the most important one. The data, with relationships receiving at least 5 percent response rate, was submitted to a factor analysis and the resulting factor solutions are presented in Table 2 a – c for the three societies. 6
(Tables 2-a, 2-b, and 2-c about here)
For the US, the final solution includes seven factors or clusters. As can be seen in Table 2-a, the first factor has three high-loading relationships: co-workers, the supervisor (“boss”), teachers and ties through a service provided to ego. We label it work relationships. The high-loading relationships for the second clusters include: parents and (not) child; the factor is identified as parent. The third factor includes
Lin, Fu and Chen 16 In-laws and spouse and is labeled as spouse. The fourth factor consisting of indirect ties and neighbors is labeled as distant ties. The fifth factor includes close friends and ordinary friends, and to a less extent, acquaintance, and is labeled as friends.
The sixth factor contains the high-loading item, previous co-workers, and is labeled as previous ties. The seventh factor has siblings and other relatives as high-loading items and is labeled as siblings\relatives.
It should be noted that the labels are arbitrary and the ordering of the factors reflect in part the presence of included items in the instrument. They do not indicate degrees of substantive significance. A full-information procedure (i.e., sum of the products of each loading coefficient and its corresponding response, a dummy variable) was used to calculate a score for each factor for each respondent. It is also interesting to note that for the US sample, three kin cluste rs have emerged: spouse, parents and siblings\relatives – representing one vertical relationship and two clusters of lateral relationships involving the nuclear and extended families..
Table 2-b presents the results and labeled factors for Taiwan. There are also seven factors : w ork ties (also teacher) , distant ties, friends, school ties\parent, neighbors\child, siblings\relatives, and previous ties. Comparing the factors of the
US and Taiwan data, they show largely similar patterns – involving clusters of work ties , friends, and distant ties. For the kin ties, however, spouse no longer constitutes a separate cluster in Taiwan; rather, it is submerged in clusters involving neighbors\child and siblings \relatives, suggesting that the kin structure in Taiwan largely implicates extended family ties and the nuclear family is not a salient and independent kin component. Also note that in Taiwan, instead of “clean” kin clusters in the US, kin ties are embedded with non-kin ties such as school ties (with parents) and neighbors (with child). It is premature to speculate the substantive meaning of this mixture until we explore their associations with social capital later.
Lin, Fu and Chen 17 Table 2-c presents the results and labeled factors for China. There are 8 factors, including distant ties, work (and teacher), (other) relatives, school ties \clan, neighbors, previous ties, Friends (not siblings) , and spouse. Compared with the US and Taiwan patterns, we note similar clusters of work and distant ties. However, the kin ties prese nt another configuration. Spouse becomes an independent cluster, suggesting the salience of the nuclear family in urban China. It reflects policy (one -child family) and ecological (small living quarters) constraints on the family structure. Again, similar to US, the kin clusters tend to be “pure,” rather than mixed as in Taiwan. In fact, the only mixing cluster shows opposite directions of friends and teachers versus siblings. Parents again take an opposite direction to other relatives. Again, its sig nificance is unclear at this point of analysis. Neighbors also become a distinctive cluster, again, we suspect, reflecting the ecological reality in urban China, where residential buildings bring neighbors close together. There is a cluster of school ties and compatriots (similar clans). School ties and clan ties (compatriots) now form a prominent cluster, representing a strong presence of institutional ties. Also note the weak presence of friends in the configuration. Ordinary friends are barely ident ifiable in two clusters (as in friends and siblings\relatives) and close friends show no significance presence in any of the clusters.
To get a clearer understanding of the substantive meanings of these varied configurations of relationships, we proceed to analyze their associations with social capital (positions representing differential rich and diverse resources) they access.
Table 3 summarizes the patterns of relationships for the three societies. We have regrouped the clusters into general categories of kin, work, and other ties.
While the major clusters seem to be similar for all three societies - work ties, friends, previous ties, distant ties and kin ties, the composition of each cluster varies. For friends, the US cluster is rather “pure”: consisting, for example, close friends,
Lin, Fu and Chen 18 ordinary friends and acquaintances. Friends in Taiwan share a cluster with kin ties.
Friends in China share a clustering with siblings but they are in opposite directions.
Some kin clusters in Taiwan are mixed wit h school ties and neighbors but not so in the US. Also, somewhat unique in China, a cluster represents school ties and clans, not seen in Taiwan or the US. But do they really carry different significance and consequences for the acquisition of social capital? A question we now turn to explore.
(Table 3 about here)
RETURNS TO SOCIAL CAPITAL
To assess the relative contributions of these tie attributes to social capital, we employed the two major indexes of social capital as the dependent variables: Upper reachability is the highest status score among all accessed occupations for each respondent, and Extensity is the number of accessed occupations. Upper reachability can be seen as an index of richness of social capital – the estimated high point among social ties ego has in the hierarchical system of the society. Extensity is an index of diversity of social capital – the estimated extent to which ego accesses resources vertically in the hierarchy. Thus the general question posed is: which of the attributes contribute relative more in accessing rich and/or diverse social capital?
First, zero-order correlations between each of the two social capital indexes and the attributes were performed for each society. Those attributes that had no or weak associations (coefficients less than .10) were eliminated and the remaining attributes became independent variables. All qualified attributes were simultaneously entered into the equation for each dependent variable. The only exceptions are the male ties and female ties. As they are substantially complementary and would create multi-colinearity biases if entered together, so we create two separate equations: one equation incorporates the male ties and the other the female ties. In most cases, both
Lin, Fu and Chen 19 male ties and female ties showed significant contributions to the dependent variables, while taking into account all other variables. We will report here only the equation with either the males or females ties whichever showed greater contributions (the other equation and results are available upon request).
Table 4 shows the ordinary linear regression equations for upper reachability for
the three societies. All coefficients are partial standardized coefficients. With the
large sample sizes, most of the coefficient s are statistically significant and the
important assessment is the relative contributions of the attributes; the standardized
coefficients effectively make such relative assessment possible.
(Table 4 about here)
For US, as can be seen in the first column in Table 4, male ties are the most
important attribute contributing to upper reachability. Friends, weak ties, white ties
and strong ties also make important contributions. Previous ties and work ties also
make some contributions. In contrast, work ties, relatives and distant ties make
little contributions.
For Taiwan, male ties also are important, followed by school ties, work ties,
distant ties and friends. Weak ties, siblings and strong ties make some
contributions as well. Friends seem important in both US and Taiwan. However,
as we noted before, friends in Taiwan cluster with relatives while they constitute by
themselves in US. It should be noted that for both US and Taiwan, weak and
strong ties almost make equal contributions to upper reac hability.
For China, male ties also are the most important. Then, work ties, relatives,
school ties and strong ties make important contributions as well. Friends appeared
in two clusters. Ordinary friends had little significance. However, close friends
and some ordinary friends, along with kin ties (other relatives) were important.
The patterns across the three societies show some discernable trends. If upper
Lin, Fu and Chen 20 reachability reflects vertical accesses to better positions in the hierarchy, male ties
are more important than female ties in all three societies. However a contrasting
pattern is discernable between US and China respondents. US respondents rely on
friends and both stronger and weaker ties; China respondents rely on stronger ties,
work and school ties, and kin ties. Taiwan seems to be in-between. Its
respondents rely on both stronger and weaker ties and friends – like those in the US,
but also on work ties, school ties and kin ties – like those in China. Also Taiwan
and China similarly s ee friends as useful only if they are part of kin tie clusters.
The data suggest, therefore, utility of social relations in the US may be shifting
away from kin ties and institutional ties and toward generalized social ties – friends
which are independent of kin and institutional ties. Utility of social relations in
China remains with kin ties and institutional ties (e.g., work, school and clan) and
some friends who may be seen as part of the family or the pseudo-family. Taiwan
retains utility of kin and some institutional ties (e.g., work and school) as well as
some utility of weaker ties.
It should be noted that in US, white ties are important. There is no comparable
measure for China and Taiwan.
Table 5 shows the results for extensity – diversity of social capital. A general
trend is that the US respondents rely on a variety of ties – male ties, strong and weak
ties, white ties, work ties, friends and even kin ties. For the Chinese respondents,
male ties, strong and weak ties, friends and, somewha t, school and kin ties are useful.
Taiwan seems to be more like US, relying on a variety of ties, ranging from male ties,
strong and weak ties, and friends. However, Taiwan respondents also rely on
distant ties such as service ties, indirect ties and old neighbors. They also rely on
school ties and parents. While male ties seem to be less important in the US, strong
ties seem to be more important. White ties are also very important.
Lin, Fu and Chen 21 (Table 5 about here)
Thus for diversity of social capital, a variety of ties are used in all three
societies – both strong and weak ties, close (close friends, kin ties) and distant ties.
CLUSTERS OF SOCIAL CAPITAL
Since social capital in the position generator is constructed from accessed occupations, it is possible to see how these accessed occupations form clusters.
These clusters inform us how likely accessing one occupation is associated with the likelihood of accessing another occupation. With the 22 occupations as dichotomous variables (accessed or not), we factor-analyzed the groupings for each society.
Again, we may label each cluster in terms of the high-loading occupations.
As can be seen in Table 6-a, there are four independent clusters for the US sample: labeled as high-level professionals (represented, with loadings higher than .40, by writers, lawyers, administrative assistants, professors and the CEOs), low -level professionals (represented by nurses, farmers, teachers, janitors and the police), production occupations (represented by production managers a nd factory operators) and service occupations (represented by baby-sitters, guards and taxi drivers). For
Taiwan, in Table 6-b, there are also four factors: organizational occupations
(personnel managers, administrative assistants, accounts, guards, production managers, computer programmers, receptionists and the CEOs), male service occupations (farmers, janitors, factory operators, taxi drivers and hotel bellboys, and the police), female service occupations (nurses, teachers, housemaids, and hair dressers), and professionals (writers, lawyers, legislators, and professors).
(Tables 6-a, 6-b and 6-c about here)
For China, in Table 6-c, the four factors are identified as organizational
Lin, Fu and Chen 22 occupations (personnel managers, administrative assistants, production managers, computer programmers, and CEOs), service occupations (housemaids, janitors, hair dressers, guards, receptionists, taxi drivers and hotel bellboys), low -level professionals (nurses, peasants, teachers, bookkeepers, and the police), and high-level professionals (writers, lawyers , computer programmers, and professors).
A comparison of the three societies, as can be seen in Table 7, shows that US and
China respondents share similar clusters : high-level and low -level professionals and service occupations. However, for US, there is a production cluster focusing on the actual managing and operating occupations in production functions while the upper-echelon managers and administrators in organizations have clustered into the high-level professionals along with professors, lawyers and writers. Thus, for US, there is a differentiation of high- and low -level personnel and by hierarchy and functions in organizations. For China as well as Taiwan, such intra-organizational differentiation is not so clear; mana gers and operators are vertically integrated into a cluster. It may suggest that the division of labor is yet to evolve – the eventual segmentation of elite administrators and the operational personnel. For Taiwan, only a cluster of high-level professionals has emerged. Low -level professionals such as nurses, teachers, factory operators and police are instead merged with service personnel such as housemaids, janitors, hair dressers, taxi drivers and hotel bellboys.
Further the merged service personnel are segregated into two clusters by gender.
(Table 7 about here)
We then explore how these clusters of occupations are associated with the two social capital indexes in each society: upper reachability and extensity and which attributes are related to these clusters. As can be seen in Table 8, for US, high-level professionals are highly associated with upper reachability and low -level professionals account for a modest degree. Production and service occupations do
Lin, Fu and Chen 23 not contribute. Thus, for a significant number of Americans, high-level professionals accessed account for rich social capital; a much smaller portion of the respondents rely on low -level professionals. For Taiwanese, rich social capital is provided by accessed high-level professionals, organizational occupations and female service personnel, while male service personnel makes little contribution. For China, both high- and low -level professionals as well as organizational occupations account for rich social capital while service occupations account for little.
(Table 8 about here)
What account for such differences? We examine the associations between tie attributes and the various clusters of accessed occupations. For the US, access to high-level and low -level professionals, the two important clusters linked to rich social capital, show two different paths, as shown in Table 9-a. For accessing high-level professionals, strong and weak ties, white ties, black ties, and friends are important.
For accessing low-level professionals, female ties, white ties, friends and relatives are important. We suspect these represent two divergent paths to rich social capital due to education, gender or race \ethnicity. We regress high-level and low-level professionals separately on gender (male), white, black, Latino (with others as the missing race\ethnic category) and education. Education, it turns out, to be the decisive variable for accessing high-level professionals (partial standardized coefficient of .36) and none of the gender and race \ethnic categories had any significant associations with high-level professionals once education is accounted for.
On the other hand, education had no bearing on accessing low -level professionals (.06) while gender (female. .09) and white (.16) account for accessing such professionals.
Therefore, education is the divide: the educated access rich social capital, through high-level professionals, and the lowly educated gain access to somewhat rich social capital if they were female and white.
Lin, Fu and Chen 24 (Tables 9-a, 9-b, and 9-c about here)
For Taiwan, as can be seen in Table 9-b, access to high-level professionals is associated with male ties but not work ties. Access to organizational occupations depends on female ties, work ties, distant ties and school ties \parents. Access to female service occupations depend on female ties. Regressions on the three significant clusters with gender and education show that access to high-level professionals depends on being male (.12) and more educated (.20). Access to organizational occupations depends on education (.38) and somewhat on being male
(.06). Access to female service occupations depends on being female (.24) and more educated (.22) . Thus, in Taiwan, education and gender jointly demarcate accesses to various clusters of important occupations through social ties.
For China, shown in Table 9-c, accessing high-level and low-level professionals depend on female ties, strong ties, school ties and relatives. Access to organizational occupations, on the other hand, depend on male ties and work ties. Associations of these clusters with gender and education shows that access to organizational occupations is associated with being male (.14) and education (.21) , whereas accessing to high- and low -level professionals are highly associa ted with education
(.31 and .32 respectively) but not with gender (.06 and .00 respectively) of the respondents. Thus, in China, access to better (higher status) occupations depends on being better educated. Further, access to organizational or hierarchic al positions also depends to being male.
As to diversity in social capital, as can be seen in Table 8, in all three societies, access to all clusters of occupations helps as expected. Some variations are observed, however. For the US, high-level professionals make the greatest contributions, whereas for Taiwan and China, organizational occupations make more important contributions. These differences reflect that the American networks have more or
Lin, Fu and Chen 25 less detached from organizational or hierarchical bases and moved toward a more professional basis.
STRUCTURAL EMBEDDING OF SOCIAL CAPITAL: TENTATIVE CONCLUSIONS
These analyses, while preliminary, provide clues about the patterns and processes by which social capital is embedded in different social structures. It seems that a detectable trend between China and US lends support to a view of transformation of social relations along the industrialization path. In the less industrial society such as China, access to rich social capital depends to a significant extent to kin ties and institutional ties (e.g., work ties, school ties, and clan ties). In the more industrial society such as the US, such access depends more on more generalized social relations – friends, rather than kin and institutional ties. Taiwan, somewhere in between in the industrialization process, also show patterns in between.
It still relies on kin and institutional ties. Friends have begun to assert its significance, but it is somewhat mixed with or undistinguished from kin ties. This is a preferred explanation to a cultural explanation, since we see differences between
Taiwan and China in the waning significance of kin ties and work ties and the emergence of the significance of friends, somehow detached from kin and institutional ties. The political-economic regime explanation does not seem to hold as Taiwan is no closer to US than it is to China in patterns.
Industrialization, as it signifies division of labor, urbanization, and social mobility, does ease up some deep-rooted spatial- and institution-bound constraints in social relations and open up social relations based on generalized social relations – friends. Friends now detached from kin and institutional ties are possible avenue for accessing better social capital and account for the meaningfulness of weaker and more distant ties in one’s social networks.
Lin, Fu and Chen 26 This differentiation perhaps offers us some more concrete clues as to what guanxi actually means in the Chinese societies (King, 1985; King, 1982/1988; Hwang,
1987; Lin, 1989; Bian, 1994; Bian & Ang, 1997; Bian, 2000). Our data would suggest that in Chinese societies, social relations principally consists of a mixture of ascribed and constructed relations, including kin, clan, school and work ties. These ties are interwoven and inter-dependent to form two core layers of relations – extended and pseudo- families – from which other ties are peripherally and indirectly connected. Friends are not independent of these relations and are in fact embedded in them. Friends gain significance if and when they are part and parcel of this expanded notion of family. Guanxi – the mixture of kin, institutional ties and friends – will evolve as the society is undergoing industrialization, as evident in the case of Taiwan. It does not go away easily and may never go away, but it nevertheless eases up in its configurations.
It would, however, be wrong to conclude that industrialization brings about more “modern” or rational social relations. In US, gender and race\ethnicity of social ties remain important – male ties and white ties present relative advantages in access better social capital. Homphily continues to be the powerful social principle that dominates all societies – regardless of the extent of industrialization. Thus, perhaps, ultimately, the universal principle of social relations will prevail over
“developmental” processes brought about by economic and political transformations.
Lin, Fu and Chen 27 Table 1. Tie Characteristics ======US Taiwan China N = 2317 N = 2406 N = 2706 ------Number of positions Accessed (mean) 7.43 ` 9.15 7.52 (standard dev.) (4.09) (5.45) (4.62)
Gender of ties Male ties (percent) 3.86 (53%) 5.48 (60%) 4.91 (65%) Female ties 3.55 3.67 2.61
Race of tie (US only ) White ties 5.17 (70%) Black ties 0.96 (13%) Lationo ties 0.76 (10%) Asian 0.16 ( 2%) Native 0.03
Strength of ties (mean) Strong 3.75 (50%) 3.72 (41%) 3.42 (45%) Weak 1.52 (20%) 1.06 (12%) 0.93 (12%)
======
Lin, Fu and Chen 28
Table 2-a. Factor Loadings of Relationships –US (for relations with >= 5 percent responses and without “Other”; principal component analysis, orthogonal rotation) Factor1 Factor2 Factor3 Factor4 Factor5 Factor6 Factor7 Relations Spouse Distant Siblings \ Work Parent Friend Previous Tie /in-law /indirect Other relative s Spouse 0.0557 -0.1264 0.6306 0.1723 -0.0195 -0.1811 -0.0105 Parents 0.0235 0.6851 0.036 0.1457 -0.1215 -0.13 0.2235 In-law 0.0772 0.1768 0.6782 -0.1512 0.0205 0.0662 0.0228 Child 0.0929 -0.5922 -0.0488 0.1157 -0.1294 -0.0503 0.3638 Sibling 0.0935 0.0877 -0.2132 -0.0892 -0.0855 -0.1954 0.6159 Other relative -0.1253 -0.0173 0.2669 0.1093 0.1781 0.1737 0.6219 Neighbor 0.0579 -0.2523 0.1137 0.6197 0.0686 -0.1008 -0.2551 Teacher 0.3006 0.2383 0.0108 0.0341 0.1668 0.4832 -0.0793 Co-worker 0.6658 0.018 0.1351 0.0094 -0.0758 -0.0474 -0.0883 Boss 0.6046 -0.1853 0.0034 -0.0116 -0.0246 -0.0322 0.1472 Previous co-worker -0.035 -0.1209 -0.0436 0.0546 -0.1457 0.7494 0.0113 Close friend 0.0237 0.0051 -0.0035 0.0533 0.7898 -0.1087 0.0142 Ordinary friend -0.211 -0.1548 0.0684 -0.1549 0.4745 0.1687 0.0602 Through service 0.5332 0.1167 -0.0105 0.0109 0.1282 0.2716 -0.0432 Acquaintance 0.2971 0.0144 -0.2675 0.0824 0.3029 -0.1515 -0.0282 Indirect tie -0.0234 0.1826 -0.0715 0.7493 -0.0048 0.1247 0.1531
Lin, Fu and Chen 29 Remarks: 1. various kin ties are independent of one another. 2. work ties form one cluster. 3. friends form one cluster, both close and general friends.
Lin, Fu and Chen 30
Table 2 – b Factor Loadings of Relationships – Taiwan (for relations with >= 5 percent responses and without “Other”; principal component analysis, orthogonal rotation) Factor1 Factor2 Factor3 Factor4 Factor5 Factor6 Factor7 Relations Siblings\other Work Distant Friend School\parent Neighbor\child Previous Tie relative Spouse -0.0712 -0.0714 -0.0897 -0.0749 0.3686 0.3737 0.2735 Parents -0.0162 -0.2796 -0.1411 0.5507 0.1669 0.0887 0.0534 Child -0.0697 -0.1738 -0.0811 -0.0036 0.5954 0.0317 0.009 Sibling -0.029 0.033 -0.0704 -0.0172 -0.0646 0.7704 0.06 Other relative 0.0612 0.1238 0.4001 0.1489 0.1786 0.5036 -0.1953 Old neighbor 0.0749 0.4196 0.1541 -0.1131 0.3334 -0.2173 0.3157 Neighbor 0.0475 0.1671 0.0504 0.0063 0.6315 -0.0464 -0.0572 Schoolmate 0.0861 0.0947 0.0831 0.5103 -0.2156 0.0644 0.2227 Teacher 0.3679 0.0483 0.0997 0.4953 -0.2117 -0.0652 0.1837 Co-worker 0.7348 0.0259 -0.1108 -0.0053 -0.0028 -0.0061 -0.1643 Boss 0.7004 -0.0325 -0.0488 0.0465 0.0185 -0.0242 0.0634 Previous co-worker -0.0715 0.0168 -0.0465 0.0952 -0.0018 0.0224 0.7950 client -0.3292 0.1216 -0.0859 0.4464 0.131 -0.2245 -0.3333 Another firm -0.2114 0.52 -0.1402 -0.0832 -0.1297 0.0715 -0.0476 Close friend -0.0209 -0.0425 0.7008 0.0016 0.0683 0.0579 -0.0649 Ordinary friend -0.1738 -0.0781 0.6646 -0.0535 -0.1358 -0.1305 0.0504
Lin, Fu and Chen 31 Through service 0.0696 0.6513 -0.091 0.0137 0.0289 0.0691 -0.015 Indirect tie -0.0259 0.4497 0.0027 0.3883 0.0446 0.1305 0.0709
Lin, Fu and Chen 32
Table 2-c. Factor Loadings of Relationships – China (for relations with >= 5 percent responses and without “Other”; principal component analysis, orthogonal rotation) Factor1 Factor2 Factor3 Factor4 Factor5 Factor6 Factor7 Factor 8 Relations Friend (not Distant tie Work Relative School/clan Neighbor Previous ties Spouse sibling) Spouse -0.009 0.0014 -0.002 -0.0182 0.0225 -0.0043 -0.0424 0.9282 Parent 0.1138 0.0103 -0.5773 0.0743 -0.0908 -0.0666 0.174 0.2616 Sibling -0.0285 -0.0152 0.1721 0.0677 -0.0351 0.0667 -0.7333 0.0968 Other relative 0.1355 0.0879 0.6272 -0.0047 0.0141 0.0494 -0.1342 0.1664 Old neighbor 0.1143 0.0598 0.1015 0.157 0.5827 -0.0036 -0.1653 -0.0781 Current neighbor -0.0679 -0.023 -0.0179 -0.0583 0.7785 0.0554 0.073 0.067 Schoolmate 0.0102 0.0283 0.0662 0.7586 -0.044 -0.0388 0.0304 -0.0267 Compatriot 0.002 -0.1067 -0.1961 0.6135 0.0977 0.0212 -0.1526 -0.0166 Teacher -0.0103 0.478 0.0872 0.2627 -0.0219 0.2186 0.3417 0.153 Coworker -0.031 0.7155 0.0246 -0.0123 -0.082 -0.1305 -0.1883 0.0088 Boss 0.011 0.6713 -0.0508 -0.0887 0.1102 -0.0049 0.1417 -0.0536 Previous 0.0277 -0.0976 0.0887 0.0091 0.1094 0.7509 0.0219 -0.0052 colleague Another firm 0.6866 -0.0841 -0.0336 -0.1364 0.0344 -0.0526 0.0412 0.0631 Close friend -0.1579 -0.1748 0.3750 0.0278 -0.1769 -0.2938 0.0934 -0.0197 Ordinary friend -0.0456 -0.1336 0.3552 -0.0056 -0.1937 0.3276 0.4186 -0.0091
Lin, Fu and Chen 33 Service 0.6491 0.0778 -0.0089 -0.0489 -0.0112 0.0399 0.0823 0.0239 Acquaintance 0.0944 -0.1323 0.3376 0.1375 0.2339 -0.4711 0.2917 0.0316 Indirect tie 0.5355 0.0082 0.1336 0.2143 0.0208 0.177 0.0104 0.0664 Remarks: 1. spouse is one dimension; 2. all other kin ties form another cluster. 3. work ties form one cluster. 4. distant and indirect ties form one cluster. 5. weak ties – acquaintances form one cluster.
Lin, Fu and Chen 34
Table 3. Clusters of Tie-Ego Relationships (Principal Component Factor Analysis with Orthogonal Rotation and Factors of Eigen Values =>1) ======US Taiwan China N = 2317 N = 2406 N = 2706 ------Relationships with ties (factors) Work Work Work Friends Friends (and Friends relatives) (not Siblings) Previous ties Previous ties Previous Distant Distant Distant
Parents Spouse Spouse Siblings \other Siblings \other relatives relatives School\parent School\ clan Neighbor\child Neighbor
======
Lin, Fu and Chen 35 Table 4. Associations of Tie Characteristics with Rich Social Capital (Upper Reachability)# ======US Taiwan China ------
Male ties .27 .34 .33 Strong ties .08 .06* .12 Weak ties .10 .09 .07 White ties .11
Work ties .05**, *** .19* .14 Friends (close, ord., acquaintance.) .15*** (close, ord., relatives) .13* (ord., not siblings) .06 Distant ties (neighbors, indirect) .03** (service, indirect, old neighbor) .15 Previous ties (pre. co-workers, teachers) .08
Siblings \relatives (siblings, other) .03**, *** .08* (other, not parent , close and ord. friends ) .14 Spouse .06 School ties (schoolmates, teachers, and parents) .20* (schoolmates, compatriots) .12 Neighbors (current, old) -.01 Quota .01
R squared adj. .33 .45 .38 N 2280 2344 2683 ======# Standardized partial regression coefficients). All coefficients are significant Except “distant ties” for US; “neighbors” for China * Significance of difference between Taiwan and China , based on metric coefficients
Lin, Fu and Chen 36 and standard deviations. ** Significance of difference between US and Taiwan, based on metric coefficients and standard deviations. *** Significance of difference between US and China, based on metric coefficients and standard deviations.
Lin, Fu and Chen 37 Table 5. Associations of Tie Characteristics and Diverse Social Capital (Extensity)# ======US Taiwan China ------Male ties .35**,*** .57 .60 Strong ties .27**,*** .11 * .19 Weak ties .15*** .10* .15 White ties .20 Work ties .10*** .08* .05 Friends (close, ord., acquaintance.) .19*** (close, ord., relatives) .18 (ord., not siblings) .16 Distant ties (neighbors, indirect) .03** (service, indirect, old neighbor) .19 Previous ties .07
Relatives (siblings, relatives) .09 .06 (relatives) .06 Spouse .04 School ties (school mates, teachers, parents) .15* (school mates, compatriots) .08 Neighbors (neighbors, child) .05 (current, previous) .04 Quota .02
R squared adj. .90 .87 .88 N 2317 2406 2706 ======# Standardized partial regression coefficients). All coefficients are significant. * Significance of difference between Taiwan and China, based on metric coefficients and standard deviations. ** Significance of difference between US and Taiwan, based on metric coefficients
Lin, Fu and Chen 38 and standard deviations. *** Significance of difference between US and China, based on metric coefficients and standard deviations.
Lin, Fu and Chen 39 Table 6-a. Factor Loadings of Accessed Positions – US (Without “bellboy”; principal component factor analysis, orthogonal rotation) Factor1 Factor2 Factor3 Factor4 Accessed Position High-level Low -level Production Service professional professional Nurse 0.2381 0.4807 -0.0338 -0.0188 Writer 0.5679 0.1152 -0.0583 -0.0733 Farmer (peasant) 0.0634 0.5756 0.2294 -0.09 Lawyer 0.5666 0.2839 -0.0756 0.0028 High (middle) school 0.2931 0.4783 -0.0252 0.0609 teacher Baby sitter (housemaid) -0.138 0.3624 0.1189 0.4629 Janitor 0.0669 0.5025 0.2192 0.2184 Personnel manager 0.3814 -0.0646 0.3562 0.3096 Administrative assistant 0.4558 -0.181 0.337 0.257 Hair dresser 0.2129 0.3416 -0.0082 0.3361 Accountant (book keeper) 0.3969 0.2376 0.1065 0.1018 Guard 0.1121 0.1543 0.0975 0.5907 Production manager 0.1389 0.0147 0.7229 0.0143 Factory operator -0.1541 0.2448 0.6608 -0.0051 Computer programmer 0.482 0.0883 0.254 0.0358 Receptionist 0.343 0.2049 0.2252 0.2947 Congressman/woma n 0.3003 0.1868 0.1697 0.2543 Taxi driver 0.0463 -0.0742 -0.0571 0.6768 Professor 0.5321 0.1932 0.0018 0.1396 Police 0.1592 0.4962 0.0827 0.1711 CEO in a big company 0.4118 -0.0061 0.3653 0.1727 Remarks: US accessing professionals, low-level professio nals, manufacturing positions, service positions
Lin, Fu and Chen 40
Table 6-b. Factor Loadings of Accessed Positions – Taiwan (principal component factor analysis, orthogonal rotation) Factor1 Factor2 Factor3 Factor4 Accessed Position Organization Service-Male Service-Female High-level Professional Nurse 0.1134 0.0256 0.6395 0.1638 Writer 0.1062 -0.0409 0.1986 0.5639 Farmer (peasant) -0.1123 0.5142 0.2528 0.1331 Lawyer 0.251 0.1281 0.173 0.6126 High (middle) school 0.2609 0.0977 0.4965 0.3056 teacher Baby sitter 0.1903 0.206 0.5757 0.1066 (housemaid) Janitor 0.1626 0.4014 0.4524 0.1072 Personnel manager 0.6424 0.0995 0.2522 0.0756 Administrative 0.7566 0.0375 0.0523 0.1769 assistant Hair dresser 0.1983 0.2549 0.5569 -0.0715 Accountant (book 0.5878 0.1744 0.3092 0.0173 keeper) Guard 0.5004 0.3301 0.2478 0.0762 Production manager 0.6662 0.2556 -0.069 0.1665 Factory operator 0.3987 0.5598 0.0098 -0.2441 Computer 0.6305 -0.0144 0.2712 0.1457 programmer Receptionist 0.6314 0.0906 0.2284 0.0701 Congressman/woman 0.1237 0.3385 -0.0236 0.6513 Taxi driver 0.0702 0.6032 0.1411 0.235 Professor 0.4033 -0.1698 0.3174 0.4803 Hotel bellboy 0.1884 0.6660 0.0826 0.1455 (carrier) Police 0.1498 0.4709 0.2748 0.2805 CEO in a big 0.5501 0.1713 -0.0442 0.4218 company
Lin, Fu and Chen 41
Table 6-c. Factor Loadings of Accessed Positions – China (principal component factor analysis, orthogonal rotation) Factor1 Factor2 Factor3 Factor4 Accessed Position Organization Service Low-level High-level professional professional Nurse 0.0373 0.0969 0.5750 0.2962 Writer 0.0031 0.2054 0.0022 0.6377 Farmer (peasant) -0.0436 0.2757 0.4834 -0.0734 Lawyer 0.2924 0.0952 0.3218 0.4647 High (middle) school 0.1227 0.0827 0.658 0.0806 teacher Baby sitter 0.0629 0.5226 0.0867 0.295 (housemaid) Janitor 0.0916 0.6849 0.0934 0.0743 Personnel manager 0.7002 0.0837 0.183 0.0405 Administrative 0.6416 0.1088 0.0207 0.3017 assistant Hair dresser 0.0948 0.5419 0.2252 0.1069 Accountant (book 0.437 0.04 0.5446 -0.0151 keeper) Guard 0.4114 0.5155 0.1788 -0.0492 Production manager 0.6742 0.2285 0.0689 0.0018 Factory operator 0.031 0.1301 -0.2592 -0.3882 Computer 0.4207 0.0999 0.0914 0.4429 programmer Receptionist 0.3227 0.4265 -0.0615 0.3318 Taxi driver 0.1707 0.4512 0.2908 0.0076 Professor 0.246 0.018 0.2279 0.5696 Hotel bellboy 0.1826 0.6185 0.0282 0.065 (carrier) Police 0.2455 0.1963 0.5508 0.1744 CEO in a big 0.5596 0.1064 0.1603 0.1626 company Remarks: China accesses high-level professional (technical?), low -level professionals, service, and organizational occupations (combining manufacturing, production and management)
Lin, Fu and Chen 42 Table 7. Clusters of Accessed Occupations (Principal Component Factor Analysis with Orthogonal Rotation and Factors of Eigen Values =>1) ======US Taiwan China N = 2317 N = 2406 N = 2706 ------
Clustering of accessed Occupations High-level prof. High-level prof . High-level prof. Low-level prof Low-level prof Production Organizational Organizational Service Service-male Service Service-female ======
Lin, Fu and Chen 43 Table 8. Zero-order Correlations Between Occupational Clusters and Social Capital USA ======Social Capital Upper-reachability Extensity ------High-professionals .63 .62
Low-professionals .27 .54
Production .03 .40
Service .06 .40 ======
Zero-order Correlations Between Occupational Clusters and Social Capital Taiwan ======Social Capital Upper-reachability Extensity ------High-professionals .41 .37
Organizational .55 .66
Service-male .00 .46
Service-female .33 .47 ======
Zero-order Correlations Between Occupational Clusters and Social Capital China ======Social Capital Upper-reachability Extensity ------High-professionals .48 .45
Lin, Fu and Chen 44
Low-professionals .57 .62
Organizational .53 .75
Service .14 .62 ======
Lin, Fu and Chen 45 Table 9-a
Associations of Tie Characteristics with Clusters of Accessed Occupations USA ======High- Low- Production Service Professional profess. ------Male ties .06 . .34 Female ties .15 .16 Strong ties .11 .04 .04 .04 Weak ties .11 .04 -.04 .03 White ties .23 .25 .14 Black ties .10 .21 Latino ties .27 Other racial\ethnic ties .09 .08 Indirect ties .09 Work ties .08 .08 .08 Friends .11 .16 -.08 .07 Previous ties .05 .03
Siblings \other relatives -.00 .12 .00 .02 Inlaws\spouse .10
R squared, adj. .37 .32 .19 .25 ======
Lin, Fu and Chen 46 Table 9-b.
Associations of Tie Characteristics with Clusters of Accessed Occupations Taiwan
======High- Organi- Service- Service- professional zational. male female ------Male ties .11 .16 Female ties .14 .21 Strong ties .04 .05 -.03 -.03 Weak ties .04 .04 -.01 .01 Distant ties .07 .17 .04 .13 Work ties -.13 .29 Friends .05 .06 .18 .21
Siblings \other relatives .20 School\parent .09 .12 .14 Neighbors\child .24 .14
R squared, adj. .23 .39 .36 .43 ======
Lin, Fu and Chen 47
Table 9-c.
Associations of Tie Characteristics with Clusters of Accessed Occupations China
======High- Low- Organizational Service Professional profess. ------Male ties .59 Female ties .12 .23 .33 Strong ties .19 .15 .11 .07 Weak ties .04 .04 .04 .14 Distant ties .08 .08 .18 Work ties .18 School ties .21 .25 -.03 .06 Neighbors .14 Friends (not siblings) .12 .05 .09
Relatives .17 .23 .02 .13 Spouse .09
R squared, adj. .25 .40 .57 .40 ======
Lin, Fu and Chen 48 Appendix. Sampled Positions in the Position Generator Instrument
Position SIOPS USA Taiwan China Legislator 85 v v Professor 78 v v v Lawyer 73 v v v Ceo 70 v v v Production manager 63 v v v Middle school teacher 60 v v v Personnel manager 60 v v v Writer 58 v v v Nurse 54 v v v Programmer 51 v v v Administrative assistant 49 v v v Accountant 49 v v v Policeman 40 v v v Peasant 38 v v v Receptionist 38 v v v Operator in a factory 34 v v v Hairdresser 32 v v v Taxi driver 31 v v v Security guard 30 v v v Maid 23 v v v Janitor 21 v v v Hotel bellboy 20 v v v ( Leading cadre in the work unit) v ( Leading cadre of the supervising work unit) v ( Regular civil servant) v
Lin, Fu and Chen 49 REFERENCES CITED
Bian, Yanjie. 1994. "Guanxi and the Allocation of Jobs in Urban China." The China Quarterly 140:971-99. ______2000. "Getting a Job Through a Web of Guanxi in Urban China." In Networks in the Global Village, edited by B. Wellma n. Boulder, CO: Westview. Bian, Yanjie and Soon Ang. 1997. "Guanxi Networks and Job Mobility in China and Singapore." Social Forces 75:981-1006. Granovetter, Mark. 1973. "The Strength of Weak Ties." American Journal of Sociology 78:1360-80. Hwang, Kwang-kuo. 1987. "Face and Favor: The Chinese Power Game." American Journal of Sociology 92(4, January):944-74. King, Ambrose Y. C. 1985. "The Individual and Group in Confucianism: A Relational Perspective." Pp. 57-70 in Individualism and Holism: Studies in Confucian and Taoist Values, edited by D. J. Munro. Ann Arbor: Center for Chinese Studies, the University of Michigan. King, Ambrose Yeo-chi. 1982/1988. "Analysis of Renqing in Interpersonal Relations (Renqi Guanxi Zhong Renqing Zhi Fensi)." Pp. 319-45 in Psychology of the Chinese (Zhongguren de Xinli), edited by K.-s. Yang. Taipei, Taiwan: Guiguan Press. Lin, Nan. 1986. "Conceptualizing Social Support." Pp. 17-30 in Social Support, Life Events, and Depression, edited by N. Lin, A. Dean and W. Ensel. Orlando, Florida: Academic Press. ______1989. "Chinese Family Structure and Chinese Society." Bulletin of the Institute of Ethnology 65:382-99. ______2005. "Social Capital." Pp. 604-12 in Encyclopedia of Economic Sociology, edited by J. Beckert and M. Zagiroski. London: Rutledge. ______2006. "A Network Theory of Social Capital." In Handbook on Social Capital, edited by D. Castiglione, van Deth, Jan and G. Wolleb. London: Oxford University Press. Lin, Nan and Mary Dumin. 1986. "Access to Occupations Through Social Ties." Social Networks 8:365-85. Lin, Nan and Walter M. Ensel. 1989. "Life Stress and Health: Stressors and Resources." American Sociological Review 54:382-99. Lin, Nan and Bonnie Erickson, eds. 2008. Social Capital:An International Research Program. 2008. Oxford: Oxford University Press.
Lin, Fu and Chen 50 Lin, Nan, Yang-chih Fu, and Ray-may Hsung. 2001. "The Position Generator: Measurement Techniques for Investigations of Social Capital." Pp. 57 - 82 in Social Capital: Theory and Research , edited by N. Lin, K. Cook and R. S. Burt. New York: Aldine de Gruyter. Son, Joon-mo and Nan Lin. 2008. "Social Capital and Civic Action: A Network-Based Approach." Social Science Research, March. Son, Joon-mo, Nan Lin, and Linda K. George. 2008. "Cross-National Comparison of Socia l Support Structures Between Taiwan and the United States." Journal of Health and Social Behavior 49(1, March):104-17. van der Gaag, Martin, and Tom Snijders. 2005. "The Resource Generator: Social Capital Quantification with Concrete Items." Social Networks 27:1-29. ______2008. "Position Generator Measures and Their Relations to Other Social Capital Indicators." Pp. 27-48 in Social Capital: An International Research Program, edited by N. Lin and B. H. Erickson. NY: Oxford University Press.
Lin, Fu and Chen 51 Notes
1 An alternative strategy is to ask the respondent to name as many occupants as she/he knows at each sampled occupation and randomly sample one of the named occupants. However, it is very inefficient (time consuming and respondent fatigue) for general community surveys. It would be very useful to conduct some intensive investigations of this alternative strategy with a limited but representative sample to estimate the bias of the efficient strategy and account for such bias in larger surveys.
2 The fieldwork was conducted from November 2004 to April 2005. When they were connected by phone , the interviewer asked for the person who fit the screening requirements. If multiple persons were qualified, the person whose birthrate was closest to July 1 was sought. If the qualified person was not available, a follow-up call was made. The interview took thirty-four minutes on average. About 30 percent of those contacted were qualified and agreed to participate; this was slightly less than but close to the rate obtained in typical telephone community surveys, about
35 percent. The response rate among those who were qualified and agreed to participate was 43 percent. Fieldwork was typically conducted from 5:30 – 8:45 p.m., Monday–Friday and Saturdays. On Sundays dialing was conducted from 10 a.m. to 2 p.m. (all local times).
3 Excluded are unemployed, laid off, retired, in school, keeping house, and with unspecified responses.
4 The actual distribution of clusters is as follows:
Shanghai 5 clusters 95 households/respondents
Beijing 4 clusters 76 households/respondents
Chongqing 4 clusters 76 households/respondents
Tianjin 3 clusters 57 households/respondents
Wuhan 3 clusters 57 households/respondents
Lin, Fu and Chen 52
Shengyang 2 clusters 38 households/respondents
Guangzhou 2 clusters 38 households/respondents
Chendu 2 clusters 38 households/respondents
159 other cities 1 cluster each 19 households/respondents each.
5 The exchange rate at the time was about RMB$8.00 = US$1.00.
6 We should note that all three questionnaires contained the same categories of responses, with the exception of “acquaintance” which was accidentally omitted in the Taiwan questionnaire.
Lin, Fu and Chen 53 Relation-specific Social Resources and their Explanation: Comparison among Taiwan, China and United States
Ray-May Hsung* Ronald Breiger**
Abstract Our study reports the relation-specific social resources from a 3-country (China, Taiwan, US) set of nationally representative surveys of position-generated networks. We show how to calculate reflected prestige scores for each contact type, on the basis of positions it accesses. There are some similarities among the countries in the kinds of tie used to access positions. For example, in each society, other relatives, current colleague, good friend, and general friend are the major ties used to access occupational positions. However, there are also important differences among the societies. For example, people in Taiwan tend to use “other relatives” to access positions. Both East Asian societies use more school and teacher ties to access positions, in comparison to the US. In fact, the prestige scores of accessed resources vary with the kind of relational channel used. The prestige score of positions accessed through “teacher” and “current boss” are consistently first- and second-highest among the three countries, and prestige scores of positions accessed through kin ties are relatively lower in each country. Likewise, prestige scores of positions accessed through good friends are higher than those accessed through general friends. This paper only focuses on the explanation of the average prestige and diversity score of accessed positions through other relatives, current colleagues, good friends, and general friends. We go on to use individual and network characteristics to explain these two dependent variables. The individual characteristics include gender, age, age square, size of household, years of schooling, employment status, and daily contacts and total number of voluntary associations engaged. The network characteristics include percentage of males using the tie, the average year, and the average degree of closeness among accessed ties, specified by relationship type. We find that years of schooling is the most important variable in predicting relation-specific prestige. Gender and the percentage of males among specific accessed ties have interaction effects.
*Professor, Department of Sociology, National Chengchi University, Taipei, Taiwan [email protected] **Professor, Department of Sociology, University of Arizona, Tucson, AZ, USA [email protected]
1
INTRODUCTION
Since Lin and Dumin (1986 ) invented the position generators to elicit an individual’s social networks, researches on social capital by position-generated networks have increased rapidly in all over the world (Lin, Cook and Burt 2001; Flap and Völker 2004; Lin and Erickson 2008; Hsung, Lin and Breiger 2008). Position-generated networks were designed by using variety of occupational positions in a society to elicit an individual’s social resources through different types of relationships with ego. Lin (2001) defined the social capital as the investment and return of the diversity of position-generated networks. In fact, the investment of the position-generated networks is a complex matrix and previous researches pay less attention to the duality nature and its characteristics of position-generated networks.
The position-generated networks include two dimensions of an individual’s social networks: distribution of occupational positions and distribution of social relationships with ego. Hsung and Breiger (2008) presented the institutional logics on the duality of the position-generated networks from the structural level. People often access specific occupational positions with specific types of relations; for example, access to lawyers through client relation, access to professors through teacher relation, etc. We have studied the association between categories of occupational positions and types of relationsships by correspondence analysis and associational model (Hsung and Breiger 2007, 2008) on the structural level, and now attempt to furthermore explore how the production of social resources through specific relationship on the individual level in this paper.
Previous literatures have used different measures to indicate the production of social capital on the individual level (Lin, Fu and Hsung 2001; van der Gaag, Snijder and Flap 2008; Erickson 2004 ; Hsung and Lin 2008). All these measures mainly focus on measuring the diversity of social contacts by aggregating all types of relations together. These researches seldom paid attention to the differentiation of position-generated networks. Are the values or social resources of accessed positions through different relations different? In Chinese society, the differentiation of Guanxi structure (cha xu ge ju) basically is determined by the web of quasi-family networks (Fei 1949). Fei emphasized that this scheme of differentiation of social contacts is ego-centric and ethical relations of familial sentiments and obligation (Bian 2001). Though this scheme is culturally ethical, it implies that different circle of relations from the ego’s extended family relations produce different social resources. Wellman
2 (1991) also proposed similar proposition that different support resources were from different types of relations (1991).
As to differentiating social capital by different types of relationships, this paper attempts to propose indexes for disaggregating social capital by types of relationships. For the accessed positions through each type of relation, we produce two measures: the average prestige of accessed positions through specific type of relation and the diversity score of accessed positions through specific type of relation.
This paper attempts to propose the following questions on the individual level: 1. Do different types of relations produce different degree of social resources? 2. Are the social resources embedded in different types of relations stratified? 3. What are the factors affecting the average prestige and diversity score of accessed positions through specific relation? 4. Are there similarities and differences on the above mentioned questions among United States, Taiwan, and China.
LITERATURE REVIEW
The Production of Social Capital
Social capital is the investment and return of accessed and mobilized social networks (Lin 2001). There are two processes in social capital in terms of accessed networks: production and return. The previous studies treat the production of social capital as the investment of accessed networks. Different researchers used different indicators to measure the values on the investment of accessed networks. Lin, Fu, and Hsung (2001) used the factor score on diversity with three indicators of position-generated networks: the number of accessed positions, the highest prestige of accessed positions, and the range between the highest and the lowest prestige among accessed positions. Flap (2002), Ven der Gaag and Snijder (2004) defined social capital as “the collection of resources owned by the members of an individual’s personal networks.” The total resources were measured by summing the human capital of the array of contacts. Erickson (2004) preferred to define social capital as the variety of social capital, measured by the number of accessed positions.
An indicator of social capital on the individual level is based on defining social capital as network diversity. This definition of social capital is to measure the potentially accessible contacts and the resource distributions along the hierarchical
3 positions. The hierarchical positions can be designed by social prestige of occupations and political authority. The network diversity indicates the important actions of modern society for an individual, to access heterophilous ties for attaining higher status. Various studies found that the large, sparse and diversified networks of an individual enhanced his/her mobility in a firm, because these networks can acquire diversified information and resources that are advantageous for work performance and promotion (Burt 1992, 1998; Podolny and Baron; Erickson 2001).
Duality of Position-generated Networks
The position-generated networks can be conceived as a dual system of resource production featuring occupational positions (the resource opportunity structure) through types of social relationships (the classification system in all societies). These two-mode networks can demonstrate the institutional logics that unify classification systems that actors use to classify their social contacts into a set of occupational positions (Breiger 1974, 2000). Institutional logics are the clustering of meanings that govern how actors arrange the system of their resource structure through types of relationships. Breiger and Mohr (2004) focused on two kinds of institutional logics for two-mode networks: logics of structural equivalence of actors’ relations, and the logic of mutual constitution between hierarchical occupational positions and types of relationships. Structural equivalence is that actors organize their relations and social activities in similar ways if they share an equivalent location within a field of action. Mutual constitution is conceived the mutually constitutive character hierarchical positions and types of relationships.
Hsung and Breiger (2007, 2008) used different data set to demonstrate the institutional logics of position-generated networks. They found that there are three structurally equivalent and mutual constituted clusters in the position-generated networks in 369 knowledge workers in four high-tech firms in Taiwan (Hsung and Breiger 2008). The first one indicates that access to the position of owner is very highly associated with having a tie to a superior, a client, or a former colleague who can access an owner. The second cluster indicates that access to a maid or cleaning person is highly associated with ties to family members or neighbors (spouse’s parents, respondent’s parents, a current or a previous neighbor) and work contacts (a subordinate, a current colleague). The third cluster indicates that the four other positions (lawyer, division head, reporter, and assemblyman/woman) are relatively equivalent to one another with respect to access. These four positions can be treated as “public-service positions.” Access to them is particularly dependent on school and
4 association connections.
They also found that the institutional logics of position-generated networks vary with different societies (Hsung and Breiger 2007). They used log linear model and correspondence analyses to figure out the institutional logics among three countries and also do the comparison of the associational models for the contingency tables between types of positions and types of relationships among China, Taiwan and the United States.
There are similar and different institutional logics of position-generated network among different societies. Hsung and Breiger (2007) found that there are some institutional logics of position-generated networks in China: sector and tie strength logics mechanize the institutional logics. The educational sector logic is constructed by the mutual constitution between teacher relation and access to professor and middle school teacher. The logic of bureaucratic sector is constructed by the mutual constitution between previous colleagues and cadre in supervisory unit. The ascribed sector is constructed by the co-constitution between other relatives and access to farmers. The logic of tie strength was found in the association between general friends and acquaintance to access hairdresser and good friends to cadre. In Taiwan, there are three clusters of sector logics. The educational sector logic is constructed by the mutual constitution between teacher relation and access to professor and middle school teacher. The logic of corporate sector is constructed by the mutual constitution between previous, current colleagues, current boss and access to personnel manager or CEO. The ascribed sector is constructed by the co-constitution between other relatives and access to farmers. In the United States, the household sector is constructed by the mutual constitution of other relatives and access to housemaid. The educational sector logic is constructed by the co-constitution between teacher and access to professor and middle school teacher. The corporate sector logic is constructed by the co-constitution between previous and current colleagues and access to personnel and production mangers. There is a mixed public service sector clustered by professor and CEO with stronger association with acquaintance, current boss and the person providing service.
From the empirical facts among China, Taiwan and the United States proposed by Hsung and Breiger (2007), there is considerable similarity in the logic of social capital (i.e., in patterns of association between relationship types and positions accessed). Then, they further used multiplicative model to find the best fitting model which indicates that there is a much weaker association of relation type to position
5 accessed in the USA than in China and Taiwan. In China, the social resources and class structure divided along two dimensions: one by market economy and the other by political authority at workplace, such as cadre position in the work unit (Bian, Breiger, Davis, Galaskiewicz 2005).
Individual’s Resource Reproduction through Specific Relations
In addition, the reproduction of dual structure was not only constructed by the institutional logics on the structural levels. On the individual level, we assumed that the resource reproduction through different types of relationships is stratified. The differentiation of social networks or social structures are not only based on the normative constraints, but also based on the resource production in the dual system of position-generated networks.
Bordieu (1983) defined the social capital as “the aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutional relationships of mutual acquaintance and recognition.” Bordieu’s definition of social capital also includes the duality of network structures: the potential resources through durable relationships. Very few researches explored the issue on whether the social resources produced from different types of relationships are different.
Lin brings social into capital definition, because he conceived that the investments of social relations create values or resources (Lin 2001). The social capital is the stock of an individual’s investment through all kinds of social relationships in daily life. Position-generated networks are the investment of social relationships with mutual recognition and institutionalized and accumulated social resources. In a capitalism society, people consciously or unconsciously meet people, some contacts become recognized by mutual sides of a relational tie, but some don’t. The social capital is the investment of those contacts with better positions (Erickson 2008). Social capital researchers tend to conceive each tie as each product or commodity which can be counted and valued as social resources. Bourdieu (1983) proposed that “A theory of capital enumerates a mechanism by which such valued resources are produced, reproduced, and accumulated.” That the valued resources through different types of relations are produced, reproduced, and accumulated is an institutionalization processes. Usually people recognize the valuable ties when we ask people to select one tie among several ties of the same relations. So, the recognized networks are usually reproduced and institutionalized, because these relationships can
6 produce better resources with more accumulation and become the stock of social capital.
Portes (1998) conceived that “social networks are not a natural given and must be constructed through investment strategies oriented to the institutionalization of group relations, usable as a reliable source of other benefits.” He reinterpreted Bordieu’s definition on social capital and proposed that social capital can be decomposable into two elements: first, the social relationship itself that allows individuals to claim access to resources possessed by their associates, and second, the amount and quality of those resources. This study attempts to empirically figure out how to calculate the amount of social resources through specific relations.
Burt (1992, p. 619) see social capital as friends, colleagues, and more general contacts through whom you receive opportunities to use your financial and human capital. Position-generated networks not only can calculate the social resources through friends, colleagues, and general friends, but also can calculate the social resources through all types of kin relationships. The average accessed prestige, a measure introduced by Campbell, Marsden, and Hurlbert (1986) is calculated as the mean of the prestige of all occupations in which the respondent knows someone.
Ven der Gaag, Snijder, and Flap (2008) found that different types of relations create different social resources. In Netherland, there is more possibility to gain diversified resources through kin relationships. They also found that there is a positive effect of access to better resources through strong ties. Family relationships gave access to more different occupations (6.39) than did acquaintance (5.19) and friends (3,35). This maybe the first research to pay attention to how many positions accessed through different types of relationships. However, they only ask respondents to identify the occupational positions which respondents know and identify the relationships among three categories: kin, acquaintance and friends. Erickson (2004) used a position generator with separate questions for a man or women in a certain position may give access to different resources, this is also an option to retrieve more specific social capital information from survey questions.
The Factors affecting social capital
Gender effects on social capital vary in different types of people or different societies or different time periods. Males seem to access more diversified resources than females in Taiwn in 1997 national survey (Lin, Fu and Hsung 2001) instead in
7 2004-2005 national survey. Erickson (2004) thought that men are thought to be better in leadership, assertiveness, decisiveness, and other traits associated with high-level position. Men are often preferred for more powerful jobs with better chances to meet people (Wellman and Wortley 1990).
A person’s own education should have a larger effect on the social resources he/she has access to (Van der Gaag, Snijders and Flap 2008). Education has been conceived as the indicators of human capital theory and cultural capital theory. Human capital has been important factor on the accessed social capital. For example individuals with higher education tend to access more diversified social capital. Education is also a kind of cultural capital, which facilitates the social contacts with similar level of education. The acquisition of social capital requires deliberate investment of both economic and cultural resources.
Erickson (2004) found that membership of associations facilitate social networks Social capital is a resource that resides in the networks and groups to which people belong.
The Network characteristics also are the sources of social capital. Gender composition of network members could be related to an occupation’s networking power. Men are instrumental resources and women are expressive functions (Wellman and Wortley 1990). Greater percentages of males of network members tend to facilitate values and diversity of social resources at work or in public world. Similarly, greater percentages of women of network members facilitate family resources or in domestic life. Erickson (2004) proposed that gender itself is a social location. Men are more active in outside work and associational activities. Women are more sociable and active in family activities.
The relation between strength of ties on social resources have been widely discussed (Granovetter 1973). The strength-of-weak-ties proposition states that weak ties rather than strong ties permit a wider reach to other parts of the social structure, thus enabling the person who uses weak ties to reach better social resources. However, Bian (1997) found that strong ties produce better social resources in China.
RESEARCH METHODS
Data used in this study were drawn from the thematic research project “Social
8 Capital: Its Origins and Consequences” sponsored by Academia Sinica, Taiwan, through its Research Center for Humanities and Social Sciences, and the Institute of Sociology. The principal investigator of the project is Nan Lin. This project called for a two-wave three-site survey to be conducted during periods between 2004 to 2005 and 2006 to 2007. The three sites selected were Taiwan, mainland China, and the United States. These sites were selected to capture two different political regimes (state socialism in mainland China and capitalism in Taiwan and the US) and two different cultural institutions (Chinese in Taiwan and mainland China and Anglo-Saxons in the US). A panel design allowed some preliminary tests of alternative causal and reciprocal effect.
This paper only used the first wave of survey data in three countries. The first wave surveys were held in Taiwan, China and the United States during 2004-2005. In Taiwan, the survey was administrated in interpersonal interviews with currently or previously employed respondents aged 21-64 in an island-wide probability sample stratified by levels of urbanization, district and household. A total of 3,281 sampled respondents completed the surveys.
In China, the survey was also administrated in interpersonal interviews with respondents aged 21-64, currently or previously employed, registered and residing in urban cities. The sampling plan was a multistage clustering probability sample. This sampling included 184 clusters from 167 cities. The total attempts of interviews are 9902, and finally this survey completed 3,535 cases. The response rate was 35.7 per cent.
In the United States, the telephone survey was administrated. The respondents are 21-64 years old and currently or previously employed. The sampling plan was a multistage clustering probability sample. This sampling includes 24 cities in the United States. The total attempts of interviews are 6,916, and finally this survey completed 3,000 cases. The response rate was 43 per cent.
The key position-generator question was: “At the time when you were starting your current job, did you know anyone who was doing the following jobs?”. Twenty-two occupations were sampled from a full list of all occupations to represent different levels of occupational prestige, for which this study follows the Standard International Occupational Prestige Scale (SIOPS) constructed by Harry B. G. Ganzeboom and Donald J. Treiman (1996).
9 The sampled occupations are grouped into three hierarchical groups: the upper class (e.g., Professor (78), Lawyer (73), CEO of a Big company (70), Congressman (85), the middle class (e.g., Middle School Teacher (60), Personnel Manager (60), Production Manager (60), Writer (58), Nurse (54), Administrative Assistant in a Large Company (53), Computer Programmer (51), Bookkeeper (45), Policeman (40), Receptionist (38), Operator in a Factory (36), Hairdresser (32), Taxi driver (31), and the lower class (e.g., taxi drivers, security guards, janitors, and hotel bell boys). Types of relations include spouse (current or previous), parents, father-in-law/mother-in-law, children, sibling, daughter-in-law, son-in-law, other relatives, old neighbor, current neighbor, school/class mate, compatriot, teacher, student, current co-worker, current boss/supervisor, current subordinate, co-worker, boss/superior, or subordinate from a previous firm, client, person working for anther firms, but known through work relations, someone from the same religious group, someone from the same association, club, or group, close friend, someone known from the internet, an acquaintance, indirect relationship (known via someone else)
In order to do comparison for the position-generated networks in three countries, we choose 21 occupational positions which show up in the surveys of three societies. Because there is no occupational position, like congressmen, in China, this paper use 21 position generators to calculate the indicators of accessible resources through specific relationships.
The general formula to compute the average prestige of accessed positions through specific relationship will be:
APik=(P1*pos1+P2*pos2+P3*pos3+……………….+P21*pos21)/Tposk Tposk=pos1+pos2+pos3+………………….+pos21
APik is the average prestige of accessed positions through kth specific relationship. Tposk is total number of accessed positions through kth specific relationship. P1, P2, P3,…………P22 are the Treiman’s prestige score for 22 occupational positions. k=1,2,3………..28 types of relationship
If respondents know anyone who engages in the following 21 occupational positions through kth specific relationship, then the variable pos1, pos2, pos3,…………, pos21 will be treated as 1; otherwise will be treated as 0.
10 We use the following example to demonstrate how to calculate the average prestige of accessed positions through specific relationship. The following example is the accessed positions through current colleagues.
lawer computer security programer guard sum Prestige (73) (51) (25) General Friends 1 0 1 2
If an individuals access these two positions through current colleagues, then the prestige of this individual is calculate as:
AP=(73*1+51*0+25*1)/2 Tposk=1+0+1=2
We also can use the above example to compute the three diversity indicators of accessed positions through specific relationship:
The total accessed positions through specific relationship (Tposk)=2 The maximum prestige of accessed positions through specific relationship is computed by the function of Max (pos1, pos2, pos3,…….pos21) through specific relationship. Maxpos=73 The minimum prestige of accessed positions through specific relationship is computed by the function of Min (pos1, pos2, pos3,…….pos21) through specific relationship. Minpos=25.
The range between the maximum and minimum prestige of accessed positions through specific relationship is
Range=Maxpos-Minpos=73-25
RESEARCH FINDINGS
The Descriptive Statistics
The Distribution of Accessed ties by Types of Relationships
11
Table 1 presents the percentages of accessed ties shared by types of relationships in the Unites States, Taiwan and China.The proportion of accessed ties through general friends is the highest in US (25.56%), Taiwan (18.01%) and China (10.10%). The proportion of accessed ties through good friends is the second highest in both US (14.50%) and China (9.50%), and that of accessed ties through other relatives (13.58%) is the second highest in Taiwan. The proportion of accessed ties through current colleagues is the third highest in both US (5.65%) and Taiwan (9.85%), and that of accessed ties through acquaintances (9.53%) is the third highest in China. It seems that general friends are the most important ties to access all types of positions in all three countries, and other relatives are more accessible and recognizable in Taiwan than those in the other two societies. This is maybe due to the dominant small and medium business which demands strong family network resources in Taiwan or the Chinese pseudo-family networking practice still prevail in Taiwan.
Table 1 about Here
The Average Prestige of Accessed Positions through Specific Relationships
Table 2 presents the average prestige of accessed positions through 28 relationships on the individual level in US, China, and Taiwan. The average prestige of accessed positions through teachers is the highest and that through current bosses is the second highest in all three societies. The average prestige of accessed positions through current neighbors is low in three countries. The average prestige of accessed positions through the same associations is the third highest in US and Taiwan; however, that through school classmate is the third highest in China. School classmates and current colleagues seem to access better resources in China compared to those in Taiwan and US. Evidently, the average prestige of accessed positions through different types of relations is stratified.
Table 2 about Here
Individual Characteristics
Table 3 indicates the difference on individual characteristics among three countries. There are more percentage of males in Taiwan sample compared to those in US and China sample. The average age is quite similar in both US and Taiwan sample,
12 and that in China sample is younger. In terms of married percentage, those in US and Taiwan sample are similar, and that in China are much higher. In terms of ethnicity composition, the majority of ethnic group in US is Whites, that in China is Han people, and that in Taiwan is Min-nan people. In terms of residence pattern, the percentages of people living in big city are similar in both US (24.0%) and Taiwan (29.0%), and that in China is higher (35.8%). The percentage of people living in suburbs is much higher in US (22.0%) than those in China (9.4%) and Taiwan (10.0%). On the average, the percentage of people living in small city and town is 36.5% in US, 30.4% in China and 25.6% in Taiwan respectively. In terms of household size, the average household size is 3.22 in US, 3.12 in China, and 4.71 in Taiwan. The percentage of children under 6 is 26.1% in US, 16.2% in China, 15.8% in Taiwan. The average years of education is 14.6 in US, 11.4 in China, and 12.0 in Taiwan. The percentage of employed people is 77.2% in US, 77.7% in China, and 73.4% in Taiwan. The average daily contacts are similar in both US and Taiwan, and that was fewer in China.
Table 3 about Here
Network Characteristics of Accessed Positions through Specific Relations
In order to do further analyses and comparison for the average prestige of accessed positions through specific relationships among three societies, we choose those relationships which share more percentages of accessed ties or more significant relationships in daily life among three societies: other relatives, current colleagues, good friends, and general friends. Table 4 presents the information on the network characteristics of accessed positions through these four types of relationships.
Table 4 about Here
Comparing the network characteristics of accessed ties through four types of relationships, we found that there are different social resources are produced through different types of relationships. We compare the following indexes: percentage of males among accessed ties through specific relationship, average closeness of accessed ties through specific relationship, maximum prestige of accessed positions through specific relationship, minimum prestige of accessed position through specific relationship, and the range between the maximum and minimum prestige of accessed positions through specific relationship.
13 It may be due to relatively lower percentage of males in US sample, so the percentages of males accessed through all kinds of ties are relatively lower compared to those in China and Taiwan. In China, the percentages of male ties through four types of relationship are greater than those of US and Taiwan. In all three societies, the percentages of males among accessed ties through other relatives are the highest compared to the percentages of males among accessed ties through the other three types of relationships. The percentages of males among accessed ties through current colleagues are the lowest in all three societies. In general, more proportion of males among other relative ties is recognized; however, less proportion of males among current colleague ties is recognized.
In all three countries, the average years of knowing other relatives are the longest (more than 20 years), and those of knowing good friends are the next longest. The average years of knowing general friends are longer than that of knowing current colleagues in the United States. However, the average years of knowing general friends are shorter than that of knowing current colleagues in both China and Taiwan. This implies that the work mobility seems to be higher in the United States than that in China and Taiwan, so the years of knowing current colleagues seems to be shorter.
The average degree of closeness to good friends is even stronger than that to other relatives in the United States. However, the average of closeness to other relatives is the strongest in both China and Taiwan. So, the kin ties are still closer than the other types of ties in Chinese culture. The average degree of closeness to general friend ties is weaker than those to current colleague ties in all three countries.
In the United States, average maximum prestige of accessed positions through general friends and good friends is higher than that through other relatives and current colleagues; however, the average maximum prestige of accessed positions through good friends and current colleagues are higher than that other relatives and general friends. The occupational range of accessed positions through general friends is the largest in all three countries. The occupational range of accessed positions through weaker ties tends to be larger. The range of accessed positions through current colleagues is the lowest in the United States; however, the range of accessed positions through current colleagues in both China and Taiwan is quite large. This implies that the American occupational system seems to be more specialized, so people tend to build up colleague ties with similar occupational positions or lower range of occupational positions. However, the jobs are not so specialized in both China and Taiwan, so current colleagues can access larger range of occupational positions.
14
In sum, different society indicates some differences in terms of relation-generated prestige and mechanisms. From the above description, some differentiation principles of access to positions through these four types of relationships are universal in three societies. Generally, weaker ties tend to access large range of occupational positions. In China, good friends (strong ties) can access higher maximum occupational positions than general friends did. In general, the network characteristics of accessed positions through four types of relationships are more similar in both China and Taiwan. This finding is also consistent with that of correspondence analysis and associational models for the contingency tables between accessed positions and types of relationships (Hsung and Breiger 2007). We summarize the average prestige and range of accessed positions through other relatives, current colleagues, good friends and general friends in three societies in order to show the stratified social resources through different types of relations (Table 4).
US Taiwan China P R P R P R Other relatives 48.5 11.9 43.8 12.1 43.8 8.1 Current colleagues 45.5 11.7 44.2 15.3 47.0 11.3 Good friends 49.7 13.5 45.5 12.0 48.3 10.7 General friends 48.2 19.7 42.1 17.4 44.2 12.7 P: average prestige of accessed positions through specific relationships R: range of accessed positions through specific relationships
We further to compute the diversity factor score by the factor analyses. Factor Analyses on diversity of accessed positions through four types of relationships are presented in Table 5. We use three indicators to figure out the factor loadings to the first factor: total number of accessed positions through specific relationship, maximum prestige of accessed position through specific relationship, range between the maximum and minimum prestige of accessed positions through specific relationships. The largest loading for the Factor I is the range of accessed positions through all four type of relationships. We also compose an index of diversity and do further regression analyses to explain it.
Table 5 about Here
15 The Factors Affecting the Social Resources through Specific Relationships
Some of our respondents reported that they had no access to any position through specific relationships. In the models of explaining the average prestige of accessed positions through specific relations, those respondents with zero accessed positions through specific relationship will be excluded from the analyses. The exclusion of these respondents suggests that there may be sample selection bias. So, we use Heckman model to estimate the regression effects. These models estimated a sample selection model and a substantive regress model and assessed the correlation between their respective error terms using the statistic ρ. If the correlation is near zero then the potential for sample selection bias is low, and the results of estimating models with and without considering sample selection will be similar (Berk 1983).
There are two dependent variables to indicate social resources of accessed positions through each specific relationship: the average prestige of accessed positions through specific relationship and the diversity score of accessed positions through specific relationship. The variables which are put into the selection model are: gender, age, age square, ethnicity, type of residence, log of household size, having children under six years of old, years of education, employed, daily contacts and total number of civic associations engaged. In addition to these variables on individual characteristics and behavior variables, we add some network characteristics variables to explain the prestige generated by specific relationships. The networks characteristics variables include percentage of males among specific relationships ties, the average years of knowing specific relation ties, the average degree of closeness to specific relational ties, the interaction between gender and the percentage of males among ties. For explaining the prestige of accessed positions through four types of relationships in three societies, we can successfully found the Heckman models to estimate the factors on the dependent variable. However, for estimating the diversity score of accessed positions through four types of relationships, some of the models can not be estimated by Heckman model, so we still use regression models to estimate them.
In order to demonstrate the factors of affecting these two dependent variables among three societies in a clear way, we summarize the findings by major factors which previous literatures have more empirical supports and discussion on them and attempt to do more dialogue with these empirical findings (from Table 6 to Table 13).
16
Gender Factors
If we measure the total number of accessed positions or diversity index through all types of relationships, then the gender differences are not significant in the United States and Taiwan. When we differentiate the social resources of accessed positions by different types of relationships, the gender differences vary with the relation-specific social resources. Men tend to create higher average prestige of accessed positions through good friends and general friends than women do in all three countries. This gender effects significantly affect the average prestige of accessed positions through current colleagues only in Taiwan. Gender did not have any effect on the average prestige of accessed positions through other relatives. Women tend to have greater probability of accessing or recognize other relatives as their accessible resources than men as most previous studies found. However, there are no differences of the quality of social resources through other relatives by gender. This implies that men recognize higher prestige of good friends and general friends than women do. Men tend to treat good and general friends as instrumental resources, so they prefer to contact or recognize those friends with better occupational positions. Access to other relatives is more ascribed process, there is no differences in recognize the quality of relative resources.
US Taiwan China P D P D P D Gender Effects Other relatives n.g. n.g. n.g. n.g. n.g. n.g. Current colleagues n.g. n.g. + n.g. n.g. n.g. Good friends + + + + + +# General friends + + + + +# +#
P: average prestige of accessed positions through specific relationships D: diversity score of accessed positions through specific relationships Life Course + significant and positive effect, - significant and negative effect, n.g. no significant effect +# p<.10
Age Factors
17 Previous literature indicated that there is a curvilinear effect of age on social capital (Erickson 2004). The social capital increased along with the increase of age until middle age. After middle age, the social capital started to fall along with the aging. When we disaggregate the social resources by types of relationships, we find that this curvilinear effect of age on the social capital only exists in the models of the average prestige and the diversity score of accessed positions through general friends in both United States and Taiwan. This curvilinear effect of age on social capital also exists in the model of average prestige and diversity scores of accessed positions through other relatives in Taiwan. However, the curvilinear effect of age on the social capital through any type of relationship was not significant.
US Taiwan China P D P D P D Age Effects Other relatives (age) n.g. n.g. + + n.g. n.g. (age square) n.g. n.g. - - n.g. n.g Current colleagues (age) n.g. n.g. n.g. n.g. n.g. n.g. (age square) n.g. n.g. n.g. n.g. n.g. n.g. Good friends (age) n.g. n.g n.g n.g n.g. n.g. (age square) n.g. n.g. n.g. n.g. n.g. n.g. General friends (age) + + + + n.g. n.g. (age square) - - - - n.g. n.g. P: average prestige of accessed positions through specific relationships D: diversity score of accessed positions through specific relationships Life Course + significant and positive effect p<.05, - significant and negative effect p<.05, n.g. no significant effect
Locality Factors
There is no significant effect of location on the average prestige and diversity score of accessed positions through other relatives in all three countries in all three societies. Respondents who live in different degree of urbanized areas recognize their other relatives randomly. In another word, respondents who are requested to recall their relatives are not constrained by the urbanization of location. The locality effect on the social resources through current colleagues exists only in the United States. Individuals in big city or suburbs tend to create higher prestige of accessed positions through current colleagues; however, the location effect on the diversity score of
18 accessed positions through current colleagues is not significant in the United States. The locality effect on the average prestige of accessed positions through good friends also exists only in the United States. Individuals in suburbs and small towns tend to produce higher prestige of accessed positions through good friends in the United State, and the small towns seem to be more advantageous to create better quality of good friends in both United States and Taiwan. Individuals in big city tend to produce greater social resources through general friends in both Taiwan and China. All the above findings imply that the representative recognition effects reported in Erickson’s study (2008). The people who live in big city and suburbs know their colleagues also working in big city and suburbs and their occupational prestige is higher than those in rural villages.
US Taiwan China P D P D P D Location Effects (contrast-rural villages) Through other relatives Big cities n.g. n.g. n.g. n.g. n.g. n.g. Suburbs n.g. n.g. n.g. n.g. n.g. n.g. Small towns n.g. n.g. n.g. n.g. n.g. n.g. Through current colleagues Big cities + n.g. n.g. n.g. n.g. n.g. Suburbs + n.g. n.g. n.g. n.g. n.g. Small towns n.g. n.g. n.g. n.g n.g. n.g. Through good friends Big cities n.g. n.g. + n.g. n.g. n.g. Suburbs + n.g. n.g. n.g. n.g. +# Small towns + n.g. + + n.g. + Through general friends Big cities n.g. n.g. + n.g. + n.g. Suburbs n.g. n.g. n.g. n.g. n.g. n.g. Small towns n.g. n.g. + +# n.g. n.g.
P: average prestige of accessed positions through specific relationships D: diversity score of accessed positions through specific relationships Life Course + significant and positive effect, - significant and negative effect, n.g. no significant effect +# p<.10
19 Education Factors
Individuals with higher education tend to have greater probability to recognize current colleagues and good friends as their accessible positions in all three countries. This is the most important factor affecting relation-specific social capital. Education significantly affects the average prestige of accessed positions through other relatives in both United States and Taiwan instead of that in China. Highly educated people recognize those relatives with better social resources in both United States and Taiwan. This implies that both United States and Taiwan have gone through modernization and capitalism for more than fifty years. However, China has transformed to market economy less than twenty years. In China, highly educated individuals often have many poorly educated relatives, especially for those relatives experiencing Cultural Revolution. Education is the most important factor predicting the average prestige of accessed positions through current colleagues in both Taiwan and China instead in the United States after taking account of selection models (Table 5). Without considering the selection bias factors between those respondents who access and not access to current colleagues, education also significantly affect the average prestige and diversity score of accessed positions through current colleagues. Education significantly affects the average prestige of accessed positions through good friends and through general friends respectively in all three countries. The effects of education on the diversity score of accessed positions through good or general friends are not so strong as those on the average prestige of accessed positions.
US Taiwan China P D P D P D Education Effects Other relatives + n.g. + n.g. n.g. + Current colleagues n.g. + + + + n.g.
Good friends + n.g. + +# + +# General friends + + + +# + n.g. P: average prestige of accessed positions through specific relationships D: diversity score of accessed positions through specific relationships Life Course + significant and positive effect, - significant and negative effect, n.g. no significant effect +# p<.10
20 Gender Composition among accessed positions
The greater percentage of males among accessed other relative ties negatively affects the average prestige of accessed positions through other relatives in three societies. This implies that females provide better social resources among other relatives. Or, alternative interpretation is that the more resourceful female relatives are more recognizable by respondents. Wellman (1991) found that women are often considered as emotional resources. Family relationships are mainly the fields dominated by emotional supports, and women with more social resources seem to play important sources among relative.
There are positive effects of percentage of males among accessed good friends and negative interaction effects between gender and percentage of males of accessed positions through good friends in both United States and Taiwan. Similar effects also exist in the models of explaining the average prestige of accessed positions through general friends in both United States and Taiwan. Specifically, female respondents with more proportion of males among accessed good friends or among general friends tend to produce better resources in both US and Taiwan. This finding indicates that more proportion of male friend ties are more advantages for women instead for men. Burt (1998) also supported the argument that women with better male social resources tend to be promoted earlier. He considered this as the borrowing social capital of women from men. Wellman (1991) also found that men often are considered as instrumental resources. . US Taiwan China P D P D P D Other relatives Percentage of males - n.g. - - - n.g. Males*Percentage of males n.g. n.g. n.g. n.g. n.g. n.g. Current colleagues Percentage of males - n.g. n.g. + n.g. + Males*Percentage of males n.g n.g. - - n.g. n.g. Good friends Percentage of males + + + n.g. - n.g. Males*Percentage of males - +# - +# - n.g. General friends Percentage of males + n.g. + n.g. -# n.g. Males*Percentage of males - - - -# n.g. n.g.
21 P: average prestige of accessed positions through specific relationships D: diversity score of accessed positions through specific relationships Life Course + significant and positive effect, - significant and negative effect, n.g. no significant effect +# p<.10
Tie Strengths
The effects of tie strength on the average prestige of accessed positions vary with different types of relations. Weaker ties seem to produce better social resources for accessing better resources of other relatives in both Taiwan and China but not in the United States. Shorter years of knowing other relative ties produce higher average prestige and greater diversity of accessed positions through other relatives in both Taiwan and China. To some extent, strong ties seem to produce better social resources through four types of relationships in China. In China, respondents who feel closer with other relatives, know current colleagues longer, feel closer with good friends, know general friends longer and feel closer with general friends tend to produce higher prestige of accessed positions through respective relationships. The greater average closeness of accessed good friend ties causes the higher average prestige of accessed positions through good friends in China. These results support findings by Bian (1997; 2008). Guanxi networks or stronger ties can access better positions. In anther interpretation, the respondents in China invest longer years of relations and more emotional feelings to their general friends and this strategy is also can be conceived as the mechanism of Guanxi networks (Hwang 1986).
US Taiwan China P D P D P D Other relatives Average years of knowing n.g. n.g. - - - - Average closeness with n.g. n.g. n.g. n.g. +# +# Current colleagues Average years of knowing n.g. n.g. n.g. -# + n.g. Average closeness with n.g. n.g. n.g. - n.g. n.g. Good friends Average years of knowing + n.g. n.g. n.g. n.g. n.g. Average closeness with -# n.g. n.g. n.g. + n.g. General friends
22 Average years of knowing n.g. n.g. n.g. n.g. + -# Average closeness with n.g. n.g. n.g. n.g. + n.g.
P: average prestige of accessed positions through specific relationships D: diversity score of accessed positions through specific relationships Life Course + significant and positive effect, - significant and negative effect, n.g. no significant effect +# p<.10
CONCLUSIONS
The contribution of position-generated networks from this study: Just like Erickson’s finding (2004) or Flap’s finding (2008), there are rules of structural differentiation rules in the social networks of daily life. Position-generated networks provide two dimensional distribution of social structure: distribution of occupational positions and distribution of types of relationships. This study creates the measures of relation-specific social resources: the average prestige of accessed positions through specific relationship, the diversity score of accessed positions through specific position. The social positions are mainly accessed through other relatives, current colleagues, good friends and general friends in the United States, Taiwan and China. In order to reduce the configuration of structural differentiation, different social resources produced by different types of relationships, we focus on the production of social resources through these four types of relationships.
The major findings and discussions of this paper are as it follows:
1. People produce different quality (average prestige or diversity score) of social resources through different types of relations in different societies (US, Taiwan, and China). Some principles of structural differentiation are universal, but some vary with different societies. Evidently, the social resources through four different types of relations are stratified. The average prestige of accessed positions through current colleagues in the United States, that through general friends in Taiwan, and that through other relatives are the lowest in respective society. The average prestige of accessed positions through good friends is the highest in three societies.
2. Education plays important role in terms of producing social resources through specific relations in three countries. The effects are higher for producing social
23 resources through good or general friends than for producing social resources through other relatives. The effect of education on the average prestige of accessed positions through current colleagues is not significant in the United States, but it is significant in both Taiwan and China. This finding suggests that the high job mobility in the United States and make the investment effects of human capital on the social capital at work is relatively not effective.
3. The patterns of relation-specific social resources and their explanations are more similar between Taiwan and China. The endogenous mechanism of gender composition and its interaction with gender on the social resources through good and general friends is similar in both US and Taiwan. Male ties are more advantageous to produce better resources through good friends and general friend ties, especially for women in both United States and Taiwan. However, female ties are more advantageous to produce better resources through other relatives.
4. The Guanxi mechanism (strong ties) seems to function better in China. The strong ties seem to produce better social resources through four types of relationships. Especially in the reproduction of social resources through general friends, knowing the general friends longer and feeling closer with general friends create higher average prestige of accessed positions. Evidently, the investment of general friends, Guanxi mechanisms seem to work better as Bian’s (1997) findings.
5. There are selection bias problems in accessed position through specific relationships (Mouw 2006). Some respondents have no access to specific relationship, then consequently those cases can not further be computed the average percentage and diversity score of accessed positions through specific relationship. In order to solve the selection bias problem, we use Heckman regression models to estimate the factors of explaining average prestige of accessed positions through all four types of relationships. However, it’s difficult to find out Heckman’s models to estimate the factors of explaining the diversity score through some specific relationships.
This paper raises some old theoretical issue on the structural differentiation of social networks across different social and cultural context. Human capital factor seems to be more important factor to determine the production of social resources in market society like United States and Taiwan. Strong ties still play important role of
24 production of social resources in China. In Chinese society, the manipulation and management of different types of relationships to produce different social resources seem to be familiar games in daily life games (Hwang 1987). Fei (1949) conceived the differentiation of Guanxi networks as a social circle which was composed by network members with different distance of extended family relationship from the ego.
REFERENCES
Bian, Yanjie. 1997. “Bringing Strong Ties Back In: Indirect Connections, Bridges, and Job Search in China.” American Sociological Review 62: 266-85.
Bian, Yanjie. 2001. “Guanxi Capital and Social Eating in Chinese Cities: Theoretical Models and Empirical Analysis.” Pp. 275-296. In Nan Lin, Karen Cook, and Ronald S. Burt (Eds.). Social Capital: Theory and Research. New York: Aldine De Gruyter.
Bian, Yanjie, Ronald Breiger, Deborah Davis, Joseph Galaskiewicz. 2005. “‘Occupation, Class, and Social Networks in Urban China.” Social Forces 84(4): 1443-1468.
Bourdieu, Pierre. 1983. “The Forms of Capital.” P. p. 241-58 in Handbook of Theory and Research for the Sociology of Education, edited by John G. Richardson. Westport, CT: Greenwood Press.
Breiger, R. L. 1974. “The Duality of Persons and Groups.” Social Forces, 53 (2):181-190.
Breiger, R. L. 2000. “A Tool Kit for Practice Theory.” Poetics 27: 91-115.
Breiger, R. L. and Mohr, J. 2004. “Institutional Logics from the Aggregation of Organizational Networks: Operational Procedures for the Analysis of Counted Data.” Computational and Mathematical Organizational Theory 10;11-43
Burt, Ronald. 1992. Structural Holes: The Social Structure of Competition. Cambridge, MA: Harvard University Press.
Burt, R. 1998. “The Gender of Social Capital.” Rationality and Society 10(1): 5-46.
25 Campbell, Karen E., Peter V. Marsden, and Jeanne S. Hurlbert. 1986. "Social Resources and Socioeconomic Status." Social Networks 8(1): 97-116.
Campbell, K. E., P. V. Marsden, and J. S. Hurlbert. 1986. “Social Resources and Socioeconomic Status.” Social Networks 8: 97-117.
Erickson, Bonnie H. 2001. “Good Networks and Good Jobs: The Value of Social Capital to Employers and Employees.” Pp. 127-158. In Nan Lin, Karen Cook, and Ronald S. Burt (Eds.). Social Capital: Theory and Research. New York: Aldine De Gruyter.
Erickson, Bonnie H. 2004. “The Distribution of Gendered Social Capital.” Pp. 27-50. In Henk Flap and Beate Voelker (Eds.). Creation and Returns of Social Capital: A New Research Program. London: Routledge.
Erickson, Bonnie H. 2008. “Why Some Occupations are Better Known Than Others.” Pp. 331-42 In Nan Lin and Bonnie Erickson (eds), Social Capital: An International Research Program. London: Oxford University Press.
Fei, Xiaotong. 1949. From the Soil, the Foundations of Chinese Society. Berkeley, CA: University of California Press. Flap, Henk and Beate Völker (eds). 2004. Creation and Returns of Social Capital. London: Routledge. Flap, H.D. 2002. “No Man is an Island.” Pp. 29-59. In O. Favereau and E. Lazega (eds.). Conventions and Structures in Economic Organization: Markets, Networks and Hierarchies. Cheltenham: Edward Elgar. Granovetter, Mark. 1973. "The Strength of Weak Ties." American Journal of
Sociology 78:1360-80. Ganzeboom, H.B.G. and Treiman, D.J. 1996. “Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations.” Social Science Research 25:201-39.
Hsung, Ray-May. Ronald Breiger, Yi-jr Lin, 2007,The Creation of Differential Social Capital: the Comparison of Taiwan, China, and United States, SUNBELT XXVII International Sunbelt Social Network Conference, 1-6 May, 2007, Corfu, Greece.
Hsung, Ray-May, Nan Lin and Ronald Breiger. 2008. Contexts of Social Capital:
26 Social Networks in Markets, Communities and Families. London: Routledge.
Hsung, Ray-May and Ronald Breiger, 2008, “Position Generators, Affiliations, and the Institutional Logics of Social Capital: A Study of Taiwan Firms and Individuals.” Edited by Ray-May Hsung, Nan Lin and Ronald Breiger. In Contexts of Social Capital: Social Networks in Markets, Communities, and Families. London: Routledge.
Lin, Nan, and Mary Dumin. 1986. “Access to Occupations Through Social Ties.” Social Networks 8: 365-85.
Lin, Nan. 2001. Social Capital: A Theory of Social Structure and Action. Cambridge: Cambridge University Press.
Lin, Nan, Yang-chih Fu, and Ray-May Hsung. 2001. “The Position Generator: Measurement Techniques for Investigations of Social Capital.” Pp. 57-84. In Nan Lin, Karen Cook, and Ronald S. Burt (Eds.). Social Capital: Theory and Research. New York: Aldine De Gruyter.
Lin, Nan and Bonnie Erickson (eds). 2008. Social Capital: An International Research Program. London: Oxford University Press.
Lin, Nan, Karen Cook, and Ronald Burt (eds). 2001. Social Capital: Theory and Research. New York: Aldine de Gruyter Pub.
Portes, A. (1998) ’Social Capital: Its Origins and Applications in Modern Sociology’, Annual Review of Sociology, 1-24.
Ven der Gaag, Martin and Tom Snijder. 2004. “Proposals for the Measurement of Individual Social Capital.” Pp. 199-218. In Henk Flap and Beate Völker (eds) Creation and Returns of Social Capital. London: Routledge.
Wellman, Barry and Scot Wortley. 1990. “Different Strokes from Different Folks: Communities and Social Support.” American Journal of Sociology 6(3): 558-588.
27 Job Search in Three Societies: Gender, Contacts, and Network Chains
Chih-jou Jay Chen and Te-lin Yu Academia Sinica Starting from the social capital model of status attainment
Education + +
Initial (parental Attained or previous) Statuses + Statuses + + Network Resources + + _ + +
Extensity of Ties Tie Strength _ Contact with Contact Status [Access to Social Capital] [Mobilization of Social Capital] Source: Lin 1999 Research Questions: Do tie-strength-effect and social-resource effect differ across gender groups in different societies?
Education +
Attained (1) Tie strength effect Statuses (2) Social resource effect
+ + (2)
Tie Strength _ Contact with Contact Status (1) [Mobilization of Social Capital] Source: Lin 1999 Data and measurements
• Data: Three-society social capital survey, wave 1 (2004-05)
• The contact or contacts – Job-search chains of multiple nodes – “The helper is my _____’s ______’s ______.” [e.g., my father’s coworker; my friend’s father’s uncle. ] The role relations and tie strength of job-search chains • Role relations: “So, the most important helper who helped you found your job is “ – My X (father, coworker, etc.) – My X’s X (father’s coworker, coworker’s friend, etc.) – My X’s X’s X (father’s coworker’s brother) • Intimacy: – “how close do you feel to the helper?”
Relationships for adjacent nodes, US 100%
80% 4. Ordinary friend 60% 3. Good/Close friend
40% 2. Work relationship
20% 1. Family member
0%
Ego and NodeEgo 1,and M Node 1, F Node 1 andNode Node 1 2,andNode M Node 2 and 2,Node FNode 2 3,and M Node 3, F Relationships for adjacent nodes, Taiwan
100%
80%
60% 4. Ordinary friend 3. Good/Close friend 40% 2. Work relationship
20% 1. Family member
0%
Ego and NodeEgo and1, M Node 1, F Node 1 andNode Node 1 andNode2, M Node 2 and 2,Node NodeF 2 and3, M Node 3, F Relationships for adjacent nodes, China
100%
80% 4. Ordinary friend
3. Good/Close friend 60% 2. Work relationship
40% 1. Family member
20%
0%
Ego and NodeEgo and1, M Node 1, F Node 1 andNode Node 1 and2,Node M Node 2 and 2,Node FNode 2 and3, M Node 3, F Adjacent Relationship between Ego and Node 1 100%
80% 4. Ordinary friend
60% 3. Good friend
2. Work relationship 40% 1. Family member
20%
0%
Male, CN Male, TWFemale, TW Female, CN Male, theFemale, U.S. the U.S. Intimacy with the Next Node, the U.S.
100%
80%
60% Not close
So so
40% Close
20%
0%
Ego and NodeEgo 1, and M Node 1, F Node 1 and NodeNode 12, and MNode Node 2 2,and F NodeNode 23, and M Node 3, F Intimacy with the Next Node, Taiwan
100%
80%
Not close
60% So so
Close 40%
20%
0%
Ego and NodeEgo 1, andM Node 1, F Node 1 and NodeNode 2,1 andM Node Node 2 2, and F NodeNode 3,2 andM Node 3, F Intimacy with the Next Node, China 100%
80%
60% Not close
So so 40% Close
20%
0%
Ego and NodeEgo 1, and M Node 1, F Node 1 and NodeNode 12, and MNode Node 2 2,and F Node 23, and M Node 3, F Intimacy between Ego and Node 1 100%
80%
Not close 60% So so 40% Close 20%
0%
Male, CN Male, TWFemale, TW Female, CN Male, theFemale, U.S. the U.S. Intimacy between ego and the last node
100%
80%
Not close 60%
So so 40% Close 20%
0%
Male, TW Male, CN Female, TW Female, CN Male, theFemale, U.S. the U.S.
Variables
• Attained status / Social Contact status: holding an executive position • Tie strength effect – Intimacy (emotional closeness) • between ego and the last helper – Role relations • between ego and the first helper in the chain – Chain length • The number of nodes of the job-search chain – Homophily • same-sex tie between ego and helper(s)
Y=attained status US Taiwan China
Total Male Female Total Male Female Total Male Female
Male ns + ns
Age + ns + + + ns + + +
Education + ns ns + + ns + + +
Social resources effect + + + + + + + + ns
Tie strength effect
Intimacy + + ns ns ns ─
Role relations (friends) ns ns ns ns ns ns ns ns ns Kin + + + ns ns ns ns ns ns Work relationship
Chain length ns ns ns ns ns ns ns ns ns
Homophily: same-sex tie ns ns ns ns ns ns ns ns ns Y=Contact status US Taiwan China
Total Male Female Total Male Female Total Male Female
Male ns + ns
Age + + ns + ns ns + + + Education + + + + + + + + +
Tie strength effect
Intimacy ─ ns ─ ns ns ns ─ ─ ns
Role relations (friends) Kin ─ ─ ─ ns ns ns + + ns Work relationship + ns + + + + ns ns ns
Chain length + + ns + + ns ns ns ns
Homophily: same-sex tie ns ns ns ─ ns ─ ns + ─ Conclusive remarks • The use of social ties of varying strengths and attribute homophily are shaped by institutional and labor market contexts. – Kinship relations play an important and effective role in China, particularly for Chinese males. – Work relationship is most effective in the U.S. and Taiwan.
• Role relations and tie-homophily are important in the mobilization of social capital. • Gender difference in the mobilization of social capital is significant in the three societies, with varying degrees and contents. Institutional Constraints and Social Capital of Individuals in the Labor Markets: Comparison among the United States, China, and Taiwan
Joonmo Son National University of Singapore
Running head: INSTITUTIONAL CONSTRAINTS AND SOCIAL CAPITAL
Word count: 10,470 Number of tables: 17
May 2008
Institutional Constraints and Social Capital of Individuals in the Labor Markets: Comparison among the United States, China, and Taiwan
Abstract
Employing the unique Social Capital Project surveys conducted in 2004-5 in the
United States, China, and Taiwan with at least 3,000 respondents from each society, I
found that institutional constraints played critical roles in determining the amount of
accessed social capital and the efficacy of activated social capital in job search process in the labor markets. Applying structural equation modeling to the identical position generator measures in the three data sets, I identified that China under the twofold institutional constraints of socialist political economy and Confucian culture retained the
least amount of accessed social capital in comparison with the other two capitalist
societies, the United States and Taiwan. In regard to the activation of social capital in job search process, I found that the effect of social capital was suppressed under the socialist control in China in comparison to the other two societies. I also found that Taiwan experienced the greatest gender inequality in status attainment process due to Confucian cultural influence.
Institutional Constraints and Social Capital of Individuals in the Labor Markets: Comparison among the United States, China, and Taiwan
The study aims to identity the interaction patterns between institutional
constraints and social capital in the status attainment process in the labor markets of the
United States, China, and Taiwan. In this paper, “social capital” of people in the labor
markets, defined “as the resources embedded in social networks accessed and used by
actors,” is the main variable of interest accounted for by the effects of the two social
institutions – political economy and culture (Lin 2001:25). The specific macrostructural
contexts take three different societal entities, the United States, China, and Taiwan. That
is, the study conducts a comparative study among the three societies. The fundamental
argument is that the macroinstitutional constraints of the political-economic and cultural systems affect the composition and utilization of social resources in the three countries.
The pattern of social tie composition within a society represents the effects of
macro institutions on its people. But the activation of social resources by the actors in a
specific socioeconomic stratum molded by a social institution is hypothesized to produce
a variation in outcomes in the labor market as well. Thus in terms of the theoretical
framework, I emphasize the structural effects of social institutions on the formation of
individual social capital; however, I do not ignore the possibility of upward or downward
mobility of individuals in the structure by optimizing their social capital, other things
being equal.
Above all, a uniform set of measures of social capital is essential for the purpose of comparative study among different countries. While social capital and its dynamic development are analyzed within the labor markets of each referred society, a consistent
1 comparison among the United States, China, and Taiwan is possible with the help of
identical measures of social capital based on the ties’ positions in the occupational
hierarchies in the three societies. This approach makes it possible to launch comparative
research on social capital in the labor markets among different countries, but has not been tried in the literature to date.
(Figure 1 about here)
Specifically, it is proposed that the advanced capitalist United States, state
socialist China, and developmental Taiwan have formed unique sets of social capital net
of the society-level differences. My first aim was to find variances in the compositions of
social capital among the countries. I hypothesize that the differences in social capital
across the three countries can be explained by the different political economic institutions
(capitalist vs. state socialist) and/or cultural institutions (Western vs. Asian) among them.
Once the possible institutional effects on the formation of social capital were
identified, the second goal of this study was to pinpoint the individual variance within each national entity, which operates through the differential utilization of social contacts in the labor markets. The utilization of social contacts to produce beneficial returns in the labor markets has been supported empirically (Lin, Dayton, and Greenwald 1978; Lin,
Ensel, and Vaughn 1981; Marsden and Hurlbert 1988; Smith 2005; Wegener 1991). One of my research aims was to reveal the differential forms of activation of social resources in the three countries, and which people are prone to use more social contacts and expect better returns in the labor market. That is, the analysis of how social resources affect upward mobility sheds light on the differential role that social resources play in different labor markets.
2 INSTITUTIONAL CONSTRAINTS AND SOCIAL CAPITAL
This study has a clear stance on the causal mechanism between institutional
constraints and social capital of individuals: institutional constraints generated from the
macro contrasts between capitalist and state socialist systems, and between Western and
East Asian cultures formed qualitatively distinct types of social capital. For a doable comparison among the United States, China, and Taiwan, the search for the causal relationship between institutional constraints and individual social capital were confined to the labor markets in the three countries.
Before going into details, I note that institutional constraints are defined as the
political economic and cultural forces in a given social structure that result in a different
set of individual social capital across societies. For instance, political regimes can form
the configuration of the educational chances and employment patterns in peculiar ways so
that overall tendency in making social networks may differ among nation-states; likewise,
strong cultural norms can divide people into segregative social relations between classes,
races, and gender groups (e.g., Caste in India, racial segregation in the United States and
Europe, and general gender inequality in most countries). This conceptualization is
critical because institutional constraints tend to form an overall framework of social
capital for people so that actors in the framework seek their own interests in the labor
market with various types of access to and activation of social resources. There have been
some studies directly or indirectly related to the theme of institutional constraints on
individual social capital.
Regarding the characteristics of social networks in the United States, Marsden
(1987) utilized the 1985 General Social Survey, which was the first survey to gather
3 personal network information from American people. Marsden found that American social networks are small, kin-centered, relatively dense, and homogeneous. In other words, a particularistic small network was characteristic of the personal relations of
Americans. In tandem with the Marsden study, McPherson, Smith-Lovin, and Brashears
(2006) utilized the 2004 General Social Survey to show that the social network of
American people had significantly shrunk in two decades with a decrease in the number of confidants.
Comparing the United States and China, Blau, Ruan, and Ardelt (1991) showed
that a comparative study of interpersonal networks is feasible. With the 1985 GSS data
used by Marsden and 1986 survey data of 1,011 Tianjin city residents in China, they
found that personal choices and relations in China, a socialist country at an early stage of
industrialization and economic development, seemed to be remarkably similar to those of
the United States, a capitalist country at an advanced stage of economic development.
The reason articulated by these authors for the similarity between the two societies is that
particularism governs all personal relations regardless of cultural, political, or economic
differences. However, their result may be strongly affected by the name generator
question they used: “Who are the people with whom you discussed matters important to
you?” This question may have a strong connotation of strong ties for respondents
naturally invoking particularistic answers about kin and nonkin ties around the
respondents, regardless of institutional differences between countries.
In a similar vein, Völker and Flap (1995) conducted a retrospective survey on 489
residents in former East Germany to compare social networks before and after the fall of the Berlin Wall. Contrary to general expectations, they found that the size of personal
4 networks of East Germans in 1993 and 1994 were smaller than those before the German
unification. The density of networks decreased, and the occupational diversity in social
network also decreased so that people’s chance to get the provision of necessities became
smaller than in the past. The authors explained that the reduction of social networks
resulted from individual reactions to the ongoing unstable institutional change. In other
words, the uncertainty of the new system compelled them to remain with their old
relations rather than absorb new ones.
Based on 1987 and 1997 surveys in Hungary, Angelusz and Tardos (2001) argued
that even after privatization general network resources of Hungarians were affected by
political involvement indirectly measured by self-reports of interest in political issues and
the frequency of political discussions. According to the authors, this may hint at the
continuing power of the old political institution on social network composition.
These empirical studies remind us of two critical issues. First, macropolitical
economic institutions within a society affect the size and composition of social networks
of individuals. Second, it may be impossible to compare social capital of individuals across various societies without a general standard. Few comparative studies of social capital have been done, partly because it was rare to see the same method of social network measurement on different societies. Even in comparative studies that used the name generator method (Blau, Ruan, and Ardelt 1991; Völker and Flap 1995), its tendency to generate information of intimate alters around an ego limited its explanatory power to private interpersonal networks without much implication about structural features of social ties in a broad scope, such as labor markets.
5 Accessed Social Capital
Against the backdrop of the macroinstitutional effects, it is essential to look into
the status attainment process of individuals in varied societies (Blau and Duncan 1967;
Sewell, Haller, and Portes 1969; Sewell and Hauser 1975). According to Lin (2001), the
status attainment process is composed of two steps: first, network resources, education,
and initial positions of parents are predicted to explain the variation of the attained status
of individuals; then contact statuses are expected to affect the attained status of egos. The
network resources retained by an ego are called accessed social capital, whereas utilized
network resources are called activated social capital. To be sure, all network resources
around an ego composed of his/her social ties are accessed social capital that is usually
measured by position or name generator, whereby a representative map of the network
and its potentiality to be utilized is captured; activated social capital is a part of the whole
network resources that is practically used for purposive actions by an ego (e.g., getting a
job or recruiting new members of a civic association). Past social network research
generally reached the conclusion that both accessed and activated social capital lead egos
to better statuses in the labor market.
Campbell, Marsden, and Hurlbert (1986) examined the associations between
accessed network resources and socioeconomic statuses based on the 1965-1966 Detroit
Area Study, and found that the resource compositions of networks were significantly
associated with attained statuses such as occupational prestige and family income.
Accessed social capital was also revealed to have positive effects on the reentry of the unemployed into the labor market. Sprengers, Tazelaar, and Flap (1988) found that those
6 who had better accessed social capital were more likely to be reemployed after a certain period of unemployment among 242 middle-aged Dutch men.
Further, accessed social capital has been found to affect income levels. Boxman,
De Graaf, and Flap (1991) found that both human capital (education) and accessed social capital, measured by work contacts in other organizations and memberships in clubs and professional associations, had direct effects on income. Recently, through mathematical simulation, Arrow and Borzekowski (2004) found that 13-15 percent of the unexplained variation in wages can be captured by accessed social capital (number of ties). Using data from the Mexican Migration Project, Aguilera and Massey (2003) found that accessed social capital of Mexican immigrants helped them find new jobs in the United
States and get higher wages, especially when they were on undocumented trips.
Activated Social Capital
As explicated above, activated social capital deals with the effect of social contacts in the process of job attainment. From data on a representative community sample in metropolitan Albany, New York, Lin, Ensel, and Vaughn (1981) found that the higher the contact status, the better the attained status of ego, controlling for parental status and education. Marsden and Hurlbert (1988) also verified that the occupational prestige and industrial sector of a contact’s status had the biggest effect on attained prestige and sector of ego, based on the 1970 Detroit Area Study.
The effect of activated social capital has also been corroborated by empirical studies done outside the United States. De Graaf and Flap (1988) conducted a comparative study of West Germany, the Netherlands, and the United States on the use of social ties in the job-search process. They found that a contact person with relatively
7 high prestige led to a job with higher prestige, especially in the Netherlands. Wegener
(1991) found that contact status significantly affected the job prestige of German
respondents. Bian and Ang (1997) compared the utilization of social ties for job changes
between China and Singapore, and found that contacts’ status had positive impact on job
changers’ attained job status in both societies.
Nonetheless, there was a major counterargument against the effect of activated
social capital: Mouw (2003) denied the effect of social contacts, showing that
occupational homogeneity between ego and contact constructed spurious causal relations.
He argued that when the same occupational ties between a job searcher and contact in terms of having the same occupational codes were excluded from the analysis, the effect of activated social capital turned insignificant, using the 1970 Detroit Area Study data used by Marsden and Hurlbert (1988). However, I found that Mouw’s argument was wrong because he failed to exclude the right cases in which ego’s “previous” job and contact’s job were in the same occupational category. Instead, he deleted the cases where ego’s “current/last” job and contact’s job were in the same occupational code, which does not make sense because job searchers utilized their contacts when they were in their previous jobs. In this study I employed the same method of deleting cases where respondents’ previous jobs and their contacts’ jobs fell in the identical occupational codes.
Although the effect of activated social capital has been generally supported in
empirical studies, the effect of weak ties has not always been found to be effective in
attaining better status. For example, based on a sample of employed adults in Chicago,
Bridges and Villemes (1986) found that the effect of weakly tied and work-related
informal contacts became insignificant once controls of human capital were included.
8 Wegener (1991) also showed that the effect of weak ties in the labor market was valid
only for individuals with high-status prior jobs, using German data of the life history of
604 adults. As an alternative to the traditional typology between strong vs. weak ties, I
introduce structural layers of binding, bonding, and belonging in this research. I
hypothesized that outer layers of belonging (participation in formal organizations) and
bonding (daily contacts) are generally better than innermost layer of binding (family-
oriented ties) in bringing better status for job searchers because the former relates people
to diverse social entities and ties outside the familiar boundary of their life. Nonetheless,
I proposed that binding is a significant source of social resources in the East Asian
societies where social relations are expanded from family relations.
DATA AND MEASURES
Data
The study used data sets from the same module administered in the United States,
China, and Taiwan, and the surveys were conducted between November 2004 and April
2005 in a synchronized fashion. The surveys were organized and funded by the
Academia Sinica in Taiwan. The project, under the title “Social Capital: Its Production and Consequences,” basically aimed to conduct a large-scale comparative study of social networks in the three societies.
The sample size was in excess of over three thousand respondents in each
country: 3,000 in the Unites States (response rate of 43.4 percent); 3,514 in China
(response rate of 40 percent); and 3,278 in Taiwan (response rate of 48 percent). The
data sets from the United States and Taiwan were nationally representative, and the
Chinese data set covered major regions and cities. Therefore, demographic
9 characteristics of the three data sets seem to be quite representative of each country,
although the Chinese data did not cover rural areas. The age of respondents uniformly
ranged between twenty-one and sixty-four in all three surveys to capture respondents currently or previously employed at the time of interview.
Measures of Accessed Social Capital
Extensity. The first indicator of accessed social capital is the number of jobs that a respondent knew from the list of twenty-one occupations. For instance, if a respondent knew nine jobs out of the list, then the extensity score is nine. Extensity can thus vary from 0 to 21.
It is desirable to check the actual distribution of extensity in the three samples of
the study. The mean and standard deviation of the measure are presented and subjected
to t-test to identify any significant differences among them at the univariate level.
(Table 1 about here)
As shown in the table, the mean extensity was the greatest in the United States
and the smallest in China, with Taiwan’s mean in the middle of the two. According to
the t-test results, the differences among the three societies are statistically significant.
Upper reachability. The remaining two indices are related to the prestige scores
attached to each occupation in the list. Upper reachability denotes the highest prestige
score among the jobs that a respondent knew and thus captures how rich the social
networks of the respondents are. For instance, if a respondent answered that he or she
knew nine job occupants out of the position generator questions, of which a CEO of a
large company had the highest prestige score, then its prestige score (70) was assigned as
the upper reachability score of the specific respondent.
10 (Table 2 about here)
At the univariate level, it turned out that the United States had higher upper reachability than the other two East Asian societies. However, China had higher upper reachability than Taiwan, according to t-test. What, then, is the reason? The reform and open policy and recent rapid marketization of the Chinese socialist government has made
it possible to introduce and increase prestigious and market-oriented jobs like CEOs in its
labor market; its people have better probability than in the past for making contacts with such jobholders, including expatriate CEOs from the West (Fernandez and Underwood
2006; Wong 2005).
Range of prestige scores. The third indicator of accessed social capital also
utilizes the prestige scores assigned to the jobs. This indicator measures particularly how
wide the network alters of a respondent are in terms of having diverse occupations on the
job list. Specifically, it is calculated by subtracting the lowest prestige score from the
highest prestige score among the jobs known to a respondent. For example, let us
suppose that a respondent knew ten jobs out of the twenty-one position generators. Next,
let us also assume that among the ten alters of the respondent a CEO of a large company had the highest prestige score (70), whereas a driver had the lowest score (31). The range of prestige scores for the specific respondent is calculated by deducting the lowest score from the highest (70-31=39).
(Table 3 about here)
The range of prestige was the greatest in the United States, followed by Taiwan
and China, according to t-test results. Since the index of range of prestige, along with
the extensity index, captures the diversity of the social network, it seems at least at the
11 univariate level that Americans enjoy the most diverse contacts in their networks,
whereas the Chinese have the least diverse ties in terms of occupational heterogeneity in
their social networks.
Factor score of the three indices. Van der Gaag and his colleagues (2007)
argued that the three indices of position generator can cause multicollinearity problems because they are positively correlated with one another. Further, upper reachability and range of prestige tend to deviate from normality and be skewed to the right. In order to exclude multicollinearity, I thus conducted factor analyses of the indices from the three data sets (principal component method, varimax rotation, a criterion of an eigenvalue equal to or greater than 1, and scoring coefficients based on varimax rotated factors).
(Table 4 about here)
A single factor score was produced in each of the three data sets. The factor score was used in the following statistical analyses. As shown, the single factor explained 81 to 86 percent of the total variance created by the three indices. The social capital indices in the three data sets had factor loadings of at least .85 on the single factor, which supports the efficacy of the factor score in the analyses.
Measures of Activated Social Capital
Presence of contact. The survey asked respondents in all three countries,
“During the process of getting your [current/last] job, how many people helped you?
[This may include all kinds of help, e.g., telling you about the job opportunities, putting in a good word for you, etc.]?” I then proceeded to make a dichotomous variable where 0 means no contact used and 1 means contact(s) used in job search.
(Table 5 about here)
12 It is interesting to see that use of contacts was most active in the United States,
while it was least used in China, with Taiwan in between. T-test results showed that the
differences in using contact(s) among the three societies were statistically significant. I conjecture that the least use of contacts in China was due mainly to the institutional intervention by the Communist government that assigned jobs and made it illegal to utilize private ties in the job search process until the mid-1990s.
Length of contacts’ chain. The survey also probed for more information about
helpers in the job search process. Specifically, the survey collected information about up
to four helpers in regard to their gender, race, and occupation. From the series of
questions, I constructed two variables to be employed in the analyses: length of helpers’
chain and contact status.
The length of helpers’ chain captures the quantity of activated contacts in the job search process. Given that we can see detailed information on who those persons were thanks to some probing questions, the variable should work as a conservative measure of the quantitative aspect of social resource activated in the job search.
(Table 6 about here)
As shown in the table above, the length of chain in China was significantly longer
than those of the other two societies. There is no significant difference found between
the United States and Taiwan, the two capitalist societies. I note that the reference
category of the variable (0) was given to no-tie users in a reconstructed variable of chain
length used in the actual analyses so that the direct-tie users without any chain took the
response category of 1 to be differentiated from no-tie users. If a respondent used two
chains, he or she thus fell into the fourth response category of 3 in the newly created
13 variable.
To examine the composition of contacts, the survey asked (1) if the respondent and their most important helpers were linked directly or indirectly, (2) what kind of
relationship they shared with the most important helpers (e.g., family, coworkers, friends,
or neighbors), and (3) how close they and their first contacts were.
(Table 7 about here)
As shown, Chinese people had the smallest percentage of direct contacts (81.5%)
to the most important helpers in comparison with the other two peoples. Next,
percentages of family ties among direct contacts of the most importance show that the
majority of direct contacts was family ties for Chinese people (57.7%), far more than
those of the other two peoples. It thus tells us that the majority of Chinese people’s most
important direct ties lie in their family members, including extended families. American
people had the smallest percentage of family members in direct contacts (21.1 percent), followed by Taiwanese people (39.2 percent). I also note that 18 percent of family ties in the most important direct contacts were parents for Chinese people, whereas it was 4.9 percent and 11.2 percent for American and Taiwanese peoples, respectively. The differences among the three societies were statistically significant according to t-tests.
Further, there was a significant gap between the United States and Taiwan in terms of using family contacts as the most important ties. The difference between the United
States and the other two East Asian societies may be due to a large extent to the
Confucian social order that centers on family values.
Contact status. The socioeconomic status of contacts should be taken into account independently of the presence of contacts and the length of contact chain because
14 the quality of contacts is strongly associated with the job search outcomes. As explicated
above, the survey asked the respondents about the occupational class of up to their fourth contact. Thus it was necessary to stratify the contacts following a consistent procedure.
To begin, the occupations of the contacts were given the occupational classification codes
following the standard occupational classifications in each society. The contact jobs of
the U.S. respondents were given the occupational codes of the 2000 census occupational
classification system, while those of the Chinese contacts were attributed to the
occupational codes of the 1995 Chinese Standard Classification of Occupations. For the
jobs of the Taiwanese contacts, the four-digit occupational codes from the Standard
Occupational Classification System of Taiwan were employed (Li et al. 1999). The
contact status variable was made with the standardized information based on the
occupational codes. Specifically, based on the occupational code information of all the contacts, I categorized contact status into four categories: low class, middle class,
professional, and executive.
(Table 8 about here)
In the next step I chose the highest occupational class among up to four contacts
of each respondent. For instance, if a respondent reported that his/her first contact’s
occupational class was low class, the second middle class, the third professional, and the
fourth executive, then I chose the occupational class of the fourth contact as the person’s
contact status since it is the highest contact resource activated in the job search process.
Thus regardless of where the contact of highest status lay within the contact chain, I
selected those specific contacts with the highest occupational class because high contact
status has been one of the most significant measures of activated social capital in the
15 literature (Boxman, De Graaf, and Flap. 1991; Campbell, Marsden, and Hurlbert 1986;
Lin 2001).
As shown in the table, it is noteworthy that the mean contact status was highest in
the United States and lowest in China, with Taiwan in between. The gaps among the three societies were statistically significant according to t-tests.
Routine job information. Routine job information or unsolicited job information
can be added as another dimension of activated social capital in the sense that job
searchers may follow up on such information to obtain new positions. Specifically, the
survey probed if someone mentioned job possibilities, openings, or opportunities to a
respondent in casual conversations (e.g., face-to-face, telephone, email, fax, etc.) during
the past year without being asked. Considering that respondents did not actively pursue
such routine job information, this measure is quite different from other activated social
capital variables related to active job search behavioral dimensions. Nonetheless, it can
be located under the umbrella of the activated social capital because routine job
information is derived from a job searcher’s network members (McDonald 2005;
McDonald and Elder 2006).
(Table 9 about here)
Routine job information was richest in the United States, where about 40 percent
of respondents obtained such information in one way or another in the past twelve
months. The Chinese lacked such information compared to the other two societies; only
22 percent of its respondents reported that they received it. The gaps of quantity in
routine job information among the United States, Taiwan, and China were statistically
significant.
16
Measures of Status Attainment
As briefly mentioned above, the outcome indices of status attainment should show
where the respondents ended up in the occupational hierarchy affected by the effects of
accessed and activated social capital, ceteris paribus. I use two indices for the outcome:
occupational class and annual income.
Occupational Class. The first of the two outcome variables from the job search
process employed in this study was the current or last occupational status of the obtained
job in the labor market at the time of interview. To begin, the occupations of the
respondents were given the occupational classification codes following the standard
occupational classifications in each society.
(Table 10 about here)
An outcome variable of occupational class was made using the occupational codes. Based on the occupational code information of each current/last job title of the
respondents, I categorized the jobs into four categories: low class, middle class,
professional, and executive. As shown in the table below, the United States had the
greatest proportion of professional and executive jobs in its occupational structure. In
contrast, the majority (69 percent) of the Chinese respondents had lower-class jobs. It is
also shown that the United States had the highest mean in the occupational classes;
Taiwan was the second, and China had the lowest mean score. The gaps in the means
were statistically significant.
Annual Income. Annual income was the other outcome variable of status attainment in the three societies. The variable was constructed to be semicontinuous with
17 at least twenty (Taiwan) to twenty-seven (United States and China) categories in the three
data sets.
(Table 11 about here)
The United States had the highest annual income level among the three countries,
China the lowest, and Taiwan in between. It appears that annual income was
underreported by U.S. and Taiwanese respondents, given that the United States recorded
$39,722 and Taiwan recorded $14,271 of the GDP per capita in 2004, respectively. The
mean annual income of the Chinese respondents was closer to its GDP per capita of
$1,315 in the same year (Department of Investment Services in Taiwan
[http://investintaiwan.nat.gov.tw/en/env/stats/per_capita_gdp.html]).
Although the response categories were not constructed identically across the three
surveys, it should not raise any fundamental obstacle in statistical analyses to capturing
the dynamics of social capital and other covariates in obtaining better financial returns in
each of the trisocietal labor markets.
Demographic Features
Age, gender, and race are key demographic features of the respondents. Note that
age is a time-varying variable that can take nonconstant values by events of interest while
the other two are time-invariant variables. To keep temporal order in the statistical models, I chose age at the starting point of the current/last positions of the respondents
rather than age at the time of the interviews.
(Table 12 about here)
Age. Chinese respondents were two to three years younger than the other two
peoples according to the mean ages. Mean ages at the starting point of current/last
18 positions were at least eight years less than those at the time of interviews in the three
societies. To maintain the temporal order in the job search models, age at the starting
point of current/last positions was used in the analyses as explicated.
Male. Males composed 46 to 52 percent of the samples in the three societies. As
hypothesized, this variable was employed to identify if there were different patterns of
gender inequality in the job search process, accessed and activated social capital, and
status attainment across the target societies.
Race. White, black, Latino, and other races were used in the models of the U.S.
data set. Whites constituted 69 percent of the sample while blacks and Latinos were 12
and 14 percent, respectively. The reference category is whites in the multivariate models.
Structural Layers of Social Relations
Social network is composed of a variety of social relations that can be differentiated in terms of strength of ties, duration of the dyadic relationship, role relational categories, or demographic homogeneity/heterogeneity. The study employs the conceptualization of concentric structural organization in the social network,
distinguishing its three layers of social relations from inside to outside (Lin, Ye, and
Ensel 1999; Son, Lin, and George 2008).
Specifically, the inner layer nearest to an ego is called binding (ties to and through spouse), while the middle layer is composed of bonding social ties that are contacted by an ego on a daily basis. Lastly, the outer layer in the social network is belonging
(community participation) measured on the number of memberships in community organizations.
19 It should be noted that binding and belonging variables were measured both when
the respondents were interviewed and when they started their current/last positions
prompted by retrospective questions, while the bonding variable was measured only at
the time of interview. Thus the multivariate analyses employ the retrospective measures
of binding and belonging to set up temporal order between explanatory and outcome
variables, although I had to compromise by using the bonding measure at the time of
interview.
(Table 13 about here)
I propose that these measures of social network structural layers are strongly
associated with the accessed social capital indices because the former functions as the
source of the latter in the sense that accessed social capital captures part of the variance in the structural layers of social network.
Binding (ties with and through spouse). The measure is composed of six
categories from 0 (not married), 1 (know almost none of spouse/partner’s friends), to 5
(know almost all friends of spouse/partner). It therefore takes persons without marital
relations as its base category to set up differential degrees of familiarity to spousal ties.
Put another way, the categories of 1 to 5 show the variations in network overlaps between
respondents and their spouses/partners. Notice that the spousal network was smaller at
the time of current/last positions than at the time of interviews in all three societies,
implying the increased degree of assimilation in spousal ties across time. There was no
systematic difference among the three countries according to the means and standard
deviations in the spouse network item.
20 Bonding (daily contact). The measure was created by the question, “On average, about how many people do you make contact with in a typical day?” As seen in the table, it had six response categories from smallest to greatest possible number of contacts.
According to the means of the variable people contacted on average twelve to fourteen persons in a typical day in the three societies. T-test results further found that the number
of people contacted daily by Chinese respondents was significantly less than those by
American and Taiwanese peoples.
Belonging (memberships in community organizations). The measure resulted
from a series of questions that asked if respondents were participating in specific types of
community organizations. Even though there were some disparities in the types and
number of organizations in the three societies, differences in the mean numbers of
memberships among them looked outstanding at first glance. Nonetheless, in order to
check if such seeming differences in the mean numbers of memberships were due to the
nonidentical set of community organizations listed in the surveys, I reduced the number
of organizations to only three identical ones (i.e., religious, leisure or sports, and other)
among the three societies at both the time of interview and the starting point of the
current/last position. The unequal distribution in the means stayed similar to the original
one; the United States had means of .77 at the time of interview and .54 when starting the
current/last position; Taiwan .30 and .15, respectively; and China .03 and .01,
respectively.
What do we see from the mean differences? First, the mean number of memberships was the greatest in the United States, while that of China was smallest, with the Taiwanese mean in the middle. The pattern holds regardless of the two time points
21 available in the data (i.e., belonging 1 and 2). Second, more specifically, the Americans
had significantly greater numbers of memberships in community organizations than the
two East Asian countries according to t-test results at .000 level. Notice that the mean
numbers of memberships were less than 1 in both Taiwan and China. Third, it should not be overlooked that there was also a statistically significant gap between Taiwan and
China in terms of number of memberships in community organizations.
The results show that the institutional and cultural differences between American
and East Asian societies were reflected in the outermost structural layer of social network.
That is, the effect of participatory civic culture in the United States seems to be reflected
in the significant difference of the mean number of memberships in community
organizations from those of the two East Asian societies, which made the United States
“a nation of voluntary associations that engage citizens in problem-solving and
governance” (Boris 1999:2). Further, the socialist deterrence against freedom of
association may have resulted in the smallest mean number of memberships in
community organizations, even significantly less than that of Taiwan although the latter had also experienced authoritarian rule. Thus the peoples in the two East Asian countries were contextually constrained to form less of a belonging structural layer in their social networks due to the lack of democratic tradition.
Socioeconomic Status
The covariates of socioeconomic status are included in the analyses in order to,
first, control their effects in the job search models, and second, identify the institutional
22 constraints reflected in different degrees of the effects of socioeconomic status in the
three societies.
Regarding the origin status, I selected three terms, say, father’s education,
mother’s education, and occupational class of father’s job when a respondent was sixteen
years old. As shown in the table below, average parental educational level of the
respondents in the United States was higher than that of the Taiwanese and Chinese
counterparts. The occupational status of father’s job in China was the lowest compared
with the other two countries.
(Table 14 about here)
With regard to the socioeconomic status of the respondents, I also chose educational and occupational achievements. Both categorical and continuous versions of
education measurements confirmed that U.S. respondents were more highly educated
than their Taiwanese and Chinese counterparts. Note that education with eight response
categories was employed in the multivariate analyses. Occupational class of first and
previous jobs consistently showed that Chinese respondents had the lowest means. I note
that both variables had “0” category for those who had not had first or previous jobs. It is
likely that most Chinese respondents tend to stay at the same jobs without progressing
through multiple careers. Lastly, it should be noted that the U.S. survey employed a
quota sampling to overcome underrepresentation of racial minorities in the sample.
However, there was no significant impact derived from the quota sampling in the
multivariate regression models. Having explicated the surveys and variables of interest
and their descriptive features, I proceed to analyze the composition and activation of
23 social capital and their impact on status attainment among the three societies with
structural equation modeling.
ACCESSED SOCIAL CAPITAL AMONG THE THREE SOCIETIES
Based on the three observed measures of accessed social capital delineated above,
it is possible to identify whether there exists intersocietal inequality among the United
States, Taiwan, and China using latent mean comparison model of the structural equation
modeling.
(Figure 2 about here)
As shown in the figure, accessed social capital is composed of extensity, upper
reachability, and range of prestige in the context of confirmatory factor analysis. For a
systematic comparison of the latent means of accessed social capital across the three societies, the three data sets were pooled together to be subjected to an empirical test of
latent mean comparison.
(Table 15 about here)
The confirmatory factor analysis generated similar factor loadings of the three
indicators of accessed social capital on the latent construct of accessed social capital
across the three societies, as shown in the table. The factor loadings indicate that the
three indicators worked as good empirical sources of the latent variable of accessed social
capital. Note that the factor loading of range of prestige was set at 1.00 as a reference
indicator in all three groups of comparison.
Then the means of latent accessed social capital were calculated based on the
factor loadings of the three indicators. Again, the latent mean of the first group (the U.S.)
was set at 0.00 to be a reference point for the other two groups. The latent mean of
24 accessed social capital in the United States was highest because those of the other two
societies took negative values. We could not be sure if the seeming differences in the
values of the latent means were statistically significant. Thus we proceeded to the next
step of the test of latent mean invariance to identify the latent group mean differences.
At the second step, the original SEM model worked as a baseline to be compared
against the subsequent models, with some constraints on the latent means of the accessed
social capital. For instance, the first model in the test of latent mean invariance forced the latent means of the three societies to be the same. The chi-squared value with the two
degrees of freedom generated by the invariance assumption was great enough to get
highly significant p-value, which indicates that the subsequent model was worse than the
baseline model. In other words, the assumption of total invariance in the latent means of
accessed social capital was rejected. Namely, there should be some significant
differences among the latent means. Likewise, the next model assumed that the United
States and Taiwan had the same latent mean of accessed social capital, which was also
rejected. It was then clear that the United States had a higher latent mean of accessed
social capital than Taiwan as was reported by the latent mean values. The third model
again confirmed that the latent mean of accessed social capital in the United States was
significantly greater than that in China. The last model shows quite different results
because the chi-squared value was drastically reduced by the imposed latent mean
invariance assumption between Taiwan and China; there was no significant gap between
the two societies. Therefore it was confirmed that the United States has significantly
greater accessed social capital compared to the two East Asian societies. However, the
25 volumes of accessed social capital in Taiwan and China cannot be differentiated from
each other.
SOCIAL CAPITAL AND STATUS ATTAINMENT
Having shown that there was unequal distribution of accessed social capital
among the three target societies, I turn to the multivariate analyses where a structural
equation model is employed with the common covariates across the three societies. The
final outcome variables of occupational class and annual income were combined into a
factor to tease out the commonality between occupational status and earnings; the latent
endogenous variable is called “status attainment.” In order to construct a simple and
universal model contact status was used as a representative indicator of activated social
capital. The structural layers of social relations were employed as the exogenous variable
in the model. I then employed a SEM multigroup path model on the three target societies.
The results are reported first, followed by the tests of parameter invariance that help
identify the differential status attainment mechanism among the three societies. Note that the multigroup path analysis results came from the reduced samples after excluding the
occupationally homogeneous cases.
Path Analysis Results
As seen in the figure, the results report that the outermost layer of belonging was
the main source of accessed social capital in the Unites States while the two other layers
were not significant. Taiwan was the only society where all three layers including
binding, the innermost layer, mattered significantly in increasing accessed social capital.
26 China showed a different pattern of association from Taiwan; the effect of binding was
not significant in China.
(Figure 3 about here)
At the next step, accessed social capital changed its role as an exogenous variable to predict activated social capital and status attainment. The structural relations between accessed and activated social capital are the first to be discussed. It is shown that accessed social capital had significant and positive relation with activated social capital in
Taiwan and China, but not in the United States. Part of the reason may be that American people use indirect ways of activating social contacts, such as routine job information in the labor market (McDonald and Elder 2006).
The effect of accessed social capital was significant and positive on status
attainment in all three societies. This tells us that even when the effect of activated social
capital is taken into account, accessed social capital has direct impact on the status
attainment outcome.
Lastly but most importantly is to see if activated social capital is a significant
predictor of status attainment. The effect was highly significant and positive, but its
magnitudes varied across the three societies; thus it is now a concern if the seeming gaps in the structural coefficients are statistically significant.
Before proceeding to the next step, I note that the SEM multigroup path analysis
results were fully saturated with control covariates imposed every step of the way to the endogenous variable of status attainment so that the structural coefficients reported in the
figure were conservative.
27 Test of Parameter Invariance
A question left from the previous step is whether the exogenous variables had
significantly different relations with the final outcome of status attainment across the
target societies. SEM has a specific test to answer such queries on structural differences
in the context of multigroup path analysis; it is called test of parameter invariance
because it assumes invariance of structural coefficients across groups (three societies in
the present study).
In particular, it is of great research interest if the effect of activated social capital
on status attainment statistically differs across the three societies. To answer this
question, the test of parameter invariance follows three steps. First, the SEM path model
presented above works as a baseline model. Second, several assumptions of parameter
invariance are imposed in turn on the alternative SEM models. Third, the chi-squared
model fit differences between the baseline and alternative models are tested to determine
if model fit stays the same or is worsened by the assumption. If the model fit does not
see any significant deterioration, then the assumption of parameter invariance is
confirmed; but if it shows a significant drop in model fit, the parameter invariance assumption is refuted, meaning that there is significant difference in the parameter between societies.
(Table 16 about here)
The second column of Table 16 shows the structural coefficients of activated
social capital from the three societies. The first test is called total invariance assumption
because it assumes that structural coefficients of activated social capital are the same
across the three societies. The chi-squared value with two degrees of freedom due to the
28 total invariance assumption generated a significant p-value, which indicates that the
assumption is wrong. Thus the structural coefficients cannot be the same across the
societies.
Next, another assumption of parameter partial invariance between the United
States and Taiwan was imposed in the alternative model. The model fit gap between the baseline and alternative models was not statistically significant, which confirmed that the effect of activated social capital can be assumed to be the same in the United States and
Taiwan. But the following test of parameter partial invariance between the United States and China confirmed that the structural coefficient of activated social capital cannot be assumed to be the same between these two societies; in other words, the United States had significantly greater effect of activated social capital on status attainment than China, as reflected in the magnitudes of the coefficients. The last test of parameter partial invariance verified in the same way that such an assumption was rejected between
Taiwan and China, verifying that the effect of activated social capital on status attainment was significantly greater in Taiwan than in China. Therefore the last column of the table reports that the United States and Taiwan had significantly greater impacts of activated social capital on status attainment than China. I suspect that a significant portion of the
Chinese deficit in activated social capital is due to the socialist constraint on use of contacts.
The next table shows the results of parameter invariance tests on all covariates having status attainment as the final outcome variable. In terms of activated social capital measures, the effects of presence of contact and routine job information were not differentiated across the three societies. But chain length of contacts had significantly
29 disadvantageous return on status attainment in the Unites States and Taiwan compared to
China, where it did not matter.
(Table 17 about here)
The effect of accessed social capital was not statistically differentiated across the
three societies according to the test results of the parameter invariance. Thus I conclude
that the Chinese socialist regime could not constrain the effect of accessed social capital
even though its tight institutional control was successful in reducing the efficacy of
activated social capital compared to the other two societies.
The parameter invariance tests on structural layers of social relations also report
interesting difference in binding layer among the societies. It is confirmed that Taiwan is the only society where binding, the innermost layer of social relations with and through spouses, worked significantly for status attainment. The family-oriented feature of
Confucian social relations may be the reason for its significance. It appears that the
Chinese socialist ideology of abolishing premodern Confucian social order was successful in reducing the power of the binding layer for status attainment. There was no significant gap found in the other two layers among the three societies.
In terms of demographic features, age showed no significant structural differences
among the three societies. However, being male revealed another noticeable difference;
being male was significantly more advantageous in Taiwan than in the United States and
China, even though males in all three societies had significantly more return on their
status attainment than the opposite gender. I suspect that much of the male advantage in
status attainment in Taiwan was due to the traditional Confucian constraint of gender inequality that formed the male-dominant labor market structure. In regard to the smaller
30 efficacy of being male in China, another Confucian society, I infer that its socialist
system has produced a negative interaction against the Confucian constraint of gender
inequality. Again, note that the socialist government in China tried from its beginning to
subvert the feudalistic Confucian social order through socialist modernization by which
redefinition of the status and role of females was made to mobilize women in production
(Croll 1983). Nonetheless, it is noteworthy that being female was still significantly
worse than being male for obtaining better returns from the Chinese labor market (Shu and Bian 2003).
Among the socioeconomic covariates, previous job and first job showed another
striking difference between American and East Asian societies. According to the test
results, first job status did not matter significantly in getting better status attainment in the
United States, but it was significant in the two East Asian societies. The differences in
the structural coefficients between the United States and the two East Asian societies
were significant on the tests of parameter invariance. However, previous job status
showed an opposite pattern: the impact was the greatest in the United States, so the
structural coefficient was significantly greater in the Unites States than in Taiwan. Thus I
conclude that in order to get better returns from labor markets, people in East Asian
societies should begin with high-status first jobs. However, what really mattered for
American people was whether they reached high-status previous jobs, regardless of
where they began their careers in the occupational hierarchy.
The other covariates such as education and origin status found no significant
differences among the three societies. In particular, it is notable that the effect of human
31 capital (education) is not differentiated across the three societies regardless of the types of institutional configuration in each society.
CONCLUSION AND DISCUSSION
With the main findings presented, I now turn to general discussion of the
comparative study of social capital and status attainment in the labor markets in the
United States, China, and Taiwan. The most notable conclusion of the present study is that institutions should be considered without preconceived ideas of individual actions, their interactions with macro structures in a society. There were variations across the three societies in terms of the effects of activated social capital on status attainment.
Activated social capital is an individual choice to maximize returns from the labor market through the influence and/or resources of social contacts. In some societies, activated social capital as an individual choice has been legitimately included in the institutional rules, norms, and rituals in labor markets. In some sense, the capacity to activate social resources is itself an asset of a job seeker that signals his/her ability and credentials endorsed by the contact persons. It thus follows that companies and organizations want new employees or members with better connections to diverse and resourceful social locations outside the organizational boundaries.
However, in other societies activating social capital goes against institutional rules and norms. Obtaining and maintaining access to diverse and rich position holders
are in most cases valued in a general social context. Nonetheless, utilization of such
resources can be illegal and prohibited in some social contexts. Not only in China as stated in the present research, but also in South Korea in the 1980s was activation of
32 social capital illegal, when the military government was in control of the country. It was common in any governmental office that beside the nameplate on the desk there was another kind of plate, inscribed, “Do not Ask for Favors through Connections.” This was because the military government believed that particularistic connections such as school ties, regional ties, and blood ties were rampantly utilized as channels of corruptive transactions between governmental officials and private parties including business owners. Despite governmental control, there occurred several large financial and political scandals involving such ties. Nonetheless, the tight control reduced the use of
contacts significantly for fear of imprisonment, in the worst cases.
Moreover, the Chinese socialist government had much tighter control on its people with its distributive system of jobs, housing, medical care, and schooling. In particular, considering that jobs were assigned by the state, it is expected that we observe less frequency and weaker power of activated social capital in the Chinese job market due to the institutional constraint against utilizing contacts than in other societies. In principle, therefore, there should be variation in the effect of activated social capital among societies in light of institutional configurations for accepting or rejecting certain type of individual choices. That mentioned, I argue that sociological research in labor market and social capital has not incorporated well the interaction between institutions and individual choices, having assumed an impractically universal social structure.
Nevertheless, the study also identified a counterreaction by individual social
capital against the institutional constraints reflected in the significant efficacy of accessed
social capital in all three societies. Specifically, accessed social capital in China did not
lose its significant power on status attainment even when activated social capital had
33 smallest impact among the three societies. I argue that it manifested the resilience of
social capital in the sense that the latent effect of accessed social capital on instrumental gains was not muted by socialist control. It is thus certain that access to more diverse and richer contacts significantly helped increase the chance of getting better returns from the labor market, regardless of whether there was tight control from outside the labor market on activated social capital.
In spite of some new findings in this study, it has limitations mainly due to the complexity of the macrocomparative research design. Given that the efficacy of social capital was examined in the labor markets of the three societies, analyses on the labor market structure and government policies in each society could increase the explanatory power of the study. For instance, macro-level indices such as female labor participation rates in the three societies could have explained more variations of the gender inequality embedded in the labor market structures. It was reported that 46 percent of the U.S. labor force, 45 percent of the Chinese labor force, and 41 percent of the Taiwanese labor force were taken by female workers in 2004 (the United States and China:
http://devdata.worldbank.org/genderstats/genderRpt.asp?rpt=profile&cty=EAP,East%20
Asia%20Pacific&hm=home, Taiwan:
http://eng.dgbas.gov.tw/public/data/dgbas03/bs2/yearbook_eng/y021.pdf). It may be that
the lower participation rate of the Taiwanese females indicates their comparative
disadvantage in the status attainment process found in the present study compared to the
females in the other two societies. More importantly, size differentials of each
occupation in the position generators by the three societies could work as controls in
34 teasing out the effect of accessed social capital. However, such specific information or
data were not available.
Next, in order to avoid using too complicated models across the three societies, I
did not include interaction variables among the variables of interest. Rather, I aimed to
examine structural differences of key covariates through their main effects. For instance,
gender inequality in activated social capital in the status attainment process was not
directly checked through an interaction term between gender and activated social capital
– I note, though, that the interaction term was not significant according to supplementary
analysis. Nonetheless the main effects of some key variables worked well in reporting
major differences among the societies.
I conducted another supplementary analysis of status attainment model on a
subsample of respondents from Beijing and Shanghai and found that activated social
capital in such urbanized cities had the same pattern of association with status attainment
measures (occupational class and annual income) identified in the whole sample. Thus
regional differences in China in terms of economic development and urbanization did not
affect the findings in the whole sample. It should also be checked in the future research using hierarchical modeling whether regional or contextual variables have any impact on
social capital formation and its efficacy on the job search process.
35 REFERENCES
Aguilera, Michael B. and Douglas S. Massey. 2003. “Social Capital and the Wages of Mexican Migrants: New Hypotheses and Tests.” Social Forces 82:671-702.
Angelusz, R. and R. Tardos. 2001. “Change and Stability in Social Network Resources: the Case of Hungary under Transformation.” Pp. 297-323 in Social Capital: Theory and Research, edited by Nan Lin, Karen Cook, and Ronald S. Burt. New York: Aldine De Gruyter.
Arrow, Kenneth J. and Ron Borzekowski. 2004. “Limited Network Connections and the Distribution of Wages.” Finance and Economics Discussion Series 2004-41, Board of Governors of the Federal Reserve System (United States).
Bian, Yanjie and Soon Ang. 1997. “Guanxi Networks and Job Mobility in China and Singapore.” Social Forces 75:981-1005.
Blau, Peter M., Danching Ruan, and Monika Ardelt. 1991. “Interpersonal Choice and Networks in China.” Social Forces 69:1037-1062.
Blau, Peter and O.D. Duncan. 1967. The American Occupational Structure. NY: Wiley.
Boris, Elizabeth T. 1999. “The Nonprofit Sector in the 1990s.” pp.1-34 in Philanthropy and the Nonprofit Sector in a Changing America, ed. Charles Clotfelter and Thomas Ehrlich. Bloomington: Indiana University Press.
Boxman, Ed A. W., Paul M. De Graaf, and Hendrik D. Flap. 1991. “The Impact of Social and Human Capital on the Income Attainment of Dutch Managers.” Social Networks 13:51-73.
Bridges, William P. and Wayne J. Villemez. 1986. “Informal Hiring and Income in the Labor Market.” American Sociological Review 51:574-82.
Campbell, Karen E., Peter V. Marsden, and Jeanne S. Hurlbert. 1986. “Social Resources and Socioeconomic Status.” Social Networks 8:97-117.
Croll, Elisabeth. 1983. Chinese Women since Mao. New York: M.E. Sharpe.
De Graaf, Nan Dirk and Hendrik Derk Flap. 1988. “With a Little Help from My Friends.” Social Forces 67:452-72.
Directorate General of Budget, Accounting and Statistics, Republic of China (Taiwan). 2008. “Table 21. Important Labor Force Status.” Retrieved April 2, 2008 (http://eng.dgbas.gov.tw/public/data/dgbas03/bs2/yearbook_eng/y021.pdf).
36 Fernandez, Juan A. and Laurie Underwood. 2006. China CEO: Voices of Experience from 20 International Business Leaders. Singapore: Wiley.
Lin, Nan. 2001. Social Capital: A Theory of Social Structure and Action. 2001. Cambridge University Press.
Lin, Nan, Paul Dayton, and Peter Greenwald. 1978. “Analyzing the Instrumental Use of Relations in the Context of Social Structure.” Sociological Methods and Research 7:149-66.
Lin, Nan, Walter M. Ensel, and John C. Vaughn. 1981. “Social Resources and Strength of Ties: Structural Factors in Occupational Status Attainment.” American Sociological Review 46:393-405.
Lin, Nan, Xiaolan Ye, and Walter M. Ensel. 1999. “Social Support and Depressed Mood: A Structural Analysis.” Journal of Health and Social Behavior 40:344-359.
Marsden, Peter V. 1987. “Core Discussion Networks of Americans.” American Sociological Review 52:122-131.
Marsden, Peter V. and Jeanne S. Hurlbert. 1988. “Social Resources and Mobility Outcomes: A Replication and Extension.” Social Forces 66:1038-1059.
McDonald, Steve. 2005. “Patterns of Informal Job Matching across the Work Career: Entry-Level, Reentry-Level, and Elite Non-Searching.” Sociological Inquiry 75:403-428.
McDonald, Steve and Glen H. Elder, Jr. 2006. “When Does Social Capital Matter? Non- Searching for Jobs across the Life Course.” Social Forces 85:521-550.
McPherson, J. Miller, Lynn Smith-Lovin, and Matthew E. Brashears. 2006. “Social Isolation in America: Changes in Core Discussion Networks over two decades.” American Sociological Review 71:353-75.
Mouw, Ted. 2003. “Social Capital and Finding a Job: Do Contacts Matter?” American Sociological Review 68:868-898.
______. 2006. “Estimating the Causal Effect of Social Capital: A Review of Recent Research.” Annual Review of Sociology 32:79-102.
Sewell, William H., A. O. Haller and A. Portes. 1969. “The educational and Early Occupational Attainment Process.” American Sociological Review 34:82-92.
Sewell, William H. and R. M. Hauser. 1975. Education, Occupation and Earnings. New York: Academic Press.
37 Shu, Xiaoling and Yanjie Bian. 2003. “Market Transition and Gender Gap in Earnings in Urban China.” Social Forces 81:1107-45.
Smith, Sandra S. 2005. ““Don’t Put My Name on It”: Social Capital Activation and Job- Finding Assistance among the Black Urban Poor.” American Journal of Sociology 111:1-57.
Son, Joonmo, Nan Lin, and Linda K. George. 2008. “Cross-National Comparison of Social Support Structures between Taiwan and the United States.” Journal of Health and Social Behavior 49:104-117.
Sprengers, Maarteen, Fritz Tazelaar, and Hendrik Derk Flap. 1988. “Social Resources, Situational Constraints, and Reemployment.” Netherlands Journal of Sociology 24:98-116.
The World Bank Group, GenderStats. 2008. “Summary Gender Profile.” Retrieved April 2, 2008 (http://devdata.worldbank.org/genderstats/genderRpt.asp?rpt=profile&cty=EAP,E ast%20Asia%20Pacific&hm=home).
Van der Gaag, Tom A.B. Snijders, and Henk D. Flap. 2007. “Position Generator Measures and Their Relationship to Other Social Capital Measures.” in Social Capital: Advances in Research, edited by Nan Lin and Bonnie Erickson. Oxford: Oxford University Press.
Völker, Beate and Henk D. Flap. 1995. “The Effects of Institutional Transformation on Personal Networks: East Germany, Four Years Later.” The Netherlands Journal of Social Sciences 31:87-110.
Wegener, Bernd. 1991. “Job Mobility and Social Ties: Social Resources, Prior Job and Status Attainment.” American Sociological Review 56:1-12.
Wong, Yiu-chung. 2005. From Deng Xiaoping to Jiang Zemin: Two Decades of Political Reform in the People’s Republic of China. Lanham, MD: University Press of America.
38 Table 1: Comparison of Extensity The U.S. Taiwan China (mean (mean (mean (S.D.)) (S.D.)) (S.D.)) Extensity 7.14 6.64 6.30 (range 0-21) (4.59) (5.04) (4.32) N 3,000 3,278 3,435 t-test The U.S. > Taiwan*** The U.S. > China*** Taiwan > China** * p < .05; ** p < .01; *** p< .001 (two-tailed test).
Table 2: Comparison of Upper Reachability The U.S. Taiwan China (mean (mean (mean (S.D.)) (S.D.)) (S.D.)) Upper reachability 69.03 62.18 64.09 (22-78 SIOPS score) (11.25) (12.83) (10.66) N 2,830 3,084 3,390 t-test The U.S. > Taiwan *** The U.S. > China *** Taiwan < China *** * p < .05; ** p < .01; *** p< .001 (two-tailed test).
Table 3: Comparison of Range of Prestige The U.S. Taiwan China (mean (mean (mean (S.D.)) (S.D.)) (S.D.)) Range of prestige 38.57 32.47 31.69 (0-56 SIOPS score) (15.97) (17.35) (16.19) N 2,830 3,084 3,390 t-test The U.S. > Taiwan *** The U.S. > China *** Taiwan > China * * p < .05; ** p < .01; *** p< .001 (two-tailed test).
39 Table 4: Factor Analysis of Accessed Social Capital when Starting Current/Last Position Variable The U.S. Taiwan China (N=2,830) (N=3,084) (N=3,390) Factor eigenvalue I 2.43 2.57 2.48 II .41 .33 .37 III .15 .11 .15 Factor loading on Factor I a Extensity .86 .89 .89 Upper reachability .89 .92 .89 Range .95 .96 .95 Factor scoring on Factor I a Extensity .35 .35 .36 Upper reachability .36 .36 .36 Range .39 .38 .38 a Principal component, minimal eigenvalue of 1, and coefficients based on varimax rotation.
Table 5: Use of Contacts in the Three Societies The U.S. Taiwan China Freq. Percent Freq. Percent Freq. Percent No contact 1,389 46.4 1,832 56.0 2,133 60.7 Contact(s) used 1,607 53.6 1,439 44.0 1,381 39.3 N 2,996 100.0 3,271 100.0 3,514 100.0
Table 6: Length of Chain in the Three Societies Length of Chain The U.S. Taiwan China Freq. Percent Freq. Percent Freq. Percent 1 882 76.9 1,054 80.8 703 63.8 2 258 22.5 218 16.7 295 26.8 3 7 0.6 30 2.3 67 6.9 4 - - 2 0.2 37 3.5 Total 1,147 100.0 1,304 100.0 1,102 100.0 Mean (S.D.) 1.24 (.44) 1.22 (.48) 1.49 (.76) t-test The U.S. = Taiwan The U.S. < China*** Taiwan < China*** * p < .05; ** p < .01; *** p< .001 (two-tailed test).
40 Table 7: Proportion of Direct Contacts and Family Ties in the Three Societies The U.S. Taiwan China Freq. Percent Freq. Percent Freq. Percent Direct contacts 1452 90.1 1314 92.3 850 81.5 Indirect contacts 159 9.9 109 7.7 193 18.5 Percent of family ties 21.1 39.2 57.7 among direct contacts (.41) (.49) (.49) (S.D.) t-test (proportion of family ties) The U.S. < Taiwan*** The U.S. < China*** Taiwan < China*** * p < .05; ** p < .01; *** p< .001 (two-tailed test).
Table 8: Occupational Class of Contacts Occupational Class The U.S. Taiwan China 1. Lower class (%) 226 238 525 (20.2) (18.3) (47.6) 2. Middle class (%) 220 614 116 (19.6) (47.1) (10.5) 3. Professional (%) 310 110 101 (27.7) (8.4) (9.2) 4. Executive (%) 365 341 360 (32.6) (26.2) (32.7) N 1,121 1,303 1,102 Mean (S.D.) 2.73 2.43 2.27 (1.11) (1.06) (1.34) t-test The U.S. > Taiwan*** The U.S. > China*** Taiwan > China*** * p < .05; ** p < .01; *** p< .001 (two-tailed test).
41 Table 9: Routine Job Information in the Three Societies The U.S. Taiwan China Freq. Percent Freq. Percent Freq. Percent 0. no information 1,661 59.3 1,753 63.1 2,455 78.0 1. got information 1,142 40.7 1,025 36.9 694 22.0 N 2,803 100.0 2,778 100.0 3,471 100.0 t-test The U.S. > Taiwan** The U.S. > China*** Taiwan > China*** * p < .05; ** p < .01; *** p< .001 (two-tailed test).
Table 10: Occupational Classes in the United States, Taiwan, and China Occupational Class The U.S. Taiwan China 1. Lower class (%) 948 872 2355 (31.7) (26.6) (68.7) 2. Middle class (%) 712 1,900 321 (23.8) (58.1) (9.4) 3. Professional (%) 979 287 524 (32.7) (8.8) (15.3) 4. Executive (%) 352 213 226 (11.8) (6.5) (6.6) N 2,991 3,272 3,426 Mean (S.D.) 2.25 1.95 1.59 (1.03) (.78) (.97) t-test The U.S. > Taiwan*** The U.S. > China*** Taiwan > China*** * p < .05; ** p < .01; *** p< .001 (two-tailed test).
Table 11: Annual Income in the United States, Taiwan, and China The U.S. Taiwan China Range (mean (mean (mean (S.D.)) (S.D.)) (S.D.)) Mean Annual 16.04 5.99 6.68 U.S., China: 1- Income (4.50) (2.71) (4.29) 27 Taiwan: 1-20 *Amount in U.S. $25,200 $10,795 $1,474 - dollar * Mean income amounts were calculated converting the specific point of the mean between its two adjacent response categories into dollar amount. The mean annual incomes in New Taiwan Dollar and Chinese Renminbi were changed in U.S. dollar using the currency exchange rates in 2004.
42 Table 12: Descriptive Statistics of Explanatory Variables: Demographic Features The U.S. Taiwan China Range/categories (mean (mean (mean of variable (S.D.)) (S.D.)) (S.D.)) Demographic Variables Age 41.48 41.02 38.37 21-64 (10.57) (11.66) (10.30)
Age 33.38 31.70 26.96 - when starting the (9.87) (9.63) (8.73) current/last position
Male .46 .52 .49 - (.50) (.50) (.50)
White .69 - - 0: nonwhite (.46) 1: white
Black .12 - - 0: nonblack (.32) 1: black
Latino .14 - - 0: non-Latino (.34) 1: Latino
Other .05 - - 0: W+B+L (.21) 1: other
N 3,000 3,278 3,514 -
43 Table 13: Descriptive Statistics of Explanatory Variables: Structural Layers The U.S. Taiwan China Range/categories (mean (mean (mean of variable (S.D.)) (S.D.)) (S.D.)) Structural layers of social network Binding 1: 2.52 2.35 2.70 0: Not married Spouse network at (2.14) (1.87) (1.59) 1: Almost none the time of 2: A few interview 3: About half 4: Most 5: Almost all Binding 2: 2.04 1.93 1.53 0: Not married Spouse network (2.15) (1.93) (1.79) 1: Almost none when starting the 2: A few current/last 3: About half position 4: Most 5: Almost all Bonding: 3.42 3.40 3.17 1: 0-4 Daily contact (1.50) (1.37) (1.18) 2: 5-9 (number of people 3: 10-19 contacted daily) 4: 20-49 5: 50-99 6: More than 100 Belonging 1: 2.01 .63 .16 U.S.: 0-10 Memberships in (1.87) (.84) (.55) Taiwan: 0-5 community China: 0-11 organizations at the time of interview
Belonging 2: 1.29 .37 .06 U.S.: 0-10 Memberships in (1.53) (.66) (.33) Taiwan: 0-5 community China: 0-11 organizations when starting the current/last position
N 3,000 3,278 3,514 -
44 Table 14: Descriptive Statistics of Explanatory Variables: Socioeconomic Status The U.S. Taiwan China Range/categories (mean (mean (mean of variable (S.D.)) (S.D.)) (S.D.)) Origin status Father’s education 4.95 3.34 3.71 1-8 (1.90) (1.50) (1.51)
Mother’s 4.97 2.55 2.88 1-8 education (1.61) (1.40) (1.53)
Occupational class 1.76 1.91 1.49 1: low of father’s job (1.02) (.76) (.95) 2: middle (using 1980 3: professional census occupation 4: executive code) Socioeconomic status Education 6.05 5.11 5.29 1-8 (Categorical) (1.28) (1.52) (1.22)
Years of education 14.84 12.25 12.05 U.S.: 0-35 (3.24) (3.71) (3.08) Taiwan: 0-34 China: 0-25
Occupational class .83 .83 .17 0: no first job of first job (1.11) (1.08) (.58) 1: low (using 2000 2: middle census occupation 3: professional code) 4: executive
Occupational class 1.53 1.42 .58 0: no prev. job of previous job (1.29) (1.03) (.97) 1: low (using 2000 2: middle census occupation 3: professional code) 4: executive
Quota sampling .44 - - 0: Non-quota (.50) 1: Quota
N 3,000 3,278 3,514 -
45 Table 15: Latent Mean Comparison of Accessed Social Capital The Unites States Taiwan China Factor Loading (Std.) Extensity .75 .78 .79 Upper reachability .80 .87 .80 Range of prestige 1.00 1.00 1.00
Latent Mean of .00 -.40 -.47 Accessed Social Capital
Test of Latent Mean Invariance US=Taiwan=China X 2 (df=2) 393.16 p-value .000*** US=Taiwan X 2 (df=1) 245.09 p-value .000*** US=China X 2 (df=1) 347.73 p-value .000*** Taiwan=China X 2 (df=1) 3.44 p-value .064
Model Fit Index CFI .98 TLI .97 SRMR .04 N 2,830 3,084 3,390 * p < .05; ** p < .01; *** p< .001 (two-tailed test).
46 Table 16: SEM Test of Parameter Invariance on Activated Social Capital The U.S. Taiwan China Activated Social Capital .22*** .20*** .09** (Contact status) Total invariance (US=TW=CH) X 2 (df=2) 11.067 p-value 0.004** Partial invariance (US=TW) X 2 (df=1) .043 p-value .836 Partial invariance (US=CH) X 2 (df=1) 7.230 p-value 0.007** Partial invariance (TW=CH) X 2 (df=1) 8.470 p-value 0.004*
Intersocietal comparison US>China Taiwan>China *Note: There appears slight difference in the significance level of accessed social capital in China from that reported in Figure 9 because the present model excluded prior steps in the path model. * p < .05; ** p < .01; *** p< .001 (two-tailed test).
47 Table 17: SEM Multigroup Analysis and Test of Parameter Invariance The U.S. Taiwan China Test of Parameter Invariance Activated Social Capital Contact status .22*** .20*** .09** US>China Taiwan>China Presence of contact -.02 -.07 -.08 US=Taiwan=China Chain length -.17** -.12** -.01 China>US China>Taiwan Routine job .01 -.02 -.01 US=Taiwan=China information Accessed Social Capital .07*** .08*** .07*** US=Taiwan=China Structural Layers Binding .03 .08*** .02 Taiwan>US Taiwan>China Bonding .03 .06*** .07*** US=Taiwan=China Belonging .07** .05** .01 US=Taiwan=China Age .01 .10 .14 US=Taiwan=China Age² -.08 -.08 -.10 US=Taiwan=China Male .07*** .16*** .08*** Taiwan>US Taiwan>China Education .47*** .54*** .46*** US=Taiwan=China Previous job .18*** .07*** .10*** US>Taiwan First job .02 .12*** .06** Taiwan>US China>US Father’s job .04 .05**** .02 US=Taiwan=China Father’s education .04 .03 .02 US=Taiwan=China Mother’s education -.04 -.07*** -.01 US=Taiwan=China
R-squared .42 .50 .32 - Model Fit Index CFI 1.00 - TLI 1.00 - RMSEA 0.00 - N 1,810 2,545 2,613 - *Note: There appear slight differences in the magnitudes of accessed social capital and in the significance level of accessed social capital in China from those reported in Figure 9 because the present model excluded prior steps in the path model. * p < .05; ** p < .01; *** p< .001 (two-tailed test).
48 Macro Micro
The U.S. Capitalist Type Capitalist of Social Taiwan Social Institution Network
State China Socialist Socialist Type of Social Institution Network
Macro Micro
Western Type Western of The U.S. Cultural Social Institution Network
China Confucian Taiwan East Asian Type of Cultural Institution Social Taiwan Network
Figure 1: Institutional Constraints on Social Capital of Individuals
49
Accessed Social Capital
Extensity Upper Range of Prestige Reachability
Figure 2: Confirmatory Factor Analysis on Accessed Social Capital
The United States
Binding .03 Activated SC -.01 .22***
Bonding .03 Accessed SC Status Attainment .06*** Belonging .23***
Taiwan
Binding Activated SC .04* .07*** .20***
Bonding .13*** Accessed SC Status Attainment .08** Belonging .11*** *
China
Binding Activated SC -.03 .08*** .09***
Bonding .09*** Accessed SC Status Attainment .07*** Belonging .05** Model Fit Index RMSEA: 0.000 * p < .05; ** p < .01; *** p< .001 (two-tailed test). SRMR: 0.081
Figure 3: SEM Multigroup Path Analysis
50 Reproduction of Inequality in the Netherlands through the Creation of and Returns to Social Capital?
Beate Völker & Henk Flap Utrecht University Dept of Sociology/ICS The Netherlands
Paper to be presented at the International Social Capital Conference Academia Sinica, Taipei, Taiwan May 29-30, 2008
(draft)
Abstract Social capital theory predicts a reproduction of inequality through accumulation and use of social capital. This is in line with previous research (e.g., Flap 1991). We follow Lin, Ao and Song (2008) and try to integrate the study of the creation and returns of social capital in one study, while inquiring into this issue in The Netherlands, using the position generator as a measure of social capital. We study the effect of resources such as one’s human capital (education and work experience) and father’s occupational prestige as well as factors that increase meeting opportunities like participation in voluntary organizations and urbanization. All these resources might be conditions for an unequal distribution of social capital in the population. Furthermore, we investigate the consequences of having different amounts of social capital on socio-economic outcomes like supervisory responsibility and income. Yet, social capital theory implies that people with more social capital are not only better off regarding earnings, but they also have a higher wellbeing in general. Therefore, we considered also satisfaction with earnings and subjective health as a return of social capital. Finally, we argue that social capital affects also attitudes that are helpful in a more heterogeneous society like having an open mind. We employ data from two waves of a national representative panel study of the networks of the Dutch (2000/2007). Our results show that human capital and prestige of the father engender social capital in the form of having access to many jobs as well as to higher prestige jobs. Living in an urban area, participating in voluntary organizations as well as having a working partner helps to acquire these forms of social capital, while in addition it also promotes social capital in the form of the range of occupations one can access. Interestingly, having a working partner works especially for men. Our empirical results show that general social capital enhances the chances of getting a supervisory position as well as earning a higher income. Unexpectedly, it affects satisfaction with earnings and subjectively reported health negatively. Last but not least, social capital, i.e. range and upper reach, do quite clearly stimulate the open-mindedness of a person.
Word count: 9.500
Keywords: creation and returns of social capital, position generator
1 1 Introduction
The growing research on social capital in the last two decennia shows that the social capital research program is flourishing and that social capital directly affects people’s life chances. Most studies focus on some types of returns of social capital, the benefits one has through the social relationships one can access and use. The value of these relationships has been demonstrated for various domains of people’s lives, amongst others for getting a job (De Graaf and Flap 1988), getting a house (Dimaggio and Louch 1998), staying healthy and having literally a longer life (Berkman and Syme 1979). Only few studies are explicitly dedicated to the creation of social capital and the resulting distribution of social capital (for an example, see Halman & Luijkx 2006: 67; see also Coleman 1990/1993). Apart from Lin, Ao and Song (2008) there are to our best knowledge no studies which aim at both, the creation as well as the returns of social capital (see Flap & Volker, 2004 for examples of separate studies on both issues). Yet, the beauty of the program as well as the major advantage of the social capital theory is that it allows non-trivial statements about the conditions under which social networks are created as well as predictions on the returns of these social networks. One reason for that gap in the existing research probably lies in the broadness of the idea – social capital is assumed and shown to have encompassing consequences for one’s life. Furthermore, another obstacle for bringing creation and returns together can be attributed to differences in measurements. Although considerable progress has been made in measuring social capital (see e.g. Van der Gaag 2004), still, many studies are rather difficult to compare because of differences in sampling, measurements as well as methods of analysis. But the major reason probably is the lack of longitudinal data sets that allow analyzing time-sequences. The purpose of this contribution is to study creation and returns of social capital simultaneously, in the same group of people and with the same measurements, using two different points of measurements. Because we want to be able to compare our results on The Netherlands with those made in other countries we follow a similar set-up as applied in Lin et al.’s study of urban citizens in China (2008).
2 Theoretical frame: social capital theory The literature on social capital has grown enormously and is still growing. This is not only due to the usefulness of that theory, but also to the broadness and vagueness of the concept of social capital. The term social capital is currently used for many phenomena even for those not exactly covered by the theory of social capital. Social capital research started by building upon the image of an individual actor, who has social capital consisting of his ties to others, the readiness of these others to help him or her, and their capability to do so, given their own resources. The main idea is that those with better social capital will be better able to realize their goals. In the latter years this idea of social networks as social capital has been overtaken by a communitarian approach on social capital (Moore et al. 2005). Impressed by Robert Putnam’s books Making democracy work (1993) and Bowling alone (2000) scholars now see social capital as a feature of larger entities, of corporate actors, like states, regions, municipalities and the like. Putnam indicates social capital by networks, general trust, norms of reciprocity and especially by memberships of voluntary organizations. In his 1993-book he makes a case for the economic an social welfare of the North of Italy and the misery of the South of Italy as flowing from the presence, respectively the absence of a regional tradition of organizational involvement. Networks at the collective level, especially so-called
2 bridging ties that interconnect different communities, promote general trust and cooperation.1 Using these lateral ties to each other, citizens force local politicians and bureaucrats to practise good government. Both perspectives developed relatively separately from each other and in different fields of the social sciences, e.g. sociology and political science. While the focus of the micro level perspective on social capital is on individual action and behavior, the focus of macro level social capital is on collective action or collective good consumption. This latter view entails that people benefit not only from accessing or using relationships - ties are not always necessary to enjoy benefits - but there are collective goods available in certain groups of people which can be accessed merely via membership to that group. For example, one can benefit from the attention neighbors pay when they watch the streets and houses in a neighborhood, without having direct ties to these neighbors. The two approaches, individual and collective level social capital, do not contradict each other, though. Rather, they are complementary. One might even apply a network interpretation of social capital at the collective level. Yet, it should be noted that the view on social capital at the macro level differs from the micro perspective in the sense that individual investments are not always necessary and returns have collective good characteristics. In our contribution, we focus on social capital at the micro level. We deliberately neglect macro aspects, because we assume that causes and consequences for this type of social capital might be different. In other words, the study of differences in the distribution of collective social capital, e.g. between neighborhoods, requires different arguments and measurements which would make the scope of this paper much too broad. Therefore, we stick to the micro perspective of social capital. In that frame, the hard core of the program is rather straightforward and consists of two statements. First, those with more and better social capital are better able to realize their goals. Second, people will invest in ties to the degree that these ties are instrumental in achieving people’s goals. In this perspective, social capital is consisting of ties to others. While the first statement of the hard core gives clues on the consequences of having and using social capital, second statement of the hard core helps to find hypotheses on the conditions under which people create social capital. In particular, this leads to the expectation that ties to people with many resources, that are ties to people who are higher on the social ladder will probably seen as most instrumental and therefore aimed to be included in a persons’ network. People who are already in the highest social strata will probably form ties to others similar to themselves, since there are no others available who provide higher access. In other words, the idea that people establish ties to others who are expected to be instrumental in the future leads to the expectation of social closure among different strata. While this reasoning assumes the importance of preferences for certain others, another type of argument can be made when acknowledging that there is also a supply side to social contacts. Personal networks and hence a person’s social capital is also affected through the chances one has to meet certain others. The social composition of the contexts in which one works, recreates etc. therefore has a substantial influence on the resulting network (see Marsden 1990, McPherson and Smith-Lovin 1987; Mollenhorst, Völker and Flap 2008a and b). Social
1 Note that in his recent work Putnam (2007) argues that diversity, in particular ethnic diversity enhances distrust and what he calls ‘turtle’ behavior, i.e. refraining from all kinds of participation and lower trust to all others, also against ethnically similar others.
3 settings are often socially homogeneously composed. 2 So, both of these views, demand as well as supply side, lead to the expectation that social capital will consolidate rather than mollify social cleavages. Our general hypotheses is that those who are already in the higher social strata will create more social capital and have again more and better returns. Further, different forms of social capital will matter for different kinds of returns. For returns related to occupational attainment, height of access to social capital in terms of prestige and social status is more important than number of others one can access or the range of social strata that is covered through ones network. For satisfaction with ones financial situation it might be more important that one sees also others who earn less, hence range of social capital can be expected to have an effect here. The same might hold for subjective health. Finally, for having an open mind (see e.g. Rokeach 1960, Laumann 1973) in particular network range as an indicator of network diversity is expected to have an effect.
3 Data and measurements a) Data The data collection for this study started in 1999/2000, in the course of the larger research program ‘Creation and Returns to Social Capital’ which has been founded by the Dutch National Science Foundation. The dataset is called SSND1 (the Survey of the Social Networks of the Dutch). The data include information on 1007 individuals between the ages of 18 and 65, representing the Dutch population. In 2006/2007 a second wave has been conducted, with the same respondents. Of the 1007 respondents of the first wave, the addresses of 850 persons have been traced after 7-8 years and 604 (71%) participated a second time in the research. Since a number of our research projects are directed to relations and social capital among neighbors and within neighborhoods, the sample is actually a neighborhood sample. Initial sampling procedure was as follows: we randomly sampled forty municipalities representing the different Dutch provinces and regions and taking into account size differences between these municipalities. Within each municipality, we randomly selected four neighborhoods. For the neighborhood identification, the Dutch zip-code system was used.3 Finally, we randomly selected two samples of 12 and 13 addresses, respectively, in each neighborhood and attempted to interview about eight respondents per neighborhood. This procedure was applied to achieve an overrepresentation of the working population. In one sample only the working individuals were interviewed and in the other, everybody who was selected was asked for an interview. In this way, we obtained one sample representing the Dutch active labor force and one representing the Dutch population in general. In the final sample, 758 respondents were employed. The response rate for both samples together was forty percent. As stated earlier, the data consists of 1007 individual respondents in 161 neighborhoods. Higher educated persons and men are slightly over- represented in the whole sample, yet the sub-sample of the working population can be considered representative for the Netherlands. Van der Gaag (2004) constructed weights for the
2 There are of course large differences regarding the type of characteristic under consideration as well as the type of context (see Mollenhorst et al. 2008a). For example, going out places are more heterogeneously composed in terms of sex than work places, or family seettings are more heterogeneously composed with regard to age than educational settings. 3 The Dutch zipcode system consist of 6 digit code, referring to geographical; areas. An area of 6 digits consists of 20-30 addresses on average, an area of 5 digits consists of 200-300 addresses and an area of 4 digits comprises 2000-3000 different addresses on average. We chose 5 digit areas in order to sample ecologically meaningful units. Note that the areas of 6 and 5 digits delineate also the route of postmen.
4 overrepresented characteristics of the population sample, yet did not find any remarkable differences in analyses with and without the weighted sample. b) Measurements SOCIAL CAPITAL. Social capital was measured using the position-generator items (Lin and Dumin 1986; Lin, Fu, and Hsung 2001; Van der Gaag, Snijders, and Flap 2008). We asked each respondent, “Do you know anyone among your relatives, friends, or acquaintances who has one of the following positions? (‘Knowing’ means that you and the person can recognize each other and also greet each other, as well as that you know this person’s first name and that you could shortly talk to him or her. A list of 30 occupational titles then followed, ranging form the lower to the upper social strata, see the list in Table 1. Next the occupational titles were recoded into prestige scores. To estimate the prestige scores the occupational titles were coded according to the Standard International Codes for Occupational Prestige Scale (ISCO) constructed by Ganzeboom and Treiman (1996). Three indices were generated from the 30 occupations: (1) Extensity - the number of occupations a respondent could access; Lin et al. call this general social capital. (2) Upper reachability - the highest occupational prestige score among accessed occupations, and (3) Range - the difference between the highest and lowest accessed occupational prestige scores. The strength of the ties that provide access to various positions is measured by an additional question to the question mentioned just above. Interpretation: “Is this person that you know in this position family, a friend or an acquaintance?”
== TABLE 1 ABOUT HERE ===
Table 1 shows that people’s access differs considerably among certain types of occupations. While about three quarters of the respondents know a nurse, only 16 percent know a union leader. Further, there are interesting differences between the strength of a tie which provides access: Access via acquaintances is by far not always the most frequent one, although this is what one would expect, since acquaintances, which are weaker ties, enhance the range of contacts. Often, most positions are accessed via strong ties to family members. In this sample of the Dutch also the summary indices show that the Dutch have ample access via family ties: the range of positions is largest for the access through family and not through acquaintances. Also the upper reach is slightly higher for family than for acquaintances. On average, the Dutch access 15 of the 30 positions provided. Most of the positions are accessed via family or acquaintances, while friends are least often mentioned.
HUMAN CAPITAL. Human capital was indicated by the following measurements: (1) education, measured on a four-level scale (secondary or lower, high school, associate college, and college and graduate); and (2) having a paid job; and (3) having a white collar job. We also inquired into the length of time one works in the current occupation, yet we did not include this indicator in our final analyses since it has no effect.
MEMBERSHIP IN VOLUNTARY ASSOCIATIONS. We included different indicators for a respondent’s activity in voluntary organizations. First, we used the number of different organizations to which one is a member, second we also used a dichotomous coding for membership. Both indicators show actually no remarkable difference in the analysis, except that the inclusion of the number of
5 organizations usually showed a somewhat higher partial correlation coefficient. Besides the count and the dummy of membership we also included whether a person gives money to charity organizations, is a blood donor, or is doing voluntary work. Remarkably, in all analyses membership in general is a more important predictor than doing voluntary work.
OCCUPATIONAL ATTAINMENT. We considered the following attainment variables: 1) Being a supervisor: measured as a dichomous variable (0= no, 1=yes). 2) Income: measured as a scale consisting of 17 categories, we used the log of income in most analyses In addition to monthly income, we also wanted to know to what degree social capital contributes to more subjective outcomes, like satisfaction with life and wellbeing. Moreover, we expect that social capital affects whether one has a closed or an open mind. We therefore included three different measures, satisfaction with one’s financial situation, self-reported subjective health and a measure for open mindedness and dogmatism.
SATISFACTION WITH FINANCIAL SITUATION: is measured via the question: “Please tell me to which degree you are satisfied with your current financial situation?”; there were seven answer categories presented, ranging from very unsatisfied to very satisfied.
SUBJECTIVE HEALTH: is measured via the question: “All in all, how do you evaluate your current health?” Seven answer categories were presented ranging from very bad to very good.
OPEN MINDEDNESS/DOGMATISM: We included a number of items form the dogmatism scale developed by Rokeach (1960) .While this scale comprises about 60 items, we only used 6 items from this scale, which show a Cronbach’s alpha of .72. The items were as follows: • Most people do not know what is good for them • Of all existing ideologies in the world there is probably just one really true. • A group of people in which many different opinions are tolerated shall not exist for long • A person who tolerates many contradicting positions has no own opinion. • With people who think different on, e.g. ,religion one should not compromise too easily in case of conflicts, and • For most questions, there is only one answer. Answer categories ranged from 1 to 7; the higher, the less agreement with the statement.
CONTROL VARIABLES. The control variables included: (1) Age; in all analyses we inquired also into the possibility of a curvilinear association of age with the outcome variable, and (2) Gender, coded as 1 = male and 0 = female. (3) Furthermore, we controlled for the degree of urbanization in the area where the respondent lived. We did so because one can argue that the number and the density of addresses in one’s living environment determine access to others in general and hence to resources and social positions in particular. We used the codes provided by the Dutch Central Bureau of Statistics (CBS 2004). Urbanization: coded as 1= less than 100 addresses per squared km, 2 = between 1000 and 1500 addresses per squared km and 3= more than 1500 addresses per squared km.
6
Take note the dependent variables were taken from the second time of measurement in 2007/2008. The independent and the control variables were taken from the 1999/2000 dataset.
Table 2 provides an overview of descriptive statistics of the other variables used in the analyses.
=== TABLE 2 ABOUT HERE ==
4 Results We analyzed our data in two steps, first, we inquired into the question of who has created more social capital and second, we studied various returns on social capital. Section 4.1 presents the regression models on the creation or distribution of social capital and section 4.2 is on the results on the consequences of social capital.
4.1. Creation of social capital
Extensity We started with a model on general socio-demographic conditions, such as gender, age, ethnic origin and the prestige of father’s job, and the degree of urbanization. The first data column in table 3 shows that those being male, and having a father with a high prestige job mention more occupational positions, i.e. have more extensive social capital. Interestingly, with increasing age, social capital first increases but decreases later on.
== Table 3 about here ==
Furthermore, we found that people in cities have more social capital at their disposal, in particular compared to people in the suburban areas. In the second model we added indicators for human capital, i.e., education, having a paid job and working in a white collar occupation. Higher educated people mention more positions and it affects also social capital extensity if one has paid work and works in a white collar occupation. The third model shows that membership has also a strong effect on social capital. Finally in the last model, we inquired into the effect of having a partner who works. Interestingly, the interaction between working partner and gender indicates that men benefit more from a working partner than women. Figure 1 illustrates this combined effect.
== Figure 1 about here ==
Upper reach Table 4 applies the same regression models to the upper reach of social capital, the highest positions one can access through one’s social network. The table shows that upper reach increases with age – note that there is no curvilinear effect found in these models – and with father’s occupational prestige. Furthermore, people in cities have more social capital than those who live in areas with a lower population density. Human capital is also a strong condition for the upper reach of social capital: higher educated people have much more access to higher strata. Furthermore, again, membership in organizations has an important effect. In the last model, we
7 again inquired into the effect of having a working partner and whether this is conditioned by the partner's sex: again, we found that men benefit in particular when their partner has a job. Not that in none of the models a main effect of sex is found.
== table 4 about here ==
This interaction effect is illustrated in figure 2.
== Figure 2 about here ==
Range Table 5 shows the models on the last indicator for social capital, the range of positions a person can access. In general it can be stated that range of positions is most difficult to explain and that the applied models are rather weak. Father’s prestige and a person’s age have no stable effects and own education is not significant at all. Only the indicators for membership in voluntary organizations and urbanization show robust significant coefficients. The former condition, membership, leads to a higher range, while living in suburban areas is associated with a smaller range of access to social positions compared to those who live in cities. Furthermore, and similar to the other models, men benefit when their partner has a paid job, more than women do. Figure 3 illustrates this finding.
== Table 5 about here ==
== Figure 3 about here ==
As to the difference between men and women in terms of social capital benefits one has through a working partner, it should be noted that if the interaction term is not included in the models, there is a significant main effect of having a working partner, indicating that in general all persons with such a partner have better access to social capital. The main effect, however, disappears when the interaction term is included in the model.
4.2 Returns of social capital We next inquired into the returns of social capital. First, we studied effects of social capital on having a supervisory position. Note that there are at least seven years between the measurement of social capital and the predicted position. Table 6 summarizes the results of the logistic regression model. We do not show the models which include upper reach of accessed positions and range of social capital because we did not find a significant effect of these two indicators on income and the other conditions did not remarkably change. General social capital does enlarge the chances of becoming a supervisor.
== Table 6 about here ==
Table 6 shows that before all men have more chance to attain a supervisory position. The odds for migrants are also above one, but the parameter is not significant. Further, there is highly significant association between the number of people mentioned in the position generator and supervisory position. The other parameters do not cross the border of significance although some of them are very close, e.g., education and ethnic origin.
8 Table 7 shows our models on monthly earnings. Again, there is a clear effect of being a man on the amount of one’s income, all other things equal. Further and unexpectedly, people in smaller villages earn more money. In addition and as one could expect, a person’s education makes for higher earnings. Also membership in voluntary organization enhances one’s income, an effect which is stable in all models. Finally, the number of positions (extensity) as well as the upper reach with regard to prestige one can access contribute to the explanation of the variation in income. Range of positions has no effect for the height of a person’s income.
== Table 7 about here ===
Table 8 shows one’s satisfaction with the earnings. Note, that we did control for actual income in these models. Here, satisfaction increases with age. Also, having a Dutch origin enhances satisfaction with monetary situation. With regard to education, in particular those who finished college are more satisfied. Furthermore, having a partner who has a paid job has also a positive effect and, not surprisingly, there is a strong positive effect of one’s current income on satisfaction. In these analyses, being merely a member of a voluntary organization did not have a strong effect, yet doing actively voluntary work makes for more satisfaction as well as being a blood donor and giving money in charity collections. Strangely, social capital indicators dampen income satisfaction.
== Table 8 about here ==
Table 9 shows the models in subjective health. Subjective health decreases first with age, yet, the effect changes its sign, indicating that older people at a certain point rate their health again as being relatively good. Furthermore, Dutch rate their health better than migrants and those who finished college feel also somewhat healthier than this whose education is less than high school. In addition, having a working partner is correlated with a better health rating, the same holds for income, while membership has no effect on subjective health. Again, those who rate their health better have actually less social capital to dispose of.
== Table 9 about here ==
Table 10, finally, shows that those with a higher range and those with a higher average range of social capital have also a more open mind. General social capital has no effect. Also quite interesting is that memberships as such did not have the expected positive effect but church members do have a more closed mind. Furthermore, as one would expect, open mindedness goes together with higher education. In addition, women are more open minded than men and open mindedness increases first with age, but decreases later on.
== Table 10 about here ==
Because of the large number of analyses, the outcomes of all tables are summarized in table 11.
== Table 11 about here ==
9 5. Conclusion Our attempt to integrate the research on the production and consequences of social capital leads to the following conclusions. As to our study of the creation or emergence of social capital, the results of our analyses show that own human capital and prestige of the father do engender own social capital in the form of having access to many jobs as well as to higher prestige jobs. Living in an urban area and participating in voluntary organizations as well as having a working partner also affect these forms of social capital positively. In addition they also promote social capital in the form of the range of occupations one can access. Having a working partner works especially for men. Considering the returns of social capital, social capital increases a person’s chances of getting a supervisory position as well as of getting a better income. More precisely, only those with more general social capital and men have better chances of becoming a supervisor. Further, men, people who have a higher education, those who are a member to more organizations, and have more social capital, do have a better income. Unexpectedly those in smaller villages earn more. Social capital has, quite unexpectedly, negative effects on satisfaction with one’s income, and also on self-reported health. Rather interesting is the empirical support that is found for the idea that an open network leads to an open mind: have more social capital, i.e., range and upper reach, increases the openness of mind. Furthermore, it should be noted that our results on supervision are quite different compared to those found by Lin et al. (2008). These authors found amongst others significant effects on supervision for age, gender (male), having college education, urbanization and memberships, while in our study only gender and general social capital, the number of accessed occupations matters. With regard to earnings the results are more similar. Most of our findings on differential access to social capital can be easily interpreted within social capital. Education and father’s prestige are ego’s resources, making them more interesting interaction partners to others, whereas memberships and urbanization increase the meeting chances. Age probably first indicates a growing work experience and other forms of social capital, later in life age indicates a shrinking shadow of the future. With regard to outcomes, it is hard to understand why having more social capital does not lead to satisfaction with one’s earnings and why it does not promote a better self-reported health. The effects are even negative. This is surprising since usually there is even in our modern societies still a small effect of father’s prestige on his children’s prestige and earnings. Having a partner with a good education and a good job is promoting the career of the main actor someone’s partner is a major form of social capital to ego, always available and strongly prepared toe help (see Bernasco, De Graaf and Ultee 1997). Further, in other studies ethnic origin has shown to have strong consequences for social resources and relationships (e.g., Volker, Pinkster & Flap 2008), while in our study effects are small or non-existent. This is probably due to the fact that we have only very few foreigners in our sample (n = 80 in the first wave). Another intriguing finding is that men benefit more from their wife’s social capital than wives do. We are not sure how we can explain this finding. It might imply that women actually are not aware of the social capital they can access via their husband. This finding deserves more inquiry. One wants to know, for example, whether this is true for all types of occupations, social status, or how it is related to the amount of hours the partners are working.
10 Also, the negative effects of social capital we have found deserve much closer attention. Maybe they are a result of social comparison processes. People who see mostly others who are similar or even better off than themselves might rate their own situation less positive. Finally, although we have employed longitudinal data, causality issues remain a problem – at the first point of our measurements people already had established a network as well as outcomes of these networks. The returns of social capital might also affect its creation. Future research should pay attention to better disentangle cause and effect. Partially this issue can be tackled through the application of structural equation models in our data.
6. Discussion Social capital research seems to have entered a new phase. In particular, it is new that studies allow for comparison of networks of citizens in different societies. European sociology seems to have a special place in this development, because sociology in Europe is catching on with the research front in various research areas, inter alia, in network research. But more so because Europe has a quite interesting institutional and cultural variation between various countries. The diffusion of multilevel-regression analysis is, of course, another stimulant of this kind of research. Our study adds to the research field by studying creation and returns of social capital and adds the case of the Netherlands to the small but growing international comparative research of social capital. Our contribution has an additional original twist, though, as it uses longitudinal data. By now there are three types of this comparative research depending on how social capital is measured. A first literature is using the name-generator annex name-interpreter methodology. Höllinger and Haller (1990) compared social networks of citizens in seven countries involved in the International Social Survey Program (ISSP) of 1986 and showed that people in Middle Europe, that is in West-Germany, Austria, and Hungary, have fewer (or no) friends than in other western countries. Using the same data set Immerfall (1997) described in more detail the differences between the networks of citizens of different western countries. Next to the size of the networks of people from different countries there are also clear differences in their composition. People from the Middle and Southern European countries include more family within their networks than people from the Northern countries, and people from Australia and the United Stated include even less family. Moreover, people in Middle Europe go more often to the same people for different kinds of help. Such differences in social capital might be caused by differences in institutions. One hypothesis is that welfare states drive out social capital incorporated in people’s personal social networks. This hypothesis is called the crowding-out hypothesis. In contrast is the hypothesis on socio-economic security: welfare states and economic prosperity enable individual citizens to engage in voluntary associations without time-pressure and to acquire a sense of belonging. Using Euro-barometer data from 1992 on 13 European countries Scheepers, Te Grotenhuis and Gelissen (2002) showed that the more developed the welfare state is, the smaller the social capital people have (‘crowding out hypothesis’). However, a recent publication by Van der Meer,
11 Scheepers and Te Grotenhuis (2008) on ISSP-data on 20 western countries from 2001 refutes the first hypothesis, social security does not have a negative impact on social participation, and actually the second hypothesis is confirmed. An interesting additional finding they make is that a corrupt state leads to less social bonds between ordinary citizens. There is another branch of comparative research that is based on measuring social capital by the degree of interpersonal trust. The volume edited by Meulemann (2008) presents multiple examples of such work, all using data from the The European Social Survey of 2002.The fact that there seem to be larger areas of multiple neighboring countries with similar network patterns suggest cultural causes. Using such a measure Halman and Luijkx (2006) studied the degree to which cultural differences have an effect on interpersonal trust. And indeed a high civic morality and a low level of individualism do promote interpersonal trust quite clearly (Halman & Luijkx 2006: 8).4 The small research literature on social capital in communist and former communist societies uses both the name-generator and the interpersonal trust questions to measure social capital (see, for example, various contributions to Badescu and Uslaner, 2003). The general upshot of what is known is that citizens in former communist countries of Central and Easter Europe have less trust in their fellow citizens than those in the West. Their associational involvement is also far lower than their Western counterparts. Their social networks do not seem to be that different form those in Western countries, though. For more empirical results on the social capital differences between transition and non-transition countries, see Kaase and Parts (2007). Lately there is yet another, a third branch of comparative research in the making, a literature that uses the position generator. Together with Lin et al. (2008) we are among the first to start this line of comparative research.5 See also various contributions to Lin and Erickson (2008). Taking all together, our exercise has shown than job status, education, gender, and age affect a person’s social capital a lot. In addition, the population density of the area where one lives has an impact on the range of positions one can access. As to the returns of social capital, we could show that even when a period of about 7 – 8 years is in between the measurements, social capital has substantial positive effects on occupational outcomes and on open mindedness. Take note, the last variant, i.e., the position generator and first variant, i.e., the name- generator basically measure the social capital in the classic network interpretation of having others that are prepared to help to a certain extent by putting their resources at the disposal of ego.
4 Next to resources social capital is the product of meeting chances. In further research we could make additional assumptions on meeting places that promote meeting specific others (see Molllenhorst et al. 2008a,b) 5 Our design is somewhat better than that by Lin et al. (2008). Lin et al. used retrospective questions to establish social capital earlier on where as we measured social capital seven years before we determined income and supervision.
12 References
Badescu, G. and E. M. Uslaner (2003) Social Capital and the Transition to Democracy. London: Routledge. Berkman, L. F. and L. Syme (1979) Social Networks: Host resistance and Mortality: a Nine- Year Follow Up Study of Alameda County Residents. American Journal of Epidemiology, 109:186-204. Bernasco, W., P. de Graaf and W. Ultee (1997) Coupled careers. European Sociological Review 14: 15-31. Burt, R.S. (1992) Structural Holes: The Social Structure of Competition, Cambridge, MA: Harvard University Press. —— (2000) “The Network Structure of Social Capital”, in R.I. Sutton and B.M. Staw (eds) Research in Organizational Behavior, pp. 345-423, Greenwich, CT: JAI Press. Coleman, J.S. (1988) “Social Capital in the Creation of Human Capital”, American Journal of Sociology, 94: S95-S121. _____ (1990) Foundations of Social Theory, Cambridge, MA: Harvard University Press. _____ (1993) “The Rational Reconstruction of Society. 1992 Presidential Address”. American Sociological Review, 58-1-15. DiMaggio, P., and H. Louch. (1998) “Socially embedded Consumer Transactions: For what kinds of Purchases do people most often use Networks?” American Sociological Review 63: 619-637. De Graaf, N.D. and Flap, H.D. (1988) “With a Little Help from My Friends”, Social Forces, 67: 452-72. Flap, H.D. (1991) “Social Capital in the Reproduction of Inequality: A Review”, Comparative Sociology of Family, Health and Education, 20: 6179-6202. Flap, H.D. (1999) “Creation and Returns of Social Capital: A New Research Program”, La Revue Tocqueville, 20: 5-26. Flap, H. and Boxman, E. (2001) “Getting Started: The Influence of Social Capital on the Start of the Occupational Career”, in N. Lin, K.S. Cook and R.S. Burt (Eds) Social Capital: Theory and Research, New York: Aldine de Gruyter. Flap, H. and Volker, B. (2004) Creation and Returns of Social Capital: A New Research Program. London and New York: Routledge. Gaag, van der, M. (2004) Measurement of individual social capital. Amsterdam: Thela thesis. Ganzeboom, H.B.G. and Treiman, D.J. (1996) “Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations”, Social Science Research, 25: 201–39. Granovetter, M. (1973) “The Strength of Weak Ties”, American Journal of Sociology, 78:1360- 80. Halman, L. and R. Luijkx (2006) “Social capital in contemporary Europe: evidence from the European Social Survey” Portuguese Journal of Social Sciences 5: 65-90.
13 Höllinger. F. and M. Haller (1990) “Kinship and social networks in modern societies: a cross- cultural comparison among seven nations” European Sociological Review 6: 103-124. Immerfall, S. (1997) Sozial Integration in den westeuropäischen Gesellschaften: Werte, Mitgliedschaften und Netzwerke. Pp. 139-173 in Stefan Hradil and Stefan Immerfall (Hrsg.) 1997, Die westeuropäischen Gesellschaften Im Vergleich. Opladen: Leske und Budrich. Kaasa, A., and E. Pats (2007) Individual-level determinants of social capital in Europe: differences between country groups. Taiwan University Press. Laumann, E.O. (1973) Bonds of Pluralism. The Form and Substance of Urban Social Networks. Wiley & Sons, Canada. Lin, N. (1982) “Social Resources and Instrumental Action”, in P.V. Marsden and N. Lin (Eds.) Social Structure and Network Analysis, Beverly Hills, CA: Sage. ____ (1999a) “Building a Network Theory of Social Capital”, Connections, 22: 28-51. ____ (1999b) “Social Networks and Status Attainment”, Annual Review of Sociology, 25: 467-87. ____ (2001) Social Capital: A Theory of Structure and Action, London and New York: Cambridge University Press. ____ (2006) “Social Capital”, in J. Beckert and M. Zagiroski (Eds.) Encyclopedia of Economic Sociology, London: Rutledge. Lin, N. and Dumin, M. (1986) “Access to Occupations through Social Ties”, Social Networks, 8: 365-85. Lin, N., Ensel, W.M. and Vaughn, J.C. (1981) “Social Resources and Strength of Ties: Structural Factors in Occupational Status Attainment”, American Sociological Review, 46: 393-405. Lin, N., Fu, Y.C. and Hsung, R.M. (2001) “The Position Generator: Measurement Techniques for Investigations of Social Capital”, in N. Lin, K.S. Cook and R.S. Burt (Eds.) Social Capital: Theory and Research, New York: Aldine de Gruyter. Lin, N., D. Ao, and L. Song 2008. Production and returns of social capital: Evidence from urban China. Forthcoming in Ray-May Hsung, Nan Lin, and Ronald Breiger (Eds.), Contexts of Social Capital: Social Networks in Communities, Markets and Organizations. Routledge, London. Lin, N; J.C. Vaughn and W. M. Ensel (1981) “Social resources and occupational status attainment”, Social Forces, 59: 1163-1181. Marsden, P.V. (1990) “Network Diversity, Substructures, and Opportunities for Contact.” Pp. 397-410. Structures of Power and Constraint. Papers in Honor of Peter M. Blau. Craig Calhoun, Marshall W. Meyer and W. Richard Scott, Editors. Cambridge University Press. Marsden, P.V. and Hurlbert, J.S. (1988) “Social Resources and Mobility Outcomes: A Replication and Extension”, Social Forces, 66: 1038-59. McPherson, M. and L. Smith-Lovin. (1987) “Homophily in Voluntary Organizations: Status Distance and the Composition of Face-to-face Groups.” American Sociological Review 52(3):370-79. Meer, van der T., P. Scheepers, and M. te Grotenhuis (2008) Does the State Affect the Informal Connections between its Citizens? New Institutionalist Explanations of Social
14 Participation in Everyday Life. In Heiner Meulemann (ed.), Social Capital in Europe: Similarity of Countries and Diversity of People? Multi-level Analyses of the European Social Survey 2002. Brill. Meulemann, H. (Ed.) (2008), Social Capital in Europe: Similarity of Countries and Diversity of People? Multi-level Analyses of the European Social Survey 2002. Brill, Leiden. Mollenhorst, G., B. Völker, B. and H. Flap (2008a) “Social Contexts and Core Discussion Networks: Using a Choice-Constraint Approach to Study Similarity in Intimate Relationships”. Social Forces, 86, 3:937-966. Mollenhorst, G., B. Völker, and H. Flap (2008) “Social Contexts and Personal Relationships: The Effect of Meeting Opportunities on Similarity for Personal Relationships of Different Strength”, Social Networks, 30:60-68. Moore, S., A. Sheila, P. Hawed and V. A. Haines (2005) “The privileging of communitarian ideas: Citation practices and the translation of social capital into public health research”, American Journal of Public Health 95: 1330-1337 Ports, A. (1998) “Social Capital: Its Origins and Applications in Modern Sociology”, Annual Review of Sociology, 22: 1-24. Putnam, R.D. (1993) Making Democracy work. Civic Traditions in modern Italy. Princeton University Press. Putnam, R.D. (2000) Bowling Alone: The Collapse and Revival of American Community, New York: Simon & Schuster. Purtnam, R.D. (2007) E Pluribus Unum. Diversity and Community in the Twenty-first Century. The 2006 Johan Skytte Prize Lecture. Scandinavian Political Studies, V30, 2: 137-174. Rokeach M. (1960) The Open and Closed Mind. New York.Basic Books. Scheepers, P., M. te Grotenhuis and J. Gelissen (2002) “Welfare states and dimensions of social capital. Cross-national comparisons in European countries” European Societies 4: 185- 207.
15 Table 1 Position Generator and Differential Access to General Social Capital (SSND1) Respondent Accessing (in percent) Position (ISCO-prestige score) Family Friend Acquaintance Total (n) Lawyer (86) 16.4 11.4 18.7 46.5 (468) Physician (84) 19.8 9.5 20.3 49.6 (500) Policy maker (82) 17.7 12.7 14.7 45.1 (454) Professional engineer (76) 36.5 13.5 15.4 65.4(659) IT expert (68) 28.0 18.3 20.1 66.4 (669) Director of firm (67) 36.5 17.1 16.9 60.5 (609) Manager (67) 34.2 18.0 13.5 65.5 (661) Union leader (66) 3.8 3.3 9.4 16.5 (167) Scientist (65) 19.3 12.0 10.7 42.0 (423) High official (64) 23.2 11.4 18.5 53.1 (536) Real estate agent (64) 6.5 6.0 18.1 30.6 (308) Technician (63) 38.8 14.2 16.0 69.0 (694) Teacher (62) 36.8 19.1 17.0 72.9 (734) Police officer (54) 11.5 8.4 22.1 42.0 (422) Bookkeeper (52) 25.2 14.0 23.5 62.8 (633) Secretary (52) 28.0 17.4 21.6 67.1 (675) Insurance agent (52) 11.2 7.7 21.2 40.1 (404) Farmer (46) 24.6 8.4 17.1 50.0 (503) Musician (45) 21.1 14.2 16.0 54.3 (547) Nurse (44) 38.8 16.3 19.9 75.0 (751) Engine driver (44) 7.8 3.2 7.5 18.4 (185) Cook (39) 16.3 11.1 18.0 45.5 (458) Barber (35) 13.2 9.5 25.4 48.3 (486) Foreman (27) 11.1 4.8 10.3 16.3 (741) Truck driver (26) 20.9 8.5 20.8 49.7 (499) Postman (26) 7.2 4.7 15.9 26.8 (279) Sales person (22) 30.4 14.2 17.2 61.9 (622) Cleaning person (20) 11.3 5.2 18.1 34.6 (349) Unskilled worker (15) 15.8 6.6 15.4 37.8 (382) Construction worker (15) 31.3 11.7 22.8 65.9 (663) Summary Indices Extensity: Mean 6.43 3.35 5.22 15.00 S.D. 3.80 3.25 4.43 5.63 Range: Mean 45.20 32.24 43.30 62.51 S.D. 19.15 22.55 22.93 11.31 Upper reachability: Mean 73.58 66.32 71.35 81.13 S.D. 12.79 15.60 15.94 8.02
16 Table 2 Summary of key variables used from SSND1, 2000, and SSND2, 2007, (Percent or Mean; n=1006 and 604 respectively) SSND1 percent mean SD Sex: Male 57.9 Female 42.1 Age 45.10 11.16 Education : Secondary or lower (1) 35.5 High school (2) 24.1 Associate college (3) 25.4 College and graduate (4) 15.0 Employment status: Having a paid job 75.5 Tenure (years) 10.87 10.21 Partner has a paid job 41.0 Kind of job for which one is educated/working in White collar (vs. blue collar and farming) 64.8 Membership in voluntary organization 86.2 1.93 1.30 Degree of urbanization: Less than 1000 addresses per km2 (1) 40.1 Between 1000 and 1500 addresses per km2 (2) 21.4 More than 1500 addresses per km2 (3) 37.9 Fathers prestige (ISCO codes) 46.86 17.58 Dutch origin 91.0 SSND2 Supervision in present/last job 42.6 Monthly net income (in analysis: 17 categories) less than 500 € pm 4.3 between 500 and 1000 € pm 13.3 between 1000 and 1500 € pm 20.4 between 1500 and 2000 € pm 20.7 between 2000 and 2500 € pm 19.6 between 2500 and 3000 € pm 7.8 between 3000 and 3500 € pm 5.1 between 3500 and 4000 € pm 3.5 4000 € pm and more 5.2 Changed job since 1999 35.1 Satisfaction with very unsatisfied 1.2 financial situation: unsatisfied 3.6 considerably unsatisfied 4.0 not satisfied/ not unsatisfied 5.1 considerably satisfied 13.4 satisfied 62.0 very satisfied 10.6 Subjective health : very bad 0.2 bad 1.7 considerably bad 5.3 not good, not bad 4.0 considerably good 14.4 good 57.1 very good 17.4 Open mindedness/dogmatism (range 9 – 30; z-score is used in analyses) 21.03 3.41
17 Table 3 Determinants of General Social Capital: number of occupations mentioned (OLS regression; source: SSND1; standardized coefficients) Added: Socio- working Added: human Added: demographic partner and capital membership characteristics interaction with gender beta beta beta beta Gender (male) .077** .050 .035 -.015 Age .948*** .611** .520** -.010 Age squared -.966*** -.602** -.550** Ns Origin (Dutch) .020 .013 -.005 -.006 Prestige of father .082** .031 .038 .033 Urbanization (ref: highest) 1000 - 1500 adr/ km2 -.109*** -.103*** -.065* -.056* > 500 adr/km2 .015 .019 .011 .018 Education (ref: Less than high school) High school .052 .049 .049 Associate college .062 .020 .015 Bachelor degree .083** .039 .036 and above Being employed .106*** .092*** .094** White collar job .078** .076** .070** Membership in voluntary .241*** .244*** association Working partner .025 Working partner*gender .105** 10.831 Constant 3.198 (2.878) 5.296 (2.948) 5.430 (2.883) (1.228) Observations 1006 965 965 965 Adjusted R-squared .031 .068 .11 .13 Note: * significant at 10%; ** significant at 5%; *** significant at 1%.
18
Table 4 Determinants of General Social Capital: upper reach through social positions (OLS regression; source: SSND 1) Added: Socio- working Added: human Added: demographic partner and capital membership characteristics interaction with gender beta beta beta beta Gender (male) .007 -.016 -.022 -.075 Age a) .138*** .149*** .134*** .143 Origin (Dutch) -.018 -.016 -.022 -.022 Prestige of father .198*** .109*** .112*** .112*** Urbanization (ref:highest) 1000 - 1500 adr/ km2 -.131*** -.111*** -.105*** -.103*** > 500 adr/km2 -.109*** -.097*** -.106*** -.111*** Education (ref: Less than high school) High school .149*** .087** .092*** Associate college .088** .162*** .161*** Bachelor degree .177*** .183*** .180*** and above Being employed .043 .037 .020 White collar job .104*** .104*** .098*** Membership in voluntary .089*** .090*** association Working partner -.004 Working partner*gender .116** 71.624 Constant 73.910 (1.639) 71.784 (1.697) 71.618 (1.692) (1.706) Observations 965 965 965 965 Adjusted R-squared .069 .14 .15 .16 Note: * significant at 10%; ** significant at 5%; *** significant at 1%. a) No effects of age squared have been found
19 Table 5 Determinants of General Social Capital: range of social positions (OLS regression; source: SSND 1) Added: Socio- working Added: human Added: demographic partner and capital membership characteristics interaction with gender beta beta beta beta Gender (male) .038 .020 .011 -.026 Age .456 .012 -.011 -.004 Age squared -.466* ns ns ns Origin (Dutch) .010 .008 -.001 -.003 Prestige of father .061* .027 .031 .032 Urbanization (ref:highest) 1000 - 1500 adr/ km2 -.122*** -.114*** -.104** -.101** > 500 adr/km2 -.004 -.001 -.016 -.023 Education (ref: Less than high school) High school .041 .040 .048 Associate college .040 .015 .017 Bachelor degree .059 .032 .028 and above Being employed .070** .060* .028 White collar job .045 .045 .036 Membership in voluntary .146*** .145*** association Working partner .056 Working partner*gender .090* 56.193 Constant 50.804 (5.890) 58.310 (2.530) 57.319 (2.531) (2.689) Observations 965 965 965 965 Adjusted R-squared .03 .04 .05 .06 Note: * significant at 10%; ** significant at 5%; *** significant at 1%.
20 Table 6 Returns of Social Capital 1: Logistic Regression on supervision Added: Added : social
membership capital Exp(B) Exp(B) Exp(B) Gender (male) 3.260*** 3.265*** 3.264*** Age a) .993 .993 .993 Origin (Dutch) 1.319 1.320 1.390 Prestige of father 1.004 1.004 1.003 Urbanization (ref:highest) 1000 - 1500 adr/ km2 .789 .788 .823 > 500 adr/km2 1.053 1.055 1.066 Education (ref: Less than high school) High school 1.004 1.004 .967 Associate college 1.164 1.167 1.171 Bachelor degree 1.594 1.601 1.585 and above Partner works .834 .834 .786 Changed job since 1999 1.010 1.016 1.005 Membership in voluntary organizations -- .995 .479 Social capital (general) -- -- 1.058*** Constant .281** .282** .139** Observations 604 604 604 Pseudo R-squared (Nagelkerke) .110 .111 .131 -2LL 508.712 508.609 500.124 * significant at 10%; ** significant at 5%; *** significant at 1% a) No effects of age squared have been found
21
Table 7 Returns of Social Capital 2: OLS Regression on monthly income added: Added: added: social Added: social social capital: membership capital: capital: upper extensity range reach beta beta beta beta beta Gender (male) .404*** .395*** .391*** .399*** .396*** Age a) .019 .003 .005 -.009 .057 Origin (Dutch) -.006 -.007 -.004 -.005 -.006 Prestige of father .056 .058 .055 .049 .057 Urbanization (ref:highest) 1000 - 1500 adr/ km2 .020 .026 .027 .032 .029 > 500 adr/km2 .087** .078** .078** .085** .078** Education (ref: Less than high school) High school .052 .049 .046 .044 .047 Associate college .272*** .256*** .254*** .242*** .255*** Bachelor degree .377*** .360*** .357*** .345*** .357*** and above Partner works .027 .025 .018 .018 .021 Changed job since 1999 .002 -.003 -.004 -.006 -.006 Membership in voluntary organizations -- .086** .070* .079** .078** Social capital -- -- .063* .076** .021 3.567 2.911 .422 2.496 Constant 3.451 (.953) (.956) (.999) (1.754) (1.230) Observations 595 595 595 595 595 Adjusted R-squared .334 .340 .350 .360 .330 * significant at 10%; ** significant at 5%; *** significant at 1% a) No effects of age squared have been found
22 Table 8: Returns of Social Capital 3 - OLS Regression on satisfaction with financial situation Added: Added: Voluntary Added: Upper work and Range reach extensity beta beta beta beta Gender (male) -.076* -.056 -.071* -.066 Age .139*** .107*** .119** .104** Age2 NS Origin (Dutch) .114** .106** .101** .100** Prestige of father .023 .029 .033 .023 Urbanization (ref:highest) 1000 - 1500 adr/ km2 -.009 -.014 -.022 -.021 > 500 adr/km2 .017 .004 -.006 .002 Education (ref: Less than high school) High school .073 .063 .063 .061 Associate college .103*** .103** .114** .099** Bachelor degree .037 .040 .051 .041 and above Partner works .156*** .158*** .155*** .155*** Changed job since 1999 -.057 -.058 -.058 -.057 Income .295*** .299*** .302*** .299*** Doing voluntary work .089** .077** .082** Giving money .094** .092*** .096** Social capital -.073* -.094** -.097** 3.350 2.322 4.088 3.535 Constant (.347)*** (.897)*** (.670)*** (.499)*** Observations 585 585 585 585 Adjusted R-squared .156 .160 .168 .160 * significant at 10%; ** significant at 5%; *** significant at 1%
23 Table 9: Returns of Social Capital 4 - OLS Regression on subjective health Added: Added: Voluntary Added: Upper work and Range reach extensity beta beta beta beta Gender (male) .013 .014 .002 .005 Age -.645* -.636* -.664** -.671** Age squared .564* .544* .586* .580** Origin (Dutch) .095** .090** .091** .090** Prestige of father -.021 -.014 -.011 -.019 Urbanization (ref:highest) 1000 - 1500 adr/ km2 .044 .045 .040 .040 > 500 adr/km2 .003 -.002 -.009 -.002 Education (ref: Less than high school) High school .060 .063 .062 .061 Associate college .114** .108** .120** .106** Bachelor degree .030 .026 .037 .028 and above Partner works .082* .092** .089** .089** Income .157** .160*** .161*** .159*** Changed job since 1999 .047 .046 .047 .049 Membership in voluntary organizations .064 .047 .055 Social capital -.092** -.081* -.093** 6.404 6.567 7.369 7.008 Constant (.7835)*** (.784)*** (.940)*** (.830)*** Observations 595 595 595 595 Adjusted R-squared .061 .070 .099 .077 * significant at 10%; ** significant at 5%; *** significant at 1%
24 Table 10: Returns of Social Capital 4 - OLS Regression on open mindedness Added: Added: Added: membership upper range and extensity reach beta beta beta beta Gender (male) -.171*** -.164*** -.155*** -.160*** Age .557* .502* .554* .551* Age squared -.652** -.581* -.659** -.640** Origin (Dutch) .063* .065* .060 .060 Prestige of father .022 .011 .006 .017 Urbanization (ref:highest) 1000 - 1500 adr/ km2 .004 .021 .028 .027 > 500 adr/km2 -.019 .004 .011 .002 Education (ref: Less than high school) High school .092** .090** .075* .078* Associate college .300*** .298*** .279*** .297*** Bachelor degree .433*** .418*** .399*** .411*** and above Partner works .019 .017 .013 .015 Income .026 .008 -.001 .021 Changed job since 1999 .071* .061 .058 .056 Church membership -.140*** -.131*** -.137*** Social capital .046 .108*** .098*** -1.513 -1.391 -2.503 -1.887 Constant (.674)** (.672)** (.795) (.701)*** Observations 597 590 590 590 Adjusted R-squared .232 .260 .273 .272 * significant at 10%; ** significant at 5%; *** significant at 1%
25
Table 11 Summary of Findings Dependent Variable Predictor Extensity Upper reach Range Supervision Income Satisfaction Subjective Open with income health mind Male + + + - - Age + + + - + Age squared - + - Education (ref: less than High school) High school + + Associate college + + + + + Bachelor degree and above + + + Origin + + Prestige of Fathers’s job + Being employed + ni White collar + + ni Working partner + + Working partner*gender + + + ni ni Urbanization (ref. > 1500 adr/km2 1000-1500 addresses / km2 - - - < 1000 adr/km2 - + Membership in voluntary associations + + + + + +church Income ni ni ni ni ni + + Social capital: Extensity ni ni ni + + - - Upper reach ni ni ni + - - + Range ni ni ni - - + Note: + indicates a positive significant coefficient; - a negative significant coefficient, NI indicates that condition has not been included in the particular analysis and en empty cell indicates no significant association.
26
Figure 1: Estimated number of positions Figure 2: Estimated upper reach of (extensity) accessed for males and females positions accessed for males and by employment status of the partners. females by employment status of the Numbers are controlled for age, partners. Numbers are controlled for employment status of the respondent and age, employment status of the urbanization. respondent and urbanization.
Figure 3: Estimated range of positions accessed for males and females by employment status of the partners. Numbers are controlled for age, employment status of the respondent and urbanization.
27
VARIETIES OF SOCIAL CAPITAL AND THEIR SOURCES
Bonnie H. Erickson and Rochelle Cote Sociology, University of Toronto
May 2008
INTRODUCTION One of the most important sources of social capital is the variety of people in a person#‘s entire social network, including weak as well as strong ties. The more varied the contacts, the more varied and rich the resources to which they potentially provide access. Past research has been almost entirely devoted to just one kind of variety: the number of different occupations in which the focal person knows anyone. But different forms of variety provide access to different ranges of resources., and, people gain different kinds of network variety in different ways. Thus we add four new kinds of network variety to the traditional occupational variety: (1) the number of middle class occupations in which the person knows anyone, (2) the number of working class occupations in which the person knows anyone, (3) the number of occupations in which the person knows a man, and (4) the number of occupations in which the person knows a woman. These forms of social capital provide access to the different kinds of resources typically controlled by higher prestige people, lower prestige people, men, and women. We then examine how unequal access to these five kinds of social capital is or is not shaped by homophily, contact opportunities, biographical constraints, social status, and position in the occupational prestige hierarchy.
FORMS OF SOCIAL CAPITAL Social capital is the resources embedded in social networks. That is, a person#‘s social capital is the array of resources belonging to the person#‘s contacts and potentially available to the person through these contacts. One important form of social capital is the variety of people known, and hence the variety of resources these people control. Variety matters for two reasons. First, the more different kinds of people one knows, the more likely it is that one of them will have the resource one needs. For example, people who know others in many kinds of occupations are more likely to have some contacts with high occupational prestige, to be able to get their help in seeking a job, and hence to get a good job (Lin 1999). Second, variety is sometimes valuable in itself. For many higher level jobs, employers prefer people with varied contacts because these contacts can be appropriated for the firm’s use in recruiting clients, scanning the environment, and so on (Erickson 2001). Varied networks also engage people in many kinds of relationships, with different kinds of people, in different settings, and this "complexity of role sets" is a source of important personal strengths including more sophisticated language and greater sense of control over one’s own life (Coser 1975). People with more varied networks also receive more varied cultural inputs from their contacts and hence develop more richly and usefully varied cultural repertoires (Erickson 1996). Thus network variety is an important advantage in life (Erickson 2003).
But there are different varieties of network variety . Theoretically, the important forms of variety are those that provide access to varied resources. Research to date relies heavily on occupational variety as measured by some form of position generator (Volker and Flap 1999, Lin, Fu and Hsung 2001). Typically, researchers develop a list of occupations which range from high to low in prestige, and ask respondents whether or not they know anyone in each of these occupations. The more occupations a person has one or more contacts in, the greater the network diversity. This simple but powerful approach works because occupation is a master role in modern societies, one which reflects important aspects of past histories (such as class background and education) and current social location. People in different occupations are different in a plethora of resources: interests, tastes, information, money, power in the workplace, skills, and so on. We call this classic form of social capital occupational diversity. For some outcomes, access to higher status positions is much more useful than access to lower status positions. For example, the higher the occupational prestige of a contact used to get a job, the better the job one gets (Lin 1999). People in higher level work often have greater control over good jobs, greater cultural resources, higher levels of information, and wider networks that provide greater indirect access to a wealth of resources. We call access to higher prestige occupations middle class diversity. Access to lower status occupations may well provide less leverage on the whole than does middle class diversity, but, may also provide access to resources of different kinds. Lower prestige people have better access to working class skills, and may be better routes to relatively desirable working class jobs, such as jobs filled by employee recruitment. Thus we also separate out and examine access to working class occupations, which we call working class diversity. Occupational variety is important because occupations differ in their resources. But occupation is not the only way in which people are divided into categories with distinctive amounts or types of resources. All major forms of inequality do this. Here, we also include gender stratification, one of the most pervasive and powerful forms of inequality and difference, and one with important resource consequences. Men and women have different tastes and are well informed about different topics (such as sports for men, but novels for women, Erickson 1996). They take responsibility for different things, such as men’s greater attention to automobile maintenance and women’s greater attention to health. Moreover, the vast gender differences in resources cannot be reduced to occupational differences. For example, the cultural contrasts that make gender boundaries are different from the cultural differences that distinguish occupations, so that gender and class each need their own analysis (Hall 1992).If social capital consists of the resources embedded in one#‘s network, and if men and women have typically distinctive resources, then contacts with men are a different form of social capital than contacts with women. It is not that ties to men are better or worse social capital, but rather that ties to men are more useful social capital for the kinds of resources men tend to control and ties to women are better social capital for the kinds of resources women tend to control. For example, men have more power in the workplace, so contacts with men are more productive of good jobs. But women know more about health, so contacts with women have a greater positive effect on health. Thus we examine both male diversity, the number of occupations in which a person knows one or more men, and female diversity, the number of occupations in which a person knows one or more women. To keep this paper and related conference presentation brief, we have combined the materials usually separated into a theory and a results section. Below we first describe the data
and data analysis procedures, then give a combined discussion of our arguments and our findings. DATA AND MEASUREMENT The data source is the 2004 federal election study in Canada. This study selected a national representative sample and administered three surveys: a telephone survey during the election, another telephone survey just after the election, and a mailed-out survey shortly afterwards. This paper uses variables from all three waves. For the main multiple regression analyses, there are somewhat over 1100 cases with complete information on all variables. The social capital measures are based on an item designed by Erickson for the 2000 federal election study (see Erickson 2004). She designed a new kind of gendered position generator (Figure 1), a variant on the original developed by Lin and Dumin (1986). To select occupations she consulted the 1996 Census of Canada to choose occupations including at least 20,000 people, since few respondents will know anyone in very small occupations. She also chose occupations with clear common-language titles in the census, to facilitate comparison between survey results and census information. Given these restrictions, she looked at several types of occupations: higher professionals, middle managers, other professionals, skilled trades, lower level service workers, semi-skilled trades, and unskilled. Within each set she chose one of the most male-dominated and one of the most female-dominated available, at roughly comparable levels of occupational prestige, while also varying prestige from very low to very high (prestige scores came from Ganzeboom and Treiman 1996). She also varied the sectors to which occupations belong, and added farmers, the largest and most typical of occupations in the agricultural sector. Table 1 shows the occupations selected, their size, gender composition, and occupational prestige. From this position generator, as repeated in the 2004 study, we constructed five social capital scales. Occupational diversity is the simple count of the number of occupations in which the respondent reported knowing anyone (whether men, women, or both). Middle class diversity is the simple count of the number of higher prestige occupations in which the respondent knew someone, with the six higher prestige occupations being lawyer, pharmacist, human relations manager, sales and marketing manager, social worker, and computer programmer. Working class diversity is the count of the number of lower prestige occupations (the remaining 11 occupations) in which a person knows someone. Male diversity is the simple count of the number of occupations in which the respondent reports knowing one or more men. Female diversity is the number of occupations in which the respondent knew any women. All the social capital scales have good reliability as measured by Cronbach’s alpha: occupational diversity = .867, male diversity = .823, female diversity = .812, middle class diversity = .697, working class diversity = .826. Most of the measure for independent variables are standard and straightforward. Age is age in years. Since age straggles up, and we expected a curvilinear effect of age on social capital, we use age and the square root of age to make a quadratic fit without excessive straggle. Gender is coded as 1 for women and 0 for men, so labelled "female.“ “Non-European” is a dummy variable that is 1 for people who report that their nations of origin are not European; this is our best approximation of non-white. Canadian born is 1 for those born in Canada, 0 otherwise. Education is an ordinal variable for level of education (not education in years of schooling). Working is 1 for those working, 0 otherwise. Prestige is the Blishen score for occupational prestige (a standard occupational prestige score for Canada). We include dummy variables for
whether or not the respondent has children under the age of 18, and whether or not the respondent is married or living with a partner. Voluntary association involvement is the number of kinds of voluntary associations in which the respondent reports being active in the past five years. We inspected the means of the social capital variables for the four main regions of Canada and found that the West and the East are consistently highest, while Ontario and Quebec are consistently lowest, so we included two dummy variables for residence in the Western provinces and in the Eastern provinces. We also inspected means for rural, mid-size, and metropolitan location. Finding that the means for social capital are all lowest in metropolitan areas, we left that as the reference category, and included dummies for rural and mid-size location. DATA ANALYSIS STRATEGY The arguments to follow predict how social capital should be related to important forms of social location, taken one at a time. But these arguments also suggest that the effects of social location are complex, sometimes direct, sometimes indirect, or both. For example if more educated people have greater social capital, is that because they met people through their educational experiences or is it because more educated people have higher levels of participation in voluntary associations? To help untangle the complex possible pathways through which social capital is formed, we conducted multivariate analysis in an approximation of life course analysis. We conducted a sequence of multiple regressions in which we first entered variables ascribed at or soon after birth (age, gender, birth in Canada or elsewhere, ethnicity) and then entered variables in the order in which they appear in lives on the average: education; adult work and family roles; and voluntary association participation. Finally, we entered geographic location to see whether regions, and rural to metropolitan areas, differ in social capital beyond what we would expect from differences in the individual attributes of their populations. While most of the regression analysis is straightforward, we needed to give special treatment to the respondent’s occupational prestige. Those respondents who are not working do not have occupations, hence do not have an occupational prestige score. Thus they would drop out of the regression if we simply used prestige as an independent variable. Fortunately, Ross and Mirowsky (1992) explain a useful way to include all respondents and to use prestige scores only for those who are working. First, recode the occupational prestige values as deviations from their mean. Second, define a dummy variable which is 1 for those working and 0 for others. Third, give those NOT working arbitrary "placeholder" values for occupational prestige, so they do not have missing data on this variable and will remain in the regression. Fourth, create a variable which is the dummy for working multiplied by the recoded prestige scores. The arbitrary placeholder values for non-working people are thus multiplied by zero and drop out, while the real prestige scores for those working are multiplied by 1 and enter the regression. Finally, include both the dummy for working and the term (working times recoded prestige) in the regression. The dummy for working shows the difference between workers with average prestige and those not working. The other term, the dummy for working or not times centered prestige scores, shows the effect of prestige only for those who are working.
THE SOCIAL SOURCES OF SOCIAL CAPITAL People differ in the amount of social capital they can get, in the kinds of social capital they can more easily get, and in how they get what they get. While the patterns are complex, as we will show, they are organized by a modest number of underlying principles. Homophily
One of these important principles is homophily (McPherson, Smith-Lovin and Cook 2001): given a choice, people prefer to meet others like themselves in ways socially defined as important. People rightly assume that they will have more in common with more similar others, and find it both more attractive and more easy to build relationships with them. Thus: Men have greater diversity of ties to men, and women have greater diversity of ties to women. Higher occupational prestige goes with greater middle class diversity and lower working class diversity. Feld (1982) points out that much of the apparent homophily in social life may not be choice but necessity. People of the same kind often end up in the same social settings, not necessarily of their own free will, as when students are placed in classrooms with children their own age. Thus we must next consider the social settings in which people participate, and the kinds of contact opportunities they provide. Social Settings as Contact Opportunities In general, people gain a greater number of chances to meet a greater variety of people if they participate in more social settings, especially social settings that differ from each other in the kinds of people in them and social settings that include an internal variety of people. The strongest example of this principle is voluntary associations. Associations of different kinds appeal to different kinds of people. For our purposes, it is especially relevant that associations differ in membership distribution of gender and occupational prestige (e.g. McPherson and Smith-Lovin 1987). Many associations also have some internal diversity of gender and occupational prestige (e.g. Erickson and Nosanchuk 1984). People who belong to more voluntary associations have more opportunities to meet a wide range of people. Further, they are well motivated to take advantage of these contact opportunities because of the homophily principle. Fellow members of the same association share a valued kind of similarity, common interest in the association’s purposes. Thus: The greater the number of voluntary associations in which a person is active, the greater the person’s level of all five forms of social capital. Participation in a social setting provides good contact opportunities for the kinds of people that populate the setting. Thus activity in voluntary associations might enhance middle class diversity more than working class diversity, because middle class people take part in associations at higher rates. However, people with modest levels of occupational prestige and/or education do become active in associations, at rates that are non-trivial even if lower than middle class rates, so others can and do meet working class people through association activity. Association activists probably meet more middle class than working class people, but still meet enough working class people to get a useful variety of working class contacts. From the social resources point of view, variety of contacts is more important than sheer number; one does not need to know a hundred carpenters to have access to the skills and information that a single carpenter can provide. The composition of a setting sets stronger limits on contact variety when the setting is very strongly dominated by one type of person. Schools are a notable example. People with higher levels of education have spent years in higher level educational institutions that are strongly dominated by fellow students who will go on to have middle class jobs. Even before entering higher level organizations like universities, the students bound for higher education are often in special tracks strongly dominated by other students also bound for educational success.
Thus, higher education provides a long history of better access to future middle class people than to future working class people. These future middle class people are as likely to be women as men, so education also promotes diversity of access to both men and women. Hence: Higher education goes with greater male, female, and middle class diversity, but not greater working class diversity. Education provides contact opportunities in itself, as people meet fellow students, but it also leads to later life experiences that themselves provide contact opportunities. Most notably, better educated people are more active in voluntary associations. Thus: The positive effects of education on diversity are weaker, but still significantly positive, after controlling for activity in voluntary associations. Where settings are not strongly dominated by one kind of person, but instead are internally diversified, people have a large range of choice among potential contacts. The effect of this range of choice depends on the setting‘s population size. If the population size is large, people have potential access to many people much like themselves, and they prefer similar others (homophily). This accounts for an initially surprising result: people living in rural areas have more diversified networks than those living in urban areas (see also Angelusz and Tardos 2001:311, for Hungary). Urban areas include people in a far greater variety of occupations, so urbanites could easily build highly varied networks. However, urban areas also include larger numbers of people in any given occupation and similar occupations. Thus urbanites more often have the option of recruiting network members from people in a narrow range of similar kinds of work (Fischer 1982: 179.) In rural areas, there are fewer occupations but also fewer people, so one is compelled to meet all or almost all of the people around, similar or not. For the same reasons, rural areas force men and women to meet each other extensively. Thus: The larger the place (from rural to mid-size town or city to metropolitan area) the lower the levels of all five forms of network diversity. Men and women move in somewhat different social circles. There is extensive gender segregation in work, with women still under-represented in higher level positions. Men and women tend to join different kinds of voluntary associations. Thus differences in contact opportunities alone would lead to women meeting more women than men do, and men meeting more men than women do. When contacts are made, homophily leads to further selection as women feel more interest in building ties with the women they meet, and men feel more in common with the men they meet. Thus men should have greater male diversity than women and women should have greater female diversity than men, even after controlling for contact opportunities including education, voluntary associations and so no. But women have somewhat less access to middle class diversity than men. Network variety should greater for those in paid employment, because working includes meeting at least some people, and past research often shows greater network diversity for those working. However, we do not find this. The sheer fact of working may not be important in an occupational system that includes large numbers of jobs with restricted contact opportunities, such as the growing number of lower level service positions like telephone soliciting. Biographical Constraints If we think of family life as a social setting that may provide contact opportunities, family ties could provide network diversity. Network variety might be greater for those with a spouse or partner, because one meets some of one#‘s partner’s contacts. Variety might be greater for those with children insofar as caring for the children leads one to meet fellow parents, childcare
workers, and others to whom children connect adults. However, family life absorbs a good deal of time, and care for younger children is especially demanding. Family life is not a social setting enhancing contact opportunities so much as it is a form of biographical constraint that can limit ability to turn any contact opportunities into new relationships. Thus: Having a partner does not add to network diversity, while having children under 18 generally reduces network diversity. Another, more powerful form of biographical effect is position in the life course. Both the youngest and the oldest adults face many biographical constraints compared to people in mid-life in North America. The young and the old have no paid work or lower level work, while those in mid-life are at the peak of their careers.. The young and the old have lower incomes, lower rates of activity in voluntary associations, and lower social standing than those in mid-life. Thus: Network diversity rises from youth to mid-life and then drops again into old age. This is true even after all controls, because we have not been able to control all of the many social advantages of mid-life. Social Status People with higher social status generally are more active in a variety of social settings, gaining more contact opportunities. Not only do higher status people encounter more varied potential network members, but these potential contacts are more willing to become network members, because higher status is attractive. Thus status advantage becomes network advantage. So powerful is the networking advantage of superior social status that position generator studies in several different parts of the world have reported it. For example in Hungary, knowing people in many occupations goes with wealth, education, working, being married, and being a landowner or self-employed (Angelusz and Tardos 2001:311). In Taiwan, occupational variety goes with education, employment, being married, and being male (Lin, Fu, and Hsung 2001:71). Other examples are plentiful. Thus we were somewhat surprised to find no great network advantage to being of European (probably white) ancestry. The lack of effect of race may well just be due to the very small number of non-white people in our sample (a study of an earlier election survey, Erickson 2004, for the 2000 election, did find a network handicap for non-whites). Being born in Canada gives people a longer time to accumulate social connections in Canada, gives people better access to the labour market, and gives greater social status. Thus the Canadian born have higher levels of social capital, and especially have better access to a range of men in many occupations and to a wider range of working class occupations. Most immigrants to Canada are highly selected for higher level education and work experience, not for working class skills such as carpentry. Immigrants do increasingly find they cannot convert their foreign qualifications into good middle class work in Canada, and many immigrants are forced to take low prestige jobs. But they are not qualified for the more highly skilled blue collar jobs, so do not often get into the more richly developed working class networks found in the skilled trades. The effect of Canadian birth fades somewhat when we control for voluntary associations, in which th4e native born are more active, and region, because immigrants overwhelmingly move to metropolitan areas where social capital is least rich on average. Occupational Prestige We initially expected occupational prestige to go with network diversity, for two reasons. First is the “Nan Lin - Peter Blau” hypothesis. Lin argues that modern occupational structures have a pyramid shape, with many people at low levels and fewer and fewer people as
one goes up the occupational ladder. Higher status people have many lower level people they can meet, while lower level people have far more restricted numbers of potential contacts above them. However, this pyramidal model no longer describes occupations in Canada well. Many working class or modest middle class jobs have been downsized out of existence or shifted outside Canada to lower wage countries. The strong growth of the service sector has brought both new lower level jobs and new higher level ones. The pyramid has become a pillar. Second, we expected that more prestigious positions would include more contact with others and more autonomy over one#‘s work including more freedom to build connections. However, many high prestige occupations can be quite socially limited. Consider highly skilled technical workers who only meet other experts at work, for example. Occupations at the same level of prestige vary greatly in the kinds of jobs they are and the kinds of networks they support. Thus we found, to our surprise, that occupational prestige does not go with network diversity. The only kind of effect is a negative one: higher prestige people have lower levels of diversity of contacts with working class occupations. The only apparent effect of high prestige positions is to insulate people from contact with lower prestige people. Future work should consider work variables more clearly linked to network formation, such as position in command hierarchies and the extent to which a job includes non-trivial work with people. A Note on Region Finally we turn to regional effects. Social capital tends to be higher in the West and the East than in Ontario or Quebec. Commenting on the high levels of social capital in the East also found in the 2000 election survey, Erickson (2004) attributed this to a long history of hardship that compelled Easterners to engage in mutual support networks, combined with low levels of in-migration that might disrupt local connections. The same applies to part of the West, especially the prairies, but not to all the West. Alberta has been an especially wealthy province for some time, and both Vancouver and Calgary are migration magnets. Thus the high levels of social capital in the West are a bit puzzling at first, but become clearer when we distinguish middle class diversity from working class diversity. The West is higher than Ontario and Quebec in middle class diversity but not working class diversity. Economic booms, the middle class in-migrants they attract, and the rising education levels they both demand and fund, all make it easier to meet socially active middle class people. The East is higher than Ontario and Quebec in working class diversity but not middle class diversity. The East#‘s history of hardship and community has helped to incorporate the disadvantaged into social networks.
REFERENCES Angelusz, Robert and Robert Tardos. (2001). #"Change and Stability in Social Network Resources: The Case of Hungary under Transformation.#" Pp. 297-323 in Nan Lin, Karen Cook, and Ronald S. Burt (eds.), Social Capital: Theory and Research. New York: Aldine de Gruyter. Coser, Rose Laub. (1975). # “The Complexity of Roles as a Seedbed of Individual Autonomy.#" Pp. 237-262 in Louis A. Coser (ed.), The Idea of Social Structure: Essays in Honor of Robert K. Merton. New York: Harcourt Brace Jovanovich. Curtis, James and Edward Grabb. 1992. # "Voluntary Association Activity in English Canada, French Canada, and the United States: A Multivariate Analysis.” Canadian Journal of Sociology 17(4): 371-388.
Erickson, Bonnie H. (1996). # "Culture, Class, and Connections.#" American Journal of Sociology 102: 217-251. Erickson, Bonnie H. (2001). # "Good Networks and Good Jobs: The Value of Social Capital to Employers and Employees." Pp. 127-158 in Nan Lin, Karen Cook, and Ronald S. Burt (eds.), Social Capital: Theory and Research. New York: Aldine de Gruyter. Erickson, Bonnie H. (2003). #"Social Networks: The Value of Variety." Contexts 2: 25-31. Erickson, Bonnie H. 2004. "The Distribution of Gendered Social Capital in Canada.#" Pp. 27-50 in Henk Flap and Beate Volker (eds.), Creation and Returns of Social Capital: A New Research Program. London and New York: Routledge. Erickson, Bonnie H. and T. A. Nosanchuk. 1984. "The Allocation of Esteem and Disesteem." American Sociological Review 49:648-658. Feld, Scott. 1982. "Social Structural Determinants of Similarity among Associates." American Sociological Review 47:797-801. Fischer, Claude S. (1982). To Dwell Among Friends: Personal Networks in Town and City. Chicago: University of Chcago Press. Ganzeboom, Harry B. and Donald J. Treiman. (1996). #"Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations.“ Social Science Research 25: 201-39. Gidengil, Elisabeth, Andre Blais, Richard Nadeau, and Neil Nevitte. (In press.) #"Women to the Left? Gender Differences in Political Beliefs and Policy Preferences.# In Manon Tremblay and Linda Trimble (eds.), Gender and Elections in Canada. Don Mills: Oxford University Press. Li, Peter S. (1992). "Race and Gender as Bases of Class Fractions and Earnings." Canadian Review of Sociology and Anthropology 29:488-510. Lin, Nan. (1999). 5Social Networks and Status Attainment.# " Annual Review of Sociology 25: 467-87. Lin, Nan and Mary Dumin. (1986). "Access to Occupations Through Social Ties.# " Social Networks 8: 365-85. Lin, Nan, Yang-Chih Fu, and Ray-May Hsung. (2001). #"The Position Generator: Measurement Techniques for Investigations of Social Capital.#” Pp. 57-81 in Nan Lin, Karen Cook, and Ronald S. Burt (eds.), Social Capital: Theory and Research. New York: Aldine de Gruyter. Hall, John R. (1992). # "The Capital(s) of Cultures: A Nonholistic Approach to Status Situations, Class, Gender, and Ethnicity.#” Pp.257-285 in Michele Lamont and Marcel Fournier (eds.), Cultivating Differences: Symbolic Boundaries and the Making of Inequality. Chicago: University of Chicago Press. Harvey, Andrew S., Katherine Marshall, and Judith A. Frederick. (1991). Where Does Time Go? Ottawa: Statistics Canada. McPherson, J. Miller and Lynn Smith-Lovin. (1987). "Homophily in Voluntary Associations." American Sociological Review 52:370-379. McPherson, J. Miller, Lynn Smith-Lovin, and James Cook. 2001. "Birds of a Feather: Homophily in Social Networks." Annual Review of Sociology 27: 415-44. McVey, Wayne W. and Warren E. Kalbach. (1995). Canadian Population. Toronto: Nelson. Mirowsky, John and Catherine E. Ross. (1999). "Well-Being Across the Life Course." Pp. 328-347 in Allan V. Horowitz and Teresa L. Scheid (eds.), A Handbook for the Study of
Mental Health. Cambridge: Cambridge University Press. Moore, Gwen. (1990). #"Structural Determinants of Men#’s and Women#‘s Personal Networks." American Sociological Review 55:726-35. Smith-Lovin, Lynn and J. Miller McPherson. (1993). "You are Who You Know: A Network Approach to Gender." Pp. 223-251 in Paula England (ed.), Theory on Gender/Feminism on Theory. New York: Aldine de Grader. Volker, Beate and Henk Flap. (1999). "Getting Ahead in the GDR.#" Acta Sociologica 37: 17-34.
Figure 1 The Election Survey Item
Here is a list of occupations. Please put a circle in the appropriate column if you know any men (column 1) or any women (column 2) in each of these occupations: Men Women Lawyer 1 2 Social worker 1 2 Carpenter 1 2 Tailor, dressmaker or furrier 1 2 Computer programmer 1 2 Security Guard 1 2 Cashier 1 2 Sales or marketing manager 1 2 Sewing machine operator 1 2 Delivery driver 1 2 Human resources manager 1 2 Janitor or caretaker 1 2 Pharmacist 1 2 Server (waiter or waitress) 1 2 Farmer 1 2
NOTE: This Figure gives the English language version of the questionnaire; a French version was also used in the survey, since Canada is bilingual. Figure 2: The Prestige and Gender Composition of Occupations in the Position Generator
OCCUPATION PRESTIGE % FEMALE
Lawyer 73 31
Pharmacist 64 56
HR Manager 60 47
Sales Manager 60 25
Social Worker 52 76
Computer Programmer 51 25
Tailor, furrier, dressmaker 40 86
Farmer 40 24 1 Carpenter 37
Cashier 34 86
9 Delivery Driver 31
Security Guard 30 20
Sewing Machine Operator 25 92
Janitor 25 32
Server 21 81
NOTE: occupational prestige scores come from Ganzeboom and Treiman (1996), and % female from the 1996 Census of Canada.
International Conference on Social Capital, May 28-30 2008
From Potential to Realization:
The Mobilization of Social Capital by Chinese Job Seekers
Yanjie Bian, University of Minnesota Xianbi Huang, University of Queensland
1 Theoretical Perspectives on Social Capital
Network membership as social capital: Putnam, Portes
Network structure as social capital: Coleman, Burt
Network resources as social capital: Lin, Erickson
2 Lin’s Theory on 2-Stage Process of Social Capital Mobilization
Access Mobilization
Initial Network Action position resources Extensity Contacts of ties (Tie strength)
3 Mobilizing Social Capital Johnson, LuAnne R. and David Knoke. 2005. “‘Skonk Works Here’: Activating Network Social Capital in Complex Collaborations.” Advances in Interdisciplinary Studies of Work Teams 10:243-262.
J SCi = ∑ R j p ji j=1
SCi = ego i’s social capital from the J alters in its ego- network pji = ego’s perceived probability of access to use alter j’s resources
Rj = total resources controlled by alter j that could be useful to ego i
4 Analytical Framework
Agency
Social Capital ? Social Capital Potential Realization
Network size Use of contact Network density Contact status Network resources Contact help
5 The Agency of Social Relations
Social capital as unintended consequences (Arrow)
Social capital from more or less a deliberate process (Granovetter, Lin, Burt)
Social Capital in Chinese Guanxi culture Cultivation: Fei’s “differentiated configuration” Maintenance: Bian’s “social eating” Adjustment: Everyday’s observation
6 Chinese Job Seekers as SR Agents
Search as a deliberate process Who: targeted persons of information & influence How: ties to connect & strategies to secure help What: information & more concrete favor
Job seekers as agents Interests: strong vs. weak Values: relationalist vs. individualist Agency variations: urgency and efforts
7 Analytic Strategy & Research Hypotheses
Two groups of job seekers
Position SC potential SC mobilization On-the-job seekers Yes Higher Lower urgency Off-the-job seekers No Lower Higher urgency Hypotheses H1: Job holders tend to have higher potential social capital than do laid-off workers H2: Laid-off workers tend to have higher mobilized social capital H3: The higher mobilized social capital, the better job search outcome 8 Research Design
Two groups of job seekers 1999 5-city household sample: about 4,752 workers from a general population sample: Probability sampling of residents Job seeking process to the last job 2000 Wuhan city laid-off worker sample: 621 laid-off workers Cluster sampling of SOE workers Job seeking process after being laid off
9 Table 1. Potential social capital
General Laid-off Work t-test Population Sample Sample
Mean S.D. Mean S.D. /Percent /Percent
Network size 27.87 39.15 20.71 24.34 >***
Network density .51 .26 .50 .27 -
Network resources 31.14 22.88 23.58 17.42 >***
Number of cases 4752 621
*p<.05, **p<.01, ***p<.001 10 Table 2. Mobilized social capital
General Population Laid-off Worker t-test Sample Sample Mean/Percent S.D. Mean/Percent S.D.
Use of contact .58 .49 .78 .42 <*** Number of contacts used .92 2.86 1.53 3.10 <*** Use of strong ties .35 .48 .51 .50 <*** Information obtained .32 .47 .45 .50 <*** Favor obtained .39 .49 .49 .50 <*** Contact status Administrator .26 .44 .30 .44 - In high work-unit rank .30 .46 .29 .46 - Professional/manager .22 .41 .14 .35 >*** Number of cases 4752 621 *p<.05, **p<.01, ***p<.001 11 Table 3. Social capital effects
Favor obtained Job match GW LW GW LW Network size -.002* -.001 -.001 -.001 (.998) (.999) (.999) (.999) Network density -.487*** -.635! -.215 -.112 (.614) (.530) (.807) (.894) Network resources .013*** .013* .008*** .006 (1.014) (1.013) (1.008) (1.006) Favor - - .318*** .732*** (1.375) (2.079) Control variables (not presented) Constant .337 -2.304 -.2.388*** 1.162 (1.400) (.100) (.097) (3.197) R square .070 .041 .105 .100 Number of cases 4385 621 4385 621 12 !p<.10, *p<.05, **p<.01, ***p<.001 Conclusions
Social capital mobilization is a deliberate process in which the agency of social relations is highly relevant The agency is reflected in the variation of urgency between an on-the-job seeker and an off-the-job seeker Urgency variation makes a difference in the amount of mobilized social capital Social capital has a higher lifting effect for laid-off workers on job matching
13 Q & A
Thank you!
14
The Effects of Network Social Capital on Personal and
Collective Outcomes in Japan: The JGSS 2003 Dataset
Ken'ichi Ikeda (Department of Social Psychology, The University of Tokyo)
[Draft] [Paper prepared for a conference on social capital held in May 29-30, 2008 at Academia Sinica, Taipei, Taiwan.]
Human beings are social animals. That is, communication is essential for social existence. Also connoted in the word "social" is that communication is buttressed by a social structural property which we call the “social network.” Thus, to study network effects is to study how human beings affect each other via communication. This is one of the essential goals of social science, and the recent focus of social capital studies. In line with this goal, this study uses an interdisciplinary approach to investigate several outcomes of social capital, especially those produced by utilizing social networks as resources. To this end, we will analyze a social network module of the JGSS2003, which is a Japanese version of nationwide General Social Survey conducted in 2003.
Social network and their outcomes Social networks are complex structures with multiple aspects, each of which can be useful for different outcomes. That is, social networks include different kinds of social capital. The research literature includes a rich array of findings linking a group of different network variables to different desirable results. However, scholars in both sociology and political science have made advances in generalizing and simplifying these findings by distinguishing two fundamentally different forms of social capital (Lin 2001; Putnam, 2000, chap 22; cf. Burt 2000). One, bonding social capital, is the capital found in stronger ties linking more similar people. The second, bridging social capital, is the capital found in weaker ties linking people who are more varied and less similar to each other. These two forms of social capital are thought to have distinctive kinds of benefits, with, for example,
1 bonding social capital a better source of social support while bridging social capital is a better source of access to scarce resources. This paper makes two contributions to the development of this topic. First, prior work is clearer and more consistent in conceptualizing types of social capital than in conceptualizing types of outcomes. Both sociologists and political scientists have clearly seen and analyzed the differences between strong and weak ties, or bonding and bridging social capital. But sociologists and political scientists have focused on different kinds of social capital outcomes. Sociologists like Nan Lin (2001a, 2001b) have stressed outcomes for individuals, distinguishing expressive from instrumental benefits. He defined social capital as “resources embedded in social networks accessed and used by actors for actions”, and especially investigated the relationship between instrumental action and social status achievement. Meanwhile political scientists have been more concerned with the capacity of social capital to develop forms of civic engagement like political participation. Among them we can discern two streams of research related to social networks. One focus has been on the informal and intimate social network that voters have around them, and the positive effects of this network on their political knowledge acquisition and on political participation via political discussion in the network (Ikeda & Huckfeldt, 2001, 2007(under review); Huckfeldt & Sprague, 1995). The other stream focuses on voluntary organizations, the origin of which goes back to the tradition of political participation studies, which was revived by Putnam (1993, 1995, 2000) and his contemporaries. It emphasizes that activity in voluntary organizations acts as a social intermediary, developing social networks which will produce a variety of individual as well as collective benefits, such as the promotion of individual well-being, the increase of political participation, the amelioration of performance of political/administrative institutions, the development of local economic activities, and increased local security and safety. We note that most of these benefits are collective in nature. Prior work usually includes just one kind of outcome and the kind of social capital thought to be most relevant. Examples include studies of strong ties and social support, weak ties and getting a job, and voluntary association activity and political participation. This approach assumes the truth of the key theoretical claims linking types of social capital to types of outcomes. This paper makes such claims testable by including different kinds of capital and different kinds of outcomes in the same study. This will be our second contribution.
2
Measuring network bonding and bridging properties and their implications We have long histories of studying different types of social network, i.e. network bonding and network bridging, which are kernels for bonding social capital and bridging social capital, respectively.
Network bonding Network bonding is essentially interaction with intimate others who provide social support, and contributes to the formation of social identity. Network bridging occurs with less intimate others (often acquaintances), who are typically from different social groups, and it serves as a social antenna for getting useful information which helps one better adaptation to the world. Empirical studies on social capital that measure bonding and bridging at the same time have not been conducted extensively yet (Burt, 2005; Some Internet community studies focus on both sides (e.g. Norris, 2004; Yuan & Gay, 2006)). Network bonding has been measured widely using a "network battery," the most typical of which is the "name generator." This method has been used in a variety of research contexts, the original version of which was mainly developed by Burt (1984). In this method, respondents are asked to name several others who are important to them (in discussing important matters, or to whom they can ask for support on important things) and then follow-up questions about their demographic or other properties, all of which are utilized for the analyses of the properties of one’s strong tie bonding network. In more applied studies in the field of political science, respondents are often asked about their political discussants using a name generator (for comparison of these two types of discussant, see Huckfeldt, 2001). There are numerous studies on what network bonding will bring about. This network provides several types of resource for social support (Plickert, Cote, & Wellman, 2007) and contributes to the well-being of the focal person emotionally or instrumentally. Especially a robust social support network is called “convoy”, which changes members of support sporadically but consistently guards people over their life course (Antonucci & Akiyama, 1987). Moreover, bonding networks often form an in-group and functions as a source of social identity. On the other hand, political studies show that bonding networks often work as an information gateway from the outer world, which forms an information environment on the focal person; i.e.
3 creates contexts of political talk and providing information bias (often ideological) for that person. As a result, bonding network is a strong determinant of political behavior (Huckfeldt & Sprague, 1995; Ikeda & Richey, 2005). Though utilizing a single network battery is useful in this line of research, there are limitations to focus only on just one type of network. As is known from a classic study by Fisher (1982), people have more than 10 bonding alters in their network (From Fisher’s finding in northern California in 1977, it is about 13 others). By only using a single network battery (and asking about 4 or 5 others), we may fail to cover a wide variety of bonding others that are contacted in daily life. Actually, people have plural bonding networks depending on their purposes; bonding others for getting important advice for life could be different from others with whom we go out with for fun; they may also be different from others who could give advice for job change. This means that one single name generator cannot cover the whole range of individual’s bonding network (cf. Burt, 1997; van der Poel, 1993). Thus, we need a survey measuring different types of network bonding at once. This multiplex network could include alters with whom individuals discuss important matters, political matters, occupational matters, etc. Ikeda & Boase (in preparation), analyzed such multiplex networks in Japan, showing that any type of bonding network brings about political participation, but that overlap between different types of networks decrease political participation.
Network bridging We have two traditions of measuring network bridging. One tradition of measuring network bridging is related to theories of civic culture and mass society, with a focus on intermediate social groups (Kornhauser, 1959; Rosenstone & Hansen, 1993; Verba, Nie, & Kim, 1978). In this tradition, social participation in voluntary associations is emphasized because it creates network bridging. By joining community associations, occupational associations, social-issue related groups, hobby groups, etc., through common interests and concerns, these intermediaries create a social space for connecting with weak ties who are typically with heterogeneous others, and for improving discussion skills, obtaining insights regarding differing opinions, and refining skills useful for organizing activities. This argument has been succeeded in recent years by social capital theories in the political science (Putnam, 1993, 2000; Ikeda & Richey, 2005; Ikeda, 2007; Park & Shin, 2005; Zukin et al., 2006). They focus on measuring intermediaries in the form of participation in voluntary associations, which is often
4 measured as the extent of activeness in participation multiplied by the number of affiliated associations. The logic of using the number of associations that an individual participates in is that the more we participate in different organizations, the more we have diverse interaction, which functions as network bridging (Kavanaugh et al., 2005). Another approach from the sociological tradition examines our reachability to people positioned in a various social strata1. Our chances of being exposed to socially heterogeneous others are not confined to those occurring in voluntary associations. Through uncountable/ unorganized situations in social life. Such contact may occur during daily consumer transactions, involvement in administrative services, sports, eating out, going to see a lawyer or home doctor, or other cultural activities. These chances in turn increase an individual’s vertical diversity of social access, as these others are often from different social strata, i.e. others are heterogeneous in the sense that your social location could be very different from their social location. The importance of this kind of bridging has been well shown through the use of position generators, developed by Nan Lin and others (Lin, Fu and Hsung 2001).
Focus on vertical social network diversity controlling other networks My task in this paper is to utilize all these different measurements of social networks at once and to try to examine their comparative functions in different social activities. The main goal of this paper is to focus on the effect of having vertical social network diversity. One reason is because the main purpose of this symposium is to focus on this point2. Yet, we have a large advantage covering the multiple types of network measurement as described above, because such coverage enables us to detect the net effects of vertical social network diversity after controlling other network indicators. In other words, if not controlled, we are not able to discern whether the effects of networks come from vertical social network diversity, the bonding nature of network, or voluntary organizational activities. By controlling them, we are able to reject alternative hypotheses. In this sense, this study has a framework for extracting the net effects of vertical social network
1 We will not mention about yet another approach which focuses on weak ties as bridging link (Granovetter, 1973; Burt, 2005), as the dataset the current paper will use does not have appropriate measurement on this point. 2 As the indicator on voluntary organization is not much elaborated in JGSS2003 survey which will be analyzed below (just asked about affiliation), detailed analyses on this area are not possible (see Ikeda & Richey(2005), and Ikeda(2007)). Also, multiplex network is closely investigated in Ikeda & Boase (in preparation).
5 diversity.
Target outcomes We will target three different types of human activities that might be affected by resources embedded in social networks. One is related to personal well-being. In social support studies, social networks are important for enhancing one's well-being, which is often discussed as effects of social networks on one’s perceived health or internal psychological conditions. Lin (2001b)’s “return of expressive actions” seems to correspond to this notion. Acock and Hurlbert (1993) targeted these outcomes using 1985 GSS data. They revealed that network bonding and bridging plays different roles on personal well-being depending on the respondent’s marital status (they included health status and life satisfaction into one well-being scale). Wellman and Gulia (1999) reviewed and concluded different kinds of social support had differentiated roles on psychological well-being (see also Johnson (1994)). Moreover, Putnam (2000, chap.20) also wrote a short review chapter on the positive association between social capital and health as well as happiness. Its unique contribution is that organizational participation is also a positive promotive factor to personal well-being. Though there are some nuanced differentiations on the effects of network bonding or bridging, we may generally assume the positive function of social network on personal well-being. Analogically to this generalization, vertical network diversity may provide some type of social support, possibly instrumental one, though we do not have any direct evidence on this type of network measurement. The second target outcome is related to individual social achievement; social networks give us a kind of capital or resource that helps us attain/elevate social status. Here, too, the effect is a benefit that individuals gain, but more of a socially competitive benefit as compared with well-being in the sense that it is ensured after (often potential) competition. In other words, it often has zero-sum property, in the sense that, in the most competitive case, once you achieved some social position, others are not able to occupy the same position. Then we may posit that social networks are more valuable for acquiring scarce resources (such as a good occupation) than they are for improving internal well-being. This conjecture is in line with Lin’s (2001b) notion that social networks are a “return of instrumental actions.” Burt (2005) showed compellingly that network bridging has a positive effect on this personal achievement. And in his reviews, Lin (2001a, Chap. 6) showed
6 the same is true for vertical social network diversity. Then in this sort of the target outcome, our task is to show the same effect as Lin’s after controlling personal network and voluntary organizational factors at once. Our final focus is on the social network as a determinant of collective output, such as political participation, which is one of the main targets of social capital studies in political science. Improvement in political participation is not an individual outcome via network capital, but rather it is a collective outcome. The fruits of social capital are not only individual outcomes such as well-being or social achievement, but also they could be collective outcomes. Political participation, for instance, is an action by individuals, but it has no clear benefit for individuals, i.e. participation does not directly improve personal political privilege nor does it directly produce economic benefit for a given individual. Instead, political participation improves general levels of political efficacy (Ikeda et al., 2008), gives higher system legitimacy, and at the same time, enables collective deliberation that is essential for democratic politics. In other words, even though participation itself is a result of personal decision, its outcome is collective in nature, contributing to social functioning. The measurement here is for each respondent, but it is an index for collective outcome 3 . Ikeda and Richey (2005) revealed Japanese “network capital” including personal network as well as involvement in voluntary organization affects positively on political participation. In another study, Ikeda & Kobayashi (2008 in press) show that there are clear positive correlations between vertical social network diversity and social/political participation. Again, Ikeda (2007) showed that vertical social network diversity contributed to increasing two different types of political participation: participation in governmental politics as well as electoral politics. This is because vertical social network diversity gives variety of routes to having a say on political process (an instrumental function). Regarding for political efficacy, as network diversity gives a bird’s eye or multifaceted viewpoint regarding societal context, the wider diversity will facilitate our understanding of the political arena, i.e. will work as a basic resource for informed citizen (informational or educational function). There is no precedent study on this point.
Based on these types of human activities, the dependent variables here will be: (1) perceived health condition (real life outcome) and life satisfaction
3 Lin (2001b) mentioned this aspect of social network outcome as “collective assets,” but did not develop its analysis.
7
(psychological outcome), referred to as personal well-being, (2) occupational status and annual income (real life outcome) as well as subjective social position in social strata (psychological outcome), referred to as social achievements, and (3) level of political participation (real life outcome) as well as political efficacy (psychological outcome), referred to as collective social achievements. When conducting the multi-variate regression analyses for the hypothesis testing below, we will use the same set of independent and control variables for the equations. This will facilitate comparisons between the results. Also it may reveal some unpredicted effect; for instance, having good health may facilitate political participation. On the other hand, we will not examine any inner causality in each type of the outcomes; i.e. causality from health to life satisfaction, from occupational status/ income to subjective social status, or political participation to political efficacy. Although these are possible, our focus was on the comparative effect of vertical social network diversity4.
Hypotheses and Research Questions Our hypotheses could be constructed based on certain types of human activities that are dependent and vertical social network diversity as independent variables. Given other factors, such as other network properties as well as demographic variables, are constant:
H1) Vertical social network diversity is positively related with personal well-being. H2) Vertical social network diversity promotes (H2-1) occupational status, (H2-2) annual income, and (H2-3) subjective status. H2) Vertical social network diversity augments (H3-1) political participation, and (H3-2) political efficacy.
Possible endogeneities There could be reverse causality regarding the two hypotheses above - the hypotheses that with the two non-psychological dependent variables - between vertical social network diversity and occupational status or political participation. Having higher occupational status may better allow one to access diverse resources in the social ladder; the higher position may make it easier to access to higher
4 In additional analyses, we checked these causal possibilities and found the effect of vertical network diversity remains the same.
8 position, as well as to lower social positions. Then the reverse causal possibility exists for H2-1. Also, those with high levels of political participation may have more chance to access people from different strata in the society than those with lower levels of participation simply because they know more weak ties. Thus again, possible reverse causality exists for H3-1. In order to examine these reversal possibilities, we will use 2SLS model approach using instrumental variables to exclude the possible endogeneity of the hypotheses (Asher, 1976). These supplemental analyses will add the validity of our analysis. In addition, we will investigate the reverse hypotheses for H2-1 and H3-1 with 2SLS models 5. It is informative to note that the above possibilities regarding reverse causality will not be relevant for psychological dependent variables, i.e. subjective social status and political efficacy. Only perceiving that one sits in a higher position in social strata seems not to be causally antecedent for expanding his/her vertical network. The same is true for perception of political efficacy. Moreover, having more annual income is also unlikely to lead directly to have vertically wider social network. Thus we will not check the possibility of reverse causality on H2-2, H2-3, and H3-2.
DATA Our data comes from the 2003 version of Japanese General Social Survey (JGSS-2003). The survey was conducted between October and November in 2003. The respondents were firstly interviewed face-to-face, and then received a self-administered questionnaire (A version or B version) that was filled in after the interview (the latter was collected a few days after by the interviewer). 3,663 respondents completed the face-to-face phase, and 1,706 completed questionnaire B version. More detailed information and the datasets is available at http://ssjda.iss.u-tokyo.ac.jp /gaiyo/0350g.html. Questions regarding occupations were asked in the face-to-face interview, and other questions in this paper were in the B version questionnaire. Included in part B were the network batteries - the name-generator measurements - that targeted three different types of personal networks: "important-matters" discussion networks, political discussion networks,
5 We may talk about another possible endogeneity between health (personal outcome) and vertical network diversity, i.e. there is a possibility that good health is facilitative to widen one’s access to diverse resource due that being healthy enables higher social activity. However, this is not our concern. What we will focus in terms of endogeneity is on the mutual causality of social factors, and not the causality between personal and social factors.
9 and occupationally related discussion networks (also measured overlaps among the networks). However, as the relationships among the network batteries are complicated due to the network member overlaps, the name generating part of the battery (i.e. to indicate four alters for each network and to point out the overlaps) was conducted in the face-to-face part of the interview. The network-batteries also had a position generator measurement as well as questions on affiliations and involvements in multiple voluntary associations in the B version.
Independent variable The vertical social network diversity was measured by a position-generator battery (Lin, Fu and Hsung 2001). In JGSS2003, we asked the respondents whether they have any acquaintance in a variety of occupations that varied by status. Out of 18 such statuses, we used 7 items which have prestige scores known in Japanese SSM (Social Strata and Mobility) nationwide survey in 1995. The occupations we chose were: “reporter, director or editor of a newspaper, a TV program or other media” (prestige score 52.2, having acquaintance in this category was 9.3% in JGSS2003), “medical doctor” (90.1, 41.9%), “small and medium-sized business entrepreneur” (68.9, 38.3%), “insurance salesperson” (44.3, 35.9%), “bank employee” (56.4, 35.4%), “factory worker” (45.2, 34.1%), and “computer programmer or data processing specialist” (66.3, 25.4%). We summed all the positive answers on these categories.
Dependent variables Health state (variable name; op5hlthz (reversed from the original)); We use a state of health variable that uses as a 5 point scale from bad health to good. Then, the higher the score, the more healthy. Satisfaction in life (satislif2): The variable was a score of the first component in a principal component analysis on five four-point scales; i.e. satisfaction in "place you live in", "your non-work activities", "your family life", "the current financial situation of your household", and "your friendships". The higher the score, the higher the life satisfaction. The correlation between health sate and satisfaction in life is 0.27. Occupational status (tp12job): A five point scale differentiated on the level of occupational status was created from the original categorical variable. The higher the score, the higher the social status. Annual income (szincomx): The higher the score, the higher the personal income (19
10
point scale). Correlation with occupational status is 0.67. Subjective social status (op10lvl (reversed from the original)): The higher the score, the higher the subjective social status (10 point scale). The exact wording of the question was: “ In our society there are groups of people that tend to be towards the top and those that are towards the bottom. Here we have a scale that runs from top to bottom. Where would you put yourself on this scale?” The correlations between this variable and occupational status/ annual income are 0.17 and 0.17, respectively. Political participation (polpartc): A scale was created by a simple count of the following activities committed by the respondent in the last 5 years; "casting a vote in an election", "participating in an activity by a residents' association or a neighbors association", "contact with an influential local person (meeting or writing a letter) by necessity", "contact with a politician or a government official by necessity", "visiting an assembly or a governmental agency for submitting a petition", "attending a meeting related to an election or politics", "assisting an election campaign (including supports of a candidate)", "participating in a civic movement or a local residents' campaign", "writing my name on a written petition", and "contributing or donating money". Political efficacy (poleff): We created political efficacy variable as a score of the first component in a principal component analysis on four four-point scales: "people like me don't have any say about what the government does", "politics and government are too complicated for me to understand what is going on", "many people vote at elections, so it doesn't matter if I don't", and "generally speaking, Diet members no longer consider the people once they are elected" (cf. Campbell et al., 1960; Rosenstone and Hansen, 1983). The higher the score, the higher the efficacious feeling. The correlation with political participation is 0.28.
Control variables Demographic variables: We used the following variables: an urban-rural scale (size of the city/town/village the respondent lives), gender (male-female), age, age squared, and education. There are three reasons to set demographic variables as control variables. First, they are indicators of one’s social location, which may affect social outcomes such as occupational achievement. For instance, gender gives socially stratified locations for both women and men, and these locations give men
11
more privilege to get better occupational position or to have more open opportunities to participate in political activities. The same is true for age and education, although we need to remember that the age is not linearly related with this advantage; elderly generations have disadvantages in their social locations. For this reason we will add to control variable age**2. Secondly, the size of cities or towns in which one lives may affect the advantage of social location. Inequality between urban residents and rural residents has been a politically serious issues in Japan for many years. Finally, as for personal well-being, demographic factors may not play an important role in social stratification differentiation, but may rather be more related to lifecycle differences, and especially age. Network variables: three types of network variables will be added as controls for excluding alternative hypotheses. (1) Size of personal bonding networks; From three name generators, we created three network size measurements (each has 0 to 4 values); size of the important matter discussion network, size of the political discussion network, and size of the occupational discussion network. These networks are obtained using the following lead sentences respectively; "I'm going to ask you about the people you often talk to. First, please think about the people with whom you talk about important matters or with whom you consult about your worries," "Now, please think of the people with whom you discuss Japanese politicians, elections or politics. You may include people with whom you occasionally talk about the above topics. You can also include the people in your replies to the earlier questions (i.e. important matter discussants)", and "Now please think of the people whom you consult about your job or whom you ask for advice concerning your job. You can also include the people in your replies to the earlier questions (i.e. important matter discussants or political discussants)" Correlations with the vertical social network diversity were 0.14, 0.21, and 0.28 respectively. By controlling these bonding network variables, we are able to examine the net effects of vertical social network diversity as a bridging network. (2) Multiplexity of personal networks: This measures the overlap of personal bonding networks. The correlation between the network diversity and the multiplexity is 0.19. (3) Voluntary associational membership: interaction that occurs in voluntary associations will provide chances to expand one’s vertical social network
12
diversity, as these interactions reach beyond daily bonding networks. We created a voluntary organization participation scale by summing up the following membership: "political associations", "trade associations", "social service group", "citizens' movement/ consumers' cooperative groups", "religious groups", sports groups and clubs", and "hobby groups and clubs (choir, photo-taking, hiking, etc.)". We will examine whether we could observe the effects of the vertical social network diversity even after controlling this large source of bridging network (the correlation between the two variables was 0.27).
Results The results in Table 1 showed consistent network bridging effects by the vertical social network diversity measurement. Firstly, the vertical social network diversity measured by the position generator has consistent positive effects on social activities. The more various acquaintances one has in the vertical social strata, the higher his/her occupational status and the higher income as well as the higher subjective social position that are attained. The same is true for the collective output; there is a positive effect on political participation. On the other hand, the result for personal outcome in terms of health or life satisfaction was not statistically significant. Thus, the result for HQ1 was not supportive; the social diversity was not related with personal outcome. Having socially broad acquaintances does not increase the personal benefit of social capital. However, H2-1 and H2-2 were both strongly supported, and H2-3 was also supported - social diversity had a clear impact on social achievement in real life as well as some impact on subjective social achievement. The results also support H3-1, but do not support H3-2. In other words, real-life collective output, i.e. political participation, was clearly affected by vertical social network diversity, but subjective collective output in terms of political efficacy was not. All in all the effect of social diversity appeared not to be strong in subjective outcome, but more in real world behavioral/ status outcome.
13
Table 1. Social Network as Social Capital and its Effects
dependent variables Personal outcome Personal social achievement Collective social achievement outcome in real psychological psychological psychological independent variables outcome in real life outcome in real life life outcome outcome outcome Life Occupation Annual Subjective Political Political health satisfaction al status income social status Participation Effica cy urban-rural 0.00 -0.04 0.08 0.15 -0.09 0.08 -0.03 gender (ma le-fema le) -0.02 0.14 -0.96 *** -3.43 *** 0.18 * -0.22 ** -0.13 * age 0.02 -0.06 *** 0.06 * 0.35 *** 0.04 * 0.11 *** 0.04 *** age**2 -0.0003 * 0.0008 *** -0.0004 -0.0035 *** -0.0003 + -0.0010 *** -0.0003 *** education 0.02 0.10 * 0.07 0.41 *** 0.23 *** -0.14 ** 0.12 *** size of important matter 0.01 0.10 * -0.01 -0.21 * 0.06 0.01 -0.04 discussion network size of political discussion 0.04 0.05 0.00 -0.08 0.07 + 0.24 *** 0.13 *** network size of occupa tiona l 0.00 0.01 0.01 0.29 ** -0.01 -0.02 -0.06 ** discussion network multiplexity of personal 0.02 0.04 0.00 -0.08 0.00 -0.01 0.03 networks voluntary organization 0.07 * 0.21 *** 0.18 *** -0.03 0.14 ** 0.62 *** 0.21 *** participation vertical social network 0.02 0.03 0.10 *** 0.18 *** 0.04 * 0.19 *** 0.02 diversity constant -2.41 *** -0.01 1.30 *** 2.51 * -7.86 *** -1.25 ** -1.32 *** N 1663 1663 748 836 1663 1663 1663 R**2 0.0657 *** 0.0673 *** 0.3267 *** 0.399 *** 0.058 *** 0.3504 *** 0.145 *** Weighted data from JGSS2003 **p<.01 *p<.05 +p<.10 All analyses are OLS. Entries are unstandardized coefficients.
Other findings not related to the hypotheses were as follows: (1) Social participation in voluntary organizations generally has consistent positive effects. It contributes to the development of personal well-being as well as status attainment (occupational and subjective). Moreover, it promotes more political participation and higher subjective political efficacy. (2) Regarding network bonding, each type of bonding has a differentiated and consistent effect. Important matter discussion network bonding is positively related with individual life satisfaction, which is in line with the point that Lin(2001a) emphasized regarding the effect of homophilous networks (important matter discussion network is known as homophilous in nature; McPerson, Smith-Lovin, & Lynn, 2001). On the other hand, annual income is a positive function of occupational discussion network ties. Political discussion network bonding has a positive impact on political participation and political efficacy.
Analyses on endogeneities The upper part of Table 2 shows the 2SLS results using the instrumental
14 variables for the vertical social network diversity. By using the variables in the dataset that are not related with the dependent variables, we created the instrumental variables for the network diversity. In this first-stage R**2 values were 0.19 when the occupational status was the dependent variables, and 0.14 when political participation was the dependent (see the upper half of the Appendix table for the details of the result6). This instrumental variable approach shows that the independent effect of the vertical social network diversity variable was genuine; the statistical effect remained constant in these analyses (though it was slightly marginally significant for the equation of occupational status; p=0.052). Going one step further, we did analyses on the reverse causal relationships, i.e. the vertical social network diversity variables as dependent and the occupational status and political participation variables as independent. Here, too, we adopted the instrumental variable approach (the first-stage R**2 values were 0.24 and 0.14, respectively; see the lower half of the Appendix table for the details of the results). The results are shown in the lower part of Table 2. By using the instrumental variable approach, the effect of occupational status remained, but the effect of political participation disappeared. Then, all in all, we can conclude; (1) The vertical social network diversity promotes occupational status positively, and vice versa. (2) However, the direction from the social diversity to political participation is one way and does not move from political participation to vertical social network diversity, possibly because political participation includes activities of a non-networking nature, such as voting act itself, or contributing money.
6 The selection of independent variables for the instruments is to find variables which correlate to the instrument but not to the 2SLS target variable; We chose these variables from a Pearson correlation matrix including the vertical network diversity, occupational status and political participation. The groups of the independent variables were from variables in demographic properties, opinions on politics and economy, personal network related variables, and interview record.
15
Table 2. 2SLS Approach to Network Diversity Effects Yellow shadow indicates that it is instrumental variable
Method OLS 2SLS OLS 2SLS dependent variables Occupa tiona l Occupational Political Political independent variables status status Participation Participation urban-rural 0.08 0.08 0.08 0.13 * gender (male-female) -0.96 *** -0.97 *** -0.22 ** -0.27 ** age 0.06 * 0.07 * 0.11 *** 0.13 *** age**2 -0.0004 -0.0006 + -0.0010 *** -0.0011 *** education 0.07 0.07 + -0.14 ** -0.14 ** size of important matter discussion network -0.01 -0.02 0.01 0.02 size of political discussion network 0.00 0.00 0.24 *** 0.27 *** size of occupational discussion network 0.01 0.01 -0.02 -0.02 multiplexity of personal networks 0.00 -0.01 -0.01 -0.03 voluntary organization participation 0.18 *** 0.23 *** 0.62 *** 0.69 *** vertical social network diversity 0.10 *** 0.11 + 0.19 *** 0.23 *** constant 1.30 *** 1.02 * -1.25 ** -1.74 *** N 748 727 1663 1566 R**2 0.3267 *** 0.296 *** 0.3504 *** 0.3097 ***
Method OLS 2SLS OLS 2SLS dependent variables Vertical social Vertical social Vertical social Vertica l social independent variables network diversity network diversity network diversity network diversity urban-rural 0.01 0.05 -0.05 -0.02 gender (ma le-female) 0.23 -0.07 -0.25 * -0.36 ** age 0.11** 0.13** 0.06** 0.13*** age**2 -0.0012 ** -0.0015 ** -0.0006 *** -0.0013 *** education 0.31** 0.32** 0.34*** 0.31*** size of important matter discussion network 0.11 0.15 0.05 0.10 size of political discussion network 0.21 * 0.22 * 0.05 0.21 *** size of occupational discussion network 0.18 * 0.18 * 0.29 *** 0.32 *** multiplexity of personal networks -0.12 + -0.14 + -0.06 -0.10 + occupational status 0.61 *** 0.63 *** political participation 0.43 *** 0.11 constant -2.58 ** -3.02 ** -0.72 -2.03 ** N 748 727 1663 1529 R**2 0.1508 *** 0.1013 *** 0.2318 *** 0.1477 *** Weighted data from JGSS2003 **p<.01 *p<.05 +p<.10 Entries are unstandardized coefficients.
Discussion and Conclusion By using a position generator as a scale for vertical social network diversity in this study we have elucidated two points that have not been well known. Firstly, the vertical diversity of social networks bring about positive collective social outcome in terms of political participation, as well as positive individual outcomes such as occupational achievement. These effects survived causal analyses. Secondly, the effects of vertical social diversity are clear even when controlling for other relevant network variables. Thus we are clearly able to affirm that these effects are not the product of bonding network size, nor size of intermediary organizational affiliations. Having a vertical social network brings people the unique social power to obtain individual and collective outcomes. More notable is that this effect is independent from that of intermediary affiliation. Analyses not shown here suggest to us that both of the variables have a mutually causal relationship. Even considering this fact, the vertical network diversity has its own effect other than that of intermediary association. In other words, whereas the core of the Putnamian social network notion was only focused on intermediary activities, that is not enough for the study of social capital. In the first place,
16 activities in voluntary association do not fill much time in our social lives. They only fill 4 to 5 hours a week, even in a country such as Holland that has relatively high levels of voluntary participation (Newton, 1999). On the other hand, it is very easy to show examples where daily contact with a diverse range of people affects our ability to expand our vertical social network, even though these activities are not based around the participation in voluntary groups. For example, in the workplace people often have lots of transactions with others who occupy upper or lower status positions. The same is true in our private life - we sporadically see doctors, lawyers, or professors from some upper strata, and we have often have the chance to contact people who are in somewhat lower strata for a variety of services, e.g. people in supermarkets, retailers, restaurants, dry cleaners, gas station etc.. This means our society is vertically stratified in the middle of everyday life, and we need this diversity to live comfortably. Moreover, those who have wide accessibility to a vertically diverse network also have a high chance of obtaining social achievement and engagement in collectively beneficial social activities. On what remains to be done in this study of these issues, two tasks can be pointed out. First, further analysis is needed that controls for network diversity within strong social ties, or controls for size among weak social ties. In this paper, we controlled only the size of strong ties as well as size or diversity of intermediary ties. These follow-up analyses will strengthen the points in this paper. Second, social networks are stratified by gender as revealed by Erickson and her colleagues (Erickson, 2004; Erickson & Miyata, 2007; Inamasu, this conference; Miyata, Ikeda & Kobayashi, 2008). This issue is lacking in this paper. We need to conduct further research on this issue with gender stratification in mind.
Acknowledgement (1) The Japanese General Social Surveys (JGSS) are designed and carried out at the Institute of Regional Studies at Osaka University of Commerce in collaboration with the Institute of Social Science at the University of Tokyo under the direction of Ichiro Tanioka, Michio Nitta, Hiroki Sato and Noriko Iwai with Project Manager, Minae Osawa. The project is financially assisted by Gakujutsu Frontier Grant from the Japanese Ministry of Education, Culture, Sports, Science and Technology for 1999-2003 academic years, and the datasets are compiled with cooperation from the SSJ Data Archive, Information Center for Social Science Research on Japan, Institute of Social
17
Science, the University of Tokyo. (2) The author expresses his thanks to Bonnie Erickson (University of Toronto) and Jeffrey Boase (University of Tokyo) for their valuable comments on the earlier version of this paper.
References Acock, Alan C., & Hurlbert, Jeanne, S. 1993 Social networks, marital status and well-being, Social Networks, 15, 309-334. Antonucci, Toni C., & Akiyama, Hiroko 1987 Social networks in adult life and a preliminary examination of the convoy model. Journal of Gerontology, 42, 519-527. Asher, Herbert B. 1976 Causal modeling. Beverly Hills, CA: Sage. Burt, Ronald S. 1984 Network items and the general social survey. Social Networks, 6, 293-339. Burt, Ronald S. 1997 A note on social capital and network content. Social Networks, 19, 355-373. Burt, Ronald S. 2000 The network structure of social capital. Research in Organizational Behavior, 22, 345-423. Burt, Ronald S. 2005 Brokerage and closure: An introduction to social capital. Oxford: Oxford University Press. Campbell, Angus, Converse, Philip E., Miller, Warren E., & Stokes, Donald E. 1960 The American voter. New York: Wiley. Erickson, Bonnie H. 2004. The Distribution of gendered social capital in Canada. (In) Henk Flap and Beate Volker (Eds.) Creation and returns of social capital. New York and London: Routledge. pp. 27-50. Erickson, Bonnie H. & Miyata, Kakuko 2007 Macro and micro gender structures: Gender stratification and social networks in Canada and Japan. mimeo. Fisher, Claude S. 1982 To dwell among friends: Personal networks in town and city. Chicago: The University of Chicago Press. Granovetter, Mark 1973 The strength of weak ties. Sociological Theory, 1, 201-233. Huckfeldt, Robert 1986 Politics in context: Assimilation and conflict in urban neighborhoods. New York: Agathon Press. Huckfeldt, Robert 2001 The social communication of political expertise. American Journal of Political Science, 45, 425-438. Huckfeldt, Robert, & Sprague, John 1995 Citizens, politics, and social communication. Cambridge, MA: Cambridge University Press.
18
Ikeda, Ken'ichi 2007 Political reality and social psychology: The dynamics of the Koizumi years. Bokutaku-sha (in Japanese). Ikeda, Ken'ichi & Huckfeldt, Robert 2001 Political communication and disagreement among citizens in Japan and the United States, Political Behavior, 23, 23-51. Ikeda, Ken'ichi & Huckfeldt, Robert 2007 (under review) Introduction (to the Section 2: Consequences of discussion and deliberation), (In) Ken'ichi Ikeda, Laura Morales, & Michael Wolf (Eds.) The role of political discussion in modern democracies in a comparative perspective. Ikeda, Ken'ichi & Kobayashi, Tetsuro (2008 in press) Making democracy work via the functioning of heterogeneous personal networks: An empirical analysis based on a Japanese election study. (In) Ray-May Hsung, Nan Lin, & Ronald Breiger (Eds.) Contexts of social capital: Social networks in markets, communities and families. London: Taylor & Francis. Ikeda, Ken'ichi, Kobayashi, Tetsuro, & Hoshimoto, Maasa (2008) Does political participation make a difference?: The relationship between political choice, civic engagement and political efficacy. Electoral Studies, 27, 77-88. Ikeda, Ken'ichi & Richey, Sean E. 2005 Japanese network capital: The impact of social networks on Japanese political participation. Political Behavior, 27, 239-260. Inamasu, Kazunari & Ikeda, Ken'ichi 2008 The effect of gendered social capital on political participation: From position generator of JES3 dataset. Paper prepared for a conference on social capital held in May 29-30, 2008 at Academia Sinica, Taipei, Taiwan. Johnson, Jefferey, C. 1994 Anthoropological contributions to the study of social networks: A review. (In) Stanley Wasserman, & Joseph Galaskiewicz (Eds.) Advances in social network analysis: Research in the social and behavioral sciences, Thousand Oaks, CA: Sage, Pp.113-151. Kavanaugh, Andrea L., & Reese, Debbie Denise, Carroll John M., Rosson, Mary Beth 2005 Weak Ties in Networked Communities. Information Society, 21, 119-131. Kornhauser, William 1959 The Politics of Mass Society. Glencoe, Ill.: Free Press. Lin, Nan 2001a Social capital: A theory of social structure and action. Cambridge, UK: Cambridge University Press. Lin, Nan 2001b Building a network theory of social capital. (In) Nan Lin, Karen Cook, & Ronald S. Burt (Eds.) Social capital: Theory and research. NewYork; Aldine de Gruyter. pp. 3-29. Lin, Nan, Fu, Yang-Chih, and Hsung, Ray-May (2001) The position generator: measurement techniques for investigations of social capital. (In) Nan Lin, Karen
19
Cook and Ronald S. Burt (Eds.) Social capital: theory and research. New York: Aldine de Gruyter, pp. 57-81. McPherson Miller, Smith-Lovin, Lynn and Cook, James M. 2001 Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415-444. Miyata，Kakuko, Ikeda, Ken'ichi and Kobayashi, Tetsuro 2008 The Internet, social capital, civic engagement, and gender in Japan, (In) Nan Lin and Bonnie H. Erickson (Eds.) Social capital: Advances in research. Oxford University Press. pp.206-233. Newton, Kenneth 1999 Social and political trust in established democracies. (In) Pippa Norris (Ed.) Critical citizens: Global support for democratic governance. Oxford: Oxford University Press. pp.169-187. Norris, Pippa 2004 The bridging and bonding role of online communities.(In) Philip N. Howard, & Steve Jones (Eds.) Society online: The Internet in context. Thousand Oaks, CA.: Sage, pp. 31-56. Park, Chong-Min, & Shin, Doh Chull 2005 Social capital and democratic citizenship: The case of South Korea. Japanese Journal of Political Science, 6, 63-85. Plickert, Gabriele, Co^te', Rochelle R., & Wellman, Barry 2007 It's not who you know, it's how you know them: Who exchanges what with whom? Social Networks, 29, 405-429. Putnam, Robert D. 1993 Making Democracy Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton University Press. Putnam, Robert D. 1995 Bowling Alone: American's Declining Social Capital. Journal of Democracy, 6, 65-77. Putnam, Robert D. 2000 Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster. Rosenstone, S. J., & Hansen, J. M. 1993 Mobilization, Participation, and democracy in America. New York: McMillan. Yuan, Y. Connie, & Gay, Geri 2006 Homophily of network ties and bonding and bridging social capital in computer-mediated distributed teams. Journal of Computer- Mediated Communication, 11(4), article 9. van der Poel, Mart G.M. 1993 Delineating personal support networks. Social Networks, 15, 49-70. Verba, Sidney, Nie, Norman and Kim, Jae-On 1978 Participation and Political Equality: A Seven-Nation Comparison. Cambridge, MA: Cambridge University Press. Wellman Barry, & Gulia, Milena 1999 The network basis of social support: A net
20
work is more than the sum of its ties. (In) Berry Wellman (Ed.) Networks in the global village: Life in contemporary communities. Boulder, CO: Westview. Pp. 83-118. Zukin, Cliff, Keeter, Scott, Andolina, Molly, Jenkins, Krista, and Delli Carpini, Michael X. 2006 A New Engagement?: Political Participation, Civic Life, and the Changing American Citizen. New York: Oxford University Press.
21
Appendix Table. Creating Instrumental Variables for 2SLS
Occupational Political 2SLS target variable 2SLS target variable status participation Vertical social Vertical social dependent variables dependent variables network network independent variables independent variables diversity diversity rural resident when R was in 15 years old -0.02 empolyment status of father when the R was 15 0.50 * labor union member 0.27 labor union member 0.18 education of father 0.10 education of father -0.10 number of family members 0.00 political ideology 0.02 political ideology 0.02 policy issue preference (welfare related) -0.03 perception of opportunity to improve family living political information exposure (scale) 0.64 *** -0.11 + standard recent changes in personal economy -0.16 recent changes in personal economy -0.18 + perception of J apanese economic climate 0.10 perception of J apanese economic climate 0.13 raio of extended family in personal network -0.76 *** ratio of job colleague in personal network 0.68 * ratio of close others in personal network 0.40 ** ratio of those who go out together in personal 0.37 * ratio of possible monetary support in personal 0.50 * ratio of possible monetary support in personal 0.49 * density in personal network (ocupational discussion) 0.51 ** density in personal network (ocupational discussion) 0.52 ** heterogeneity of gender in personal network 0.29 heterogeneity of gender in personal network -0.32 + heterogeneity of hobby in personal network -0.28 + educational heterogeneity in personal network 0.83 *** job heterogeneity in personal network 0.77 *** job heterogeneity in personal network 1.00 *** cooperativeness in the interview -0.01 constant 1.82 ** constant 1.45 * N 1566 N 1566 R-squared 0.1885 *** R-squared 0.1368 *** Vertical social Vertical social 2SLS target variable network 2SLS target variable network diversity diversity
dependent variables Occupational dependent variables Political independent variables status independent variables Participation
marital status 0.36 *** marital status 0.77 *** number of family members 0.08 ** size of city lived in 15 years old 0.05 rural resident 0.25 * perceived possibility to lose one's job 0.14 ** opinion on able political party (no such party) -0.47 *** opinion on a priority: personal benefit versus societal perceived degree of ease to find a similar job 0.16 * -0.14 * benefit support LDP (consevative political party) 0.10 support DPJ (liberan political party) -0.01 no political party support (3point scale) -0.07 support CGP (religeous political party) 0.21 support JCP (communist political party) 0.62 *** no political party support (3point scale) -0.16 * ratio of people who have managerial job status in 0.76 *** frequency of talk in personal network -0.18 *** personal network ratio of friends in personal network -0.31 + ratio of non-voters in personal network -0.45 + ratio of males in personal network 1.01 *** frequency of talk in personal network -0.16 ** cooperativeness in the interview -0.16 *** constant 0.61 + constant 2.77 *** N 727 N 1529 R-squared 0.237 *** R-squared 0.1446 *** note: All the equations are without multicollineality problem.
22
The effect of gendered social capital on political
participation: Using the Position Generator method
on the JES3 Dataset
Kazunori Inamasu & Ken’ichi Ikeda (Department of Social Psychology, The University of Tokyo)
[Draft] [Paper prepared for a conference on social capital held in May 29–30, 2008 at Academia Sinica, Taipei, Taiwan]
1. Introduction The importance of relationships with weak ties has been cited repeatedly since Granovetter’s famous work showed their advantages in job searches and other activities (Granovetter, 1973, 1974). For instance, Burt (1992) tried to capture the importance of weak ties by developing the “Structural Holes” concept. Putnam (2000) classified social capital as bonding (close relationship within groups) or bridging (weak ties between groups), and emphasized how the latter connection brings people in contact with heterogeneous others. However, relative to close relationships, weak ties are difficult to measure in social surveys. Respondents may be puzzled if they are asked about an acquaintance they occasionally meet, while they may easily imagine intimate friends. Considering this problem, Lin and colleagues (Lin & Dumin, 1986; Lin, Fu & Hsung, 2001) devised one of the most effective methods for measuring weak ties: the Position Generator. This method asks whether members of a variety of occupations are acquainted with respondents (prestige scores of these occupations are known in advance). The utility of the method has been tested in various countries, including the United States (Lin & Dumin, 1986), Taiwan (Lin et al, 2001), Canada (Erickson, 2004a), Hungary (Angelusz & Tardos, 2001), and Japan (Erickson & Miyata, 2004). Erickson (2004b) claimed that the Position Generator was a more effective measurement method than either the “resource generator” (Van der Gaag & Snijders, 2004), which asks whether there are acquaintances who offer some kind of resources, or the “network size” method,
1 which solicits the number of acquaintances. She argued that, in the latter case, it is difficult to measure network size accurately without excluding redundant networks; and, the former method is limited as well, because it lacks the most fundamental indicator of resources, information about the location of people in the social structure. On the other hand, because one’s occupation reflects one’s social role and resources in society, the position generator is a more general method for measuring the social capital of weak-tie relationships. One use of the Position Generator is to search for factors that influence people’s network formations. In such work the Position Generator is a dependent variable, and it is often predicted by demographic variables such as age, gender, education, work, or family. Also, because the Position Generator accompanies a hierarchy of occupational prestige scores, it may be regarded as a direct measure of social “capital” (Lin, Fu & Hsung, 2001) and used as an independent variable to explain personal or economic achievements, such as salary increases or promotions. It makes sense, for instance, to suggest that people with access to a higher class via their social networks are more likely to attain good positions than those without such access. Beyond personal achievements, a good measure of social capital, such as the Position Generator, can be used to examine a broad range of human behaviors, including political behaviors. Early work by Granovetter (1973) had already described the relationship between weak ties and the integration of communities. In recent studies, social capital has attracted the attention of political scientists as public goods. As “politics is a sideshow in the great circus of life” (Dahl, 1961), there are limited chances for ordinary people to participate in the political process. The Position Generator can measure such opportunities effectively, as it measures the chance structure of political participation by mobilization. Miyata, Ikeda and Kobayashi (2008), for instance, have used the Position Generator to explain participation in voluntary associations that support democracy (de Tocqueville, 1848/1945; Putnam, 2000). In this study, we examine the direct effects of network diversity on political participation, but not on the participation in intermediary groups that mediate between people and politics (Putnam, 2000).
2. Political participation in Japan Even though democracy requires the participation of people in the political process, turnout in Japanese national elections is not high at all, with about 60%
2 of eligible voters participating. Most postindustrial countries have a similar problem. Moreover, other forms of political participation are much less popular than voting. Rosenstone and Hansen (1993) remark that the majority of Americans do not participate in politics other than casting their votes in national elections, which is also true for Japan. However, this does not always mean that the voter is negligent. As Downs (1957) points out, the benefits of political participation are so few and insignificant that it is reasonable not to invest time, money, or mental energy in it. Apart from low levels of overall participation, differential levels of participation have been observed in such demographic factors as age, socioeconomic status, and gender. Specifically, there are gender differences in political behavior consistent with the long-lasting gender-role expectation that “politics is not an activity that women should do” (Pharr, 1981). She observed that most political activities are dominated by men, though voter turnout has been almost same in gender. However, because both men and women are largely uninterested in politics, the gender gap in political participation has been almost ignored publicly, even though sexism in offices has been often discussed. As a result, this issue is not ordinarily considered by voters, except when a Cabinet starts and the number of female cabinet members becomes a popular topic.
3. Political participation and social networks The concept of social capital can be used to address this issue. It is difficult to create political networks, acquire appropriate information, and engage in political activities during the course of a busy daily life. However, most people have informal social networks that are part of their daily lives. If their networks support them by keeping a lid on costs to access information, having contact with the government, or participating in political groups, they can easily get involved in politics. It may be that people’s nonpolitical networks act as a shortcut to political involvement, even if they often do not know they are indirectly participating in politics. Previous research has shown the effect of daily networks on political participation. Verba, Schlozman and Brady (1995) insisted that the influence of everyday contact with friends was greater than that of recruitment efforts by strangers in political organizations, whereas Richardson (1991) treated “koenkai”, candidate support clubs, as interpersonal networks and emphasized the mobilizing effect they have in Japanese politics. Huckfeldt’s work has
3 revealed that political social capital is accumulated as a byproduct of daily interactions, and it increases participation in politics (Huckfeldt & Sprague, 1993, 1995; La Due Lake & Huckfeldt, 1998). And, in a similar vein, after controlling for participation in the koenkai, Ikeda and Richey (2005) showed the positive effects of social networks on political engagement.
4. The effect of diversity of weak ties on political participation Weak social ties also affect political participation. Kotler-Berkowitz (2005) examined the effects of the diversity of networks in terms of race, class, and religion. His study shows that network diversity has a positive effect on political participation, independent of the influence of network size. In fact, the effect became clearer when network diversity was standardized by size. Because this survey asked respondents whether they had any acquaintances from different social categories, it is similar to the Position Generator in terms of measuring the diversity of weak ties. However, because differentiation according to race, class, and religion is not clear in Japanese society, we cannot use this technique in the same way. Therefore, we consider the Position Generator to be the best measure of weak social ties in Japan, especially as the scale has already been tested (see above). Moreover, we have another reason to use the Position Generator in this study. When researchers use the method, they often ask about male and female acquaintances separately, which allows us to address the problem of distribution of gendered social capital. In fact, Erickson has used the Position Generator to examine gender differences over the life course (Erickson, 2004a; Erickson & Miyata, 2004). Hence, the position generator method is a good approach to explore the gender gap in political participation.
5. Research questions and hypotheses The purpose of this paper is to demonstrate how people can get involved in politics through weak ties created in everyday life. More specifically, we examine how women who are alienated from politics owing to traditional constraints are encouraged to participate in political processes. Given that voters lack commitment to politics and the inequality in political participation threatens the legitimacy of democratic government, we maintain that it is crucial to reveal the role of daily networks in civic engagement. We examine participation in all political activities except for voting. We
4 divide participation into two types, electoral and governmental, following the argument presented by Rosenstone and Hansen (1993). Electoral participation includes activities related to elections, such as helping in an electoral campaign or participating in assemblies related to an election. Governmental participation refers to less-direct activities, such as contacting a local influential person or politicians, participating in civic movements, signing a petition, or making a donation. Because of the nature of these activities, it is likely that the diversity of occupational ranks of one’s informal social networks will have a positive effect on governmental participation rather than on electoral participation, which is limited to the narrow sense of electoral politics. In particular, we expect that female network diversity will have an impact only on governmental participation. Female network diversity, as defined here, does not mean network diversity of female respondents, but the occupational diversity of female acquaintances (of both male and female respondents). Of course, gender differences of the respondents should not be ignored. An earlier study found that same-sex networks are more diverse than cross-sex networks (Erickson, 2004a), and since network diversity is related to increased political participation, these findings suggest that male network of females versus of males, and also that by the female network of females versus of males should be examined separately. These considerations lead us to three hypotheses.
Hypothesis 1: Network diversity (as measured by the Position Generator) mainly promotes governmental political participation. Hypothesis 2: The diversity of female networks (as measured by the Position Generator) mainly increases governmental participation. Hypothesis 3: Dividing the sample by sex will make the effects of female networks on political participation clearer.
6. Method 6.1. Survey data Our study is based on one of the largest Japanese panel sampling surveys covering the whole country, JES3（Japanese Election Studies 3）. We use the 8th and 9th waves for this analysis. The surveys were conducted on September 1–10 and September 15–26, i.e. before and after the General Election, in 2005. Both of them employed face-to-face interviews and total of 2,134 respondents were
5 interviewed for the 8th wave. The response rate was 71.1% and effective sample size was 1,517. As for the 9th wave, the sample size was 1,735 (panel sample from the 8th wave, except for those who expressed strong objections to being re- interviewed). The response rate for the sample was 87.1% (N=1511). The JES3 project is supported by Specially Promoted Research on Science (Ministry of Education & Sciences), school years 2001–2005 (Principal investigator: Ken’ichi Ikeda, Investigators: Yoshiaki Kobayashi and Hiroshi Hirano).
6.2. Measurement Political participation We asked whether respondents had participated in any of 13 political activities in the last five years. There were five electoral participation items: “Participated in assemblies pertaining to an election or politics”, “Helped in an election campaign (such as supporting a candidate)”, “Tried to persuade others to vote for a particular candidate”, and “Helped the election campaign (related to a particular party)”. Governmental participation comprised eight items: “Contacted a local influential person”, “Contacted a politician/official”, “Went to an assembly or public office to petition or lobby”, “Participated in civic/resident movement”, “Signed a petition”, “Made donation or participated in fund-raising”, “Participated in a demonstration”, and “Spoke out about politics on the Internet”. We used categorical principal component analysis to classify the items and adopted the first and second principal component scores as measures of electoral and governmental participation, respectively. The Position Generator method Based on previous studies that used the Position Generator, we listed 23 occupations and asked respondents whether they have acquaintances who are in these occupations. Then, we used 16 items that have prestige scores from earlier research in Japan (Naoi & Seiyama, 1991; Umino, 2000) to create an additive scale. Details are described in the Results section (Table 1). Control variables Control variables were divided into three categories. The first category consists of demographic variables such as gender, age, education, year of residence, and city size. Each of these variables has been related to political participation in previous studies. The second set of variables is often used to explain participatory behavior
6 in the field of political science: political knowledge, political interest, access to political information in TV news and newspapers, and affiliation to koenkai. Political knowledge was measured by counting the number of governmental ministries that the respondents were able to list without assistance. To assess “political interest” we asked, “How regularly do you pay attention to the political situation?” and coded from 1 for “Never” to 4 for “Almost always”. Access to political information in TV news and newspapers was measured by counting the number of programs or papers that respondents usually watch or read. Affiliation to koenkai, Japanese traditional candidate support groups, was coded as a dummy variable, using 1 for participation and 0 for none. The third type of control variables were network variables other than diversity of weak ties determined by the Position Generator. To clarify the impact of the independent variable, the effect of strong ties needs to be distinguished from the effect of weak ties. Diversity of strong ties was measured by the number of occasions that respondents met close friends. To measure the number of weak ties, respondents were asked to report the number of New Year’s cards that they wrote last New Year, including those sent by e-mail.
7. Results 7.1. Descriptive statistics of Position Generator Descriptive statistics show large gender differences in networks (Table 1). Although various occupations are listed in male networks, female networks are largely limited to a few occupations, such as officer of local community, teacher of elementary school, nurse, owner of retail store, or waitress.
Table 1 Frequency table of position generator, sorted by gender
Male acquaintance Female acquaintance Prestige score Male female male female Physician 90.1 48.7% 42.5% 6.2% 8.5% Lawyer 87.3 14.1% 7.0% 0.5% 1.0% Owner of large company 87.3 11.8% 4.8% 0.5% 0.7% Member of the diet 74.9 15.2% 5.7% 1.2% 0.7% Officer of local community 67.2 67.6% 54.6% 33.5% 50.1% Manager of government office 67.2 4.6% 0.7% 1.2% 0.4% Teacher of elementary school 64.6 29.3% 25.4% 16.1% 27.9% Reporter 63.2 7.4% 3.5% 1.3% 1.7% Manager of local government office 60.4 33.2% 13.3% 7.0% 5.0% Nurse 59.7 10.6% 8.5% 32.4% 45.6% Police 57.9 26.9% 15.5% 0.8% 1.6% Programmer 54.6 12.5% 5.7% 3.4% 1.7% Owner of retail store 51.3 49.5% 36.1% 18.3% 33.8% Postman 46.2 23.7% 20.5% 1.6% 2.5% Guard 39.9 15.9% 8.6% 1.6% 1.6% Waiter/Waitress 38.1 8.9% 5.4% 8.9% 14.0%
7 In addition, it was confirmed that same-sex networks are richer than cross-sex networks, as in Erickson’s study (2004a), which we established by measuring “extensity” (Lin et al., 2001), that is, a count of the number of acquaintances in different occupations. The results in Table 2 show that the extensity of same-sex networks is higher than that of cross-sex networks. In this research, we employ extensity as a measure of weak-ties network diversity.
Table 2 Extensity of gendered networks Male Female Male networks mean 3.8 2.0 SD 3.2 1.7 Ｎ 744 767 Female networks mean 1.3 2.6 SD 1.7 2.5 Ｎ 744 767
7.2. Political participation Table 3 shows the percentage of respondents who participate in each political activity, according to gender. The findings do not show the traditional bias of males being more involved in politics than females. For instance, more females engage in some political activities, such as civic/resident movements. Signing petitions or making donations are as common among males as females. In other words, the proportion of governmental participation is relatively close in males and females, while there are large gender differences in that of electoral participation.
Table 3 Frequency table of political participation Items male female Contacted a local influential person 15.3% 6.8% Contacted a politician/official 9.4% 4.2% Went to an assembly or public office to petition or lobby 7.1% 2.6% Participated in assemblies pertaining to an election or politics 19.9% 14.6% Helped in an election campaign (such as supporting a candidate) 10.0% 6.1% Participated in civic/resident movement 6.6% 19.3% Signed a petition 18.7% 16.2% Made donation or participated in fund-raising 17.5% 16.2% Participated in a demonstration 1.1% 0.9% Spoke out about politics on the Internet 1.9% 0.5% Try to persuade others to vote for a particular candidate 20.6% 13.7% Went to the election rallies and street oratories 13.4% 9.5% Helped the election campaign (related to a particular party) 3.1%2.1%
8 The electoral/governmental participation scales were constructed with a principal component analysis using a tetrachoric correlation matrix because the items are binary coded. When we rotated the results with promax method, two types of political participation were extracted, as we predicted (Table 4). Then, we employed the principal component scores to create our dependent variables. The Pearson’s correlation coefficient of the scales is r=.54 (Component 1 is “electoral participation”, and component 2 is “governmental participation”).
Table 4 Principal component analysis for political participation (promax) Items component 1 component 2 Contacted a local influential person 0.28 0.10 Contacted a politician/official 0.32 0.06 Went to an assembly or public office to petition or lobby 0.14 0.23 Participated in assemblies pertaining to an election or politics0.370.06 Helped in an election campaign (such as supporting a candidate) 0.39 0.00 Participated in civic/resident movement 0.07 0.40 Signed a petition 0.01 0.38 Made donation or participated in fund-raising -0.05 0.43 Participated in a demonstration -0.14 0.53 Spoke out about politics on the Internet -0.03 0.35 Try to persuade others to vote for a paticular candidate 0.35 -0.03 Went to the election rallies and street oratories 0.41 -0.16 Helped the election campaign (related to a particular party) 0.45 -0.14 eigenvalue 4.75 3.06 Proportion 0.37 0.24 Comulative Proportion 0.37 0.60 N=1511
7.3. Does the diversity of weak ties affect political participation? The results of multiple regression analyses in Table 5 show positive effects of the extensity of male networks on electoral and governmental participation (p<.001). These results do not support Hypothesis 1 (network diversity measured by the Position Generator mainly increases governmental participation), because electoral participation is also clearly affected by extensity. However, female networks have marginal positive effects on governmental participation only (p<.1). This is consistent with Hypothesis 2.
Table 5 Regression analysis for two types of political participation
9 Electoral Governmental Dependent ：Political participation B B Gender -0.02 0.00 Age 0.00 0.00 Education -0.03 * -0.01 Residence years 0.01 -0.01 City size 0.00 -0.02 * Political knowledge 0.00 0.01 + Political interest 0.09 *** 0.05 *** Political information: News paper 0.04 0.06 * Political information: TV news 0.00 0.01 Candidate support club 0.47 *** 0.08 Extensity of male network 0.04 *** 0.02 *** Extensity of female network 0.01 0.02 + Diversity of strong ties 0.06 *** 0.02 * Size of weak ties 0.00 0.00 constant -0.27 ** -0.09 N 1373 1373 R-squared 0.29 0.17 +<.1 *<.05 **<.01 ***<.001 Demographic variables do not have strong effects on the two types of political participation. Higher education decreases electoral participation (p<.05) and the size of the residents’ city has a negative effect on governmental participation (p<.05). This former finding does not match previous studies in the United States and Europe, in which high levels of education are related to political participation. However, the results indicate that less-educated people who live in the countryside may participate in politics because of the influences of their community, which is typically observed in Japan (Kabashima, 1988; Richardson, 1991). In fact, the interaction term between education and city size has a marginal, positive effect on electoral participation (p<.1). Though Kabashima (1988) found a nonlinear relationship between age and political participation in Japan, we did not use the square of age as a control variable because it did not have significant effects on participation in our data. Among the second set of control variables, political interest significantly promotes both electoral and governmental participation (p<.001), whereas affiliation with candidate support clubs only increases electoral participation (p<.001). Richardson (1991) emphasized the impact of candidate support clubs as a particularly Japanese type of mobilization, however, in this study, participation in candidate support clubs only influenced electoral participation. This indicates that governmental participation is beyond traditional political activities. Although the diversity of strong ties increases both types of participation, after controlling for the extensity of weak ties, this extensity of weak ties does
10 not have any significant effect on participation. This result supports the idea that the Position Generator is a better measure of network as social capital than simply measuring network size. On the other hand, we recognize that the number of New Year’s cards one sends might not be an appropriate measure of weak ties, so we tested this with another multiple regression analysis using personal achievement (annual income) as a dependent variable (Table 6). We consider this a valid analysis because the more weak ties one has, the greater benefit of one’s network (Burt, 2005). This analysis shows that while the diversity of male networks promotes income (p<.01), the size of weak ties (number of New Year’s cards) also has a positive effect (p<.001). Thus, the reason the number of New Year’s cards fails to influence political participation is not because of a problem with the scale itself. To explain political participation, then, the occupational diversity of weak ties is superior to a simple measure of network size.
Table 6 Regression analysis for annual income Income B Gender 0.16 Age -0.03 *** Education 0.32 *** Residence years 0.16 ** City size 0.05 Extensity of male network 0.09 ** Extensity of female network 0.03 Diversity of strong ties 0.07 Size of weak ties 0.005 *** constant 2.34 *** N 992 R-squared 0.29 +<.1 *<.05 **<.01 ***<.001
Next, we divided the sample by sex and ran multiple regression analyses, similar to the ones presented in Table 5, to test Hypothesis 3. Table 7 shows effects of female network diversity that did not emerge in analysis of the overall sample. In the sample of females, we found not only the male networks but also the female networks have positive effects on electoral (p<.05) and governmental (p<.01) participation. This result supports Hypothesis 3. Even though females are said to be excluded from politics, there are ways to participate in politics through the same-sex networks.
Table 7 Regression analysis for two types of political participation, sorted by
11 gender Electoral participation Governmental participation male female male female BBBB Age 0.00 0.00 0.00 0.00 Education -0.04 + -0.02 -0.02 0.02 Residence years -0.01 0.02 -0.03 * 0.01 City size 0.00 0.00 -0.01 -0.02 * Political knowledge 0.00 0.00 0.01 0.00 Political interest 0.10 *** 0.07 *** 0.06 ** 0.01 Political information: News paper 0.04 0.04 0.07 ** 0.04 + Political information: TV news -0.01 0.01 0.00 0.01 Candidate support club 0.48 *** 0.49 *** 0.15 ** -0.01 Extensity of male network 0.04 *** 0.03 *** 0.03 *** 0.02 ** Extensity of female network 0.00 0.03 * 0.01 0.03 ** Diversity of strong ties 0.08 *** 0.03 * 0.02 + 0.01 Size of weak ties 0.00 0.00 0.00 0.00 + constant -0.38 ** -0.23 * -0.09 -0.14 N 671 ** 704 671 704 R-squared 0.32 0.25 0.19 0.17 +<.1 *<.05 **<.01 ***<.001
In addition, political interest and affiliation with candidate support clubs do not have any effect on governmental participation for females, which is interesting, because these variables have been associated with increased political participation in past studies. This indicates that governmental participation by females is related to activities that are believed to be nonpolitical.
7.4. Are there routes to politics through nonpolitical networks? Because occupations in the Position Generator are generally used to assess vertical social access, we used occupations with a variety of prestige scores. To examine more directly whether the informal social networks of daily life promote participation in politics, we constructed scales of occupations to distinguish between the effects of political and nonpolitical networks. To make measures of political and nonpolitical network diversity, we distinguished between seven political occupations, including “Executive above director in municipal office”, “Executive above director in prefectural office”, “Executive above director in governmental department”, “Mayor/village chief”, “Member of local assembly”, “Member of national assembly”, and “Manager of support group for politicians” (these are not included in the Position Generator). A list of nonpolitical occupations was made by excluding the seven political occupations and four others that are difficult to classify as either political or nonpolitical. This left 12 occupations from 23 on the original list. We used the numbers of acquaintances in the different occupations for each category as the measures of political and nonpolitical diversity.
12 Table 8 Effects of political or nonpolitical network on political participation Electoral Governmental Dependent ：Political participation B B Gender 0.02 0.02 Age 0.00 0.00 Education -0.04 ** -0.01 Residence years 0.01 -0.01 City size -0.02 -0.03 ** Political knowledge 0.00 0.00 Political interest 0.07 *** 0.04 ** Political information: News paper 0.04 0.06 * Political information: TV news 0.00 0.01 Candidate support club 0.34 *** 0.01 Extensity of political network 0.15 *** 0.07 *** Extensity of nonpolitical network 0.00 0.01 * Diversity of strong ties 0.05 *** 0.02 + Size of weak ties 0.00 0.00 constant -0.09 -0.02 N 1373 1373 R-squared 0.37 0.20 +<.1 *<.05 **<.01 ***<.001
The results in Table 8 show that the diversity of nonpolitical networks has a positive effect only on governmental participation (p<.05), while that of political networks has effects on both electoral and governmental participation (p<.001). Though it is natural that political networks have effects on political participation, the effects of nonpolitical networks (after controlling for the effects of political networks) is noteworthy. This indicates that even in the context of nonpolitical social networks, diversity of networks promotes governmental participation. With respect to gendered social capital, we asked whether the effects of female networks described above are due to political or nonpolitical networks. The results presented in Table 9 indicate that female networks are fairly nonpolitical. Gender differences in political occupations are larger than in our model shown in Table 1. The respondents having acquaintances in each occupation are fewer, except for the case of male networks for males.
Table 9 Frequency table of gendered political networks
Male network Female network Male female Male female Executive above director in municipal office 33.2% 13.3% 7.0% 5.0% Executive above director in prefectural office 12.1% 3.3% 2.4% 1.6% Executive above director in government department 4.6% 0.7% 1.2% 0.4% Mayor/village chief 16.0% 7.0% 0.5% 0.8% Member of local assembly 30.7% 15.5% 2.8% 2.6% Member of national assembly 15.2% 5.7% 1.2% 0.7% Manager of support group for politicians 14.7% 6.4% 1.3% 3.1%
13 The following results are similar to the results presented in Table 7. When we split samples by sex, nonpolitical participation increases governmental participation only for females (p<.01). This suggests that the effects of female networks represent the effects of nonpolitical networks.
Table 10 Effects of political or nonpolitical network on political participation, sorted by gender Electoral participation Governmental participation male female male female BBBB Age 0.00 0.00 + 0.00 0.00 Education -0.04 * -0.02 -0.03 + 0.03 Residence years -0.01 0.02 -0.03 * 0.01 City size -0.02 -0.01 -0.03 * -0.03 ** Political knowledge 0.00 0.00 0.01 + 0.00 Political interest 0.07 ** 0.07 ** 0.05 ** 0.01 Political information: News paper 0.04 0.04 0.07 ** 0.04 + Political information: TV news 0.00 0.01 0.01 0.01 + Candidate support club 0.37 *** 0.33 *** 0.09 * -0.08 + Extensity of political network 0.14 *** 0.16 *** 0.08 *** 0.07 ** Extensity of nonpolitical network -0.01 0.00 0.01 0.02 ** Diversity of strong ties 0.06 *** 0.03 * 0.01 0.02 + Size of weak ties 0.00 0.00 0.00 + 0.00 + constant -0.07 -0.08 0.07 -0.08 N 671 704 671 704 R-squared 0.40 0.33 0.24 0.18 +<.1 *<.05 **<.01 ***<.001 8. Discussion The diversity of weak ties invites people into politics. Our analyses revealed that the diversity of weak ties influences political participation, even after controlling for other variables that are routinely used in political science. Consistent with previous research demonstrating the importance of interpersonal networks as political social capital, our study shows that the diversity of weak ties (as measured by the Position Generator method) also promotes political participation. On the other hand, the results also demonstrated the dark side of social capital. Whereas Pharr (1981) examines female political participation by focusing on norms of gender roles, this study reveals that one reason why females are less active than males in political participation in Japan may be the biased distribution of social capital. Only the male networks have a positive effect on electoral participation. Although Putnam (2000) pointed out that bonding social capital has a negative externality, this research shows that gender differences in weak ties, such as bridging social capital, also lead to negative outcomes. In general, interpersonal networks give people an easy route to political participation. However, it is harder for females to participate in politics because their political interpersonal networks are poorer than males’.
14 The findings reported here, that female or nonpolitical networks promote governmental participation, suggest a way to narrow the gender gap in political behavior. If we consider routes to political participation as a one-way street, as Milbrath (1965) suggests, huge gender biases in the number of professional politicians means that females have fewer opportunities to participate in the political process. However, the effects of female or nonpolitical networks on governmental participation indicate that there may be an easy way for women to increase their participation in politics: increase their political social capital by having more female as well as male contacts in their social networks. In addition, this nonpolitical route should be effective for all voters who are alienated from politics. By increasing the diversity of one’s weak-tie relationships, participation in politics could be greatly facilitated. Thus, an increase in social capital among disaffected citizens could be an antidote to the poor level of political participation in Japan. Finally, whereas participation via a nonpolitical route (with little thought to politics) is advantageous if increased participation is only had through this nonpolitical network route, it would likely be insufficient to support democratic institutions. Further research should examine whether nonpolitical networks can become triggers for a wider range of political activities.
Reference Angelusz, Robert & Tardos, Robert 2001 “Change and stability in social network resources: the case of Hungary under transformation.” (In) Nan Lin, Karen Cook & Ronald S. Burt (Eds) Social Capital: Theory and Research. New York: Aldine de Gruyter, pp. 297–323. Burt, Ronald S. 1992 Structural Holes: The Social Structure of Competition. Cambridge, MA: Harvard University Press. Burt, Ronald S. 2005 Brokerage and Closure: An Introduction to Social Capital. Oxford, UK: Oxford University. Press Dahl, Robert A. 1961 Who Governs? Democracy and Power in an American City. New Haven, CT: Yale University Press. Downs, Anthony 1957 An Economic Theory of Democracy. New York: Harper & Brothers. Erickson, Bonnie. H. 2004a “The distribution of gendered social capital in Canada.” (In) Henk, Flap & Beate, Völker (Eds) Creation and returns of Social Capital. London: Routledge, pp. 27–50.
15 Erickson, Bonnie H. 2004b “A report on measuring the social capital in weak ties.” A Report Prepared for the Policy Research Initiative, Ottawa, Canada, March 31. Erickson, Bonnie H. & Miyata, K. 2004 “Macro and micro gender structures: gender stratification in Canada and Japan”. Paper presented at the annual meeting of the American Sociological Association. San Francisco, CA, Aug. 14. Granovetter, Mark. S. 1973 “The strength of weak ties,” American Journal of Sociology. 78, 6, 1360–1380. Granovetter, Mark. S.1974 Getting A Job: A Study of Contacts and Careers. Cambridge, MA: Harvard University Press. Huckfeldt, Robert & Sprague, John 1993. “Citizens, Contexts, and Politics,” (In) A.W. Finifter (Ed.), Political Science: The State of the Discipline II. Washington, D.C.: American Political Science Association, pp. 281–303. Huckfeldt, Robert & Sprague, John 1995 Citizens, Politics, and Social Communication: Information and Influence in an Election Campaign. New York: Cambridge University Press. Ikeda, Ken’ichi & Richey, Sean E. 2005 “Japanese network capital: The impact of social networks on Japanese political participation.” Political Behavior, 27, 239–260. Kabashima, Ikuo 1988 Political participation. Tokyo: The University of Tokyo Press (in Japanese). Kotler-Berkowitz, Laurence 2005 “Friends & politics: Linking diverse friendship networks to political participation.” (In) Alan S. Zuckerman (Ed.) The Social Logic of Politics: Personal Networks as Contexts for Political Behavior. Philadelphia, PA: Temple University Press, pp.152–170. La Due Lake, Ronald & Huckfeldt, Robert 1998 “Social capital, social networks, and political participation.” Political Psychology, 19, 567–584. Lin, Nan & Dumin, M. 1986 “Access to occupations through social ties.” Social Networks 8, 365–385. Lin, Nan, Fu, Yang-chih & Hsung, Ray-May 2001 “The position generator: measurement techniques for investigations of social capital.” (In) Nan Lin, Karen Cook & Ronald S. Burt (Eds) Social Capital: Theory and Research. New York: Aldine de Gruyter, pp. 57–81. Milbrath, Lester W. 1965 Political Participation: How and Why Do People Get Involved in Politics? Chicago: Rand McNally. Miyata，Kakuko, Ikeda, Ken'ichi and Kobayashi, Tetsuro(2008) “The internet,
16 social capital, civic engagement, and gender in Japan.” (In) Nan Lin and Bonnie H. Erickson (Eds.) Social Capital: An International Research Program. Oxford University Press. Pp.206-233. Naoi, Yu & Seiyama, Kazuo 1991 Class Structure of Contemporary Japan. Tokyo: The University of Tokyo Press (in Japanese). Pharr, Susan, J. 1981 Political Woman in Japan: The Search for Place in Political life. LA: University of California Press. Putnam, Robert D. 2000 Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster. Richardson, Bradley M. 1991 “The Japanese voter: Comparing the explanatory variables in electoral decisions.” (In) Scott Flanagan, Shinsaku Kohei, Ichiro Miyake, Bradley M. Richardson & Joji Watanuki The Japanese Voter. New Haven: Yale University Press, pp. 369–430. Rosenstone, Steven. J. & Hansen, John M. 1993 Mobilization, Participation, and Democracy in America. New York: Macmillan. de Tocqueville, Alexis 1848/1945 Democracy in America. New York: Vintage Books. Umino, Michio 2000 Sense of Unfairness and Political Consciousness. Tokyo: The University of Tokyo Press (in Japanese). Van der Gaag, Martin & Snijders, T. 2004 Proposals for the measurement of individual social capital. (In) Henk, Flap & Beate, Völker (Eds) Creation and returns of Social Capital. London: Routledge, pp. 199–217. Verba, Sidney, Schlozman. K. L. & Brady, H. E. 1995 Voice and equality: Civic voluntarism in American politics. Cambridge, MA: Harvard University Press.
17 各項「接觸的社會資本」測量對現職地位、工作類別、收入
與階級認同之影響
The Effect of the Measures of Access to Social Capital on Occupational
Status, Job Classification, Income, and Class Identification
黃毅志*
1
*台東大學教育系(所)教授
各項「接觸的社會資本」測量對現職地位、工作類別、收入
與階級認同之影響
摘要
在台灣社會中，就個人工作生涯的地位取得而言，除了職業地位之外，公私 部門、是否高科技機構、非典型工作與老闆階級所代表的工作類別也很重要，這 都可能影響到收入與階級認同；至於各項「接觸的社會資本」之測量，這包括林 南(Lin,2001)以及熊瑞梅與黃毅志(Hsung and Hwang,1992)所提出的測量，對現職地 位、工作類別與收入、階級認同有何影響？何者影響比較大？則是先前國內研究 所沒探討的。本研究運用 2004 年進行調查的「社會資本的建構與效應」全國性 大樣本資料做路徑分析，以探討上述問題；研究發現顯示：這兩項測量對現職地 位、進高科技機構工作機率、收入與階級認同都有正向影響，對於家屬工作、為 不同機構工作的非典型工作與失業的機率都有負向影響；不過熊瑞梅與黃毅志的 測量之影響都大於林南的測量，特別是在對收入的直接影響與對階級認同的影響 上。
關鍵詞：接觸的社會資本測量、工作類別、階級認同
2
The Effect of the Measures of Access to Social Capital on Occupational Status, Job Classification, Income, and Class Identification
Abstract
In terms of career status attainment in Taiwan, in addition to occupational status, public/private section, whether hi-tech institution, nonstandard job, and boss class or not are important as well. They might have an effect on income and class identification. There is no research in Taiwan to explore the effect of the measures of access to social capital that Lin (2001), and Hsung and Hwang (1992) developed on occupational status, job classification, income, and class identification and to analyze which one has more effect. The author uses the national data, the Construct and Effect of Social Capital Survey, to explore mentioned-above questions by path analysis. The main findings of this study show that both measures have positive effects on occupational status, the probability entering hi-tech institution, income, and class identification. By contrary, they have negative effects on the probability of household, nonstandard job, and unemployment. However, the effect of the measure Hsung and Hwang developed is higher than Lin’s, especially on the direct effect of income and on the effect of class identification.
Keywords: measures of access to social capital, job classification, class identification
3
自從林南提出社會資本對職業地位取得影響的理論，將結構面的社會網絡變
項帶入地位取得模式，補足傳統地位取得研究過度強調屬於個人特質的教育對職
業之影響後(Lin,Ensel and Vaughn,1981;Lin,2001)，從深具社會學特色的社會網
絡，來分析地位取得之因果機制，早已在歐美社會的階層化研究成為熱潮；在台
灣也曾做過不少研究( 孫清山、熊瑞梅，1985;熊瑞梅、黃毅志，1992；Hsung and
Hwang,1992;孫清山、黃毅志，1994，1995；陳至柔，1995;黃毅志,1996;Lin,Fu and
Hsung,2001)。然而相較於最近歐美社會與中國大陸的研究在這方面理論與方法
之持續發展(Lin and Bian, 1991;Lin, 2001,2003; Lai, Leung and Lin,1989; Bian,1997；Bian and Ang,1997；Baldi, 1998；Podolny and Baron ,1997； Fernandez and Weinberg , 1997；Fernandez, Castilla and Moore,2000； Fernandez and Fernandez-Mateo,2006; Castilla,2005；Smith,2005; Wegener, 1991;Yakubovich,2005)，最近台灣的研究顯得沉寂多的。在台灣社會中，就個 人工作生涯的地位取得而言，除了職業地位之外，公私部門、是否高科技機構、 非典型工作與老闆階級所代表的工作類別也很重要，這都可能影響到收入與自認 的主觀社會地位(相當於階級認同)（黃毅志，2001a，2001b）。至於各項「接觸 的社會資本」之測量，這包括 Lin(2001:76)以及熊瑞梅與黃毅志(Hsung and Hwang,1992)所提出的測量，對職業地位、工作類別與收入、階級認同有何影響？ 何者影響比較大？則是先前國內研究所沒探討的，這也就是本研究所要探討的研 究問題。本研究運用 2004 年進行調查的「社會資本的建構與效應」全國性大樣 本資料做路徑分析，以探討上述研究問題。
一、 文獻檢討
（一）勞力市場的工作類別 筆者(黃毅志，2001a，2001b)曾綜合國內外文獻檢討，首先根據「公私部門」， 將勞力市場分為兩類；接著針對私人部門就業機構員工數所代表的「公司大小」， 機構所在的行業為核心或邊陲「經濟部門」，以及在台灣很重要的是否為老闆之
4 「階級位置」(謝國雄，1989)，將私人部門的工作分為八類。這也就將所有就業 民眾的第一個(最初)工作，與現在工作的類別分為以下九類，以分析台灣勞力市 場上的代內工作生涯流動所涉及的勞力市場分隔現象 :
1.核心部門老闆階級 5.邊陲部門老闆階級 2.核心部門小公司受雇者 6.邊陲部門小公司受雇者 3.核心部門中公司受雇者 7.邊陲部門中公司受雇者 4.核心部門大公司受雇者 8.邊陲部門大公司受雇者 9.公家部門受雇者
上述研究採用 1992 年台灣地區社會變遷階層組資料做分析，研究發現顯 示：與私人部門相較，在公家部門工作，不但主觀社會地位、工作滿意度都較高， 並可能普遍存有內部勞力市場，工作穩定、福利優厚，教育年數對於職業的影響 也大，具有初級市場的特色，為許多人所嚮往；相對而言，私人部門則具次級市 場的特色；而且九類勞力市場間的流動障礙，主要存在於公私部門之間，也就構 成公私雙元分隔的勞力市場。不過最初在公家部門工作者，現為老闆的比率仍達 到 14%；最初在大公司工作的受雇者，現為老闆的比率更高達四分之一。這可歸 因於台灣有著許多規模不大，而容易開創與維持的小老闆就業機會 (謝國雄， 1989；許嘉猷、黃毅志，2002) ，特別是在服務業(熊瑞梅、黃毅志，1992；許 嘉猷、黃毅志，2002 )。當老闆，就算是小老闆，不但工作具自主性，又可免除 被雇主剝削，而有高收入(謝國雄，1989；孫清山、黃毅志，1995) ；即使某些 部門，如公家部門或某些大公司，存有內部勞力市場，也不一定能防止員工轉業 當老闆。 上述研究（黃毅志，2001b）發現並顯示，在公家部門工作者自認的主觀社 會地位較高，除了因為教育、職業地位與工作收入較高而提高主觀社會地位之 外，即使控制了教育、職業地位與收入，在公家部門工作者的主觀社會地位仍較 高，主觀社會地位當也反映社會大眾對自己社會地位（或聲望）的評價(Kluegel, SingletonⅡ and Starnes,1977:600；Weber,1978:302-307) ，也是重要的報酬 (return)(Lin,2001：76)，顯示公私部門本身即代表著一項階層區分；不過在控 制教育年數與現職地位後，公家部門的收入與私人部門沒有顯著差異。然而章英
5 華與黃毅志（2007）用 2005 年台灣地區社會變遷資料做分析，發現在控制教育 年數與現職地位後，公家部門的收入仍高於私人部門，更加凸顯公家部門的優 勢。綜合過去研究發現，或許意謂著，最近十幾年，公家部門呈現越來越佔優勢 的趨勢。 上述研究的工作分類並沒考量非典型工作（nonstandard job）的存在。根 據 Kalleberg et al.(2000)，對於美國近年快速成長的非正規典型之研究，非 典型工作的成長，意謂著原先有許多典型工作(standard job)存在，它指的是「為 固定就業機構所雇用的全職、持續性工作」；從事這些工作的員工，在雇方控制 下，以勞力換取薪資；而且往往受到政府保護，而提高就業安全，享有健康保險、 退休金------等福利；這很類似雙元勞力市場理論所說的初級市場之好工作 （Hodson and Kaufman, 1982:728-732; Smith,1990）。 Kalleberg et al.(2000)所研究的非典型工作的工作者，包括為固定就業機 構，所暫時雇用的全職、兼職員工；經常要隨工作來源不同，而為不同機構工作， 並受到控制的受雇者；以及實際上自己就是老闆，能控制自已工作的獨立工作 者，這主要是自營作業者。非典型工作往往是暫時、兼職的，而且欠缺政府的就 業安全保護，福利也不足，工作收入又低；這很類似這很類似雙元勞力市場理論 所說的次級市場之壞工作（Hodson and Kaufman, 1982:728-732; Smith,1990）。 Kalleberg 等人(2000)探討了美國社會中，各類非正規工作的成長，然而台 灣有許多為家裏事業工作，且往往沒領固定薪資的「家屬工作者」(章英華、傅 仰止， 2003:176)，收入可能較低，不過 他們是否大多屬於非正規工作，仍有 待進一步研究釐清。至於 Kalleberg 將自營作業之小老闆，也當做是非正規工作 者，然而在台灣當小老闆，往往有許多好處，而為許多人嚮往，看來不應視為非 正規工作。
至於近年台灣產業，包括工業與服務業都迅速轉型邁向資本、技術密集與
規模經濟，如工業中電機、資訊高科技業與高度專業服務業的發展（王振寰，
2007；林季平，2005；蔡明璋，2005），提供許多電機、資訊工程師的就業機會，
對高級人力的需求日增(章英華、黃毅志，2007)，促進高科技機構的興起；高科
技機構工作者的教育、現職地位、收入與階級認同很可能都高出非高科技機構工
作者許多；不過在控制教育年數與現職地位後，高科技機構工作者的收入與階級
6 認同是否仍高於非高科技機構工作者，仍有待進一步研究之釐清。
(二)社會資本對工作生涯的地位取得之影響 傳統的地位取得研究者將「教育是影響職業取得的最重要因素」之發現，詮 釋為階層化過程依照普同主義(universalism)運作，任用人才主要根據才能、學 識 (Blau and Duncan，1967)，這很接近經濟學人力資本論之論點：將教育視為 一項很重要的人力資本，認為高教育者是由於工作專業強、效率高、貢獻大，而 得到高報酬(Smith,1990)。不過這些論點都強調屬於個人特質的教育之影響，被 批評為結構面的分析不足 (Hodson and Kaufman, 1982；黃毅志，1992) 。 林南(Lin,2001)的社會資本理論，則將結構面的機制納入地位取得分析，所 強調的是：影響個人取得好工作，如高職業地位的工作之因素，除了教育之外， 結構層面的社會網絡中所存有的社會資本亦很重要；而社會資本指的是：一個人 透過社會網絡，所能直接或間接接觸到，並可能動用起來幫助達成行動目標之人 際資源；一個人社會資本越多，越能從他的親朋好友中得到幫助，越可能取得好 工作。社會資本還可進一步區分為「接觸的社會資本(access to social capital) 」與「動員的社會資本(mobilization of social capital)」；一個人 所結交的人越多越廣、社會地位越高(即接觸的社會資本越多) ，找工作時，往 往越可能動用高地位的人(即動員的社會資本越多)，而找到好工作。依此理論， 階層化之過程，絕非完全依照普同主義運作，講究社會關係，廣結網絡，對於找 到好工作也有很大的幫助；在找到工作後，對於未來的進一步發展，以及工作報 酬，如工作收入、工作福利與主觀社會地位，也可能有所助益。 不論在歐美或台灣，先前探討社會資本對於工作生涯的地位取得影響之研 究，大都以職業地位取得為分析焦點。然而就地位取得而言，除了職業之外，公 私部門與階級所代表的工作類別往往也很重要，這影響到工作收入、工作福利與 主觀社會地位------等等工作報酬(Lin and Bian,1991 ； Lin,2003 ； Bian,1997；黃毅志，2001a，2001b；孫清山、黃毅志，1995;)；不過先前台灣 卻只有少數研究探討社會資本對工作類別及工作報酬之影響（熊瑞梅、黃毅志， 1992;孫清山、黃毅志，1995;Lin,Fu and Hsung,2001），而且這些少數研究所探 討的工作類別都侷限於是否為老闆階級，而沒探討對公私部門、高科技機構、非
7 典型工作以至於沒工作的失業之變項，而近年台灣的失業率高漲（行政院主計 處，2008；曾敏傑，2001），探討社會資本對失業率的影響當是重要的研究問題。 而且先前台灣研究在探討社會資本對職業地位取得之影響時，大都探討「動員的 社會資本」之影響( 孫清山、熊瑞梅，1985; Hsung and Hwang,1992;孫清山、 黃毅志，1994；陳至柔，1995 ；黃毅志,1996;Lin et al.,2001)，只有少數研 究探討「接觸的社會資本」之影響(孫清山、黃毅志，1995；Hsung and Hwang,1992;Lin et al.,2001)。
(三)「接觸的社會資本」的理論概念之重要性與不同測量方法之比較 林南最初將結構面的社會資本納入地位取得分析之理論，把社會資本對職業 取得的影響歸成「社會資源命題」、「地位強度命題」與「連繫強度命題」這三個
命題（Lin，1982，1990，1995，2001）；當初的社會資源即林南後來所說的「動 員的社會資本」，主要指找到工作者介紹人（或幫助者）的職業地位(Lin,2001)。 這三個命題所能探討的研究對象，僅限於靠師長親友介紹或幫忙找到工作之民 眾，例如連繫強度命題所指出「用弱連繫較易於動用高地位之職業介紹人」之適 用範圍，僅限於靠人介紹取得職業之民眾，社會資源命題與地位強度命題亦是如
此。以熊瑞梅和黃毅志 (Hsung and Hwang,1992)之研究作說明，靠介紹取得初 職者，祇佔取得初職的樣本總數之 51.3%；靠介紹取得現職者更下降到佔現職樣 本 27.3%，下降的原因主要是：有許多人變成老闆（大都是小老闆），求職時用 不著介紹人。除了所能分析的對象有限外，能分析的樣本太小，也增添統計推論
的不確定性。而台灣要進入公家部門就職，往往要透過考試(黃毅志，2001a)， 並不需要親友介紹或幫忙。
「接觸的社會資本」所能探討的研究對象，就不限於靠介紹或幫忙找到工作
之民眾。熊瑞梅和黃毅志（1992）就曾探討「接觸的社會資本」對於台中都會區
初職為製造業受雇者（即黑手），在調查時為老闆（即頭家，大多為小老闆）機
率之影響，發現「接觸的社會資本」較高者，當老闆的機率較高。至於在台灣「接
觸的社會資本」是否會影響進公家部門的機會，還沒有看到有研究做過分析；不
過由於要進入公家部門就職，往往要透過考試，看來「接觸的社會資本」較高，
8 並不會提高進公家部門的機會。
至於「接觸的社會資本」之測量方法，林南(Lin,2001)在回顧許多文獻後，
建議用定位（position generator）法，請受訪者針對問卷上所呈現的許多職業名
稱圈選他們在求職時，有所認識的人任職之職業；「接觸的社會資本」之測量含
三個變項：1、有認識人任職的職業數(extensity of embedded resources)；2、認識人
最高的職業地位(upper reachability)；3 認識人最高職業地位與最低地位的差距
(heterogeneity)。認識人最高的職業地位很高，代表認識資源（財富、權力、聲望）
很高的人，而有助於提昇自己的社會階層；所選職業的數目、職業最高地位與最
低地位的差距很大，代表可得到各式各樣的人之幫忙。
林南等人（Lin et al.,2001）的研究用 1997 年的台灣地區社會變遷調查網絡
組資料作分析，先對上述三個變項做因素分析而得到一個因素，並命名「接觸的
社會資本」，並用其因素分數做進一步的迴歸分析，分析「接觸的社會資本」對
職業地位、收入之影響；研究發現顯示：就男性樣本而言，「接觸的社會資本」
越高，職業地位、收入越高；上述「接觸的社會資本」對收入的正向影響，除了
有「接觸的社會資本」越高，職業地位越高，進而提高收入之間接影響外，控制
職業地位後，仍有「接觸的社會資本」越高，收入越高的直接影響；筆者認為即
使兩人的職業一樣，「接觸的社會資本」較高者，仍可能由於「接觸的社會資本」
較高，有助於事業發展，而得到高收入。就女性樣本而言，「接觸的社會資本」
越高，收入越高（控制職業），但「接觸的社會資本」對於職業地位並無顯著影
響。林南等人並進一步分析就不同性別的老闆、家屬工作者與受雇者共 6 組樣本
而言，「接觸的社會資本」對收入的影響，不過由於各組的樣本往往太小，「接觸
的社會資本」對收入的影響只有在兩組達到顯著；而林南等人並沒分析應該也很
重要的「接觸的社會資本」對當老闆、家屬工作者或受雇者的機會之影響。
不過前述熊瑞梅和黃毅志（1992）所採用的測量與林南等人（Lin et al.,2001）
有所不同。熊瑞梅和黃毅志的測量考量到社會資本的另一向度，即 Flap(1989)所
言的「網絡中可能動員的資源總體」。如果有Ａ、Ｂ、Ｃ三個人，每人有認識人
9 任職的職業數都是５（參下表），Ａ認識人最高的職業地位及最高地位與最低地
位的差距顯然不如Ｂ、Ｃ；不過雖然Ｂ和Ｃ認識人最高的職業地位及最高地位與
最低地位的差距相等，但是很顯然的Ｃ的網絡成員中有許多高地位者，可能動員
的網絡資源總體最大，這卻是林南(Lin,2001)所建議的三個變項測量都顯現不出
來，因而熊瑞梅和黃毅志設計一個新的「接觸的社會資本」測量以代表 Flap 的
「一個人網絡中可能動員資源的整體」：可能動員網絡資源總體＝有認識人任職
的職業社經地位分數之總和，依此「接觸的社會資本」之測量，C 為
296(=75+70+70+50+31)而高出 B(205)許多。然而熊瑞梅和黃毅志在分析「接觸的
社會資本」對當小老闆機率的影響時，仍只用「有認識人任職的職業數」。
最高地位 第二 第三 第四 第五
A 35 分 34 33 32 31
B 75 34 33 32 31
C 75 70 70 50 31
熊瑞梅和黃毅志（Hsung and Hwang,1992）同年的另一篇論文則以有認識人
任職的職業社經地位分數之總和測量「接觸的社會資本」，並發現這項測量對於
所取得職業地位的正向影響比林南等人（Lin et al.,2001）的測量大一些；一個具
有建構效度(construct validity)的「接觸的社會資本」測量，與在理論上有重大關
連的變項，如所取得職業地位（Lin,2001），在實證分析當能呈顯出很清楚的關連
性 (依 Carmines and Zeller, 1979:22-24），此即熊瑞梅和黃毅志的測量之建構效度
要比林南等人（Lin et al.,2001）的測量高一些。Gaag,Snijders 與 Flap(2004)則建
議今後研究可用有認識人任職的職業聲望之總和來測量「接觸的社會資本」，他
們雖說是用職業聲望之總和來測量，然而實際分析是卻用到職業社經地位；此
外，Gaag 等人並沒對他們所建議的測量之建構效度做分析。
10
二.研究假設
依前面的文獻檢討，可提出下列研究假設（參見圖一）： 1.工作收入越高，階級認同越高；職業地位越高，階級認同越高(控制收入)；公 家部門的階級認同較高(控制收入，黃毅志，2001b)。 2.職業地位越高，工作收入越高；相較於在私人部門固定工作，老闆、公家部門 者，有較高的工作收入，家屬工作、非典型工作者有較低的工作收入(孫清山、 黃毅志，1995；章英華、黃毅志，2007；黃毅志，2001b；Kalleberg et al.,2000）。 3.接觸的社會資本接多者，越可能找到好工作，包括職業地位越高，越可能到高 科技機構工作、當老闆，越不可能從事家屬、非典型工作，越不可能失業 (Lin,2001)。 4.在控制職業地位與工作類別後，接觸的社會資本對工作收入仍有直接正影響 (Lin et al.,2001)。
職業地位
接觸的社會資本 工作收入 階級認同
工作類別
圖一:本研究的因果模型
說明：控制變項包括父親職業、族群、性別、出生年次等背景變項 以及現工作地都市化程度、教育年數。
11
三.研究方法
（一）資料來源 根據筆者所加入的跨國社會資本比較追縱調查研究整合型計劃，在台灣地區 的第一次調查（2004 年）所蒐集之全國代表性大樣本（N=3281）資料作分析， 以檢證圖一之因果模型。
(二)變項測量 1. 接觸的社會資本 根據 Position Generator 法先選出 22 項職業納入問卷中，再請受訪者 回答在找現在職位時有認識的人就職之職業，這 22 項職業的職業聲望與社經地 位，依 Ganzeboom 與 Treiman(1996)之國際量表，分別是下列括弧外與括弧內 的數字：
(1)護士 54（43） (2)作家 58（65） (3)農民 38(23) (4)律師 73（85） (5)中學老師 60（69） (6)褓母 23（25） (7)清潔工 21（16） (8)人事主管 60（69） (9)大公司行政助理 (10)美髲師 32（29）(11)會計 49（51） (l2)警衛 (保全人 49（54） 員)30（40） (13)生產部門經理 63 (14)工廠作業員 (15)電腦程式設計師 (16)櫃檯接待 38（52） （67） 34（31） 51（71） (17)立法委員 64（77） (18)計程車司機 (19)大學教授 78（77）(20)搬運工 20（30） 31（30） (21)警察 40（50） (22)大企業老闆 70（70）
至於測量「接觸的社會資本」方法，首先參考林南等人（Lin et al.,2001）， 先測量(1)有認識人任職的職業數，(2)認識人最高的職業地位，以及(3)認識人最 高職業地位與最低地位的差距(即認識人職業地位全距)，這三個變項，再對這三 個變項做探索性因素分析而得到一個因素，再用其因素分數測量「原接觸的社會 資本（1）」做進一步的迴歸分析；因素分數為標準分數，平均數0，標準差 1。
12 林南等人原先用職業聲望測職業地位，不過 Ganzeboom 與 Treiman 的職業聲望 量表之建立，雖採用 1988 年的職業分類，然而各類職業的聲望分數，卻仍根據 早年的 Treiman（1977）聲望分數，而得到許多不合理的結果，如計算機程式設 計師聲望 51 分，低於計算機助理（53），本研究也就用社經地位測量職業地位； 本研究所做的進一步分析也顯示用職業聲望測量職業地位，所得到前述三個變項 的因素分數在迴歸分析中對職業地位、工作類別、階級認同的整體影響低於用社
經地位測量職業地位（由於篇幅限制，表格省略）。 本研究也根據熊瑞梅和黃毅志（Hsung and Hwang,1992）以有認識人任職的
職業社經地位分數之總和測量「原接觸的社會資本（2）」，整體樣本在這項測量
最高分 1124(N=16)，即 22 項職業都有認識的人就職者，最低分 0，即 22 項職業
都沒有認識的人就職者(N=150)，平均數為 323.6，標準差 265.9；用社經地位測
「原接觸的社會資本（1）」，也便於比較它與「原接觸的社會資本（2）」的建構
效度，若用職業聲望測「原接觸的社會資本（1）」，若發現它的建構效度低於「原
接觸的社會資本（2）」，這有可能只是因為職業聲望測量不佳所致。本研究並參
考 Gaag 等人(2004)的建議，以有認識人任職的職業聲望之總和來測量「原接觸
的社會資本（3）」，這項測量最高分 1036，最低分 0，平均數 315.0，標準差 245.4；
由於職業聲望測量不佳，可預期它的建構效度低於「原接觸的社會資本（2）」。
這項調查也用相同的 22 項職業問受訪者「現在」，而非「找現在職位時」有
認識的人就職之職業，本研究也用上述方法得到「現接觸的社會資本（1）」、「現
接觸的社會資本（2）」這兩項測量；「現接觸的社會資本（2）」最高分 1124（N=23），
即 22 項職業都有認識的人就職者，最低分 0（N=114），即 22 項職業都沒有認識
的人就職者，標準差為 285.1，平均數為 407.2，比找現在工作時的平均「原接觸
的社會資本（2）」(323.6)高出不少；並在迴歸分析中比較這兩項測量對工作收入、
階級認同的影響；由於這兩個變項發生在找現在工作前，所以不能用來分析對現
職地位、工作類別的影響。又由於研究結果顯示：「原接觸的社會資本（3）」測
量與其它變項的關連低於「原接觸的社會資本（2）」，這當可歸因於職業聲望測
13 量不佳，也就不再分析「現接觸的社會資本（3）」的影響，以簡化分析。
2.六類工作類別 根據文獻檢討將接受調查時台灣地區民眾的現在工作類別分成以下六類： (1)老闆：在 2005 年時，台灣的老闆有 64%沒雇用員工，31%雇用員工 1-9 人， 而絕大多數是小老闆(章英華、黃毅志，2007)。 (2)在私人部門固定工作 (3)在公家部門固定工作 (4)家屬工作 (5)隨工作來源不同而為不同機構工作的非典型工作 (6)沒有工作的失業者：本研究為了探討社會資本對找工作的影響，也將失業者 納入分析。 本研究並對台灣地區民眾的現在工作類別依是否屬於高科技機構做區分： 3.高科技機構：請受訪者填答「您現在工作的公司（機關）屬於高科技嗎」、「您 現在工作的公司（機關）有沒有內部網路系統」，以第一題答是且第二題答有 者為高科技機構，否則為非高科技機構；如此的高科技機構測量與教育年數、 現職地位與階級認同的關連性(Eta)，都比只以第一題答是來界定高科技機構 的測量高出不少(表略)，而建構效度較高。很可能有許多受訪者雖主觀認定 現在工作的公司（機關）屬於高科技，然而事實上公司（機關）並沒有內部 網路系統，而並不屬於高科技，因而只以第一題來界定高科技機構的測量之 建構效度也就較低。 4.現職地位：根據 Ganzeboom 與 Treiman( 1996)的國際職業量表之四碼社經地 位做測量。 5.工作收入：由於收入採等級尺度的問卷，本研究也就依各組的組中點測量收 入，如 9-10 萬元組的組中點為 9.5 萬，使其合乎迴歸分析的尺度設定；不 過收入最高的等級為 30 萬元以上，並無組中點，也就以前兩級每級差距 5 萬元做測量依據，由於 30 萬元以上這組收入比前一級 25-30 萬元(組中點
14 27.5 萬)多一級，收入也就算是差 5 萬元， 而以 32.5 萬元計，收入 30 萬元 以上者只佔 0.3%，應不會給收入測量帶來多少誤差。在迴歸分析時，參考 一般作法，如 Liu 與 Sakamoto(2002)，Zeng 與 Xie（2004），對收入取對數做
為測量；收入取對數的 R2 比未取對數要高出許多（表略）。 6.階級認同：以「假如把社會上所有的人分成上層、中上層、中層、中下層和 下層階級，您認為自己是屬於哪個階級」的問卷，依回答階級的高低，分別 給5到1等，數字越高，代表階級認同越高。 7.工作福利數：依受訪者對「您現在工作有沒有提供下列福利給您？ (1)退休金（2）勞/健保（3）分紅配股」之回答，以提供的福利數做測量。 8.控制變項 (1)父親職業：同現職地位測量。 (2)族群：由於原住民樣本太少(N=13)而當做缺失值(missing value)處理，並 不納入分析，而族群分台灣閩南、台灣客家與外省人，迴歸分析時以台灣閩 南為對照組。 (3)性別：迴歸分析時以女性為 0 當對照組。 (4)出生年次：以受訪者所填出生年次(民國)做測量。而依據相關研究，出生 年次與工作收入具有先升後降的二次函數關係（孫清山、黃毅志，1995； 章 英華、黃毅志，2007），因此本研究進一步納入出生年次的平方項作分析。 (5)現工作地都市化程度：用行政院主計處（1983）所編的「中華民國統計地 區標準分類」作測量，將工作地的鄉鎮市依都市化程度分成八個等級，都市 化最高者給 8，最低給 1。
(6)教育年數：而為了估計教育年數對因果模型中各變項之影響，將受訪者教
育程度轉換成大專以上教育年數與中小學教育年數兩個教育年數變項；例如
大學畢業者，接受的中小學教育年數為 12 年，大專以上教育年數為 4 年，高
中畢業者，中小學教育年數為 12 年，大專以上教育年數為 0 年；若大專以上
教育年數的影響(b)不等於中小學教育年數的影響，則教育年數就具有非直線
影響(Featherman and Hauser,1978；Zeng and Xie,2004) 。 （三）分析方法 本研究根據因果模型，進行迴歸分析，並檢證相關假設，此即路徑分析；當
15 依變項為順序尺度以上，如收入時，進行 OLS 迴歸分析；當依變項為名目尺度， 如工作類別時，則進行邏輯迴歸分析。在 OLS 迴歸分析中，為了要比較各自變項 的影響，也就列出標準化迴歸係數(β)，Zhou(2005)也是用標準化迴歸係數(β) 做分析，Messner 等人（2004）則用結構方程模式(SEM)的標準化係數(β)做路 徑分析。 四、研究發現
（一）「接觸的社會資本變項」與階層變項關連雙變項分析 1. 「接觸的社會資本變項」與階層變項相關分析
從表 1 可看到所有「接觸的社會資本變項」與主客觀地位取得階層變項的 相關（r）都達顯著（P<.05）。「原接觸的社會資本（1）」「原接觸的社會資本（1）」 的三個成份：(1)「原認識人最高的職業地位」，以及(2)「原認識人職業地位全 距」，(3)「原認識人任職的職業數」，與「原接觸的社會資本（1）」的相關（r） 分別為.926.966、.874，以「原認識人任職的職業數」的相關最低；這三個成份 與「原接觸的社會資本（2）」的相關分別為.706、.769、.984，反而以「原認識 人任職的職業數」的相關最高，還高過「原接觸的社會資本（1）」與「原接觸的 社會資本（2）」的相關（.883）；同樣以認識人職業地位總合做測量的「原接觸 的社會資本（2）」與「原接觸的社會資本（3）」的相關更是高達.987。三項「原 接觸的社會資本」測量與職業地位、工作收入、階級認同的相關，都以「原接觸 的社會資本（2）」最高，由於職業聲望測量不佳，「原接觸的社會資本（3）」的 相關也就都低於「原接觸的社會資本（2）」；「原接觸的社會資本（1）」與職業地 位的相關（.374）略高於「原接觸的社會資本（3）」（.359），不過與收入、階級 認同的相關（.320、.287）略低於「原接觸的社會資本（3）」（.332；.292）。「原 認識人任職的職業數」與「原接觸的社會資本（2）」的相關雖然很高，不過它與 職業地位、收入、階級認同的整體相關，都低於「原接觸的社會資本（1）」、「原 接觸的社會資本（2）」不少。 由於「原接觸的社會資本（3）」的職業聲望測量不佳，進一步的迴歸分析
又顯示它對於現職地位的影響（β）低於「原接觸的社會資本（2）」(見表4)， 對於工作類別、收入與階級認同的影響也都低於「原接觸的社會資本（2）」（表
16 略），本文也就不再分析「現接觸的社會資本（3）」與現職地位、工作類別、收 入與階級認同的關連性，以及「原接觸的社會資本（3）」與工作類別的雙變項關 連性，以簡化分析。 從表 1 還可看到「現接觸的社會資本（2）」與收入、階級認同的相關分別 為.376、.352，仍都高於「現接觸的社會資本（1）」的.324、.312不少。 綜合以上分析，「原接觸的社會資本（2）」與現職地位、收入、階級認同的相 關都高於「原接觸的社會資本（1）」；「現接觸的社會資本（2）」與收入、階級認 同的相關都高於「現接觸的社會資本（1）」；就此初步的雙變項分析而言，「接觸 的社會資本（2）」的建構效度高於「接觸的社會資本（1）」。
17 表 1 各項「接觸的社會資本」變項測量與階層變項相關係數（r）表
原認識人 原認識人 原認識人 原接觸的 原接觸的 原接觸的 現認識人 現認識人 現認識人 現接觸的 現接觸的 現職 月收入 階級 最高職業 職業地位 任職的 社會資本 社會資本 社會資本 最高職業 職業地位 任職的 社會資本 社會資本 地位 （萬元） 認同 地位 全距 職業數 （1） （2） （3） 地位 全距 職業數 （1） （2）
原認識人 1 最高職業地位
原認識人 .892* 1 職業地位全距 （N=3089）
原認識人 .660* .772* 1 任職的職業數 （3073） （3073）
原接觸的 .926* .966* .874* 1 社會資本（1） （3073） （3073） （3073）
原接觸的 .706* .769* .984* .883* 1 社會資本（2） （3073） （3073） （3223） （3073）
原接觸的 .695* .774* .983* .881* .987* 1 社會資本（3） （3073） （3073） （3223） （3073） （3223） .402* .323* .323* .374* .376* .359* 現職地位 1 （3084） （3084） （3217） （3068） （3217） （3217） 現認識人 .735* .665* .522* .698* .552* .548* .371* 1 最高職業地位 （3062） （3062） （3111） （3046） （3111） （3111） （3153）
現認識人 .665* .727* .596* .721* .594* .601* .320* .899* 1 職業地位全距 （3062） （3062） （3111） （3046） （3111） （3111） （3153） （3160）
現認識人 .584* .660* .795* .731* .785* .789* .354* .676* .782* 1 任職的職業數 （3085） （3085） （3217） （3069） （3217） （3217） （3264） （3157） （3157）
現接觸的 .713* .737* .680* .773* .687* .689* .370* .930* .968* .880* 1 社會資本（1） （3060） （3060） （3109） （3044） （3109） （3109） （3150） （3157） （3157） （3157）
現接觸的 .620* .660* .789* .742* .802* .797* .405* .716* .776* .984* .886* 1 社會資本（2） （3085） （3085） （3217） （3069） （3217） （3217） （3264） （3157） （3157） （3271） （3157） .306* .278* .309* .320* .344* .332* .447* .287* .281* .342* .324* .376* 月收入（萬元） 1 （3056） （3056） （3185） （3040） （3185） （3185） （3229） （3121） （3121） （3227） （3119） （3227） 階級認同 .288* .237* .271* .287* .305* .292* .383* .293* .262* .318* .312* .352* 294*. （2932） （2932） （3035） （2919） （3035） （3035） （3071） （2986） （2986） （3070） （2984） （3070） （3042） 1 * 表 P < .05
18 2.不同工作類型與其他階層變項、各項「原接觸的社會資本變項」 關連均數比較分析
從表 2 可看到不同工作類型與其他主客觀地位取得變項，這包括教育年 數、現職地位、收入、福利數、階級認同，以及「原接觸的社會資本（1）」、「原 接觸的社會資本（2）」的關連都達顯著（P<.05）。 在公家部門固定工作者的平均教育年數、現職地位、收入、福利數、階級 認同，以及「原接觸的社會資本（1）」、「原接觸的社會資本（2）」都最高，然而 在控制教育年數後，「原接觸的社會資本（1）」、「原接觸的社會資本（2）」是否 仍對在公家部門固定工作的機率有顯著正影響；在控制教育年數、現職地位與收 入後，在公家部門固定工作者階級認同是否仍最高，仍有待進一步迴歸分析來釐 清。 老闆階級的平均教育年數、現職地位、福利數與階級認同雖低於總平均， 平均月收入（4.95 萬元）則僅次於在公家部門固定工作者(5.16 萬元)，「原接觸 的社會資本（1）」、「原接觸的社會資本（2）」也都高於總平均。家屬工作者與為 不同機構工作的非典型工作者之教育年數、現職地位、收入、福利數、階級認同， 以及「原接觸的社會資本（1）」、「原接觸的社會資本（2）」都是所有工作者中最 低者，而且除了家屬工作者的階級認同（2.49）略高於沒工作的失業者（2.35） 之外，也都低於失業者；失業者的平均教育年數、職業地位、收入、福利數、階 級認同，以及「原接觸的社會資本（1）」、「原接觸的社會資本（2）」則都低於在 私人部門固定工作者與老闆。在此必須說明的是：根據本研究所用調查資料的問 卷設計，每週工作 29 小時以下就可算是失業者，並沒問其現在工作的現職地位、 收入與福利數，而所問的是最後一個工作之職業地位、收入與福利數，其職業地 位、收入與福利數仍都低於總平均許多；如果能問到其在工作，很可能有許多現 在沒工作或工時很短，而收入與福利數又更低。 至於上述工作類別與「原接觸的社會資本（2）」的關連性（Eta=.187），則 低於與「原接觸的社會資本（1）」的關連性（Eta=.208）；就此初步的雙變項分 析而言，「接觸的社會資本（2）」的建構效度低於「接觸的社會資本（1）」，然而 在控制其它變項後，工作類別與「接觸的社會資本（2）」的關連性是否仍低於「接 觸的社會資本（1）」，仍有待進一步迴歸分析來釐清。
19 表 2 不同工作類別者的階層變項 與各項「原接觸的社會資本」測量均數比較分析 教育 現職 月收入 福利數 階級 原接觸的 原接觸的 年數 地位 （萬元） 認同 社會資本（1） 社會資本（2）
40.85 4.95 .86 2.66 .087 351.7 老闆 11.08 （Ｎ＝486） （487） （471） （398） （460） （449） （475）
在私人部門 12.73 44.56 3.87 1.50 2.73 .14 362.7 固定工作 （1341） （1341） （1329） （1332） （1283） （1293） （1330）
在公家部門 14.08 51.79 5.16 1.78 2.93 .32 397.3 固定工作 （322） （321） （321） （322） （313） （316） （321）
10.12 35.51 2.28 .80 2.49 -.34 251.2 家屬工作者 （193） （192） （190） （167） （179） （177） （192）
為不同機構 9.98 31.15 2.58 .45 2.16 -.63 165.2 工作 （61） （61） （60） （53） （56） （49） （59）
10.69 36.35 2.84 .83 2.35 -.26 255.2 失業 （214） （214） （214） （207） （199） （197） （207）
12.16 43.11 4.00 1.31 2.68 .07 343.6 總平均 （2617） （2616） （2585） （2479） （2490） （2481） （2484）
Ｅta .339* .332* .263* .419* .190* .208* .187* * 表 P < .05
20 從表 3 可看到在高科技機構工作者的平均教育年數、現職地位、收入、福利 數、階級認同，以及「原接觸的社會資本（1）」、「原接觸的社會資本（2）」都高 出在非高科技機構工作者許多，而且差異都達顯著；而是否高科技機構工作者與 「原接觸的社會資本（2）」的關連性（Eta=.152），則高於與「原接觸的社會資 本（1）」的關連性（Eta=.135）。
表 3 高科技與非高科技機構工作者的階層變項 與各項原「接觸的社會資本」測量均數比較分析
教育 現職 月收入 福利數 階級 原接觸的 原接觸的 年數 地位 （萬元） 認同 社會資本（1） 社會資本（2）
非高科技 11.43 41.23 3.50 1.13 2.63 -.05 309.8 機構 （N=2935） （2934） （2895） （2772） （2745） （2738） （2883）
高科技 14.01 50.82 4.95 2.04 2.87 .39 441.1 機構 （340） （340） （339） （338） （332） （335） （340）
11.69 42.22 3.65 1.23 2.65 .00 323.7 總平均 （3275） （3274） （3234） （3110） （3077） （3073） （3223）
Eta .203* .206* .143* .325* .085* .135* .152* * 表 P < .05
21 （二）各項「接觸的社會資本」測量對現職地位、工作類別 影響之迴歸分析
以下根據迴歸分析來比較各項「接觸的社會資本」測量對主客觀地位取得變 項的影響，並比較各項「接觸的社會資本」測量之建構效度；在以下分析中，性 別、族群、出生年次、父親職業等背景變項以及現工作地都市化程度、教育年數 只做為控制變項，若無必要就不會對它們的影響多做說明。 從表 4 可看到各項「原接觸的社會資本」測量對現職地位的影響以「原接觸 的社會資本（2）」最大（β=.119），略高於「原接觸的社會資本（1）」（β=.112）、 「原接觸的社會資本（3）」（β=.113），「接觸的社會資本（3）」影響不如同樣以 職業地位總合做測量的「原接觸的社會資本（2）」，當可歸因於前者的職業聲望 測量不佳，不過三者的影響都達顯著(p<.05)。前面的相關分析亦顯示「原接觸的 社會資本（3）」與收入、階級認同的相關都低於「原接觸的社會資本（2）」，以 下的迴歸分析也就不再分析「原接觸的社會資本（3）」之影響，以精簡分析。 關於以上分析有些讀者可能會質疑，找到現在前原先職業地位不但會正向影 響「原接觸的社會資本」，也會正向影響現職地位，以上分析在未控制原先職業 地位的情況下，估計「原接觸的社會資本」對現職地位之影響，可能會高估影響。 本研究在進一步的分析中(表略)，也就以剛進現在公司（機關）的職業地位做為 原先職業地位的指標，在控制原先職業地位的情況下，估計「原接觸的社會資本 （1）」、「原接觸的社會資本（2）」、「原接觸的社會資本（3）」對現職地位之影響， 三者的影響（β）分別為.032、.035、.035，影響都達顯著，仍以「原接觸的社 會資本（2）」的影響最大；不過影響都不大，這可歸因於有 85.9%%的受訪者原 先四碼職業與現在職業相同，原先職業地位與現職地位相關(r)高達.92，「原接 觸的社會資本」難以發揮影響所致；此外，已控制原先職業地位，也就無法呈顯 取得原先職業前之「接觸的社會資本」透過原先職業地位對現職地位的間接影 響，這也會低估「接觸的社會資本」對現職地位的影響。因而在表 4 中，本研究 也就在沒控制未控制原先職業地位的情況下，估計「原接觸的社會資本」對現職 地位之影響，林南等人(Lin et al.,2001:72)也是這樣做。
22 表 4 各項「原接觸的社會資本」測量對現職地位影響比較迴歸分析
現職地位（1） 現職地位（2） 現職地位（3） b （β） b （β） b （β） 男性 -.16 （-.006） -.09 （-.003） -.08 （-.003） 本省客家 -.37 （-.009） -.48 （-.012） -.48 （-.012） 外省 -.18 （-.004） -.19 （-.004） -.15 （-.003） 出生年次 -.20*（-.160） -.19*（.-.155） -.19*（-.156） 父親職業 .05*（.050） .05*（.054） .06*（.055） 現工作地都市化 .30*（0.37） .30*（0.37） .32*（.039） 大專以上教育年數 3.32*（.431） 3.32*（.430） 3.35*（.433） 中小學以上教育年數 1.53*（.285） 1.50*（.288） 1.51*（.290） 原接觸的社會資本（1） 1.61*（.112） 原接觸的社會資本（2） .006*（.119） 原接觸的社會資本（3） .007*（.113） 常數 29.3* 26.8* 26.6* R2 .404* .410* .410* N 2763 2892 2892 * 表 P < .05 ** 族群對照組為本省閩南
23 從表 5A、B的多項式邏輯迴歸分析，可看到「原接觸的社會資本（1）」與「 原 接觸的社會資本（2）」對現各工作類別相對機率的影響大致都很接近，都一致呈 現「原接觸的社會資本」越高，現為家屬工作者、為不同機構工作的非典型工作 者與失業者的相對機率顯著越低之現象，而對現為老闆機率都無顯著影響；本研 究所做的進一步分析亦顯示：熊瑞梅與黃毅志(1992)採用的「有認識人任職的職 業數」對現為老闆機率亦無顯著影響（表略）；不一樣的是：「原接觸的社會資本 （1）」對於在公家部門固定工作的相對機率並無顯著影響，「原接觸的社會資本 （2）」則對於在公家部門固定工作的相對機率有顯著負影響。多項式邏輯迴歸分 析無法計算代表各別自變項對於依變項影響大小的標準化係數，如OLS迴歸的β 值；不過由於在表 5A、B的多項式邏輯迴歸分析中，除了「原接觸的社會資本」 的測量不一樣外，所有變項都一樣，仍當可用整個模型的解釋力（R2）來比較兩 項「原接觸的社會資本」對現工作類別的影響大小，由於採用「原接觸的社會資 本（2）」的表 5B之R2（.254）高於採用「原接觸的社會資本（1）」的 表 5 A（ . 2 4 8 ）， 「原接觸的社會資本（2）」的影響當大於「原接觸的社會資本（1）」。
24 表 5A 原接觸的社會資本（1）對現工作類別影響多項式邏輯迴歸分析
在公家部門 老闆/ 家屬工作者/ 為不同機構工作者/ 失業/ 固定工作者/ 在私人部門 在私人部門 在私人部門 在私人部門 在私人部門 固定工作者 固定工作者 固定工作者 固定工作者 固定工作者 b （S.E.） b （S.E.） b （S.E.） b （S.E.） b （S.E.） 男性 .523（.131）* -.356（.142）* -.643（.178）* -.075（.411） .203（.174） 本省客家 -.158（.176） -.055（.199） -.103（.247） .887（.451）* .178（.237） 外省 -.148（.212） -.225（.226）* -.812（.441） .201（.783） .099（.286） 出生年次 -.066（.007）* -.086（.008）* -.052（.010）* -.015（.023） .002（.010）
父親職業 .002（.005） -.004（.005） -.010（.008） -.017（.020） .012（.007） 現工作地 -.025（.036） -.126（.040）* -.083（.051） -.077（.116） .072（.051） 都市化 大專以上 -.190（.043）* .308（.040）*-.285（.078）* -.157（.159） -.143（.060）* 教育年數
中小學以上 .013（.031） .241（.053）* .004（.039） .059（.107） -.148（.041）* 教育年數 原接觸的 .085（.067） -.040（.083） -.207（.091）* -.489（.214）* -.283（.091）* 社會資本（1）
常數項 2.5* .4 .7 -1.7 .8 -2Log Likelihood 5437* Nagelkerke R2 .248 N 2227 * 表P<.05 ** 族群對照組為本省閩南
25 表 5B 原接觸的社會資本（2）對現工作類別影響多項式邏輯迴歸分析
在公家部門 老闆/ 家屬工作者/ 為不同機構工作者/ 失業/ 固定工作者/ 在私人部門 在私人部門 在私人部門 在私人部門 在私人部門 固定工作者 固定工作者 固定工作者 固定工作者 固定工作者 b （S.E.） b （S.E.） b （S.E.） b （S.E.） b （S.E.） 男性 .519（.127）* -.305（.140）* -.588（.172）* -.001（.367） .212（.170） 本省客家 -.182（.174） -.058（.199） -.162（.245） . 684（.434） .174（.233） 外省 -.163（.209） -.151（.222） -.714（.411） .424（.652） .038（.284） 出生年次 -.066（.007）* -.086（.008）* -.051（.010）* -.018（.020） .001（.009）
父親職業 .002（.005） -.005（.005） -.010（.008） -.019（.018） .013（.007） 現工作地 -.018（.035） -.125（.039）* -.068（.049） -.016（.104） .062（.050） 都市化 大專以上 -.193（.043）* .334（.039）*-.295（.077）* -.161（.155） -.153（.060）* 教育年數
中小學以上 .001（.030） .242（.051）*-.007（.038） -.055（.084） -.147（.039）* 教育年數 原接觸 .000（.000） -.001（.000）* -.001（.000）* -.004（.001）* -.001（.000）* 社會資本（2）
常數項 2.5* .6 1.0 .0 -.4 -2Log Likelihood 5707* Nagelkerke R2 .254 N 2317 * 表 P<.05 ** 族群對照組為本省閩南
26 從表 6 的二項式邏輯迴歸分析可看到，不論是用「原接觸的社會資本（1）」 或「原接觸的社會資本（2）」做測量，都有「原接觸的社會資本」較高，會顯著 提高進高科技機構工作機率的現象；不過用「原接觸的社會資本（2）」的R（.145）2 大於「原接觸的社會資本（1）」（.137），仍顯示原接觸的社會資本（2）」的影響 大於「原接觸的社會資本（1）」。
表 6 各項「原接觸的社會資本」測量 對進高科技機構工作影響二項式邏輯迴歸分析
進高科技 進高科技 機構工作 機構工作 b （S.E.） b （S.E.） 男性 .583（.132）* .569（.131）* 本省客家 .427（.169）* .419（.168）* 外省 .669（.186）* .640（.185）* 出生年次 .027（.007）* .029（.007）*
父親職業 -.003（.005） -.003（.005） 現工作地都市化 -.078（.036）* -.084（.036）* 大專以上教育年數 .167（.035）* .161（..035）* 中小學以上教育年數 .188（.059）* .199（.059）* 原接觸的社會資本（1） .306（.080）* 原接觸的社會資本（2） .001（.000）*
常數項 -4.3* -4.9* -2Log Likelihood 1736* 1763* Nagelkerke R2 .137 .145 N 2767 2897 * 表 P<.05 ** 族群對照組為本省閩南
27 （三）各項「接觸的社會資本」測量對收入、階級認同影響 之迴歸分析
從表 7 式（1）、（2）可看到「原接觸的社會資本（2）」對工作收入的正向
影響（β=.143）略高於「原接觸的社會資本（1）」（β=.137），而且兩者的影響 都達顯著(p<.05)。式（3）、（4）控制了現職地位與工作類別，「原接觸的社會資
本（2）」對工作收入的影響（β=.097）仍略高於「原接觸的社會資本（1）」（β
=.086），兩式都一致顯示現職地位越高，收入顯著越高，相較於在私人部門固定 工作者，老闆、在公家部門固定工作者的收入顯著較高，家屬工作者、失業者的 收入顯著較低；然而為不同機構工作的非典型工作者收入與在私人部門固定工作 者並無顯著差異，高科技機構工作者的收入與非高科技機構工作者也無顯著差 異，表 2 在私人部門固定工作者收入高於為不同機構工作的非典型工作者，表 3 高科技機構工作者的收入高於非高科技機構工作者，當可歸因於在私人部門固定 工作者、高科技機構工作者的教育年數與現職地位較高。 式（3）、（4）控制了現職地位與工作類別，估計「原接觸的社會資本（2）」 與「原接觸的社會資本（1）」對現在工作收入的直接影響（見圖 1），這可能會 低估 「接觸的社會資本」對收入的影響，因為與現在工作收入關連性較大的應 是現在的「現接觸的社會資本」，而非找到現職時的「原接觸的社會資本」；式（5）、 （6）也就控制了現職地位與工作類別，估計「現接觸的社會資本」對工作收入
的直接影響，「現接觸的社會資本（2）」與「現接觸的社會資本（1）」的影響（β） 分別是.135 與.107，果然都大於「原接觸的社會資本（2）」與「原接觸的社會資 本（ 1）」之影響，而「現接觸的社會資本（2）」的影響比「現接觸的社會資本（1）」 高出不少。
28 表 7 各項「接觸的社會資本」測量對工作收入影響迴歸分析
工作收入 工作收入 工作收入 工作收入 工作收入 工作收入 (取對數)（1） (取對數)（2） (取對數)（3） (取對數)（4） (取對數)（5） (取對數)（6） b （β） b （β） b （β） b （β） b （β） b （β） 男性 .261*（.245） .259*（.242） .244*（.229） .243* （.228） .250*（.235） .246*（.231） 本省 .035 （.024） .036 （.024） .042 （.028） .043 （.028） .042 （.028） .039 （.024） 閩南 外省 .026 （.015） .024 （.014） .026 （.015） .024 （.014） .009 （.005） .017 （.010） 出生 .078*（1.562） .078*（1.573） .068*（1.371） .069* （1.379） .068*（1.375） .066*（1.330） 年次 出生年次 （ ） （ ） （ ） （ ） （ ） （ ） 平方 -.001* -1.750 -.001* -1.753 -.001* -1.508 -.001* -1.509 -.001* -1.508 -.001* -1.449 父親 .000 （-.001） .000*（-.001） .000 （-.012） .000 （-.012） .000 （-.007） .000 （-.012） 職業 現工作地 （ ） （ ） （ ） （ ） （ ） （ ） 都市化 .018* .061 .016* .055 .016* .055 .015* .049 .015* .050 .014* .049 大專以上 （ ） （ ） （ ） （ ） （ ） （ ） 教育年數 .074* .263 .075* .263 .032* .114 .032* .113 .032* .113 .029* .103 中小學 以上 .050*（.218） .051*（.225） .033*（.142） .033* （.147） .030*（.131） .030*（.133） 教育年數 原「接觸 的社會資 .074*（.137） .047*（.086） 本」（1） 原「接觸 的社會資 .000*（.143） .000* （.097） 本」（2） 現「接觸 的社會資 .059*（.107） 本」（1） 現「接觸 的社會資 .000 （.135） 本」（2） 現職 .010* （.263） .010* （.259） .010*（.266） .009*（.254） 地位 老闆 .075* （.055） .067* （.049） .062*（.046） .058*（.042） 在公家 部門 .093* （.060） .093* （.059） .070*（.045） .085*（.054） 固定工作 家屬 -.284* （-.141） -.263* （-.132） -.287*（-.143） -.265* （-.132） 工作者 為不同 （ ） （ ） （ ） （ ） 機構工作 .077 -.016 -.101 -.023 -.104 -.022 -.102 -.023 失業 -.147* （-.074） -.149* （-.076） -.138*（-.069） -.138* （-.070） 高科技 .052 （.033） .051 （.032） .045 （.028） 機構 常數 -1.2* -1.3* -1.1* -1.2* -1.1* -1.2* R2 .341* .342* .426* .425* .420* .427* N 2207 2293 2204 2290 2256 2309 * 表 P<.05 ** 族群對照組為本省閩南，現工作類別對照組為在私人部門固定工作者
29 從表 8 式（1）、（2）可看到「原接觸的社會資本（2）」對階級認同的正向
影響（β=.140）比「原接觸的社會資本（1）」（β=.106）高出不少，不過兩者的 影響都達顯著(p<.05)。式（3）、（4）控制了現職地位與工作類別，「原接觸的社
會資本（2）」對階級認同的影響（β=.115）仍比「原接觸的社會資本（1）」（β
=.082）高出不少，不過兩者的影響都達顯著；兩式都一致顯示現職地位越高， 階級認同顯著越高，在私人部門固定工作者的階級認同則與其它工作類型者，包 括在公家部門固定工作者大都無顯著差異，除了在式（3）顯著高於失業者之外， 不過兩者的階級認同只差了.131 等（b=-.131），並無多大實質意義，只是因為 樣本大差異才達顯著；高科技機構工作者的階級認同與非高科技機構工作者也無 顯著差異。式（5）、（6）又控制了工作收入，收入越高，階級認同顯著越高，在 私人部門固定工作者的階級認同在公家部門固定工作者仍無顯著差異，「原接觸
的社會資本（2）」對階級認同的影響（β=.098）仍比「原接觸的社會資本（1）」
（β=.065）高出不少，不過兩者的影響都達顯著。 式(3)、(4)控制了現職地位與工作類別，估計「原接觸的社會資本（2）」與 「原接觸的社會資本（1）」對階級認同的直接影響（參見圖 1），這可能會低估 「接 觸的社會資本」對階級認同的直接影響，因為與現在階級認同關連性較大的應是 現在的「現接觸的社會資本」；式（7）、（8）也就控制了現職地位與工作類別， 估計「現接觸的社會資本」對工作收入的直接影響，「現接觸的社會資本（2）」
與「現接觸的社會資本（1）」的影響（β）分別是.168 與.142，果然都大於「原 接觸的社會資本（2）」與「原接觸的社會資本（1）」之影響，而「現接觸的社會 資本（2）」的影響仍比「現接觸的社會資本（1）」高出不少。式(9)、(10)又控 制工作收入，「現接觸的社會資本（2）」與「現接觸的社會資本（1）」對階級認
同的影響（β）分別是.146 與.124，也都大於式（5）、（6）「原接觸的社會資本（2）」 與「原接觸的社會資本（1）」之影響，「原接觸的社會資本（2）」對階級認同的 影響仍比「原接觸的社會資本（1）」高出不少，不過兩者的影響都達顯著。
30
表 8 各項「接觸的社會資本」測量對階級認同影響迴歸分析
階級認同（1） 階級認同（2） 階級認同（3） 階級認同（4） 階級認同（5） 階級認同（6） b （β） b （β） b （β） b （β） b （β） b （β） 男性 -.175* （-.103） -.177*（-.104） -.177*（-.104） -.179*（-.105） -.255*（-.151） -.250*（-.148） 本省 .093 （.039） .087 （.036） .108 （.046） .105*（.044） .088 （.037） .087 （.036） 閩南 外省 -.056 （-.021） -.065 （-.024） -.039（-.014） -.047（-.017） -.035（-.013） -.041 （-.015） 出生年次 .029 （.362） .029 （.360） .018（.228） .018（.224） .002（-.026） -.001 （-.013） 出生年次 （ ） （ ） （ ） （ ） （ ） （ ） 平方 .000 -.477 .000* -.471 .000 -.312 .000 -.303 .000 -.024 .000 -.307 父親職業 .001 （.024） .002 （.026） .001（.014） .001（.015） .001（.018） .001（.018） 現工作地 （ ） （ ） （ ） （ ） （ ） （ ） 都市化 .017 .035 .013 .027 .014 .029 .010 .020 .009 .020 .006 .013 大專以上 （ ） （ ） （ ） （ ） （ ） （ ） 教育年數 .107* .240 .107* .238 .069* .156 .068* .152 .055* .125 .055* .124 中小學 以上 .074* （.192） .071*（.188） .057*（.149） .055*（.146） .046*（.120） .045*（.119） 教育年數 原「接觸 的社會資 .093* （.106） .072*（.082） .057*（.065） 本」（1） 原「接觸 的社會資 .000*（.140） .000*（.115） .000*（.098） 本」（2） 現職地位 .012*（.196） .011*（.194） .008*（.144） .009*（.146） 老闆 .048（.022） .046（.021） .018 （.008） .018 （.008） 在公家 部門 -.040（-.016） -.028（-.011） -.069（-.028） -.053（-.022） 固定工作 家屬 .020（.006） .001（.000） .124（.038） .091（.028） 工作者 為不同 （ ） （ ） （ ） （ ） 機構工作 -.113 -.015 -.258 -.036 -.085 -.011 -.230 -.032 失業 -.131*（-.041） -.117（-.037） -.083（-.026） -.070（-.022） 高科技 -.018（-.007） -.028（-.011） -.032（-.013） -.040（-.016） 機構 工作收入 （ ） （ ） (取對數) .321 .201 .295 .184 常數 1.2* 1.1* 1.2* 1.1* 1.5* 1.4* R2 .175* .185* .202* .212* .223* .230* N 2140 2213 2137 2210 2118 2189 * 表 P<.05 ** 族群對照組為本省閩南，現工作類別對照組為在私人部門固定工作者
31
表 8 各項「接觸的社會資本」測量對階級認同影響迴歸分析(續)
階級認同（7） 階級認同（8） 階級認同（9） 階級認同（10） b （β） b （β） b （β） b （β） 男性 -.167*（-.098） -.167*（-.098） -.238*（-.141） -.234*（-.138） 本省 .108*（.045） .107*（.045） .088（.037） .088（.037） 閩南 外省 -.055（-.020） -.055（-.020） -.044（-.016） -.047（-.017） 出生年次 .013（.170） .013（.159） -.005（-.063） -.004（-.049） 出生年次 .000（-.242） .000（-.224） .000（.020） .000（.010） 平方 父親職業 .001（.018） .001（.016） .001（.021） .001（.020） 現工作地 .006（.013） .006（.012） .003（.006） .003（.006） 都市化 大專以上 .068*（.152） .064*（.142） .055*（.124） .052*（.117） 教育年數 中小學 以上 .053*（.136） .049*（.128） .043*（.110） .040*（.105） 教育年數 現「接觸 的社會資 .128*（.142） .111*（.124） 本」（1） 現「接觸 的社會資 .000*（.168） .000*（.146） 本」（2） 現職地位 .011*（.189） .011*（.186） .008*（.142） .008*（.144） 老闆 .026（.012） .020（.009） .000（.000） -.005（-.002） 在公家 部門 -.049（-.020） -.037（-.015） -.069（-.028） -.058（-.023） 固定工作 家屬 .015（.005） .002（.001） .110（.034） .087（.027） 工作者 為不同 -.204（-.027） -.259（-.036） -.171（-.023） -.232（-.032） 機構工作 失業 -.111（-.034） -.102（-.032） -.068（-.021） -.059（-.019） 高科技 -.028（-.011） -.041（-.016） -.039（-.016） -.050（-.020） 機構 工作收入 .292*（.183） .269*（.169） (取對數) 常數 1.4* 1.2* 1.7* 1.5* R2 .214* .221* .232* .236* N 2180 2228 2159 2206 * 表P<.05 ** 族群對照組為本省閩南，現工作類別對照組為在私人部門固定工作者
32 五、結論與討論
本研究運用 2004 年進行調查的「社會資本的建構與效應」全國性大樣本資 料做路徑分析，探討各項「接觸的社會資本」之測量，這包括林南(Lin,2001)以及 熊瑞梅與黃毅志(Hsung and Hwang,1992)所提出的測量，對現職地位、工作類別與 收入、階級認同有何影響？何者影響比較大？研究發現顯示：這兩項測量對現職 地位、進高科技機構工作機率、收入與階級認同都有正向影響，對於家屬工作、 為不同機構工作的非典型工作與失業的機率都有負向影響；不過熊瑞梅與黃毅志 的測量之影響都大於林南的測量，特別是在對收入的直接影響與對階級認同的影 響上。至於用兩項測量所得到的不同研究發現主要是：熊瑞梅與黃毅志的測量對 於在公家部門固定工作的機率有顯著負影響，林南的測量無顯著影響。
關於以上發現，可以做進一步討論如下： （一）「接觸的社會資本」影響主客觀地位取得的因果機制 不論是林南或熊瑞梅與黃毅志的「接觸的社會資本」測量都一致顯示：「接 觸的社會資本」越高者，現職地位越高，進而提高收入與階級認同；「接觸的社 會資本」越高者，越可能進高科技機構工作，越不可能成為家屬工作、為不同機 構工作的非典型工作與失業者，越不可能成家屬工作與失業者，有提高收入、階 級認同的間接影響；即使控制了現職地位與工作類別，仍有「接觸的社會資本」 越高者，收入越高的直接影，可能由於「接觸的社會資本」較高，有助於事業發 展，而得到高收入；即使控制了現職地位、工作類別與收入，仍有「接觸的社會 資本」越高者，階級認同越高的直接影響，這卻是本研究的因果模型所沒顯示的。
(二)林南和熊瑞梅與黃毅志的「接觸的社會資本」測量影響不同之討論 林南和熊瑞梅與黃毅志的「接觸的社會資本」測量對以上各項主客觀地位 取得的影響顯得一致，用不同測量做分析得到一致的發現，加強「接觸的社會資 本」的影響當很廣汎之說服力，而本研究發現「接觸的社會資本」對於階級認同 有不小的正向影響，筆者還沒看到先前國內外研究有如此之發現，不過兩項測量 的影響仍有下列不同：
33 1. 不過熊瑞梅與黃毅志的「接觸的社會資本」測量之影響都大於林南的測量， 特別是在對收入的直接影響與對階級認同的影響上，此即熊瑞梅和黃毅志的 測量之建構效度要比林南的高一些。熊瑞梅與黃毅志的測量對收入、階級認 同的直接影響較大，可用文獻檢討中提到的 B、C 做說明；雖然Ｂ和Ｃ有人 認識的職業數、認識人最高的職業地位及最高地位與最低地位的差距相等， 而在林南的測量上兩者相等，但是顯然的Ｃ的網絡中有許多高地位者，可能 動員的網絡資源總體較大，在熊瑞梅和黃毅志的測量上較高，而收入較高， 這卻是林南的測量所無法顯現的，林南的測量較不易呈現「接觸的社會資本」 較高，收入也較高的現象，對收入的直接影響也就較小。
至於為何熊瑞梅與黃毅志的測量對階級認同的直接影響較大，部份的
原因可能是他們的測量可顯現Ｃ可能動員的網絡資源總體較大，而在本研
究無法測量的客觀地位上較高，C 的階級認同也就較高的現象；如Ｂ和Ｃ
在本研究所用 Ganzeboom 與 Treiman( 1996)的四碼社經地位測量上都是大
專教師而同為 77 分，然而 C 卻是台大正教授，而實際客觀地位高於台東大
學助理教授的 B，C 的階級認同也就較高；而台大正教授實際客觀地位高於
台東大學助理教授，乃本研究的職業地位所無法測量，C 因為本研究無法
測量的客觀地位較高而階級認同較高，在熊瑞梅與黃毅志的測量上也較
高，這會提高熊瑞梅與黃毅志的測量對階級認同的直接影響；林南的測量
則無法顯現 C 的接觸的社會資本較高，階級認同也較高的現象，所估計對
階級認同的直接影響也就較小。此外，亦有可能 C 認識許多高職業地位的
人，而屬於高聲望的地位團體(status group,Weber,1978:302-307)，B 認識許
多低職業地位的人，而屬於低聲望的地位團體，使得 C 的階級認同高於 B
（Kluegel etal.,1977;Hodge and Treiman,1968）。 2. 熊瑞梅與黃毅志之「接觸的社會資本」測量對於在公家部門固定工作的機率 有顯著負影響，林南的測量無顯著影響，這都符合文獻檢討中所提到「要進入公 家部門就職，往往要透過考試，看來接觸的社會資本較高，並不會提高進公家部 門的機會」之論點；不過「接觸的社會資本」較高者，應較可能在私人部門中找
34 到好工作，而相對降低進公家部門就職的機會，看來根據熊瑞梅與黃毅志之「接 觸的社會資本」測量所得到的發現更加合理。
(三)對假設檢證結果之討論 本研究所提出的假設大多得到研究發現有力的支持，得不到支持者如下： 1.「接觸的社會資本」對當老闆（大多是小老闆）相對機率無顯著影響，可能的 原因是近 8 年來，當老闆在收入上的優勢下降許多(章英華、黃毅志，2007)，雖 然當老闆在收入上仍佔一些優勢，然而當老闆的風險大，可能有多關場(店)或倒 閉(章英華、黃毅志，2007)，如此的收入優勢不足以吸引許多人當老闆，「接觸 的社會資本」較高者，當老闆的相對機率也就沒有提高；這與過去研究(熊瑞梅、 黃毅志，1992)的發現有所不同。 2.在控制教育、現職地位與收入後，在私人部門固定工作者的階級認同與在公家 部門固定工作者無顯著差異，顯示現在公家部門固定工作者的階級認同較高，是 因教育、現職地位與收入較高所致，公私部門本身已不能代表著一項階層區分； 這有別於過去研究採用 1992 年台灣地區社會變遷調查資料做分析的發現(黃毅 志，2001b)，不過現在公家部門在收入的優勢倒是增強。 3、在控制教育、現職地位後，為不同機構工作的非典型工作者的收入並沒顯著 低於在私人部門固定工作者，這可能是如此的非典型工作者樣本太小(參見表 2) 所致。
(四)對於未來研究建議 本研究發現「接觸的社會資本」對階級認同有直接正向影響，並推測這有可
能是「接觸的社會資本」高者，屬於高聲望地位團體所致；然而不論是熊瑞梅與
黃毅志或 Lin 的「接觸的社會資本」測量，所依據的都是有認識人認職之職業，
單是認識並不代表就是強連繫，與所認識的人並不一定是強連繫，而屬於同一地
位團體(Weber,1978:302-307)；不過「接觸的社會資本」高者仍可能由於認識高
職業地位的人多，而與其中許多高職業地位的人成好友，乃至於結婚而是強連
繫，而屬於同一地位團體，進而提高階級認同(Lin,1982)；「接觸的社會資本」對
階級認同的影響很可能以屬於強連繫的親密網絡成員之職業地位為中介，親密網
35 絡成員之職業地位可根據受訪者所回答的最要好朋友們之職業地位做測量，然而
本研究所用資料並沒調查受訪者最要好朋友們的職業地位，也究無法檢證「接觸
的社會資本對階級認同的影響以親密網絡成員之職業地位為中介」之假設，建議
未來社會資本調查仍可調查受訪者最要好朋友們的職業地位，以檢證上述假設。
參 考 書 目
王振寰 2007 "從科技追趕到創新的經濟轉型：南韓、台灣與中國"，台灣社會研 究季刊第六十八期，頁 177-226。 中華民國行政院主計處
1983 中華民國統計地區標準分類。 2008 人力資源調查統計結果。
林季平 2005 "台灣的人口遷徙及勞工流動問題回顧：1980-2000"，台灣社會學 刊第三十四期，頁 147-209。 章英華、黃毅志 2007 “台灣地區教育分流對階層位置影響之變遷：1997 年與 2005 年的 比較”，發表於中央研究院社會學研究所主辦「台灣的社會變遷」 研討會。 章英華、傅仰止 2003 台灣地區社會變遷基本調查計劃 四期三次執行報告，中研院社會 所。 許嘉猷、黃毅志 2002 "跨越階級的界限？：兼論黑手變頭家的實證研究結果及與歐美社 會之一些比較"，台灣社會學刊，27，1-59。
孫清山、黃毅志 1994 "社會資源、文化資本與職業取得"，東海學報 ，35，127-150。 1995 "教育、收入與社會資源和階級取得過程之關連"，收錄於林松齡、 王振寰主編，台灣社會學研究的回顧與前瞻，頁 151-182。台中： 東海大學社會系。 孫清山、熊瑞梅 1985 台灣製造業勞工社會網絡與離職、轉業行為研究。國科會研究計劃
36 報告。
陳至柔 1995 "族群、生涯發展、和工作流動---經濟部門流動過程的族群差異"， 於中央研究院民族學研究所， 變遷中的臺灣社會：第二次台灣地 區社會變遷基本調查國際研討會發表。 黃毅志
1992 "地位取得: 理論與結構分析"，思與言，30(4)，131-168。 1996 "台灣地區民眾地位取得之因果機制:共變結構分析"，東吳社會學 報第五期，頁 213-248。
2001a "台灣地區的勞力市場分割之探討:流動表分析"，台灣社會學刊， 25，157-200。
2001b "台灣地區多元勞力市場的事業成就之比較分析"，國科會人文社會 彙刊，11(4)，356-370。 曾敏傑
2001 "中高齡勞工失業現況與變遷：1982 與 1996 失業潮之比較"，台灣 社會學刊，25：243-279。 熊瑞梅、黃毅志
1992 "社會資源與小資本階級"，中國社會學社刊，16，107-138。
蔡明璋 2005 "台灣的新經濟：文獻的回顧與評述"， 台灣社會學刊第三十四期， 頁 221-247。 謝國雄
1989 "黑手變頭家─台灣製造業中的階級流動"，台灣社會研究季刊， 2(2)，11-54。
Blau, P. M. and O. D. Duncan 1967 The American Occupation Structure. New York, Wiley.
Baldi, S. 1998 "Normative Versus Social Constructivist Processes in the Allocation of Citations: A Networks-Analytic Model." American Sociological Review 63:829-846. Bian,Y. 1997 "Bringing Strong Ties Back in Indirect Ties, Network Bridges, and Job
37 Searches in China." American Sociological Review 66:366-385.
Bian,Y. and S. Ang 1997 "Guanxi Networks and Job Mobility in China and Singapore." Social Forces 75(3):981-1005.
Carmines, E. G. and R. A. Zeller
1979 Reliability and validity assessment. Beverly Hills : Sage Publications. Castilla, E. J. 2005 "Social Networks and Employee Performance in a Call Center." American Journal of Sociology 110(5):1243-83. Featherman, D. L. and R. M. Hauser 1978 Opportunity and Change. New York: Academic Press. Fernandez, R. M. and I. Fernandez-Mateo 2006 "Networks, Race, and Hiring. " American Sociological Review 71:42-71. Fernandez, R. M. and N. Weinberg 1997 "Sifting and Sorting: Personal Contacts and Hiring in a Retail Bank" American Sociological Review 62:368-385. Fernandez,R. M., E. J. Castilla and P. Moore 2000 “Social Capital at Work: Networks and Employment at a Phone Center.” American Journal of Sociology 105(5):1288-356. Flap, H. l989 "Social Capital in the Reproduction of Inequality: A Review." Paper Presented at the Meeting of Research Committee 28 on Social Stratification of the International Sociological Association. Utreht April. Ganzeboom, H. B. and D. J. Treiman 1996 “Internationally comparable measures of occupational status for the 1988 internation standard classification of occupations. Social Science Research, 25: 201-239. Hodson, R. and R. L. Kaufman 1982 "Economic Dualism: A Critical Review." American Sociological Review 47:727-739. Hodge, R. W., and D. J. Treiman 1968 "Class identification in the United States." American Journal of Sociology, 73, 535-47.
38 Hsung, R. M. and Y. J. Hwang 1992 "Job Mobility in Taiwan: Job Search Methods and Contacts Status." Paper Presented at the XII International Sunbelt Social Network Conference, San Diego, C. A., U.S.A. Kalleberg, A. L., B. F. Reskin and K. Hodson 2000 "Bad Jobs in America: Standard and Nonstandard Employment Relations and Job Quality in the United States." American Sociological Review 65:256-278. Kluegel,J.R., R. SingletonⅡ and C. S. Starnes 1977 "Subjective Class Identification," American Sociological Review 65.
Lai, G. W. F., S. Y. Leung and N. Lin 1989 "Network resources, Contact Resources and Status Attainment" Social Network 20(2):159-178. Lin, N. (林南) 1982 "Social Resources and Instrumental action", P.p.131-146 in P. Marsden and Nan Lin(ed.), Social Structure and Network Analysis, Beverly Hills:Sage. 1990 "Social Resources and Social Mobility: A Structural Theory of Status Attainment",Pp.247-271 in Ronald L. Breiger(ed.), Social Mobility and Social Structure. Cambridge University Press. 1995 "Social Resources: A Theory of Social Capital." the Revue Francaise de Sociologie. 2001 Social Capital：A Theory of Social Structure and Action. Cambridge: Cambridge University Press. 2003 "Job Search in Urban China :Gender ,Network Chains ,and Embedded Resources",To appear in: Henk Flap and Beate Volker,eds., Creation and Return of Social Capital, Prager. Lin, N. and Y. J. Bian 1991 "Getting Ahead in Urban China." American Journal of Sociology 97(3):657-688.
Lin, N., W. M. Ensel and J. C. Vaughn 1981 "Social Resources and Strength of Ties: Structural Factors in Occupational Status Attainment." American Sociological Review 46:393-405. Lin, N., Y.C. Fu and R. M. Hsung
39 2001 "The Position Generator: Measurement:Techniques for Investigations of Social Capital", in N. Lin, K.S. Cook and R.S. Burt (eds) Social Capital: Theory and Research, New York: Aldine de Gruyter. Messner, S. F., E. P. Baumer, and R. Rosenfeld 2004 "Dimensions of Social Capital and Rates of Criminal Homicide." American Sociological Review 69:882-903. Podolny, J. M. and J. N. Baron 1997 "Resources and Relationships: Social Networks and Mobility in the Workplace." American Sociological Review 62:673-692. Liu,J. and A. Sakamoto 2002 “The Role of Schooling in Taiwan’s Labor Market:Human Capital,Screening,or Credentialism?Taiwanese Journal of Sociology,29,1-56. Smith, M. R. 1990 "What is New in New Structuralist Analyses of Earnings？" American Sociological Review 55:827-841. Smith, S. S. 2005 "Don’t put my name on it: Social Capital Activation and Job-Finding Assistance among the Black Urban Poor. " American Journal of Sociology 111(1):1-57. Treiman, D. J. 1977 Occupational prestige in comparative perspective. New York: Academic Press. Gaag, V. D. M., T. A. B. Snijders and H. D. Flap 2004 "Position Generator measures and their relationship to other Social Capital measures." This research is part of the Research Program “ Creation of and returns to social capital; Social Capital in education and labor markets.”(SCALE), a combined project of the universities of Utrecht (UU), Groningen (RuG), and Amsterdam (UvA), funded by the Dutch Organization for Scientific Research, project number 510-50 - 204. Weber, M. 1978 Economy and society. Berkeley: University of California Press.
Wegener, B. 1991 "Job Mobility and Social Ties: Social Resources, Prior Job, and Status Attainment." American Sociological Review 56: 60-71.
40
Yakubovich, V. 2005 "Weak Ties, Information, and Influence: How Workers Find Jobs in a Local Russian Labor Market. " American Sociological Review 70:408-421. Zeng, Z. and Y. Xie 2004 "Asian-Americans’ Earning Disadvantage Reexamined: The Role of Place of Education." American Journal of Sociology 109 (5): 1075-1108. Zhou, X. 2005 "The Institutional Logic of Occupational Prestige Ranking: Reconceptualization and Reanalyses " American Journal of Sociology 111（1）:90-140.
41 Your Body Knows Who You Know:
* Social Capital and Well-Being Inequality in the United States
Lijun Song, PhD candidate Department of Sociology Duke University
Revised May 23, 2008
Word Count: 6,722 Tables: Four
* Direct correspondence to Lijun Song, Department of Sociology, Box 90088, Duke University, Durham, NC 27708 ([email protected]). An earlier version of this article was presented at the Annual Meeting of ASA in New York, 2007, the Annual Meeting of SSS in Atlanta, 2007, The Mid-Winter Meeting of the American Sociological Association Methods Section, “Network and New Methods for Global Health”, in Durham, 2008, and the International Sunbelt Social Network Conference in St. Pete Beach, 2008. Data used in this article were drawn from the thematic research project “Social Capital: Its Origins and Consequences", sponsored by Academia Sinica, Taiwan, through its Research Center for Humanities and Social Sciences, the Institute of Sociology, and Duke University. The principal investigator of the project is Nan Lin. The author thanks Linda K. George, Mary Elizabeth Hughes, Kenneth C. Land, and Nan Lin for their helpful comments.
1 ABSTRACT
Does social capital, resources embedded in social relationships, influence health and
well-being? This research examines whether social capital impacts depressive symptoms
and marital satisfaction over and above the effects of personal and family resources,
using a national U.S. sample. Results demonstrate the direct effect of social capital on both outcomes net of individual and family socioeconomic status. Results also show that social capital mediates the effect of individual education on depression, and that social capital moderates the effect of gender on depression, and the effects of gender, individual education and occupational prestige on marital satisfaction. This research suggests that social capital not only contributes to health and well-being independently, but also interplay with other structural risk factors, and that that socioeconomic resources can flow through social relationships to produce disparities in health and well-being.
Key Words: social capital, depression, marital satisfaction
2 Social stratification of health and well-being is one unique contribution of medical
sociology (Bird, Conrad and Fremont, 2000). A huge body of evidence exists on how
social inequalities produce disparities in health and well-being. Among other social
stratification factors, socioeconomic resources at multiple levels (i.e., individual, family,
community) have received substantial attention. Their impact on social patterning of
health and well-being has been well-established over time and space (Link and Phelan
1995; Robert and House 2000). We are now challenged to explore other dimensions of
socioeconomic resources by drawing on recent mainstream sociological theories.
Prominent among these theories is the notion of social capital. Social capital has a broad
range of definitions (Coleman 1988; Bourdieu 1983/1986; Lin 2001; Putnam 1993). This
article does not attempt to reconcile these debates. Instead it contributes to this
burgeoning literature by asking whether social capital theory proposed by Lin (2001)
applies to inequalities in health and well-being.
Social capital is resources embedded in social networks, indicated by network
members’ socioeconomic positions (Lin 1999b, 2001). Social capital has gathered
burgeoning attention in the sociological literature (for reviews, see Lin 1999a; Portes
1998). However, whether it influences health and well-being over and above the effects
of other forms of socioeconomic resources remained under-explored. The major purpose of this research is to examine whether social capital decreases depressive symptoms and increases marital satisfaction over and above the effects of individual and family socioeconomic resources in the United States, and further whether social capital mediates and moderate the effects of other establish structural risk factors for health and well- being.
3 This paper is organized as follows: First, I introduce social capital theory, and
distinguish social capital from other dimensions of socioeconomic resources. Then I
derive main-effect, mediating-effect, and moderating-effect hypotheses from social
capital theory. Next, I conduct analyses using a national U.S. sample, and describe the
results. I conclude with implications of this study for future research.
SOCIAL CAPITAL AS A NETWORK RESOURCES THEORY
Social capital theory is grounded in the classic tradition of capital theories (e.g., Marx’s
capital theory, human capital theory, cultural capital theory) that explain how various
types of capital bring returns. This theory defines social capital as “resources embedded in a social structure that are accessed and/or mobilized in purposive actions” with expected returns (Lin 2001: 29). 1 It presupposes a hierarchical social structure in the
shape of a pyramid, where resource allocation depends on structural positions (e.g.,
education, occupation, authority). It specifies social capital as resources embedded in
one’s social networks (Lin 1999b), that is, socioeconomic positions of one’s network
members.
Social capital measures can be derived from two instruments, the name generator
and the position generator. The name generator is not as useful and efficient as the
position generator to capture social capital (Van der Gaag, Snijders, and Flap 2008). The
name generator maps personal network (McCallister and Fischer 1978). It typically asks respondents to name a fixed number of contacts (usually five) with whom they discuss important matters (Burt 1984). Social capital is indicated by socioeconomic status of named contacts. The name generator captures networks characterized by strong ties,
4 small size, higher socioeconomic homogeneity, bounded contents and locations
(Campbell and Lee 1991; Marsden 1987). Further, it focuses on individuals rather than structural positions, and fails to capture the full range of resources embedded in social ties (Lin 2006).
In contrast, the position generator maps positional network (Lin and Dumin 1986;
Lin, Fu and Hsuang 2001). It asks respondents to identify contacts associated with a representative sample of ordered structural positions. If respondents know several people who are in that type of position, they are usually asked to name the one that occurs to them first. Social capital is indicated by the distribution of accessed positions. The position generator maps networks not limited by strong ties, locations, contents, and homogeneity (Lin et al. 2001; Lin forthcoming). It proves to be flexible, reliable, valid, and economical in describing access to social capital across societies (Lin 1999a; Van der
Gaag et al. 2008).
Social capital is a resource locator embedded in social relationships. It is relevant to but distinct from another three resource locators, individual-level, family-level, and community-level socioeconomic resources. Individual and family socioeconomic resources are fundamental causes of disease and illness (for a review, see Link and
Phelan 1995). Community socioeconomic resources also prove to impact health (for a review, see Robert and House 2000). By contrast, social capital comes from a relational perspective instead of from an individual or contextual perspective. As described above, social capital especially that measured through the position generator not only goes beyond the control of individuals, but also goes above the contexts of families and geographic communities. Social capital is possessed by individuals’ network members
5 rather than by individuals themselves, their families, or their neighborhoods. One can access and use social capital only through his or her social ties with network members.
Social capital theory proposes hypotheses for both the sources of social capital and its returns (Lin 2001). For the sources of social capital, this theory states that the current access to social capital is positively associated with prior personal and family socioeconomic positions. For the returns of social capital, this theory argues that social capital as network members’ resources contributes to successful actions through four mechanisms: information, influence, social credentials, and reinforcement. These mechanisms apply to health and well-being outcomes. First, network members’ resources offer valuable, timely, and updated information about health sources that is otherwise not available. Such information includes knowledge about health risks, health lifestyles and behaviors (e.g., smoking, drinking, exercise, and health check-ups), medical care, health insurance, health welfare, health service, and so on. This type of information can exert direct salutary effect for institutionalized and noninstitutionalized populations.
Second, network members’ resources exert influence that is otherwise not available. Individually possessed power and social ordering have direct effects on health- related decisions and policies, on capitalizing and controlling new health information, on exposure and vulnerability to risks of disease, and on the interaction with the dominant health care system (Adler et al. 1994; Williams and Collins 1995; Abrums 2000). Social capital such as network members’ power and authority could work in the same way.
Third, network members’ resources act as social credentials. These credentials indicate individual access to resources through social networks and relations. Social credentials could directly promote health; for example, through intervening in the
6 medical care process. A case study shows how the care and attitude from the hospital changed dramatically for a hospitalized black woman near death after her ex-husband, a physician, served as her advocate (Abrums 2000).
Fourth, network members’ resources reinforce group identification. Reinforced identification is directly related to the maintenance of mental health (Lin 2001a). Also, subjective status (e.g., class identification) protects health net of objective socioeconomic status (for a review see Schnittker and McLeod 2005). Social capital enhances subjective status (Hodge and Treiman 1968; Jackman and Jackman 1973). Social capital can thus exert indirect health effect through shaping subjective status.
In addition to these four mechanisms, I propose a material rationale for the impact of social capital on health and well-being. Individually possessed material resources (e.g., income, and wealth), primarily determined by their occupational positions, improve health through controlling access to health care and insurance, nutrition, housing, schooling, and recreation, minimizing the exposure to undesirable financial life events, and enhancing financial coping resources (Adler et al. 1994; Williams and Collins 1995;
Adler and Newman 2002). Similarly material resources of one’s network members can maintain and promote his or her health, for example, through financial support.
Considerable studies have examined social capital theory across societies. They find consistent evidence for structural sources of social capital, including personal socioeconomic status, and for the positive role of social capital in status attainment in the job market (Lin 1999a; Marsden and Gorman 2001). Much less is known about the protective effects of social capital on individual health and well-being. An exceptional study is praiseworthy as the first attention to health and well-being benefits of social
7 capital (Acock and Hurlbert 1993). It uses the name generator. It calculates the average
educational level of named contacts to indicate social capital. It finds that social capital
enhances life satisfaction and reduces anomia net of personal education and family
income. We are still not certain whether social capital measures derived from the position
generator impact health and well-being, and whether social capital interplays with other
risk factors for health and well-being.
Available data contain the position generator, and offer the unique opportunity to
examine social capital theory using measures derived from the position generator. The
goal in the present study is limited: to demonstrate the effect of social capital on health
and well-being, while controlling for as many other concepts as the data set permits. The
main purpose is to explore whether social capital impacts health and well-being over and
above the effects of personal and family socioeconomic resources in the United States, and whether social capital mediates or moderates the relationships of other structural factors with health and well-being.
HYPOTHESES
In this analysis I propose main-effect, mediating-effect, and moderating-effect hypotheses based on my theoretical interests and available data. Social capital theory expects the explanatory power of social capital for health and well-being to be independent of personal and family socioeconomic resources, due to its unique mapping of network resources. The main effect hypothesis postulates that:
H1: Social capital measured through the position generator exerts positive impact
on health and well-being net of personal and family socioeconomic status.
8 Furthermore, social capital is a function of personal socioeconomic resources which are fundamental causes of disease. The mediating effect hypothesis states that:
H2: Social capital mediates the positive impact of personal socioeconomic status
on health and well-being.
In addition, I explore the interactions between social capital and other structural stratification factors. There are few theoretical guidelines permitting me to specify specific patterns of interactions. I propose two alternative sets of hypotheses, based on conventional conceptualizations. One conceptualization, called the compensation effect, states that social capital benefits health and well-being to a greater degree for disadvantaged individuals, including females, nonwhites, and those with lower personal and family socioeconomic status, because disadvantaged individuals may be more motivated to resort to social capital to maintain and promote health and well-being (H3).
The alternative conceptualization, called the snowballing effect, contends that social capital contributes to health and well-being to a greater degree for advantaged individuals,
including males, whites, and those with higher personal and family socioeconomic status,
because advantaged individuals may be more likely to have greater social capital (H4).
H3: Social capital has a greater positive effect on health and well-being for
disadvantaged individuals (females, nonwhites, and those with lower
socioeconomic status) than for advantaged individuals (males, whites, and those
with higher socioeconomic status).
9 H4: Social capital has a greater positive effect on health and well-being for
advantaged individuals (males, whites, and those with higher socioeconomic
status) than for disadvantaged individuals (females, nonwhites, and those with
lower socioeconomic status).
DATA AND METHODS
Data
Data are drawn from the thematic research project “Social Capital: Its Origins and
Consequences,” sponsored by Academia Sinica, Taiwan, through its Research Center for
Humanities and Social Sciences, and the Institute of Sociology, and Duke University.
The US data are from a national sample of respondents aged twenty-one to sixty-four, currently or previously employed (see Lin and Dan 2008). They were collected in early
2005 using a random-digit dialing telephone survey. During the survey process, another sampling criterion was imposed to recruit more minorities (i.e., African Americans and
Latinos) so that the sample represents the census ethnic distribution. The response rate among those who were qualified and agreed to participate was 43 percent. The final sample consists of 3,000 respondents. Listwise deletion of cases with missing values in the variables of interest would incur a loss of 19 percent of the total sample. Thus I employ a recently developed multiple imputation method for missing values in the independent variables. I generate an imputed version of the raw data based on ten imputations using Stata software.2 The final analysis sample for depression includes
2,939 cases, and that for marital satisfaction includes 1,917 cases. A summary of sample characteristics is shown in Table 1.
10 Table 1 about Here
Dependent Variables
Depression. It is measured by 13 items from the CES-D scale (Radloff 1977) (see
Appendix A). The summated total score ranges from 0 to 39, with higher values
indicating more depressive symptoms. The sample mean is 6.18. The distribution of
depressive symptoms is rightly skewed. I apply a logarithmic transformation to normalize
this variable.
Marital satisfaction (applicable only to married and cohabiting respondents). It is
rated on a four-point scale (1= very satisfied, 2=moderately satisfied, 3= a little
dissatisfied, 4=very dissatisfied). I reversed the order of these responses so that the higher
the value, the more satisfied the respondent. 76 percent of married and cohabiting
respondents were very satisfied with their marriage life.
Explanatory Variable
I use the position generator to measure social capital (Lin and Dumin 1986; Lin et al.
2001). This instrument samples and assesses occupational prestige of one’s contacts in a
hierarchical society. Each respondent was asked, “Next, I am going to ask some general
questions about jobs some people you know may now have. These people include your
relatives, friends and acquaintances (acquaintances are people who know each other by
face and name). If there are several people you know who have that kind of job, please tell me the one that occurs to you first.” A list of 22 jobs was presented to respondents:
from hotel bellboy up to professor. I code the occupational prestige of each job using the
11 Standard International Occupational Prestige Scale (SIOPS) (Ganzeboom and Treiman
1996). I count the total number of occupations in each of which respondents identified one contact, and summate the prestige scores of these occupations respondents have access to. Then I generate one social capital index, average reachability, the average prestige score of the accessed occupations. It equals the summated prestige scores divided by the total number of accessed occupations. The values of average reachability range from twenty-one (i.e., respondents knew one contact in only one of the twenty-two listed positions: janitor) to seventy-eight (i.e., respondents knew one contact in only one of the twenty-two listed positions: professor). Table 2 shows the distribution of occupational positions in the position generator and average reachability.
Table 2 about Here
Control Variables
I control three groups of factors in all analyses: sociodemographic variables, personal and family socioeconomic characteristics, and subjective social status.3 Sociodemographic variables include gender (1=female), age, four dummy variables for race (white, black,
Latino, others), marital status (1=marriage, 0=non-married for analyzing depression;
1=married, 0=cohabiting for analyzing marital satisfaction). I also control for “quota”.
This dummy variable indicates whether the interview was conducted after a quota for recruiting more minorities was imposed during the survey process.
Personal socioeconomic characteristics include education and occupational prestige. Education is an ordinal variable: 1) high school or less, 2) associate program and
12 degree, 3) college degree, and 4) master’s degree or beyond. A dummy variable is created
for each category. The first category, high-school or less, is the reference group in data
analyses. Occupational status of the current or the last job is a continuous variable, based
on the NORC/GSS Occupational Prestige scores for the 1980 Occupational Classification
(Nakao, Hodge and Treas 1990; Nakao and Treas 1990). Family socioeconomic
characteristic is indicated by annual family income. It has 28 ordinal ranges. I calculate
medians of all ranges, and then take their natural logs for a normal distribution of income.
Subjective social status is indicated by self-reported class. Each respondent was
asked, “If the society is divided into upper class, upper-middle class, middle class,
middle-lower class, and lower class, which one do you think you belong to?” We
reversed the order of these five responses so that the higher the score, the higher the respondent’s subjective social status.
Analytic Strategy
I use multivariate ordinary least squares (OLS) regression models to examine the main, mediating, and moderating effects of social capital on depression and marital satisfaction.4 For each outcome, I first construct the basic model only containing control
variables, then enter social capital into the basic model to examine its main and mediating
effect, and finally enter interaction terms to examine its moderating effects. Additionally,
I use the instrumental variables method to identify the causality in the relationship of
social capital with wellbeing and health based on my cross-sectional data.
RESULTS
13 Social Capital and Depression
I first conduct OLS regression analyses of depression (see Table 3). The basic model only contains control variables (see Model 1). Consistent with previous studies, women, the younger, Latinos, and the non-married are more depressed than men, the elder, whites, and the married. Respondents, who achieved less education and lower annual family income, and perceived themselves to belong to lower classes , are more depressed than those who attained more education, higher annual family income, and identified with higher classes.
Table 3 about Here
Model 2 adds social capital to the basic mode. Net of control variables, social capital is significantly and negatively related to depression (p<.001). Respondents with higher average reachability have fewer depressive symptoms than those with lower average reachability. The educational influence on depression decreases after the entry of social capital. The coefficients for two educational categories, having an associate degree and having a master degree, become insignificant. The significance level for the coefficient for another educational category, having a college degree, also decreases from the level of .05 to that of .10. These results imply that social capital medicate the relationship between education and depression. Also note that the sizes of coefficients for gender, race, and subjective social status decrease slightly after the addition of social capital.
14 Furthermore, I explore the moderating effect of social capital through other stratification variables on depression. I constructed five groups of product terms of social capital separately with gender, race, education, occupational prestige of the current or last job, and annual family income (log). I mean-centered the occupational prestige of the current or last job and annual family income (log) before creating their interaction terms to avoid multicollinearity and to make coefficients more interpretable. I added each product term separately into Model 2. Only one product term increases the explained variance, the interaction term of social capital with gender. As Model 3 shows, the negative effect of social capital on depression is bigger for females than for males (p<.10).
Social Capital and Marital Satisfaction
Next I explore the impact of social capital on marriage satisfaction (see Table 4). The basic model only contains control variables (see Model 1). Males, the younger, whites, and the married are more satisfied with their marriage than females, the elder, blacks, and the cohabitants. Respondents achieving higher occupational prestige in their current or last job and perceiving themselves to be in higher classes are more satisfied with their marriage than those obtaining lower occupational prestige in their current or last job and
identifying with lower classes.
Table 4 about Here
Model 2 enters social capital into the basic model. Net of control variables, social
capital is positively associated with marriage satisfaction at a marginal significance level
15 (p<.10). Respondents with higher average reachability are more satisfied with their
marriage than those with lower average reachability. Also note that the magnitudes of
coefficients for gender, race, marital status, and subjective social status decrease slightly after the entry of social capital.
Additionally, I explore the moderating effect of social capital through other
stratification variables on marital satisfaction in the same way I analyze depression. I
enter each product term separately into Model 2. Three groups of product terms increase
the explained variance, the interaction terms of social capital with gender, education, and
occupational prestige. As Model 3 shows, the positive effect of social capital is stronger
for females than for males (p<.05). As Model 4 shows, the positive influence of social
capital is bigger for respondents receiving higher school education or less than for those
achieving an associate degree (p<.10). As Model 5 shows, the positive impact of social
capital decreases across levels of occupational prestige of the current or last job (p<.10).
Finally, I employ the instrumental variable method to identify the causality in the
social capital-wellbeing relationship based on our cross-sectional data. I use two
instrumental variables, household size (log), and voluntary participation (1=yes, 0=no),
for the prediction of each outcome. Theoretically these two instruments have direct
associations with the establishment of social ties and the accumulation of social capital,
but no direct relationship with each health outcome. Statistically these two instruments
are significantly correlated with social capital, but not with each health outcome. The
Hausman test results fail to reject the null hypothesis that OLS regression estimates and
IV regression estimates are equal at the significance level of .01 for both outcomes
(Prob>chi2=1.000). 5 This evidence allows us to comfortably assert that the relationship
16 between social capital and well-being is not an artifact of reverse causation or incidental
association.
CONCLUSIONS
This research extends social capital theory into disease and illness. I examine whether
social capital impacts depressive symptoms and marital satisfaction over and above the
effects of personal and family socioeconomic status. Main results from OLS regression analyses of both outcomes are consistent. First, findings support the main effect hypothesis (H1). Social capital exerts direct effect net of personal and family socioeconomic resources. Second, findings support the mediating effect hypothesis (H2).
It holds only for the relationship between personal education and depression. Personal education decreases depression through advancing social capital. Third, for the moderating effect hypotheses, findings consistently support the compensation effect hypothesis (H3) instead of the snowballing effect hypothesis (H4). The compensation effect hypothesis holds for multiple relationships. Social capital moderates the effect of gender on depression, and the effects of gender, individual education and occupational prestige on marital satisfaction. Additionally, findings from instrumental variable analyses provide strong evidence for the causal flow running from social capital to health.
This study extends relevant literature theoretically in four ways. First, its findings advance our understanding of the social dynamics by which social capital influences health and well-being, although this study does not add statistically high explanatory powers to health and well-being stratification. Social capital is one of the most acknowledged contributions from sociology to other fields of social science during the
17 past two decades (Portes 1998). Within sociology this concept has received little attention
in the health and well-being literature. This study shows that social capital is not only a social determinant of health and well-being but also interplay with other structural risk
factors. This study sheds light on a new research area of social capital studies and on a
new social source of persistent disparities in health and well-being. Second, its findings
enhance the significance of social relationships in the social production of health and
well-being. Social relationship is one of the most established social antecedents of
disease (House 2001). This study indicates that social capital is another unique
relationship-based risk factor, and that socioeconomic resources can flow through social
relationships to produce health and well-being disparities. Furthermore, social capital
theory is embedded in the traditional stratification literature in sociology. The
establishment of the relationship of social capital with health and well-being contributes
to relating health and well-being disparities with general stratification theories in
sociology. Additionally, our study strengthens the structural perspective in medical
sociology, one unique sociological approach to health and well-being (Bird, Conrad, and
Fremont 2000). Social capital can be another important approach to social structure,
health, and well-being because network members’ structural positions reflect “structural
arrangements in which individuals are embedded” (Pearlin 1989: 241).
This study is only a beginning effort in applying social capital theory to health
and well-being stratification. Future studies are needed to confirm and generalize the
contribution of social capital to the social processes of well-being production in two
directions. First, given our results that social capital plays a stronger protective role for
depressive symptoms than for marital satisfaction, future studies should examine the link
18 between social capital and other well-being outcomes. Evidence from multiple outcomes
will help to testify a persistent causal association between social capital and well-being
(Link and Phelan 1995). Second, I speculate on five potential mechanisms between social
capital and well-being. Available data only allow us to explore social capital as a
hierarchical structural constraint and an upstream cause of well-being inequality, and
examine the effect of social capital on well-being rather than to testify mechanisms
directly. Future studies need to explore the downstream causes, that is, potential
mediators and moderators between social capital and well-being, for a better
understanding of how this unique social factor shapes well-being. In addition, I
distinguish social capital from personal, family, and community socioeconomic resources
in terms of conceptualization and operationalization. Available data only allow me to
examine the direct effect of social capital net of personal and family socioeconomic
resources, and its mediating and moderating effect on the relationship between personal
and family socioeconomic resources. Future research should analyze the effect of all
these four resources locators in a single study, and explore their causal relationships.
There are also two data limitations that should be kept in mind. First, the data are
based on a cross-sectional research design. Both social capital and well-being outcomes
were measured at the survey time. Despite the fact that our instrumental variables
analyses support the social causation argument that social capital primarily influences
well-being rather than vice versa, a process of social selection is possible. Physical or
mental illness may prevent individuals from knowing or contacting others with higher
social positions. Future studies should use longitudinal data to examine the competing
arguments of social selection and social causation. Additionally, our data are from a
19 national sample of respondents aged twenty-one to sixty-four, currently or previously employed. Future studies need a national sample of respondents from all ages and all employment backgrounds for the purpose of generalizability.
20 NOTES
1. A growing body of health literature equates social capital with other established concepts derived from general social networks and community social networks, such as social integration, social support and social cohesion (Baum and Ziersch 2003; Carpiano
2006; Kawachi, Subramanian, and Kim 2008; Whitley and McKenzie 2005). A detailed discussion on the distinction between social capital as Lin proposes and the other relationship-based social concepts is beyond the scope of this study. In brief, Social integration is the extent of participation in social networks, indicated by active engagement in social roles and social activities, and cognitive identification with network members (Brissette, Cohen, and Seeman 2000). Social support is the assistances from social networks, indicated by the quantity and quality of perceived or received help from network members (Lakey and Cohen 2000; Pearlin 1989). Social cohesion is the degree of social bonds and social equality within social networks, indicated by trust, reciprocity, and the lack of social conflict (Kawachi and Berkman 2000). By comparison, social capital uniquely captures structural positions possessed by individuals’ network members, which differs from individuals’ own social participation, their network members’ assistance, or equality, trust and reciprocity between them and their network members.
2. See Royston (2004, 2005a, 2005b) for documentation of two Stata user-written programs, Ice and Micombine.
3. Some may argue that the relationship between age and depression is nonlinear, and that the duration of marriage or cohabitation and the presence of children may influence marriage satisfaction. In parallel analyses, I find no significant correlations between age
21 and depression, and between years of marriage or cohabitation, the presence of kids and marital satisfaction. I choose not to control them in my multivariate analyses.
4. Parallel analyses, using generalized ordered logit/partial proportional odds models to predict marital satisfaction as an ordinal variable, found similar results. I report results from OLS for the sake of simplicity.
5. Results of IV analyses are available upon request.
22 REFERENCES
Abrums, Mary. 2000. "‘Jesus Will Fix it After Awhile’: Meanings and Health." Social
Science and Medicine 50:89-105.
Acock, Alan C. and Jeanne S. Hurlbert. 1993. "Social Networks, Marital Status, and
Well-being." Social Networks 15:309-34.
Adler, Nancy E., Thomas Boyce, Margaret A. Chesney, Sheldon Cohen, Susan Folkman,
Robert L. Kahn, and S. Leonard Syme. 1994. "Socioeconomic Status and Health:
the Challenge of the Gradient." The American Psychologist 49:15-24.
Adler, Nancy E. and Katherine Newman. 2002. "Socioeconomic Disparities in Health:
Pathways and Policies." Health Affairs 21:60-76.
Bird, Chloe E., Peter Conrad, and Allen M. Fremont. 2000. "Medical Sociology at the
Millennium." Pp. 1-32 in Handbook of Medical Sociology, edited by C. E. Bird, P.
Conrad, and A. M. Fremont. Upper Saddle River, NJ: Prentice Hall.
Burt, Ronald S. 1984. "Network Items and the General Social Survey." Social Networks
6:293-339.
Bourdieu, Pierre. [1983]1986. "The Forms of Capital." Pp. 241-58 in Handbook of
Theory and Research for the Sociology of Education, edited by J. G. Richardson.
Westport, CT: Greenwood Press.
Campbell, Karen E. and Barrett A. Lee. 1991. "Name Generators in Surveys of Personal
Networks." Social Networks 13:203-21.
Coleman, James S. 1988. "Social Capital in the Creation of Human Capital." American
Journal of Sociology 94:95-121.
23 Ganzeboom, Harry B.G. and Donald J. Treiman. 1996. "Internationally Comparable
Measures of Occupational Status for the 1988 International Standard
Classification of Occupations." Social Science Research 25:201-239.
Hodge, Robert W. and Donald J. Treiman. 1968. "Class Identification in the United
States." American Journal of Sociology 73:535-47.
House, James S. 2001. "Understanding Social Factors and Inequalities in Health: 20th
Century Progress and 21st Century Prospects." Journal of Health and Social
Behavior 43:125-42.
Jackman, Mary R. and Robert W. Jackman. 1973. "An Interpretation of the Relation
between Objective and Subjective Social Status." American Sociological Review
38:569-82.
Lin, Nan. 1999a. "Social Networks and Status Attainment." Annual Review of Sociology
25:467-88.
------. 1999b. "Building a Network Theory of Social Capital." Connections 22:28-51.
------. 2001. Social Capital: A Theory of Social Structure and Action. Cambridge:
Cambridge University Press.
------.2006. "Social Capital." Pp. 604-12 in Encyclopedia of Economic Sociology,
edited by J. Beckert and M. Zafirovski. London: Routledge.
------. Forthcoming. "A Network Theory of Social Capital." In Handbook on Social
Capital, edited by D. Castiglione, J. V. Deth, and G. Wolleb. Oxford: Oxford
University Press.
Lin, Nan and Mary Dumin. 1986. "Access to Occupations through Social Ties." Social
Networks 8:365-85.
24 Lin, Nan, Yang-Chih Fu, and Ray-May Hsung. 2001. "The Position Generator: A
Measurement Technique for Investigations of Social Capital." Pp. 57-81 in Social
Capital: Theory and Research, edited by N. Lin, K. Cook, and R. S. Burt. New
York: Aldine de Gruyter.
Lin, Nan and Dan Ao. 2008. “The Invisible Hand of Study: An Exploratory Study.” Pp.
107-32 in Social Capital: Advances in Research, edited by Nan Lin and Bonnie
Erickson. New York: Oxford University Press.
Link, Bruce G. and Jo C. Phelan. 1995. "Social Conditions as Fundamental Causes of
Disease." Journal of Health and Social Behavior Extra Issue: 80-94.
Marsden, Peter V. and Elizabeth H. Gorman. 2001. "Social Networks, Job Changes, and
Recruitment." Pp. 467-502 in Sourcebook on Labor Markets: Evolving Structures
and Processes, edited by I. Berg and A. L. Kalleberg. New York: Kluwer
Academic/Plenum Publishers.
Marsden, Peter V. 1987. "Core Discussion Networks of Americans." American
Sociological Review 52:122-31.
McCallister, Lynn and Claude S. Fischer. 1978. "A Procedure for Surveying Personal
Networks." Sociological Methods & Research 7:131-48.
Nakao, Keiko, Robert W. Hodge, and Judith Treas. 1990. "On Revising Prestige Scores
for All Occupations." NORC, Chicago.
Nakao, Keiko and Judith Treas. 1990. "Computing 1989 Occupational Prestige Scores."
NORC, Chicago.
Pearlin, Leonard I. 1989. "The Sociology Study of Stress." Journal of Health and Social
Behavior 30:241-56.
25 Portes, Alejandro. 1998. "Social Capital: Its Origins and Applications in Modern
Sociology." Annual Review of Sociology 24:1-24.
Putnam, Robert D.1993. Making Democracy Work: Civic Traditions in Modern Italy.
Princeton, N.J.: Princeton University Press.
Radloff, Lenore Sawyer. 1977. "The CES-D Scale: A Self-Report Depression Scale for
Research in the General Population." Applied Psychological Measurement 1:385-
401.
Robert, Stephanie A. and James S. House. 2000. "Socioeconomic Inequalities in Health:
Integrating Individual-, Community-, and Societal-Level Theory and Research."
Pp. 115-35 in Handbook of Social Studies in Health and Medicine, edited by Gary
L. Albrecht, Ray Fitzpatrick, and Susan C. Scrimshaw. Thousand Oaks: SAGE.
Royston, Patrick. 2004. "Multiple Imputation of Missing Values." The Stata Journal 4:
227-41.
------. 2005a. "Multiple Imputation of Missing Values: Update." The Stata Journal 5:
88-102.
------. 2005b. "Multiple Imputation of Missing Values: Update of ICE." The Stata
Journal 5: 527-36.
Van der Gaag, Martin P. J., Tom A. B. Snijders, and Henk D. Flap. 2008. "Position
Generator Measures and their Relationship to Other Social Capital Measures." Pp.
27-48 in Social Capital: Advances in Research, edited by N. Lin and B. Erickson.
New York: Oxford University Press.
Wayman, Jeffrey C. 2003. "Multiple Imputation for Missing Data: What Is It and How
26 Can I Use It?" Paper presented at the 2003 Annual Meeting of the American
Educational Research Association, Chicago, IL, April.
Williams, David R. and Chiquita Collins. 1995. "US Socioeconomic and Racial
Differences in Health: Patterns and Explanations." Annual Review of Sociology
21:349-86.
27 Table 1 Summary of sample characteristics Analysis Sample One Analysis Sample Two Mean or Percent SD Mean or Percent SD Dependent variables Summed depressive symptoms 6.18 5.66 Marriage satisfaction (%) Very dissatisfied 1.46 A little dissatisfied 3.44 Moderately satisfied 18.99 Very satisfied 76.11 Control Variables Gender (1=female) (%) 54.07 56.13 Age (mean) 37.49 10.53 38.00 9.82 Married (1=married 0=unmarried) (%) 64.04 Married (1= married 0=cohabiting) (%) 98.17 Race (%) White 69.31 73.87 Black 11.77 7.82 Latino 13.20 13.20 Other race 5.72 5.11 Quota (%) 43.52 43.92 Education (%) High school 39.16 38.86 Associate degree 20.93 20.66 College degree 25.14 24.88 Master degree 14.77 15.60 Occupational prestige (mean) 45.68 13.97 46.29 14.10 Annual family income (%)* (N=2,556) (N=1,636) Less than 35,000 (USD) 25.55 17.54 35,000-60,000 27.66 25.00 60,000-90,000 23.67 27.81 90,000 and more 23.12 29.65 Subjective social status (%) Lower class 5.83 4.00 Middle-lower class 16.06 13.86 Middle class 57.93 58.93 Upper-middle class 17.80 20.81 Upper class 2.39 2.39 Household size 3.21 2.01 3.65 2.03 Voluntary participation (%) 74.45 77.26 Number of observations 2,939 1,917 Note: Depression and marriage satisfaction are respectively the dependent variable in
analysis sample one and analysis sample two; means or percents of most variables are
28 based on imputed data (Wayman 2003); we imputed the missing values for the natural log of annual family income, thus we report the distribution of grouped annual family income in the raw data for simple description.
29 Table 2 Distribution of occupational positions in the position generator and social
capital indices
Analysis Sample One Analysis Sample Two Position (SIOPS) Respondent Accessing Respondent Accessing (Percent) (Percent) Professor (78) 37.12 38.60 Lawyer (73) 55.39 56.34 CEO (70) 20.11 20.81 Congressman (64) 12.08 12.00 Production manager (63) 16.94 18.15 Middle school teacher (60) 48.89 50.18 Personnel manager (60) 33.04 33.18 Writer (58) 21.67 21.02 Nurse (54) 70.26 71.10 Computer programmer (51) 48.96 48.77 Administrative assistant (49) 31.51 30.52 Accountant (49) 31.27 32.81 Policeman (40) 50.49 52.01 Farmer (38) 42.60 45.38 Receptionist (38) 38.86 39.07 Operator in a factory (34) 25.55 27.75 Hairdresser (32) 60.26 61.50 Taxi driver (31) 8.81 7.88 Security guard (30) 24.57 23.11 Housemaid (23) 27.15 29.21 Janitor (21) 28.72 29.84 Hotel bellboy (20) 2.72 2.40
Social capital index Average reachability Mean 48.66 48.65 S. D. 6.60 6.46 Range of scores 21-78 23-78 Number of observations 2,939 1,917 Note: Depression and marriage satisfaction are respectively the dependent variable in
analysis sample one and analysis sample two; occupational prestige scores in parentheses.
30
31 Table 3 OLS regressions of depression on control variables, social capital, and interaction term (N=2,939) Model 1 Model 2 Model 3 Gender - Female .133*** .123** .574* (.036) (.036) (.259) Age -.005** -.005** -.004** (.002) (.002) (.002) Race (base: white) Black .041 .031 .035 (.058) (.058) (.058) Latino -.174** -.170** -.169** (.057) (.057) (.057) Other race -.025 -.019 -.025 (.077) (.077) (.077) Marital status -.191*** -.196*** -.197*** (.038) (.038) (.038) Quota -.000 -.000 -.000 (.000) (.000) (.000) Education (base: high school or less) Associate -.089† -.073 -.071 (.048) (.048) (.048) College -.123* -.082† -.084† (.048) (.049) (.049) Master -.143* -.096 -.099 (.060) (.061) (.061) Occupational prestige (current/last job) -.001 -.001 -.001 (.001) (.001) (.001) Annual family income (log) -.106*** -.106*** -.105*** (.027) (.027) (.027) Subjective social status (self-reported class) -.123*** -.116*** -.115*** (.024) (.024) (.024) Social capital (average reachability) -.012*** -.007† (.003) (.004) Social capital*female -.009† (.005) Constant 3.443*** 3.955*** 3.705*** (.276) (.300) (.332) Adjusted R-squared .0600 .0656 .0663 Note: Standard errors in parentheses; † p≤ .10; * p≤ .05; ** p≤.01; *** p≤.001 (two-tailed
tests).
32 Table 4 OLS regressions of marital satisfaction on control variables, social capital, and interaction terms (N=1,917) Model 1 Model 2 Model 3 Model 4 Model 5 Gender - Female -.066* -.062* -.492* -.061* -.063* (.028) (.028) (.212) (.028) (.028) Age -.004** -.004** -.004** -.004** -.004** (.001) (.001) (.001) (.001) (.001) Race (base: white) Black -.110* -.107* -.111* -.107* -.105* (.053) (.053) (.053) (.053) (.053) Latino -.072 -.076† -.077† -.075 -.077† (.046) (.046) (.046) (.046) (.046) Other race .048 .042 .049 .050 .048 (.063) (.063) (.063) (.064) (.063) Marital status .291** .288** .276** .278** .284** (.103) (.103) (.103) (.103) (.103) Quota .000 .000 .000 .000 .000 (.000) (.000) (.000) (.000) (.000) Education (base: high school) Associate -.008 -.014 -.015 .497† -.017 (.038) (.038) (.038) (.287) (.038) College -.037 -.050 -.048 -.140 -.054 (.038) (.039) (.039) (.303) (.039) Master -.008 -.023 -.019 .415 -.018 (.047) (.048) (.048) (.335) (.048) Occupational prestige (current/last job) .002† .002† .002† .002† .016* (.001) (.001) (.001) (.001) (.007) Annual family income (log) .033 .033 .031 .032 .031 (.021) (.021) (.021) (.021) (.021) Subjective social status .082*** .078*** .077*** .079*** .078*** (self-reported class) (.020) (.020) (.020) (.020) (.020) Social capital (average reachability) .004† -.001 .007* .003 (.002) (.003) (.003) (.002) Social capital*female .009* (.004) Social capital*associate -.011† (.006) Social capital*college .002 (.006) Social capital*master -.009 (.007) Social capital* occupational prestige -.0002† (.0002) Constant 2.903*** 2.730*** 3.003*** 2.599*** 2.886*** (.237) (.255) (.288) (.275) (.262) Adjusted R-squared .0310 0322 .0339 .0333 .0337
33 Note: Standard errors in parentheses; † p≤ .10; * p≤ .05; ** p≤.01; *** p≤.001 (two-tailed tests)
34 APPENDIX A
CES-D scale. The thirteen items used are:
Next, I am going to read a list of the ways you might feel. For each description
that I read to you, please tell me how often you have felt this way during the past
week……
1. I did not feel like eating; my appetite was poor
2. I felt like everything I did was an effort
3. My sleep was restless
4. I felt depressed
5. I felt lonely
6. People are unfriendly
7. I felt sad
8. I could not get going
9. I was bothered by things that usually do not bother me
10. I felt I could not shake off the blues even with the help of my family/friends
11. I felt fearful
12. I had crying spells
13. I felt that people disliked me.”
These indicators are rated on a four-point scale (0= rarely or none of the time:
less than one day in the past week; 1= some or little of the time: 1-2 days in the past week;
2= occasionally or moderate amount of time: 3-4 days in the past week; 3= most or all of the time: 5-7 days in the past week).
35
Structural and Individual Determinants of Social Capital --- A
Multilevel Analysis1
Shanhui Wu
Department of Sociology
Duke University
December 2nd, 2007
Words: 8660
1 Direct all correspondence to Shanhui Wu, Department of Sociolgoy, Duke University, Box 90088, Durham, NC 27708 ([email protected]) . Data used in this article were drawn from the thematic research project “Social Capital: Its Origins and Consequences", sponsored by Academia Sinica, Taiwan, through its Research Center for Humanities and Social Sciences, and the Institute of Sociology. The principal investigator of the project is Nan Lin. I gratefully acknowledge Professor Nan Lin, Philip Morgan, Xueguang Zhou, Angela O’Rand, and Kenneth Land for their helpful advices and comments. An early version of this article was presented at the 2005 American Sociological Association annual meeting.
1 Abstract:
The significance of social capital for stratification has attracted research interest
in the unequal distribution of social capital among individuals. The factors investigated
by prior studies, however, are mostly at the individual level. Using a 2005 multilevel data
set from 167 cities in China, this study identifies the macro-micro linkages that determine
individuals’ social capital. The results from multilevel linear and negative binomial regressions reach three conclusions. First, individuals’ social capital is influenced by
socioeconomic development at the city level. People who live in more developed cities
access more social capital, measured as the number of social contacts, secondary-group
ties, and high-status persons reachable through network connections. Second, cross-city
geographic mobility and father’s network extensity positively affect individuals’ social capital. Migrants and people whose fathers have extensive networks access more social
capital. Third, there are significant interactions between structural and individual factors
in determining social capital. The positive effects of economic development on social
capital are smaller for people from families with better networks, and the advantages
created by father’s network extensity are moderated by economic development.
2 Over the last three decades, researchers have demonstrated that social capital ---
social resources available through personal networks --- provides competitive advantages
for social achievements. In the labor market, social capital helps individuals find a high-
status and well-paid job, enhances access to power and authority, aids in receiving early promotion, or assist in starting up an enterprise (Granovetter 1974; Ensel 1979; Lin et al.
1981; Marsden and Hurlbert 1988; Sprengers et al. 1988; Boxman et al. 1991; Burt 1992;
Erickson 1996; Bian 1997; Fernandez and Weinberg 1997; Volker and Flap 1999;
Petersen et al. 2000; Lin 2001; Renzulli and Aldrich 2005; McDonald and Elder 2006).
The significance of social capital for stratification has attracted extensive research
interest in the unequal distribution of social capital among individuals. This literature reveals that an individual’s family background, personal characteristics, and achieved status influence the size and composition of his/her social network and therefore the
embedded social resources that constitute social capital. However, the factors and
mechanisms examined by this literature are exclusively at the individual level. Few
studies have systematically explored the impact of social structure on people’s social
capital. The purpose of this article is to investigate this connection.
By social structure, I refer to the social units that contain individuals.
Neighborhoods, communities, villages, and cities are examples of these social units.
Individuals pursue social capital under certain social structures (Frank and Yasumoto
1998), and macro level social contexts define the opportunity that precludes or makes
possible various kinds of social interactions and contacts (Blau 1977). As Kadushin
(2004) argues, a community with a strong culture of trust and mutual aid will promote
social contacts among residents, while an anomic community may hinder social
3 interactions. Combining structural and individual factors should further our
understanding of the mechanisms that generate unequal distribution of social capital among individuals.
The social units under examination here are cities. Using 2005 survey data of 167
China cities and multilevel modeling techniques, this article specifies the micro-macro
relations through which city-level and individual-level factors interact with each other to
determine people’s social capital. Specifically, I examine socioeconomic development of
a city at the macro level, and family background in social networks and cross-city
geographic mobility and migration at the micro level. I argue that: 1) the social capital
one can access is influenced by city-level social contexts; 2) individuals differ in
accessing social capital according to their migrant status and family background in social
networks; and 3) there are interaction effects between structural and individual
determinants on social capital. The theoretical framework is presented in Figure 1.
(Figure 1 about here)
In addition to contributing to social capital literature, this study may also advance
the understanding of social stratification in China. First, Chinese society is well known
for its emphasis on social networks or guanxi in economic and social organizations
(Walder 1989; Cheng and Rosett 1991; Smart 1993; Bian 1997). Social capital may play
a greater role in stratification in China than in many other societies. Second, as a result of
the development strategy adopted by the government in the reform period, huge
development discrepancies exist among regions and cities in China (Hauser and Xie
4 2004). If social capital is determined by city-level social contexts, the differences in
socioeconomic development between cities will generate unequal distribution of social
capital among their citizens, and ultimately create and reinforce regional social inequality in general.
INDIVIDUAL DETERMINANTS OF SOCIAL CAPITAL
Concepts of social capital suggest the values of social interactions and relations
among individuals, families, and communities (Coleman 1988; Putnam 1993; Lin 1999).
At the individual level, social capital is defined as the resources embedded in social
networks (Bourdieu 1986; Lin 2001). The volume of social capital available to an
individual depends on the size and composition of his/her social network, and the social
resources held by network members (Bourdieu 1986; Flap 2002). The bigger the size and the more diverse and resourceful its members, the more social capital a network contains.
Social capital creates advantages in the labor market because it provides a number
of functions: (1) it offers opportunities for accessing and obtaining information, (2) it
provides venues for social ties to exert influence, (3) it augments one’s credentials, and (4)
it reinforces personal identity. With these functions, individuals need not even activate
their social capital (e.g., use social contacts in job search) to receive the benefits. 2
People with more social capital tend to receive unsolicited job information from personal contacts through routine conversations (Granovetter 1995; Lin 2004; McDonald and
2 Indeed, using personal contacts to search for jobs does not necessarily provide benefits (Granovetter 1973; Lin 1999; Mouw 2003).
5 Elder 2006), and social capital increases job seekers’ attractiveness to employers
(Erickson 2001). As a result, social capital helps people find high-status and well-paid
jobs (Sprengers et al. 1988; Boxman et al.1991; Erickson 1996; Lin 2004); gain more power and influence within organizations, communities, and nations (Moore 1979; Brass
1985; Laumann and Knoke 1987; Miller 2006); and get promoted more quickly (Burt
1992; Podolny and Baron 1997). Social capital also exerts a positive effect on business start-ups. Renzulli et al. (2000) show that a social network with a high proportion of kin and homogeneous ties, which contain less social capital, is a critical disadvantage for owning a business.
Numerous studies have connected individuals’ social capital with their family
background, personal characteristics, and achieved status. First, people whose parents
have higher educational and occupational status are found to access more social capital
(Lin and Dumin 1986; Barbieri 1996; Volker and Flap 1999). Second, females have less
social capital than males because their social networks are smaller in size (Fischer 1982;
Campbell 1988), consist of more localized contacts with kin and neighbors (Marsden
1987; Moore 1990; Scott 1996; Renzulli et al. 2000), are confined to smaller and
peripheral organizations (McPherson and Smith-Lovin 1982), and include less powerful
and resourceful members (McPherson and Smith-Lovin 1986; Scott 1996; Lin 2001).
Third, the size of a social network increases with one’s age until the individual’s
resources and social involvements begin to shrink at a certain stage (Morgan 1988;
Erickson 1996; Pugliesi and Shook 1998). Finally, it is observed that educational and
work-related status influences social capital through altering the size and composition of
social networks. Moore (1990) demonstrates that network size and proportion of nonkin
6 ties are positively related to education, and individuals with full-time jobs and who occupy managerial or professional occupations had larger social networks and less kin members in their networks. Therefore, an individual’s achieved status in terms of education and occupation helps create larger and more diversified social networks that contain more social capital (Lin and Dumin 1986).
Two additional individual determinants are investigated here. One is family background in terms of social networks. Individuals from families that are richer in social networks are exposed to the effectiveness of social capital earlier and more frequently, and are better educated in the social skills for making social connections.
More importantly, these individuals can even directly “inherit” social capital from family members by being introduced to their social networks and transferring family members’ friends to their own friends. This mechanism of transferring social capital is especially significant within the context of China. In Chinese cultural traditions, family is the core of the social structure and the original source of social relations. Individuals build social connections with outsiders and turn such connections into highly intimate ties through the web of family members and kinship (Fei 1992 [1949]; Fried 1969 [1953]; King 1985; Lin
1989).
H1: Individuals from families that are richer in social networks have more social
capital.
7 The other factor under consideration is cross-city geographic mobility and migration. Researchers have debated the effect of geographic migration on personal networks and social capital. Some view geographic migration as a disruptive force to individuals’ networks of social relations. They contend that migrants sever their social connections in areas of origin when they move away. As strangers in areas of destination who face a different environment and culture, migrants may experience stressful lives and be isolated from local social networks. Therefore, migrants’ social networks tend to be smaller and contain fewer resources (McLanahan and Sandefur 1994; South and
Haynie 2004). Other researchers argue that geographic migration can be beneficial to migrants’ social networks. After a short adjustment, migrants are able to establish social ties in the new environment, and advanced communication and transportation systems enable migrants to maintain social connections to their areas of origin (Litwak 1960;
Kaufer and Carley 1993; Fischer 2002; Bidart and Lavenu 2005). Considering that migrants are presented with opportunities to meet diverse people and build social relations in their new locations, while still keeping some of their old social contacts in their areas of origin, I expect that cross-city migration will enhance individuals’ personal networks and social capital in the long run.
The positive effect of geographic mobility applies to people who have temporary cross-city mobility (e.g., a short-term visit to another city during a business trip). While temporary movers gain the opportunities of meeting new people through their physical mobility, they do not encounter the problems of environmental and cultural differences that permanent migrants do in their areas of destination. Therefore, the opportunities
8 provided by short-term mobility across cities may be a pure addition to the social capital of these people.
H2: Permanent migration across cities enhances an individual’s social capital.
H3: Individuals who have temporary mobility across cities have more social capital.
STRUCTURAL DETERMINANTS OF SOCIAL CAPITAL
The size and composition of personal networks that constitute social capital depend on the local contexts that define the opportunity structure for social interactions and contacts. These contexts can be the size and distribution of population. Mayhew and
Levinger (1976) argue that the density of social interactions among people increases as local population size grows. Population growth also promotes the social division of labor in an area. Consequently, individuals’ social networks tend to become larger and more diverse when local population size increases. Blau’s (1977) structural opportunity theory suggests that heterogeneity of population distribution and inequality in the distribution of social resources in a community or society create structural opportunities for social associations across different social groups. To the extent that intergroup contacts link an individual to more diverse people, the structural opportunities created by local contexts promote individuals’ social capital.
9 The macrostructural parameters I will examine in this article are economic and
educational development at the city level. From a broader perspective of Simmel (1958
[1908]), individuals’ social networks tend to be more diverse in members as a society or
community becomes more developed and complex. Specifically, I argue that
socioeconomic development at the city level enhances individuals’ social capital through three mechanisms: by increasing resource abundance, creating structural opportunities, and inducing institutional changes that affect social relations.
The most direct way that socioeconomic development influences individuals’ social
capital is to increase the resource abundance within a locality. As a city becomes more
affluent and its residents occupy more resources, the social capital one can obtain through
his/her personal network members increases. Individuals living in a more developed area
also have more resources to invest in social networking. For example, the advanced
systems of communication and transportation associated with socioeconomic
development make social contacts and interactions easier and more frequent (Kaufer and
Carley 1993).
Second, socioeconomic development promotes structural opportunities for social
contacts by creating new social spaces where people can interact with one another. A developed area is more capable of providing resources to support the development and maintenance of community organizations and activity programs in which residents come
together to form and sustain social relations. Jencks and Mayer (1990) reveal that in
wealthy communities, after-school programs increase the social interactions of children
and their parents alike. In China, it has been observed that new institutions, such as night
classes, training organizations, clubs, and voluntary organizations, emerge when an area
10 becomes more developed, thus providing more places for individuals to meet and form
social ties with others (Ruan et al. 1997). Because individuals now build social
connections through more differential social dimensions, their networks increase not only
in size but also in heterogeneity.
Third, socioeconomic development helps people access more social capital by
inducing institutional changes that alter the composition of individuals’ personal
networks. One of the institutional changes during the development process is family
structure. As family sociologists argue, the cultural values concerning coresidence of
married children with parents will be undermined by the development process. At the
same time, when individuals gain economic independence and new institutional
arrangements (e.g., school, service industry, and voluntary organization) emerge to take
over the traditional tasks of family in socialization and social support, parents and adult
children will have less need to live together. These changes in cultural values and
material conditions make extended families less commonplace in modern societies (Levy
1949; Goode 1970; Treas and Wang 1993; Logan and Bian 1999). As a result, people
tend to spend less of their lifetimes with family members and kin than with others outside
their family clans. Secondary-group relations form an increasing part of individuals’
social networks as an area becomes more developed (Ruan et al. 1997). In the sense that
secondary groups are more heterogeneous in social resources than family members and kin, the change of network composition leads individuals to more social capital.
11 H4: Individuals who live in a city that is more developed economically have more
social capital.
H5: Individuals who live in a city that is more developed educationally have more
social capital.
INTERACTIONS OF STRUCTURAL AND INDIVIDUAL EFFECTS
In addition to their direct effects, structural and individual determinants may
interact with each other to influence social capital. The effects of local socioeconomic
development may vary for individuals with different attributes, and the effects of individual factors may also be moderated by socioeconomic development. My theoretical interest in the interaction effect centers on the question of who will benefit more from socioeconomic development in terms of social capital. In other words, will socioeconomic development enlarge or reduce the social capital inequality created by the factors at the individual level?
I suspect that there are negative interactions between socioeconomic development
and the two individual-level factors of family background in social networks and cross-
city migration. In a less-developed city where material resources and structural
opportunities for social interactions are comparatively limited, family is the major
channel through which individuals develop their personal networks. Individuals whose
families have better social networks therefore gain advantages in social capital. These
12 advantages generated by family networks will be reduced as alternative channels for social contacts are created in the development process. Moreover, because individuals from families that are richer in social networks spend more time and energy on interacting with family networks, they may be less likely to take advantage of the new opportunities for social interactions created by socioeconomic development. Therefore, socioeconomic development may be more beneficial for those whose family members have limited social networks.
Similar patterns exist for the interaction between permanent cross-city migration and socioeconomic development. As strangers who face a new environment and culture and possible discrimination in the places of destination, migrants are less likely than original residents to be integrated into local social networks and organizations (Jencks and Mayer 1990), and therefore are less capable of receiving the benefits from local development. Socioeconomic development also moderates the advantages of migration in providing new opportunities for social interactions, as advanced communication systems take the place of physical mobility in promoting and maintaining social contacts over long distance. 3
H6: there are negative interaction effects of socioeconomic development and family
networks on social capital.
H7: there are negative interaction effects of socioeconomic development and
permanent cross-city migration.
3 The interaction effects do not apply to temporary cross-city movers because they return to the areas of origin after a short-term visit to other cities. They are not newcomers in the cities that they lived in.
13 DATA AND METHODS
Data
The individual-level data comes from a 2005 national random sample of adults
residing in urban China. Multistage systematic probability sampling techniques are used
to draw the sample. At the initial stage, households in all urban cities were sequenced.
Clustering of nineteen consecutive households was the basic unit of sampling at this stage.
An interval was thus established and a random start number was drawn. Starting from
the top of the list of all households by cities, clusters of households were sampled,
resulting in 184 clusters from 167 cities. At the second stage, all qualified respondents in
each sampled household were identified, and the one whose birth date was closest to June
30 was designated as the sample respondent. A personal interview was conducted by
professional interviewers between November 2004 and March 2005. The response rate is
about 40 percent.4 The total sample size is 3529. City-level data on socioeconomic
development in 2003 is compiled from ACMR All China Data Center. After deleting the
cases with missing value, the samples include around 2700 respondents in 159 cities. 5
Measuring Dependent Variables
I adopt the position generator proposed by Lin and Dumin (1986) to measure individuals’ social capital. It uses a sample of occupational positions with different
prestige scores and asks respondents whether he/she has social contacts with people who
occupy these occupations. In comparison with the name generator that emphasizes strong
4 The relatively low response rate was due to several factors. We held the sampled respondents rigidly and without replacement. Respondents who could not be contacted and interviewed after the initial attempt and follow-ups were counted as lost. We also found that more and more urban residents in China are becoming less inclined to be interviewed, and we made no effort to force their participation. 5 The sample size slightly differs for the three measures of social capital specified below.
14 tie relations, this method is thought to be better in measuring the overall networks of
individuals (Campbell and Lee 1991; Lin 2001).
From the information collected, I construct three indicators of social capital that
measure the size, diversity, and embedded resources of respondents’ personal networks.
The first measure is the number of positions with which a respondent has social contacts.
The second measure is the proportion of secondary-group contacts (i.e., nonfamily and nonkin ties) in the total social contacts of a respondent. The third measure is the prestige score of the highest position that a respondent can reach through his/her social network.
Measuring Independent Variables
Economic development is the GDP per capita (ten thousand yuan) for each city.
Educational development is measured by the percentage of educational expenditure in the
total expenditure of local government.6 Family background in social networks is measured by the extensity of father’s social network. It is a binary variable that is scored
1 if father’s network is “very extensive” or “extensive,” and scored 0 otherwise.
Permanent cross-city migration is measured by the number of cities that respondents have lived in for more than six months. I measure one kind of temporary cross-city mobility,
namely the short-term visit to other cities through training programs. To construct this
measure, I first identify if a respondent participated in any training program in the past. If
the answer is yes, I further ask whether the location for the training is within or outside
the city that the respondent lives in. The variable is then coded 1 for respondents who
participated in training programs held in another city, and 0 for respondents who did not
6 In China, government is the major source of educational investment. The expenditure of local government on education should capture the educational development in a city.
15 participate in any training program or the training program was held in the city in which
they lived. 7 I also construct the interaction terms between socioeconomic development measures and both cross-city migration and father’s social networks.
Measuring Control Variables
At the city level, population density is measured by the number of persons per
squared kilometer (ten thousand per sq. kilometer). Industrialization is the percentage of
the industrial GDP in the total GDP. Unemployment rate is measured by the percentage
of unemployed labor in the total labor force. At the individual level, gender is a binary variable that equals 1 for males and 0 for females. Age is measured in years. Marital
status is a binary variable that is scored 1 if the respondent is married and 0 otherwise.
Education is the years of formal schooling. Training is a binary variable that is scored 1 if
the respondent participated in any training program and 0 otherwise. Occupational status
is measured by the number of people the respondent supervises. Communist party
membership is a binary variable that is scored 1 for a communist party member and 0 for
others. Family background control variables include father’s educational and occupational status. The former is an ordinal variable with seven levels and the latter is an ordinal variable with eight levels.
The descriptive statistics of the measures are presented in Table 1. On average,
people have contacts with 6.91 positions. Among these contacts, 73 percent are
connected to people in secondary groups. The highest position that an individual can
reach through his/her personal network has a prestige score of 65.
7 This variable may measure training as well as temporary cross-city mobility. Therefore, I construct a variable that measures training and put it in each of my analysis for control.
16
(Table 1 about here)
Methods
The research design of structural and individual’s effects in this study is addressed by using multilevel models that account for the nonindependence of observations within cities (Snijders and Bosker1999; Raudenbush and Bryk 2002; Etile and Etile 2003; Luke
2004). Multilevel models estimate two equations simultaneously: a level-1 (individual- level) model and level-2 (city-level) model. The level-1 model is either a linear model
(for the highest position contacted and the proportion of secondary-group contacts) or a negative binomial model (for the number of contacted positions). The level-1 linear model can be written as:
Yij = β0j + ∑ βqXqij +εij
where Yij is the highest position of social contacts or the proportion of secondary-group
contacts for respondent i in city j; β0j is the intercept; Xqij is the value of individual-level
covariate q; and βq is the partial effect of the covariate on the highest position. The error
term εij is assumed to be independently and normally distributed with constant variance
2 σ . The level-1 negative binomial model views the count variable Yij, as sampled from a constant dispersed Poisson distribution. It replaces Yij with the natural long link, ŋij = log
(λij), where λij is the count number and ŋij is the log of the count number.
17 The level-2 model is the same in both the linear and negative binomial cases. The
intercept from level-1, β0j, is allowed to vary randomly across cities:
β0j = γ00 + γ01Z01 + γ02Z02 + γ03Z03 + γ04Z04 + γ05Z05 + θW + U0j
where γ00 is the average value of the outcome variable across cities, Z01 through Z05 are
city-level predictors, γ01 through γ05 are the regression coefficients of the effects of the
city-level variables, W are the interaction terms between socioeconomic development,
father’s social networks, and permanent cross-city migration, θ are the regression
coefficients of the interactions, and U0j is the city-level error term, assumed to be
normally distributed with a variance of τ.
RESULTS
The results of multilevel analyses for the three measures of social capital are
presented in tables 2 through 4. In the analyses for each measure, model 1 examines only
the main effects of the covariates and model 2 further enters interaction terms.
Table 2 presents the results from multilevel negative binomial models for the first
measure of social capital, namely the number of contacted positions. According to the
main effect model, father’s network extensity, permanent and temporary cross-city geographic mobility, and economic and educational development of a city significantly
affect the number of positions with which the respondent has social contacts. Holding
other variables constant, individuals whose fathers occupy extensive social networks
18 have social contacts with 6 percent (exp (.06)-1) more positions than those whose fathers’ networks are not extensive. Compared to their nonmobile counterparts, people make permanent or temporary cross-city mobility tend to connect with more positions. 8 Net of
the individual-level attributes, a unit increase in GDP per capita and ten percent
increment in educational expenditure of local government enlarges the scale of an
respondent’s social network by 7 percent (exp (.07)-1) and 11 percent (exp(.10)-1),
respectively. These results provide evidence for hypotheses 1 through 5.
(Table 2 about here)
With regard to the interaction effects between structural and individual factors
(model 2), only hypothesis 6 is supported. Although the coefficients for all interaction
terms are negative as expected, only that for the interaction between local economic
development and father’s network extensity achieves significance. It appears that original
residents do not receive significantly more benefits from local socioeconomic
development than migrants. The effect of economic development increases to .07 for
individuals with limited family networks (compared to the overall effect of .05 in model
1 for all respondents), and the coefficient for father’s network increases to .18 for
individuals living an underdeveloped city (compared to the overall effect of .06 in model
1 for all respondents). 9
Figure 2 demonstrates the two-way interaction effects between economic
development and father’s network extensity. On the one hand, the difference in the slopes
8 The effect of permanent cross-city migration on the number of contacted positions is marginally significant (p<.10). 9 Statistically, the effect of father’s network in model 2 is for individuals living in a city with 0 GDP per capita. I interpret it as an underdeveloped city.
19 of the two lines suggests that the effect of economic development on social capital is
differentiated for the two social groups divided by father’s network extensity. Increase in
GDP per capita is more beneficial for individuals whose family networks are less extensive in expanding social connections. On the other hand, economic development moderates the effect of father’s network on social capital. The importance of father’s network declines as a city gets more developed. While an extensive family network is an advantage for accessing social capital in a less-developed city, it becomes a disadvantage in a more-developed city.
(Figure 2 about here)
Table 3 reports the multilevel linear regression coefficients on individuals’ social
capital measured as the proportion of secondary-group contacts. As hypothesis 2 predicts,
the results in model 1 show that permanent cross-city migration helps people make social
contacts with secondary groups instead of family members and kin. In comparison to
their nonmobile counterparts, migrants have a higher percentage of nonkin ties in their
social networks, holding other variables constant. Father’s network extensity and
temporary cross-city mobility do not significantly influence the composition of
individuals’ social networks in this regard. At the city level, economic and educational
development promotes secondary-group social relations as well. A unit growth in GDP
per capital and 1 percent increment in educational expenditure of local government
increase the proportion of nonkin ties by 1.65 and 0.45 percents, respectively. These
results confirm hypotheses 4 and 5.
20
(Table 3 about here)
In addition, the negative and significant coefficient of the interaction term
between father’s network extensity and economic development suggests that these two
factors interweave with each other to determine the proportion of nonkin ties in one’s
social network, although this seems to be a one-way interaction. Economic development
does not significantly moderate the effect of father’s network extensity, but individuals
from families with more extensive social networks benefit less from the opportunities
created by economic development for developing social ties with secondary groups.
While the effect of economic development on second-group ties for all respondents is
1.65 (in model 1), this effect climbs to 2.57 (in model 2) for individuals whose father’s
network is not extensive. Figure 3 further illustrates the negative interaction between economic development and father’s network extensity. Such interaction effect provides some support for hypothesis 6.
(Figure 3 about here)
The results of multilevel linear regression on the third measure of social capital,
the highest position one can reach through his/her network relations, are presented in
Table 4. The expected main effects of all structural and individuals’ factors except for
local educational development are confirmed by the significant coefficients shown in
model 1. Permanent and temporary cross-city mobility, father’s network extensity, and
21 local economic development promote respondents’ social connections with people who
occupy high positions in a society. Controlling for other variables, an additional
permanent migration across cities increases the prestige score of the highest position in
one’s social network by 0.44 points. The highest position reachable through network relations is 3.14 points higher in prestige score for individuals who experience short-term
mobility to other cities through training programs than for those who do not have such
temporary cross-city mobility. A unit increment of in GDP per capita generates an
increase of 0.82 points in prestige score for the highest position in a respondent’s social
network. The coefficients in model 2 indicate that there are no significant interaction
effects between the structural and individual factors.
(Table 4 about here)
The analyses on the three measures of social capital also confirm most findings of
prior studies. Males tend to have larger social networks than females in that they are
connected to more positions, and they are more capable of reaching people in higher
positions. Their social networks contain a higher proportion of nonkin ties than those of
females. Individuals’ educational and occupational status also fosters their social
connections with people in different positions, people with high social status, and people
from secondary groups. The nonlinear relations between social capital and age are found
for two measures of social capital, the number of contacted positions and the proportion
of secondary-group contacts. One factor that deserves special attention within the
Chinese context is political power. The significant positive effects of Chinese Communist
22 Party membership on all three measures of social capital suggest that political power in
China is not only an important source of stratification but also a critical factor in social
relations.
DISCUSSION AND CONCLUSIONS
Using a national survey data consisting of 3529 respondents from 167 China cities
and multilevel modeling techniques, this article aims to identify the macro-micro linkages
through which structural and individual factors interact with each other to shape the contour of social capital inequality. Specifically, I investigate the research questions of how economic and educational development at the city level, permanent or temporary geographic mobility across cities and father’s network extensity at the individual level, and the interactions between the macro and micro level factors determine the distribution of social capital among individuals.
For the three measures of social capital used in this study, the results from
multilevel analyses suggest positive effects of socioeconomic development, permanent
and temporary cross-city migration and mobility, and family background in social
networks on individuals’ social capital. At the city level, economic development helps
individuals expand social contacts with people in different positions, promotes social
connections with people occupying positions of high status in a society, and increases
secondary-group ties. Educational development also exerts positive effects on the number
of positions with which individuals are connected and the proportion of nonkin ties in
23 their social networks, but it does not significantly help individuals reach high-status
people. These results suggest that the social capital embedded in one’s social network is
determined by the contextual structures in the social units that contain individuals. By
increasing material affluence, creating new social spaces, and inducing institutional
changes, socioeconomic development in a city provides more resources and opportunities
for social contacts and interactions. Residents of a more-developed city, therefore, access
more social capital than their counterparts in a less-developed city by being able to build social networks that are larger in size, and more diverse and resourceful in network members. These findings extend Blau’s (1977) structural theory of intergroup relations in that not only population size and distribution but also socioeconomic development and its
associated social processes define the opportunity structures for social contacts and
interactions.
At the individual level, in addition to people’s personal characteristics, family
background in social status, and achieved status suggested by prior studies, father’s
network extensity and permanent or temporary mobility across cities are three other
factors that influence people’s social capital. Individuals whose fathers have extensive
social networks tend to interact with people in a lager number of positions and are
connected to people with higher status in a society. These connections can be achieved by
earlier awareness of the importance of social capital and better education on the skills for
social networking that usually happen in families with broad social networks. These
people can also turn part of their family’s social capital, if not all, into their own by
joining the networks of family members. That father’s network extensity enhances
children’s ability to reach people in higher positions indicates that children are very likely
24 to be introduced to the most significant members in their father’s networks. This finding
of the significance of father’s network in children’s social capital contributes to the literature of transgenerational stratification. Not only inequality in human capital and occupational status (Blau and Duncan 1967), but also inequality in network relations and social capital are transferable from one generation to another.
Geographic mobility and migration also enhance individuals’ social capital. The
results form multilevel regressions demonstrate that cross-city migrants tend to have
social networks consisting of social ties with people in more positions, people having
high status, and people from secondary groups. Temporary mobility to other cities
through training programs also increases the number of contacted positions and help
individuals to connect with high-status persons. These findings contribute to the debate
on the relationships between geographic mobility and social networks (McLanahan and
Sandefur 1994; Fischer 2002; South and Haynie 2004; Bidart and Lavenu 2005).
Through physical mobility, cross-city movers are presented with better resources (Fischer
2002) and additional opportunities otherwise unavailable for social networking with
people in the new locations. Advanced transportation and communication systems also
enable movers to maintain at least part of their social connections in areas of origin
(Kaufer and Carley 1993). In comparison to their nonmobile counterparts, therefore,
cross-city movers will have larger, more diverse and resourceful networks that contain
more social capital in the long run.
Furthermore, structural and individual factors do not influence social capital
independently. Rather, they are interacting with each other to determine the distribution
of social capital among individuals. My analyses indicate negative interactions between
25 economic development and father’s network extensity in determining social capital
measured by the number of contacted positions and the proportion of secondary-group
contacts. For the number of contacted positions, this negative interaction is two-way. On
the one hand, individuals whose family networks are not extensive are better at taking advantage of the benefits and opportunities created in the development process. On the other hand, economic development moderates the positive effect of father’s network extensity. The advantages that father’s network provides are reduced as a city becomes more developed. For secondary-group contacts, economic development is also more beneficial for disadvantaged individuals whose fathers have limited social networks.
These negative interactions suggest that the process of economic development reduces social capital inequality among individuals based on described status and the transgenerational inequality in social capital discussed above. The coefficient of the interaction between economic development and father’s network extensity on the highest position one can reach though network relations, though negative, is not significant, suggesting that the advantages in reaching high-status people created by father’s network extensity is the most difficult ones to be moderated by economic development.
In addition to the theoretical contributions, the findings in this study also have
some policy implications for social stratification in China. The strong effect of
socioeconomic development on personal networks and social capital suggests that
regional disparity in socioeconomic development (Hauser and Xie 2004) will generate
inequality in social capital among residents of different areas and cities. Given the crucial
role of social networks on status attainment in Chinese societies (Walder 1989; Cheng
and Rosett 1991; Smart 1993; Bian 1997), the unequal distribution of social capital will
26 reinforce and aggravate regional social inequality in general. Therefore, it is important to
adjust the strategies and policies that give some regions priority for socioeconomic
development. It is also critical for Chinese government to loosen the sanctions on
geographic mobility so that individuals, especially those living in an underdeveloped area,
can break through the constraints of local conditions and enhance their social capital by physical mobility across areas.
The substantial roles that socioeconomic development plays on personal networks
and social capital found in this study calls for the need of taking social structures into
account when explaining social capital inequality. In addition to socioeconomic development, political institutions and cultural traditions may also affect social relations and interactions among people. For example, the significant effects revealed in my analyses of Chinese Communist Party membership on the three measures of social capital indicate that political power is not only an important source of accessing material resources (Walder 1995; Bian and Logan 1996; Zhou 2000), but also a critical factor for building social connections under a socialist regime where political authority is at the center of economic and social life. As Kadushin (2004) suggests, a culture of trust and mutual aid in a community affects the social interactions among its residents. Although the analyses in this study are based on data from China, the factors and mechanisms identified can be applied to other countries and societies. It will be very interesting to conduct comparative studies across societies through which we can discover how structural and individual factors interact with each other to shape the unique contour of social capital inequality.
27 Table 1: Descriptive Statistics for Measures of Dependent, Independent, and Control Variables
Mean N Dependent variables Number of contacted positions 6.91 (4.39) 2729 Proportion of secondary-group contacts (%) 72.73 (26.62) 2679 Highest contacted position 64.89 (11.11) 2702
Independent variables Economic development - GDP per capita (ten 2.20 (1.41) 2729 thousand yuan) Educational development (% of local government 14.95 (6.28) 2729 expenditure on education ) Father’s social networks (1=extensive; 0=not .39 (.49) extensive) Number of permanent cross-city migrations .52 (1.19) 2729 Temporary cross-city mobility (1=have mobility; .06 (.24) 2729 0= have no mobility)
Control variables Male .49 (.50) 2729 Age 38.38 (10.39) 2729 Marital status (married=1; other=0) .83 (.38) Years of education 11.41 (3.18) 2729 Number of supervisees 9.03 (54.85) 2729 Communist party membership .22 (.41) 2729 (1= party member; 0=other) Father’s education (ordinal variable: 7 levels) 3.71 (1.46) 2729 Father’s supervisees (ordinal variable: 8 levels) 1.98 (1.62) 2729 Training (1=receives training; 0=does not receive .27 (.44) 2729 training) Population density (ten thousand per sq. kilometer) .13 (.13) 2729 Industrialization (% of industrial GDP) 50.24 (10.69) 2729 Unemployment rate (%) 6.64 (4.16) 2729
Note: Standard deviations are in the parentheses.
28 Table 2: Multilevel Negative Binomial Regression Coefficients on the Number of Contacted Positions
Model 1 Model 2 Individual-level variables Age .20 (.09) * a .20 (.09) * a Age squared -.03 (.01) ** b -.03 (.01) ** b Male .09 (.02) *** .09 (.02) *** Married -.04 (.35) a -.07 (.35) a Education .03 (.00) *** .03 (.00) *** Number of supervisees .07 (.01) *** b .07 (.01) *** b Communist party membership .11 (.03) *** .13 (.03) *** Father’s educational level .01 (.01) .05 (.08) a Father’s supervisory level .01 (.01) .07 (.07) a Father’s network-extensive .06 (.02) * .18 (.09) # Number of permanent cross-city .16 (.08) # a .03 (.03) migrations Training .12 (.03) *** .12 (.03) *** Temporary cross-city mobility .12 (.04) ** .12 (.04) **
City-level variables Population density -.01 (.18) -.01 (.18) Unemployment rate .02 (.05) a .02 (.05) a Industrialization -.02 (.03) a -.02 (.03) a Economic development .05 (.02) * .07 (.02) ** Educational development .10 (.04) * a .10 (.05) * a
Interaction terms Economic development * father’s -.04 (.02) * network Educational development* father’s -.02 (.78) b network Economic development * permanent -.02 (.04) a cross-city migration Educational development * -.09 (.17) b permanent cross-city migration Constant .97 (.25) *** .94 (.25) ***
Notes: (1) N=2729 for individual-level variables; N=159 for city-level variables. (2) Coefficients reported are log-event counts. (3) Standard errors are in the parentheses. (4) a Coefficients and standard errors multiply by 10; b Coefficients and standard errors multiply by 100. (5) # p<0.10; * p<0.05; ** p<0.01; ***p<0.001.
29 Table 3: Multilevel Linear Regression Coefficients on the Proportion of Secondary- Group Contacts
Model 1 Model 2 Individual-level variables Age .87 (.38) * .85 (.38) * Age squared -.13 (.04) ** a -.13 (.04) ** a Male 5.73 (.98) *** 5.78 (.98) *** Married -4.05 (1.60) * -4.11 (1.60) * Education .72 (.19) *** .73 (.19) *** Number of supervisees .25 (.09) ** a .25 (.09) ** a Communist party membership .65 (1.30) .53 (1.30) Father’s educational level .02 (.38) -.04 (.39) Father’s supervisory level -.04 (.32) -.06 (.32) Father’s network-extensive .60 (1.10) 5.35 (4.22) Number of permanent cross-city 1.12 (.41) ** 2.92 (1.65) # migrations Training .42 (1.23) .52 (1.26) Temporary cross-city mobility .20 (2.22) .30 (2.22)
City-level variables Population density -.06 (.07) b -.05 (.07) b Unemployment rate .05 (.21) .05 (.21) Industrialization .01 (1.00) .01 (1.00) Economic development 1.65 (.79) * 2.57 (.90) ** Educational development .45 (.17) ** . 53 (.19) **
Interaction terms Economic development * father’s -1.87 (.84) * network Educational development* father’s -.03 (.20) network Economic development * permanent -.21 (.36) cross-city migration Educational development * -.10 (.08) permanent cross-city migration Constant 40.08 (10.18) *** 37.93 (10.32) ***
Notes: (1) N=2679 for individual-level variables; N=159 for city-level variables. (2) Standard errors are in the parentheses. (3) a Coefficients and standard errors multiply by 10; b Coefficients and standard errors multiply by 100. (4) # p<0.10; * p<0.05; ** p<0.01; ***p<0.001.
30 Table 4: Multilevel Linear Regression Coefficients on the Highest Position in One’s Personal Network
Model 1 Model 2 Individual-level variables Age .22 (.15) .21 (.15) Age squared -.02 (.02) a -.02 (.02) a Male .99 (.39) * -1.32 (.63) * Married -1.31 (.63) * -4.11 (1.60) * Education .91 (.07) *** .91 (.07) *** Number of supervisees .09 (.03) * a .09 (.03) * a Communist party membership 1.90 (.51) *** 1.87 (.51) *** Father’s educational level .45 (.15) ** .44 (.15) ** Father’s supervisory level .06 (.12) .06 (.12) Father’s network-extensive 1.31 (.43) ** 3.15 (1.65) # Number of permanent cross-city .44 (.16) ** 1.04 (.65) migrations Training 1.57 (.49) ** 1.60 (.49) ** Temporary cross-city mobility 3.14 (.88) *** 3.16 (.88) ***
City-level variables Population density .04 (.26) b .03 (.26) b Unemployment rate .01 (7.65) b -.08 (7.65) b Industrialization -.03 (.04) -.03 (.04) Economic development .62 (.29) * .82 (.34) * Educational development .07 (.06) . 11 (.07)
Interaction terms Economic development * father’s -.47 (.33) network Educational development* father’s -.05 (.08) network Economic development * permanent -.01 (.14) cross-city migration Educational development * -.04 (.03) permanent cross-city migration Constant 45.72 (3.89) *** 44.78 (3.95) ***
Note: (1) N=2702 for individual-level variables; N=159 for city-level variables. (2) Standard errors are in the parentheses. (3) a Coefficients and standard errors multiply by 10; b Coefficients and standard errors multiply by 1000. (4) # p<0.10; * p<0.05; ** p<0.01; ***p<0.001.
31
Figure 1: Structural and Individual Effects on Social Capital
City-Level Context
Individual’s Social Capital
Individual Attribute
32 Figure 2: Effect of Economic Development on the Number of Contacted Positions: Comparison between Two Groups Divided by Father’s Network
9 8.5 8 7.5 7 6.5 Predicted NumberContacted of Postions 6
0 2 4 6 8 10 GDP Per Capita
father's network-not extensive father's network-extensive
33 Figure 3: Effect of Economic Development on the Proportion of Secondary-Group Contacts: Comparison between Two Groups Divided by Father’s Network
90 80 70 60
Predicted Proportion of SecondaryGroup - Contacts 0 2 4 6 8 10 GDP Per Capita
father's network-not extensive father's network-extensive
34 References:
Bian, Y. 1997. “Bring Strong Ties Back in: Indirect Connection, Bridges, and Job Search
in China.” American Sociological Review 62:366-85.
⎯⎯⎯. and J. Logan. 1996. “Market Transition and the Persistence of Power: The
Changing Stratification Order in Urban China.” American Sociological Review
61:739-58.
Bidart, C. and D. Lavenu. 2005. “Evolutions of Personal Networks and Life Events.”
Social Networks 27:359-76.
Blau, P. 1977. Inequality and Heterogeneity: A Primitive Theory of Social Structure.
New York: Free Press.
⎯⎯⎯ and O. Duncan. 1967. The American Occupational Structure. New York: Wiley.
Bourdieu, P. 1986. “The Form of Capital.” Pp. 241-58 in Handbook of Theory and
Research for the Sociology of Education, edited by J.G. Richardson. New York:
Greenwood Press.
Boxman, E., P. De Graaf and H. Flap. 1991. “The Impact of Social and Human Capital
on the Income Attainment of Dutch Managers.” Social Network 13:51-73.
Brass, D. J. 1985. “Men’s and Women’s Networks: A Study of Interaction Patterns and
Influence in an Organization.” Academy of Management Journal 28:327-43.
Burt, R. 1992. Structural Holes: The Social Structure of Competition. Cambridge, MA:
Harvard University Press.
Campbell, K. E. 1988. “Gender Differences in Job-Related Networks.” Work and
Occupations 15:179-200.
35 ⎯⎯⎯ and B. A. Lee. 1991. “Name Generators in Surveys of Personal Networks. Social
Networks 13: 203-21.
Cheng, L. and A. Rosett. 1991. “Contract with a Chinese Face: Social Embedded Factors
in the Transformation from Hierarchy to Market, 1978-1989.” Journal of Chinese
Law 5:143-244.
Coleman, J. S. 1988. “Social Capital in the Creation of Human Capital.” American
Journal of Sociology 94:S95-S121.
Ensel, W. M. 1979. Sex, Social Ties, and Status Attainment. Albany: State University of
New York at Albany.
Ericson, B. H. 1996. “Culture, Class and Connections.” American Journal of Sociology
102:215-51.
⎯⎯⎯. 2001. “Good Networks and Good Jobs: The Value of Social Capital to
Employers and Employees.” Pp. 127-58 in Social Capital: Theory and Research,
edited by N. Lin, K. Cook, and R. Burt. New York: Aldine De Gruyter.
Eitle, D. and T. Eitle. 2003. “Segregation and School Violence.” Social Forces 82:589-
615.
Fei, X. 1992 [1949]. From the Soil, the Foundations of Chinese Society. Berkeley, CA:
University of California Press.
Fernandez, R. M. and N. Weinberg. 1997. “Sifting and Sorting: Personal Contacts and
Hiring in a Retail Bank.” American Sociological Review 62:883-902.
Fischer, C. S. 1982. To Dwell Among Friends. Chicago, IL: Chicago University Press.
⎯⎯⎯. 2002. “Ever-More Rooted Americans.” City & Community 1:177-98.
36 Flap, H. 1991. “Social Capital in the Production of Inequality.” Comparative Sociology of
Family, Health, and Education 20:6179-6202.
⎯⎯⎯. 2002. “No Man Is an Island.” Pp. 29-59 in Conventions and Structures in
Economic Organization: Markets, Networks and Hierarchies, edited by O. Favereau
and E. Lazega. Cheltenham, UK: Edward Elgar.
Frank, K. and J. Yasumoto. 1998. “Linking Action to Social Structure within a System:
Social Capital Within and Between Subgroups.” American Journal of Sociology 104:
642-86.
Fried, M. 1969 [1953]. Fabric of Chinese Society: A Study of the Social Life in a Chinese
Country Seat. New York: Octagon Books.
Goode, W. J. 1970. World Revolution and Family Patterns. New York: Free Press.
Granovetter, M. 1973. “The Strength of Weak Ties.” American Journal of Sociology 78:
1360-80.
⎯⎯⎯. 1974. Getting a Job. Cambridge, MA: Harvard University Press.
⎯⎯⎯. 1995. Getting a Job: A Study of Contacts and Careers. Chicago: University of
Chicago Press.
Jencks, C. and S. Mayer. 1990. “The Social Consequences of Growing Up in a Poor
Neighborhood.” Pp. 111-86 in Inner City Poverty in the United States, edited by L.
Lynn Jr. and M. McGeary. Washington, D. C.: National Academy Press.
Hauser, S. and Y. Xie. 2004. “Temporal and regional variation in earnings inequality:
Urban China in Transition between 1988 and 1995.” Social Science Research 34:44-
79.
Kadushin, C. 2004. “Too Much Investment in Social Capital?” Social Networks 26:75-90.
37 Kaufer, D. S. and K. M. Carley. 1993. Communication at a Distance: The Effect of Print
on Socio-Cultural Organization and Change. Hillsdale, NJ: Lawrence Erlbaum.
King, Ambrose Y.C. 1985. “The Individual and Group in Confucianism: A Relational
Perspective.” Pp. 57-70 in Individualism and Holism: Studies in Confucian and
Taoist Values, edited by D. Munro. Ann Arbor: Center for Chinese Studies,
University of Michigan.
Laumann, E. O. and D. Knoke. 1987. The Organizational State: Social Choice in
National Policy Domains. Madison: University of Wisconsin.
Levy, M. 1949. The Family Revolution in Modern China. Cambridge, MA: Harvard
University Press.
Lin, N. 1989. “Chinese Family Structure and Chinese Society.” Bulletin of the Institute of
Ethnology 65:382-99.
⎯⎯⎯. 1999. “Social Networks and Status Attainment.” Annual Review of Sociology 25:
467-87.
⎯⎯⎯. 2001. Social Capital: A Theory of Structure and Action. Cambridge and New
York: Cambridge University Press.
⎯⎯⎯. 2004. “The Invisible Hand of Social Capital.” Presented at the Sunbelt Social
Networks Conference, Redondo Beach, California, February 18.
⎯⎯⎯, W. M. Ensel and J. C.Vaughn. 1981. “Social Resources and Strength of Ties:
Structural Factors in Occupational Status Attainment.” American Sociological Review
46:393-405.
⎯⎯⎯ and M. Dumin. 1986. “Access to Occupations through Social Ties.” Social
Networks 8:365-85.
38 Litwak, E. 1960. “Geographic Mobility and Extended Family Cohesion.” American
Sociological Review 25:385-94.
Logan, J. R. and F. Bian. 1999. “Family Values and Coresidence with Married Children
in Urban China.” Social Forces 77:1253-82.
Luke, D. A. 2004. Multilevel Modeling. Thousand Oaks, CA: Sage Publications.
Marsden, P.V. 1987. “Core Discussion Networks of Americans.” American Sociological
Review 52:122-31.
⎯⎯⎯ and J. S. Hurlbert. 1988. “Social Resources and Mobility Outcomes: A
Replication and Extension.” Social Forces 66:1038-59.
Mayhew, B. and R. Levinger. 1976. “Size and Density Interaction in Human
Aggregates.” American Journal of Sociology 82:86-110.
McDonald, S. and G. Elder Jr. 2006. “When Does Social Capital Matter? Non-Searching
for Jobs across the Life Course.” Social Forces 85:521-49.
McLanahan, S. and G. Sandefur. 1994. Growing Up with a Single Parent: What Hurts,
What Helps. Cambridge, Mass: Harvard University Press.
McPherson, M. and L. Smith-Lovin. 1982. “Women and Weak Ties: Differences by Sex
in the Size of Voluntary Organizations.” American Journal of Sociology 87:883-904.
⎯⎯⎯. 1986. “Sex Segregation in Voluntary Associations.” American Sociological
Review 51:61-79.
Miller, J. 2006. Pathways in the Workplace: The Effects of Gender and Race on Access to
Organizational Resources.” New York: Cambridge University Press.
Moore, G. 1979. “The Structure of a National Elite Network.” American Sociological
Review 44:672-92.
39 ⎯⎯⎯. 1990. “Structural Determinants of Men’s and Women’s Personal Networks.”
American Sociological Review 55:726-35.
Morgan, D. L. 1988. “Age Differences in Social Network Participation.” Journal of
Gerontology 43:S129-37.
Mouw, T. 2003. “Social Capital and Finding a job: Do Contacts Matter?” American
Sociological Review: 68:868-98.
Podolny, J. M. and J. N. Baron. 1997. “Resources and Relationships: Social Networks
and Mobility in the Workplace.” American Sociological Review 62:673-93.
Peterson, T., I. Saporta and M. L. Seidel. 2000. “Offering a Job: Meritocracy and Social
Networks.” American Journal of Sociology 106:763-816.
Pugliesi, K. and S. L. Shook. 1998. “Gender, Ethnicity, and Network Characteristics:
Variation in Social Support Resources.” Sex Roles 3-4:215-38.
Putnam, R. D. 1993. Making Democracy Work: Civic Traditions in Modern Italy.
Princeton, NJ: Princeton University Press.
Raudenbush, S. and A. Bryk. 2002. Hierarchical Linear Models. Thousand Oaks, CA:
Sage Publications.
Renzulli, L and H. Aldrich. 2005. “Who Can You Turn To? Tie Activation within Core
Business Discussion Networks.” Social Forces 84:323-41.
⎯⎯⎯, H. Aldrich and J. Moody. 2000. “Family Matters: Gender, Networks, and
Entrepreneurial Outcomes.” Social Forces 79:523-46.
Ruan, D., L. C. Freeman and X. Dai. 1997. “On the Changing Structural of Social
Networks in Urban China.” Social Networks 19:75-89.
Simmel, G. 1958 (1908). Soziologie. Berlin: Duncker & Humblot.
40 Smart, A. 1993. “Gifts, Bribes, and Guanxi: A Reconsideration of Bourdieu’s Social
Capital.” Cultural Anthropology 8:388-408.
Scott, D. B. 1996. “Shattering the Instrumental-Expressive Myth: The Power of
Women’s Networks in Corporate-Government Affairs.” Gender and Society 10:232-
47.
Snijders, T. and R. Bosker. 1999. Multilevel Analysis: An Introduction to Basic and
Advanced Multilevel Modeling. London: Sage Publications.
South, S. and D. Haynie. 2004. “Friendship Networks of Mobile Adolescents.” Social
Forces 83:315-50.
Sprengers, M., F. Tazelaar and H. D. Flap. 1988. “Social Resources, Situational
Constraints, and Reemployment.” Netherlands Journal of Sociology 24:98-116.
Treas, J. and W. Wang. 1993. “Of Needs and Contracts: Filial Piety Perceived in
Contemporary Shanghai.” Pp. 87-98 in The Changing Contract across Generations,
edited by V. L. Bengston and W. A. Achenbaum. New York: Aldine De Gruyter.
Volker, B. and H. D. Flap. 1999. “Getting Ahead in the GDR: Social Capital and Status
Attainment under Communism.” Acta Sociologica 41:17-34.
Walder, A. G. 1989. “Social Change in Post-Revolution China.” Annual Review of
Sociology 15:405-24.
⎯⎯⎯. 1995. “Local Governments as Industrial Firms.” American Journal of Sociology
101:263-301.
Zhou, X. 2000. “Economic Transformation and Income Inequality in Urban China:
Evidence from Panel Data.” American Journal of Sociology 105:1135-74.
41 Appreciating the Complexity:
Ethnic Diversity and Generalized Trust in the Melting Pot
Wenhong Chen
Postdoctoral Researcher Department of Sociology, Duke University E-Mail:[email protected] Telephone: 1-919-660-5604
Work in Progress
Please don’t cite or distribute without the author’s permission
Abstract
Existing research has shown a monotonic negative relation of ethnic diversity and social capital. For instance, Putnam (2007) found that ethnic diversity in the neighborhood put a damper on various forms of social capital. In this paper, we merged a national representative survey on social capital with the US census data to explore the relation of ethnic diversity and trust. Departing from existing studies, our results suggest a non-linear relation of ethnic diversity and generalized trust. Moreover, the relation is contingent on class, race, and community type.
Acknowledgements
Data used in this article were drawn from the thematic research project "Social Capital: Its Origins and Consequences", sponsored by Academia Sinica, Taiwan, through its Research Center for Humanities and Social Sciences, and the Institute of Sociology. The principal investigator of the project is Nan Lin.
1 Appreciating the Complexity:
Ethnic Diversity and Generalized Trust in the Melting Pot
Since Putnam (1995) published his provocative article “Bowling Alone” demonstrating a decline of social capital in communities across America, the search of the smoking guns has begun. TV watching, Internet surfing, and women’s participation in the labor force have been interrogated as culprits. More recently, the attention has been directed to ethnic diversity as a prime suspect. Putnam (2007) highlighted a monotonic, negative relation of ethnic diversity and social capital and argued that at least in the short to medium run immigration and ethnic diversity put a damper on various forms of social capital. Although Putnam’s work may have received the most media coverage, numerous studies have reported a monotonic, negative relation of ethnic diversity and social capital since a while (Rice and Steele 2001; Costa and Kahn 2001, 2003; Saguaro Seminar 2001; Alesina and Ferrara 2002; Anderson and Paskeviciute 2006; Coffe and Geys 2006; Leigh 2006; Paxton 2007). Many studies have adopted an all-embracing definition of social capital (Putnam 1995; 2000), which may include generalized trust, participation in voluntary associations, community attachment, to civic engagement. Critics have pointing out such an approach tends to be “suspiciously protean”, misleading, and lack analytical clarity (Kay and Johnston 2007: 3). In this paper, we focus on the relation of ethnic diversity in the neighborhood and generalized trust at the individual level as an important component of social capital. Political scientists, sociologists, and economists have increasingly paid attention to trust and its positive impact as an important component of social capital. Generalized trust is based on "a belief in the benevolence of human nature in general" (Yamagishi and Yamagishi 1994: 139). In other word, generalized trust is the trust in the trustworthiness of people with whom we may or may not have direct social interaction. If social capital is the glue that holds a society together, trust is the lubricant that smoothes cooperation (Putnam 2000). Nations with more trusting citizens have stronger economic growth (Knack and Keefer 1997). Trust enhances community safety (Sampson, Raudenbush, and Earls 1997). Trust supplements formal social-legal institutions and reduces opportunism and malfeasance. It facilitates economic transactions especially when uncertainty is high (Mizruchi and Stearns 2001). Critics also point out the paradox that social capital tends to be the highest in less urbanized communities with small,
2 homogenous populations struggling with economic stagnation and out- migration (Florida 2002). Parallel to the growing literature on trust, diversity has also been a topic that has gained currency (Arneil 2006). First, ethnic diversity is a fact of life that cannot be written off in many societies including not only traditional immigrant-receiving countries such as US, Canada, and Australia but also European and Asian countries. How to create a new “we” is the central challenge for modern multiethnic immigrant-receiving societies (Putnam 2007). Second, diversity has positive returns too. At the aggregated level, the influx of immigration compensates the aging populations in most developed societies. Ethnic diversity (at the metropolitan level) is positively related to the development of high-tech industry and population growth (Florida 2002). The literature on workgroup diversity shows that diversity nurtures creativity. Multiple perspectives, diverse skill-sets, and boundary-crossing networks encourage innovation (Knippenberg and Schippers 2007). An ethnically diverse population enhances national comparative advantage in an increasing globalized world. Third and perhaps most important, diversity - ethnic or otherwise - is ultimately about social integration based on equality of access to opportunities, a fundamental principle in democratic societies. Thus, as McClain puts it, if there is a tradeoff of diversity and social capital, “maybe it is worth the price” (2003: 101). The tradeoff of ethnic diversity and social capital has been dubbed as “one of the most troubling findings” in social capital studies (Hooghe 2007: 709). It touches enough a nerve in the post 9-11 world. Hero pointed out that social capital thesis is “socio-psychologically comfortable…because it emphasizes and resonates with noble sentiments of community, consensus, and connectedness” (2007: 165). A romanticization of Tocquevillian America as the golden days of large-sized voluntary associations tends to overlook the oppression and exclusion of racial and cultural minorities (Arneil 2006). Both social capital and diversity have positive and negative impact. Blau argued that macrosocial integration rested on “extensive intergroup relations, not on strong ingroup bonds” (1977:11). The history of nation building in the US, Canada, and Australia has shown that people of different race and ethnicity can integrate. The assimilation theory describes the process in which immigrants shed their ethnic-cultural baggage and embrace the common culture and identity of the host society (Gordon 1964). Although the canonical assimilation theory has been criticized for oversimplifying the process as one-way and linear (Kivisto 2004), mainstream society and culture have been profoundly diversified in the process of immigrants becoming American (Alba and Nee 1997). We
3 probably should go beyond asking why social capital does not embrace diversity or why diversity erodes social capital. The question is how to strike a balance of diversity and trust, a balance of social integration and cohesion. In this paper, we are interested in whether, when, and how ethnic diversity is related to generalized trust. Is there a negative relation across the board, as many existing studies have suggested? Or does it vary along the familiar fault lines of social inequality such as class and race? To answer these questions, we merged a national representative survey data on social capital with the US census data. Our analysis includes both objective and subjective measures of ethnic composition in the neighborhood as we use ethnic diversity at the census tract level and self-reported neighborhood type. Unlike existing studies, our analysis suggests a significant non-linear relation of ethnic diversity and generalized trust. Trust is high when ethnic composition in the neighborhood is homogenous. It declines with increasing ethnic diversity but rebounds once it reaches a tipping point, controlling subjective neighborhood type, neighborhood characteristics, and individual characteristics. Moreover, we go beyond the one-size-fits-all analysis and find that the relation of ethnic diversity to generalized trust is contingent. Splitting the whole sample along race, class, and community types, the relation of ethnic diversity and trust remains significant only among Whites, middle class, and small town residents, respectively. Compared to Whites, generalized trust among Blacks and Latinos is less responsive to ethnic diversity in the neighborhood. Compared to middle class, generalized trust among upper and lower class are not as sensitive to ethnic diversity in the neighborhood. Compared to residents in other types of community, residents in small towns react more strongly to ethnic diversity in their neighborhood. The rest of the paper is organized as follows. We review existing literature on ethnic diversity and generalized trust and develop hypotheses in the first section. The second section discusses the data and variable construction. Results are presented in the third section. We conclude with a discussion of future research.
Literature Review
Ethnic Diversity and Generalized Trust: a Monotonic Negative
Relation?
4 Early research tended to be dominated by the positive impact of trust. Recent research has provided a more subtle and layered picture of the return to trust. First, trust is neither necessary nor sufficient for cooperation. Informal social mechanism such as reputation and the potential of ostracism can sustain cooperation. A healthy dose of distrust has motivated the design of institution and organization that deters opportunism and malfeasance (Cook, Hardin, and Levi 2005). Second, the impact of trust, as that of social capital in general, is non-linear and may have a dark side. For instance, there is a curvilinear relation of trust and innovation. Innovation is nurtured by the support but constrained by the pressure of social conformity from strong, trusting ties (Ruef 2002). Third, the return to trust changes across groups, just as the value of social capital varies for different purposes and different populations. Hero (2007) found that Blacks were actually poorer compared to Whites in states with high levels of social capital than in states with low levels of social capital. In other words, Blacks do not benefit as much as Whites from the return of aggregated social capital in their states. If the return to trust can be curvilinear and contingent, how about the production side of trust? Trust has been found to be related to a confluence of factors. Trust is unevenly distributed along the lines of social inequalities. Existing studies have shown that trust often comes from affluence, security, life satisfaction, and good health. People who are better off and better educated are more trusting since their power, status, and asset make it safe to trust (Glaeser 2001; Leigh 2006). People who are on the wrong side of social stratification (e.g., less educated, low income earners, unemployed, and racial/ethnic minorities) tend to be less trusting (Paxton 2005). Disadvantages at both the individual and neighborhood level generate mistrust (Ross, Mirowsky, and Pribesh 2001). Although Putnam’s paper (2007) may have received the most media coverage, a monotonic negative relation of ethnic diversity and social capital has been reported in numerous studies since a while. Scholars have used terms such as ethnic fragmentation, heterogeneity, or diversity to capture ethnic composition (at levels ranging from neighborhood to metropolitan area to nation state). In the literature review, we use the term used in the work cited. At the national level, comparative studies suggest that citizens in countries with higher levels of ethnic homogeneity - measured by the proportional size of the largest ethnic group in the total population (Paxton 2007) or the Herfindahl index (Anderson and Paskeviciute 2006) - have higher levels of generalized trust.
5 Most research has examined the relation of ethnic heterogeneity and trust at the aggregated metropolitan or county level. Costa and Kahn (2001) found that rising racial heterogeneity at the metropolitan level accounted for the decline of social capital from mid 1970s to the end of 1980s in the US. Iowan towns with higher levels of white ethnic diversity have lower levels of community attachment (Rice and Steele 2001). Using data in Flemish municipalities, Coffé and Geys (2006) reported that aggregated social capital at the municipality level was negatively related to ethnic heterogeneity. Using national representative data in Sweden, Gustavsson and Jordahl (2008) found that trust was negatively related to the proportion of migrants at the county level. Concentrated socioeconomic disadvantage and concentrated immigration at the neighborhood level explain a large part of neighborhood variation in social cohesion and trust (Sampson, Raudenbush, and Earls 1997). For Whites in Seattle, ethnic diversity in the neighborhoods is negatively related to the perception of neighbor relation as trusting (Guest, Kubrin, and Cover 2008). Yet, studies have also shown that residents in more racially diverse neighborhood have lower levels of interracial prejudice (Oliver and Wong 2003). Using a national representative survey in Canada, Aizlewood and Pendakur (2007) found that contextual-level indictors of ethnic-cultural diveristies at the neighborhood level have limited impact on individual’s political and interpersonal trust. Why is there a tradeoff between ethnic diversity and generalized trust? Does diversity make it difficult to build trust or does diversity destroy existing trust? The existing literature has employed several theses to explain the relation of ethnicity diversity and social capital in general and trust in particular: the social categorization theory, the contact hypothesis, and the conflict hypothesis. Homophily is a powerful rule that affects social interaction. Individuals tend to interact with others who are similar to themselves. What would happen when members of different ethnic group interact? The social categorization theory predicts that individuals tend to attribute positive traits to ingroup members but negative characteristics to outgroup members (Tajfel 1981). Thus, growing diversity increases the number of people with negative characteristics, which in turn decreases trust because people trust less those with undesirable characteristics (Uslaner 2002). The contact hypothesis however argues that mistrust comes from the lack of informed judgment and bringing different ethnic groups into contact reduces prejudices and stereotypes (Bown 1995). A widely cited example is Stouffer’s study (1949) of American soldiers during the Second World War: white soldiers who had direct contact with black soldiers were more supportive to the idea of
6 mixing white and black platoons. By contrast, the conflict hypothesis extends the social categorization theory and argues that contact may increase conflict as interethnic competition of economic, political, or symbolic resources grows (Quillian 1995). In a similar vein, the resource competition theory argues that resource competition increases distrust and discrimination against ethnic competitors (Horowitz 1995). The increase of immigrants, especially newcomers of different ethnic and linguistic background may generate distrust. For instance, the attitude of Israeli Kibbutz members toward new immigrants from former Soviet Union was negatively associated with the availability of resources and opportunities (Isralowitz 1998). Organization scholars point out that diverse teams have difficulties in communicating expectations, developing group identification, setting goals, and coordinating, which may increase tension and hinder performance (Reagans, Zuckerman, and McEvily 2004; Gibson and Gibbs 2006). Putnam (2007) reported that both inter- and intra-ethnic trust were lower in ethnically diverse communities and labeled it as “constrict theory” whereas people in multiethnic neighborhoods choose withdrawal and isolation rather than interaction or confrontation. They, using Putnam’s term, “hunker down” to shield themselves from information overload associated with ethnic diversity. Put it more straightforwardly, people in ethnically diverse neighborhoods simply tune off. Yet, if ethnic diversity is the source of information overload, the constrict theory does not explain why people living in ethnically diverse neighborhoods turn their backs to both in- and out-group contact. Thus, whether contact with outgroup members leads to trust or distrust is contingent on the context, including institutional arrangements, ongoing political, economic, and demographic development, the nature of the contact, and group and individual characteristics. Allport (1954) maintained that four optimal conditions that would facilitate positive outcomes of intergroup contact: the parties involved have equal status, shared goals, intergroup cooperation, and institutional support (see also Pettigrew 1998). In a meta-analysis of 515 studies on intergroup contact theory, Pettigrew and Tropp (2006) found that intergroup contact helped to reduce prejudice when Allport’s optimal conditions were present. Pettigrew further outlined four interrelated socio-psychological processes that explain why contact can lead to positive changes in intergroup attitude: “learning about the outgroup, changing behavior, generating affective ties, and ingroup reappraisal” (1998: 70). First, learning about the outgroup increases informed judgment. Second, growing contact with outgroup members makes it
7 more routine and acceptable. Gradually, people adjust their intergroup attitude to resolve the dissonance between increased intergroup contact and negative intergroup attitude. Third, intergroup friendships that help to reduce prejudice are more likely to be formed in mixed neighborhoods. Fourth, contact with outgroups increases the exposure to new perspectives and reevaluation of ingroup norms and values. As people’s view becomes more cosmopolitan, their attitude becomes less parochial. Stein, Post, and Riden (2000) showed that the contextual and behavioral measures of contact have opposite impacts on Whites’ racial attitude toward minorities. A growing proportion of ethnic minority at the county level increases Whites’ negative racial attitude but it also increases frequent interethnic contacts, which in turn helps to reduce Whites’ negative racial attitude. Therefore, we argue that there may be a curvilinear relation of diversity and trust. Although initial contact with members of other ethnic groups may reduce trust, people may become more accustomed to differences, accumulate more knowledge about outgroup, develop interethnic contacts, and adjust their attitude once ethnic diversity has increased to a certain point in the neighborhood. By these arguments, Hypothesis 1: the relation of ethnic diversity and trust is not monotonically negative.
Group and Context Contingency
Previous studies have provided hints that the relation of ethnic diversity and social capital may vary among contexts and groups. Linguistic heterogeneity significantly decreases interpersonal trust in less democratic countries but have no significant impact on trust in more established democracies (Anderson and Paskevicuite 2006). The negative relation of racial composition of the neighborhood and trust is less influential to people who have regularly contact with their neighbors (Stolle, Soroka, and Johnston 2008). Although Putnam did not find significant gender, racial, and generational variations in the negative relation between ethnic diversity and social capital, he called for a better understanding of conditions under which the relation “is strong, weak, or even non-existent” (2007: 163). We argue that the relation of ethnic diversity and trust may be contingent on class, ethnicity, and community type. First, there may be a racial difference in the relation between ethnic diversity and trust. Research suggests that members of majority groups are more likely than those of minority groups to feel their
8 group identity threatened by the presence of ethnic heterogeneity (Forbes 1997). The concentration of subordinate groups can spur hostility from the dominate groups (Key 1949). Although workers in racial and ethnic homogenous workgroups have in general higher productivity, commitment, and cohesiveness, African-American workers’ attitude do not correspond to the racial composition of their workgroups (Riordan and Shore 1997). Using pooled GSS data from 1974 to 1994, Alesina and Ferrara (2002) reported that racial fragmentation at the metropolitan level significantly decreased trust among Whites but not that among Blacks. In a similar vein, Marshall and Stolle (2004) found that racial heterogeneity at the neighborhood level increased the propensity of trust among Blacks but not that among Whites. Using a national representative sample in Canada, Soroka, Helliwell and Johnston (2007) found that the proportion of visible minority in the neighborhood increased the likelihood of interpersonal trust among visible minorities but decreased that among members of the majority.1 Given these findings in the exiting literature, Hypothesis 2a: the relation of ethnic diversity and trust varies among racial groups. Second, the relation of diversity and trust may vary along the class line. Compared to the attention to race and ethnicity here have been few studies on the class variation in the relation of ethnic diversity and social capital. The Social Capital Community Benchmark Survey reported that Class divide in access to social capital was magnified in ethnically diverse communities: In terms of civic activity, there is not much difference between a high- tech executive in Houston and a high-tech executive in Nashua (New Hampshire), but there is a very substantial difference between an auto mechanic in Houston and an auto mechanic in Nashua (Saguaro Seminar 2001:7). Thus, Hypothesis 2b: the relation of ethnic diversity and trust varies among class groups. Third, the relation of ethnic diversity and trust may vary across community types: big cities, suburbs, small towns, and rural area. Simmel (1903 [1950], 1922[1955]) analyzed how population concentration in big cities enabled critical masses and personal freedom that encouraged urbanites to develop socially more intersecting and geographically more dispersed group affiliations than small- town people. Yet, the constantly changing environment in big cities also creates
1 Sorkoa and colleagues measured trust here using the wallet question (2007: 98).
9 stress. As individuals’ mental capability of perception and information process is limited, urbanites develop a blasé attitude to protect themselves from the “intensification of nervous stimulation”. Simmel’s thesis has been further developed by the Chicago School and beyond. Wirth (1938) argued that urbanism as a way of life meant both autonomy and loneliness. Network composition of urban and small-town dwellers are markedly different whereas urbanites have fewer contacts drawn from “the ‘traditional’ complex of kin, neighborhood, and church” and more contacts drawn from “more modern and more voluntary contexts of work, secular associations, and footloose friendship”( Fischer 1982: 258). While urbanites interact with diverse contacts in intersecting social circles with increased autonomy, they also have fragmented solidarity and partial commitments to their networks and vis-à-vis (Wellman 2002; Pescosolido and Ruin 2000). Fischer (1982) further argued that urban life decreases trust among people in larger communities, both because of crime and the sub-cultural variety. Aizlewood and Pendakur (2007) found that community size was an important predictor of the level of interpersonal trust and other type of social capital among Canadians. However, residents in urban and suburban areas may be less sensitive to ethnic diversity in the neighborhood as cities have historically been the major destination of internal and international migration. Park pointed out that “great cities have always been melting-pots of races and of cultures Out of the vivid and subtle interactions of which they have been the centers, there have come the newer breeds and the newer social types” (1925: 40). Compared to residents in small towns or rural areas, urbanites and suburbanites’ greater exposure to people of a variety of racial, ethnic, and cultural backgrounds may wear out the negative relation of ethnic diversity and trust thank to blasé or a more cosmopolitan, tolerant attitude. By contrast, in ethnically homogenous small towns or rural areas, residents may react more strongly to the growing presence of racial and ethnic minorities. Therefore, Hypothesis 2c: the relation of ethnic diversity and trust varies across community types.
Method and Data Analysis
The data have two parts. The first part came from a national, random-digit-dial telephone survey of American adults aged 22-65 who were currently or previously employed. Conducted from November of 2004 to March of 2005, the
10 survey was the US leg of a large-scale comparative study of social capital in the US, China, and Taiwan. It took an average of 35 minutes and 3,000 interviews were completed (hereafter the SC-USA data). The response rate is 43%. The second part of the data consisted of selected variables at the census tract level from the US census data 2000. Merging the two datasets together allows us to include in our analysis neighborhood characteristics measured at the census tract level from the census data and generalized trust, self-reported neighborhood type, and individual characteristics from the SC-USA data. The Dependent Variable The dependent variable is generalized trust. Respondents in the SC-USA survey were asked “Do you think most people can be trusted?” It is similarly formulated as the trust question in the World Values Survey (World Value Study Group 1994). Respondents were asked to choose their answer from a 1-4 scale where 1 means “almost all of the people cannot be trusted”, 2 “Most people cannot be trusted”, 3 “Most people can be trusted”, and 4 “Almost all of the people can be trusted”. Slightly more than 5% of the respondents thought that almost all of the people could not be trusted, 14% thought that most people could not be trusted, 65% thought that most people could be trusted, while 16% thought that almost all of the people could be trusted. Independent Variables Ethnic diversity. We measure ethnic diversity through the ethnic diversity at the census tract level and self-reported neighborhood type. We construct ethnic diversity in a census tract k by the Herfindahl index (Hipp 2007). It is calculated as follows:
where Pi is the proportion of the population of ethnic group i out of a total population of I raical/ethnic groups. Five racial/ethnic groups are included: white, black, Latino, Asian, and others. White, throughout this paper, refers to non-Hispanic white. The greater the value of the index, the more ethnically diverse the neighborhood. Self-reported neighborhood type. While census tract has been widely used as a proxy of neighborhood, it may differ from the subjective neighborhood experienced by individuals in their everyday life (Putnam 2007). The SC-USA data has a subjective measure of experienced neighborhood type. Respondents
11 were asked the question: "Do most of your neighbors use English or another language at home” and given the following answers: 1) mostly English, 2) mostly some other language(s), and 3) a mixture of English and other language(s). Three types of neighborhood are constructed: English-speaking neighborhood where most of the respondent’s neighbors speak mostly English, ethnic enclave neighborhood where most neighbors speak some other languages, and mixed neighborhood where both English and other languages are used. On average, 83% of the respondents live in English-speaking neighborhood, 6% in ethnic enclaves, and 11% in mixed neighborhood. Neighborhood characteristics. We take into account neighborhood characteristics by controlling the following contextual variables drawing from the US census data 2000: the percentage of recent immigrants in the neighborhood measured by the share of foreign-born residents who entered the country since 1990, the percentage of unemployed, the percentage of people living below poverty, the percentage of renter, the percentage of people who had BA degree, the percentage of people aged 65 years and older, and the average commuting time for workers aged 16 years and older not working at home. These variables are included to capture the exposure of the neighborhood to recent immigrants and the socioeconomic resources and demographic characteristics of the neighborhood. Individual characteristics. It is important to control individual factors that may be related to generalized trust: gender, age, homeownership, education, whether English was spoken at home, nativity, household income, and residential tenure measured as the decades the respondent had been living in the neighborhood. The analysis also takes into account race, class, and community type. Respondents were asked the question: “If the society is divided into upper class, upper-middle class, middle class, middle-lower class, and lower class, which one do you think you belong to?” Based on answers to the question, the respondents are categorized into three classes: upper class (including both upper class and upper-middle class), middle class, and low class (including both middle-lower class and lower class). Respondents are grouped into Whites, Blacks, Latinos, and others based on self-reported racial/ethnic background. A cross-tabulation of class and race indicates that Blacks and Latinos are disproportionably distributed in low class and vice versa. Nonetheless, there are considerable proportions of both low class Whites and upper and middle class racial minorities. Based on respondent’s answer, community type includes big cities, suburbs, small towns, and rural area.
12 Results
Across the Board
Table 2 reported a series of OLS regression on generalized trust of the whole sample Binary and multinomial logistic regression models are also employed to examine the relation of ethnic diversity and generalized trust. The results are similar to the OLS regression models. Results suggest a significant non-linear relation of ethnic diversity in the neighborhood and generalized trust at the individual level. While individuals’ generalized trust is negatively related to ethnic diversity in the neighborhood, it bounces back as the level of ethnic diversity passes a tipping point. The pattern holds after self-reported neighborhood type, individual characteristics including race and class, and community type are controlled. People living in ethnic enclaves have a significantly lower level of generalized trust. Residents of mixed neighborhood also have a lower level of trust compared to those living in English-speaking neighborhood, although the difference is not statistically significant. Thus, hypothesis 1 is supported that the relation of ethnic diversity in the neighborhood and generalized trust at the individual level is nonlinear. [INSERT TABLE 2 ABOUT HERE] The percentage of recent immigrants in the neighborhood has no significant relation to generalized trust. That is, the influx of new immigrants into the neighborhood does not have significant negative impact on generalized trust, when other factors are controlled. The unemployment rate in the neighborhood has a significant negative relation with generalized trust. A higher level of unemployment in the neighborhood is associated with a lower level of generalized trust. In comparison, the proportion of residents with BA degree in the neighborhood significantly increases generalized trust at the individual level but loses statistic significance when individual class status is controlled. The average commuting time at the neighborhood level also has a role to play in affecting individuals’ generalized trust. Residents in neighborhoods where people have longer commuting time have lower levels of generalized trust. However, the relation loses significance when race is controlled. Other neighborhood characteristics have no significant relation with generalized trust at the individual level. In sum, low levels of some dimensions of socioeconomic resources in the neighborhood are negatively associated with generalized trust.
13 Similar to previous studies (Putnam 2007; Aizlewood and Pendakur 2006), individual characteristics are found to be more important to generalized trust than neighborhood characteristics. The results are in tandem with findings in existing research that disadvantaged individuals tend to trust less. Generalized trust significantly increases with age, education, and speaking English at home. Nativity also significantly boosts trust, although its significance disappears when race is controlled. Latinos and Blacks are significantly less trusting than Whites and other groups. Low class Americans are significantly less trusting than their upper and middle class counterparts. It seems that there is no significant relation between generalized trust and community types.
By Race and Ethnicity
Table 3 reported a series of OLS regression on generalized trust among three racial groups: Whites, Blacks, and Latinos. Asians and others were not included due to their relatively small size in the sample. The significant nonlinear relation between ethnic diversity at the neighborhood level and generalized trust at the individual level only holds for Whites. Generalized trust among Blacks and Latinos is not significantly affected by ethnic diversity in the neighborhood. Nonetheless, compared to other ethnic groups, Latinos living in ethnic enclaves have a significantly lower level of generalized trust. Hypothesis 2a is supported that the relation of ethnic diversity and trust varies across racial and ethnic groups. [INSERT TABLE 3 ABOUT HERE]
By Class
In table 4 we split the whole sample into three subgroups along the class line: upper class, middle class, and low class. We have also used current SEI as the indicator of class and trichotomized it into three subgroups. The results are similar to the categorization based on self-reported class identification. The significant nonlinear relation between ethnic diversity at the neighborhood level and generalized trust at the individual level identified in the whole sample only holds among middle class. At the upper and the lower end of the class strata, the relation of ethnic diversity and generalized trust is insignificant. The results suggest that neither upper nor low class is as sensitive as middle class to ethnic diversity in the neighborhood. Moreover, compared to the whole sample, living in ethnic enclave neighborhood significantly decreases trust among upper and middle class but has a positive yet insignificant relation to trust among low class.
14 Therefore, Hypothesis 2b can be accepted that the relation of ethnic diversity and trust varies markedly across class. [INSERT TABLE 4 ABOUT HERE]
By Community Type
In table 5 the sample is split based on community type: big cities, suburbs, small towns, and rural areas. The results show that the nonlinear relation of ethnic diversity and generalized trust remains significant only for residents in small towns. Ethnic diversity in the neighborhood has a positive but non-significant relation with generalized trust among urbanites and suburbanites. Ethnic diversity in the neighborhood has a negative yet insignificant relation with generalized trust among residents in rural area. Living in ethnic enclaves significantly decreases generalized trust among suburbanites. Thus, hypothesis 2c is supported that the dynamics between ethnic diversity and generalized trust play out differently in different types of communities. [INSERT TABLE 5 ABOUT HERE]
Summary and Discussions
The results demonstrate a significant non-linear relation of ethnic diversity in the neighborhood and generalized trust at the individual level. Individuals’ generalized trust bounces back as the level of ethnic diversity passes a tipping point, controlling self-reported neighborhood type, individual characteristics including race and class, and community type. Moreover, the relation of ethnic diversity and generalized trust varies considerably across class, race, and community type. First, generalized trust among upper and lower class are not as responsive to ethnic diversity in the neighborhood as that among middle class. Second, compared to Whites, generalized trust among Blacks and Latinos is less responsive to ethnic diversity in the neighborhood. Third, compared to urbanites, suburbanites, or rural residents, residents in small towns react more strongly when ethnic diversity in their neighborhood increases. Ethnic diversity measured by the experienced neighborhood also has a role to play in generalized trust. In general, residents in ethnic enclaves have significantly less trust than those living in neighborhood where most residents speak English. The negative relation varies considerably along class, race, and community type whereas living in ethnic enclave neighborhood significantly decreases generalized trust among upper and middle class, Latinos, and suburbanites. Living in mixed neighborhood has no significant relation to generalized trust.
15 After or even before race and class are controlled, most neighborhood characteristics have no significant relation to generalized trust. In the whole sample, the influx of recent immigrants to the neighborhood has no significant relation to trust. It however significantly decreases generalized trust among suburbanites and Latinos. In rural area, the most important factor related to generalized trust is the influx of recent immigrants, which actually increases generalized trust among rural residents. In the whole sample, a higher level of unemployment is significantly related to a lower level of generalized trust. It also significantly decreases trust among Upper class, Latinos, and small town residents, respectively. In overall, the results suggest that increased contact with immigrants does not have significant negative impact on generalized trust but the lack of some dimensions of socioeconomic resources in the neighborhood significantly reduces generalized trust. Individual characteristics are more critical to generalized trust than neighborhood characteristics. For the whole sample, generalized trust significantly increases with age, education, and speaking English at home. This general pattern only holds true for Whites but not for Blacks and Latinos. In overall, education is significant to generalized trust except among upper class, blacks, and in rural area. Speaking English at home significantly increases generalized trust among whites, middle and low class, and people living in big cities. Compared to socioeconomic status captured by education, homeownership and income have limited effect on generalized trust. Generalized trust is significantly affected by race and class. Generalized trust is significantly affected by race and class whereas Latino, black, and low class people are significantly less trusting. A closer look of subgroups along racial and class lines suggests that the racial gap in generalized trust primarily exists among middle and upper class and the class divide in generalized trust among Whites. That is, race trumps class in affecting generalized trust among Blacks and Latinos. Unlike Whites, higher class status does not translate into more trust among middle and upper class Blacks and Latinos. Low class people, regardless race, are less trusting than middle and upper class people. That is, race is not important to generalized trust among people of disadvantaged class status. There is a clear class divide in trust among Whites and higher levels of trust go with higher class status. For Blacks, the relation is reversed but not significant. For Latinos, the relation of class and trust is nonlinear. The level of generalized trust among upper class Latinos is significantly lower than that among their middle and low class co-ethnics. Future research needs to explore why the relation of class and trust among Whites is opposite to that among racial minority groups.
16 Race and class also affect generalized trust differently across community types. In big cities and suburbs, Latinos have significantly less trust than members of other racial group. In small towns and rural area, Blacks have significantly less trust than other racial groups. The class divide in generalized trust is most evident in small towns where as higher class status means higher levels of generalized trust. Middle class suburbanites have a significant higher level of trust than both upper and low class suburbanites. The class-based trust divide in big cities and rural area is fragmentary as there is a lack of statistic significance and a lack of uniformed direction. Sensitivity analysis The census data that provide variables on neighborhood characteristics were collected in 2000 and the SC-USA data including the variable about generalized trust in 2004/2005. Thus, to some extent, we can argue that ethnic diversity at the neighborhood level affects individual’s generalized trust in a significant non- linear manner. In analysis not presented here, we have also examined language diversity in the neighborhood and the results are similar to that of ethnic diversity. We have also used the proportions of various racial groups in the neighborhood instead of the Herfindahl index but did not find any significant relation between the proportion of any racial group in the neighborhood and generalized trust. What is relevant to generalized trust seems to be the overall composition rather than the relative size of any particular racial group in the neighborhood. The results suggest that the factors that affect generalized trust work differently between groups and strata. The same model often has dramatically different explanation power across groups and strata. For instance, the model is extremely weak in predicting generalized trust among Blacks. It reveals that we need to develop better ways to incorporate context and history into the analysis of ethnic diversity and social capital (Hero 2007). Many studies refer to the contact hypothesis but few have actually tested it when examining the relation of diversity and trust. In analysis not presented here, we have used participation in neighborhood organizations and the number of neighbors in the respondent’s social networks as proxy of contact within the neighborhood. While participation in neighborhood organizations is positively related to generalized trust, the statistic significance disappears when other factors are controlled. The number of previous and current neighbors in social networks also has no significant relation with generalized trust. Self-selection
17 Ethnic diversity in the neighborhood can be the product of complicated social processes. It is a debating issue whether and the extent to which ethnic homogeneity in the neighborhood can be attributed to ingroup preference and structural exclusion. Ethnic homogenous neighborhoods may include both the “excluded” - deprived inner-city racial minority enclaves -and the “exclusionary” - high-end suburban gated communities. Residents in the “excluded ghettos” tend to be segregated minority members with fewer economic opportunities (Massey and Denton 1993). Magee, Fong and Wilkes (2007) reported that Chinese immigrants living in neighborhoods with higher level of co-ethnics were more likely to report discrimination than their counterparts living in less Chinese concentrated neighborhoods. By contrast, residents of exclusionary enclaves are more likely to be self-segregated to “protected themselves from a perceived danger from below” (Marcuse 1997: 314). In a similar vein, mixed neighborhoods are diverse in both ethnic and socioeconomic terms (Fong and Shibuya 2005). They may include upscale gentrified urban neighborhood, affluent ethnoburbs - a new form of suburban multiethnic community (Li 1998), and inner-city twilight zone where low-income immigrants of various ethnic background live side by side. Race/ethnicity and socioeconomic status are the most important factors on neighborhood attainment, after a variety of factors are controlled (Sampson and Sharkey 2008). Oliver and Wong (2003) found that self selection was primarily a white phenomenon. Self-segregation is less an issue for racial minority and people with modest socioeconomic means as they have less latitude in choosing the neighborhood they prefer. Yet, for Whites and privileged people who have the luxury of self-selection, how does it factor into the relation of ethnic diversity and trust? Putnam (2007) argue that it is not very plausible that “white flight” has left behind less trusting people in neighborhoods “invaded” by ethnic minorities. It is also not very plausible that more trusting people select ethnically homogenous neighborhoods and less trusting people ethnically diverse neighborhoods. Yet, is the relation between ethnic diversity and generalized trust spurious because disadvantaged people tend to distrust and they also tend to live in ethnic diverse neighborhoods? Alesina and Ferrara (2002) argued that the aggregated level of generalized trust in ethnically heterogeneous neighborhood may be dragged down by a higher proportion of distrusting residents who were disadvantaged and a vicious circle began as distrust diffused. Our results however show that the relation of ethnic diversity and generalized trust is not spurious as it remains significant after controlling socioeconomic resources in the neighborhood and individual socioeconomic status, race, and class.
18 Conclusions and Contributions
There has been much ado about the relation of ethnic diversity and social capital. Diversity is not a liability that spoils social capital. Trust is based on bonding ties and diversity provides the opportunity for developing bridging ties. To tap into the diversity that nurtures creativity, that transcends a narrowly defined “we” based on ascribed traits such as race and ethnicity, it takes more than the appreciation of ethnic cuisine. One of the first steps is to appreciate a complex amalgam of class, race, and community type that mediate the relation of diversity and trust. Hero argued that existing studies on social capital had adopted an “color-blind’ approach, although “social outcomes look quite different when disaggregated by racial groups than when lumped together” (2007: 155). Indeed, the same approach has also aggregated class and community type. In this paper, we advance existing studies on ethnic diversity and trust through examining their relation across class, race, and community types. It is a small step from merely treating them as control variables. Yet, we have made a few discoveries. First, the paper casts doubt on a monolithic picture of a negative relation of ethnic diversity and social capital. Ethnic diversity, measured either objectively by neighborhood characteristics at the census tract level or subjectively by respondent’s self-reported neighborhood type, shows a non-linear relation with generalized trust contingent on race, class, and community type. Ethnic diversity in the neighborhood has a negative relation with generalized trust only up to a threshold point. Second, the paper demonstrates that the production of social capital is contingent. The relation of diversity and trust changes markedly across race, class and community type. That is, the mechanisms behind a negative relation between ethnic diversity and social capital work differently across racial/ethnic groups, class, and communities. Once splitting the whole sample along race, class, and community types, the parabolic relation of ethnic diversity and trust only remains significant among Whites, middle class, and small town residents. Living in ethnic enclaves decreases generalized trust in the whole sample and the relation holds among Latinos, middle and upper class, and suburbanites, respectively. Third, there is a clear race and class divide in generalized trust. While Latinos, Blacks, and people of low class tend to be less trusting, a closer look into subgroups reveals that the racial gap in trust primarily exists among middle and upper class and class divide in trust among Whites.
19 Why is there a nonlinear relation of ethnic diversity at the neighborhood level and generalized trust at the individual level? Why does trust bounce back when ethnic diversity increases after a tipping point? Our results provide support to the contact hypothesis that growing contact between racial and ethnic groups help to increase mutual understanding and promote trust (Pettigrew 1998). The paper has several limitations. First, the sample, although nationally representative, is not a cluster sample design. Thus, the data structure does not allow us to conduct a multilevel analysis. Second, Sturgis, Patulny, and Allum (2007) reported that many factors that significantly affect trust in a cross- sectional model lose their significance in panel data analysis. Thus it is important to use panel data, which will also offer a better understanding of the casual order. Third, our data do not differentiate in-group and out-group trust. Fourth, our broad-brush categorization of racial groups may gloss over many of the sub- ethnic group differences and the perception and reaction to the presence of different sub-ethnic groups in the neighborhood may vary considerably. Fifth, better measures of social contacts within and beyond neighborhood, for instance the ethnic diversity of people’s daily contacts, are needed. In this paper, we have focused on ethnic diversity in the neighborhood and generalized trust at the individual level as an important component of social capital. Various type of social capital including neighborhood attachment, civic participation, and memberships in voluntary associations are found to be related to trust (Brehem and Rahn 1997; Li, Pickles and Savage 2005; Sturgis, Patulny, and Allum 2007; Paxton 2007). We argue that future research needs to explore the relations of diversity and trust through a better conceptualization of trust and diversity and a network approach. First, we call for more research on relational trust, which Cook defined as “the kind of specific trust that emerges between individuals in specific types of social relations” (2005: 9). What the WVS generalized trust question is actually measuring has been increasingly questioned. Is trust normative, behavioral, relational, or all of above? Uslaner (2002) distinguished moralistic trust from strategic trust whereas the former was a moral predisposition and the latter built on previous experience. Based on results that education and financial situation were the only two factors that significantly affected interpersonal trust in panel data, Sturgis, Patulny, and Allum (2007) speculated that trust was a normative value learned in the early childhood and remained relatively stable over the life course. Soroka and colleagues (2006) argued that generalized trust was likely to be the product of cultural learning as they found that it was significantly related to immigrants’ country of origin.
20 Second, we call for a more comprehensive understanding of diversity as a multidimensional construct. Race and ethnicity is only one of the many dimensions of diversity. We know little about the roles of diversities of class, gender, religion, occupation, or sexual orientation in the production of social capital. Simmel (1908 [1971]) argued that geographic closeness did not necessarily mean social closeness. Although neighborhood has been and still is one of the most important foci where people develop social relationships, neighbor ties may only present a small and shrinking part of people’s social networks (McPherson, Smith-Lovin, and Brashears 2006). The vast majority of white Americans do not live in racially diverse neighborhoods and have no direct competition of neighborhood resources with minority groups (Oliver and Mendelberg 2000). In Britain, Letki (2008) found that racial diversity in the neighborhood had little impact on social interaction with neighbors but was negatively related to trust in neighbors, which led her to speculate that interracial attitude was affected by implicit and explicit racial stereotyping in the wider society beyond factors at the neighborhood level. We need go beyond diversity in the neighborhood by including diversity at places where people work, worship, and play. Third, if both social capital and diversity is relational, the changing structures of social network and the ways in which people interact with one another may hold the key to the understanding of diversity and social capital. For Coleman (1990), social capital is a mechanism that generates and preserves resources through a densely knit, close network structure. For Burt (1992), social capital comes from bridging otherwise disconnected groups. Lin (1999) emphasized resources embedded in social networks - the socioeconomic status of network members - as an important factor related to access and mobilization of resources. Future research needs to incorporate these insights and draw on refined network measures to facilitate a deeper understanding of the relation of diversity and social capital.
21 References
Aizlewood, Amanda and Ravi Pendakur. 2007. "Ethnicity and Social Capital in Canada." Pp. 169-198 in Social Capital, Diversity, and the Welfare State, edited by F. M. Kay and R. Johnston. Vancouver: UBC Press. Alba, Richard and Victor Nee. 1997. "Rethinking assimilation theory for a new era of immigration." International Migration Review 31:826-874 Alesina, Alberto and Eliana La Ferrara. 2000. "Who Trusts Others?" Journal of Public Economics 85:207-234. Allport, Gordon W. 1954. The Nature of Prejudice. Cambridge: Addison-Wesley Anderson, Christopher J., and Aida Paskeviciute. 2006. "How Ethnic and Linguistic Heterogeneity Influence the Prospects for Civil Society: A Comparative Study of Citizenship Behavior." The Journal of Politics 68:783-802. Arneil, Barbara. 2006. Diverse Communities: The Problem with Social Capital. Cambridge: Cambridge University Press. Blau, Peter M. 1977. Inequality and Heterogeneity: A Primitive Theory of Social Structure. New York: The Free Press. Bown, Rupert. 1995. Prejudice: its social psychology. London: Blackwell Scientific Brehm, John and Wendy Rahn. 1997. "Individual-Level Evidence for the Causes and Consequences of Social Capital." American Journal of Political Science 41:999-1023. Burt, Ronald. 1992. Structural Holes. Chicago: University of Chicago Press. Coffé, Hilde and Benny Geys. 2006. "Community Heterogeneity: A Burden for the Creation of Social Capital?" Social Science Quarterly 87:1053-1072. Coleman, James S. 1988. “Social Capital in the Creation of Human Capital.” The American Journal of Sociology 94: S95-S120. Cook, Karen S., Russel Hardin, and Margaret Levi. 2005. "Cooperation Without Trust?" New York: Russell Sage. Cook, Karen. S. 2005. "Networks, norms, and trust: The social psychology of social capital 2004 Cooley Mead Award Address." Social Psychology Quarterly 68:4-14. Costa, Dora L. and Matthew E. Kahn. 2001. "Understanding the Decline in Social Capital, 1952-1998." National Bureau of Economic Research.
22 Costa, Dora L. and Matthew E. Kahn. 2003. "Civic Engagement and Community Heterogeneity: An Economist's Perspective." Perspectives on Politics 103-111: Working Paper 8295 Fischer, Claude. 1982. To Dwell Among Friends. Berkeley: University of California Press. Florida, Richard. 2002. The Rise of the Creative Class. New York: Basic Books. Fong, Eric and Kumiko Shibuya. 2005 "Multiethnic Cities in North America." Annual Review of Sociology 31:285-304. Forbes, H. D. 1997. Ethnic Conflict: Commerce, Culture, and the Contact Hypothesis New Haven: Yale University Press. Gibson, Cristina B., and Jennifer L. Gibbs. 2006. “Unpacking the Concept of Virtuality: The Effects of Geographic Dispersion, Electronic Dependence.” Administrative Science Quarterly 51:451-495. Glaeser, Edward L., David I. Laibson, José A. Scheinkman, and Christine L. Soutter. 2000. "Measuring trust." Quarterly Journal of Economics 115:811-846. Gordon, Milton M. 1964 Assimilation in American Life: The Role of Race, Religion, and National Origins. New York Oxford University Press. Guest, Avery M., Charis E. Kubrin, and Jane K. Cover. 2008. "Heterogeneity and Harmony: Neighbouring Relationships among Whites in Ethnically Diverse Neighborhoods in Seattle." Urban Studies 45:501-526. Gustavsson, Magnus and Henrik Jordahl. 2008. "Inequality and trust in Sweden: Some inequalities are more harmful than others." Journal of Public Economics 92:348-365. Hero, Rodney E. 2007. Racial Diversity and Social Capital: Equality and Community in America. Cambridge: Cambridge University Press. Hipp, John R. 2007. "Block, Tract, and Levels of Aggregation: Neighborhood Stucture and Crime and Disorder as a Case in Point." American Journal of Sociology 72:659-680. Hooghe, Marc. 2007. "Social Capital and Diversity: Generalized Trust, Social Cohesion and Regimes of Diversity." Canadian Journal of Political Science 40:709-732. Horowitz, Donald L. 1985. Ethnic groups in conflict. Berkeley: University of California Press.
23 Isralowitz, Richard E. 1998. "Attitudes of Israeli Kibbutz members toward immigration." The Journal of Social Psychology 138:259-262. Kay, Fiona M. and Richard Johnston. 2007. "Social Capital, Diversity, and the Welfare State." Vancouver: UBC Press. Kivisto, Peter. 2001. “Theorizing transnational immigration: a critical review of current efforts.” Ethnic and Racial Studies 24:549-577. Knack, Stephen and Philip Keefer. 1997. "Does Social Capital have an Economic Payoff? A Cross-country Investigation." The Quarterly Journal of Economics 112:1251-1288 Knippenberg, Daan van and Michaéla C. Schippers. 2007. "Workgroup Diversity." Annual Review of Psychology 58:515-541 Leigh, Andrew. 2006. "Trust, Inequality and Ethnic Heterogeneity." The Economic Record 82:268-280. Letki, Natalia. 2008. "Does Diversity Erode Social Cohesion? Social Capital and Race in British Neighbourhoods." Political Studies 56:99-126. Li, Wei. 1998. "Anatomy of a New Ethnic Settlement: The Chinese Ethnoburb in Los Angeles." Urban Studies 35:479-501. Li, Yaojun, Andrew Pickles, and Mike Savage. 2005. "Social Capital and Social Trust in Britain." European Sociological Review 21:109-123. Lin, Nan. 1999. "Social Networks and Status Attainment." Annual Review of Sociology 25:467-487. Magee, William, Eric Fong, and Rima Wilkes. 2007. "Neighbourhood Ethnic Concentration and Discrimination." Journal of Social Policy 37:1-25. Marcuse, Peter. 1997. "The Ghetto of Exclusion and the Fortified Enclave: New Patterns in the United States." American Behavioral Scientist 41:311-326. Marschall, Melissa J. and Dietlind Stolle. 2004. "Race and the city: Neighborhood context and the development of generalized trust." Political Behavior 26:125-153. Massey, Douglass and Nancy Denton. 1993. American apartheid: Segregation and the making of the underclass. Cambridge, MA: Harvard University Press. McCain, Paula D. 2003. "Social Capital and Diversity: An Introduction " Perspectives on Politics 1:101-102. McPherson, Miller, Lynn Smith-Lovin, and Matthew E. Brashears. 2006. “Social isolation in America: Changes in Core Discussion Networks over Two Decades.” American Sociological Review 71:353-375.
24 Mizruchi, Mark S. and Linda Brewster Stearns. 2001. “Getting Deals Done: The Use of Social Networks in Bank Decision-Making.” American Sociological Review 66: 647-671. Oliver, Eric and Tali Mendelberg. 2000. "Reconsidering the Environmental Determinants of White Racial Attitudes." American Journal of Political Science 44:574-589. Oliver, Eric J. and Janelle Wong. 2003. "Intergroup Prejudice in Multiethnic Settings." American Journal of Political Science 47:567-582. Park, Robert Ezra. 1925. "The City." Pp. 1-46 in The City, edited by R. E. Park, E. W. Burgess, and R. D. McKenzie. Chicago: University of Chicago Press. Paxton, Pamela. 2005. "trust in decline?" Contexts 4:40-46. Paxton, Pamela. 2007. "Association Memberships and Generalized Trust: A Multilevel Model Across 31 Countries." Social Forces 86:47-76. Pescosolido, Bernice A., and Beth A. Rubin. 2000. “The Web of Group Affiliations Revisited: Social Life, Postmodernism, and Sociology.” American Sociological Review 65:52-76. Pettigrew, Thomas F. 1998. "Intergroup Contact Theory." Annual Review of Psychology 49:65-85. Pettigrew, Thomas F. and Linda R. Tropp. 2006. "A Meta-Analytic Test of Intergroup Contact Theory." Journal of Personality and Social Psychology 90:751- 783. Putnam, Robert. 1995. "Tuning In, Tuning Out: The Strange Disappearance of Social Capital in America." PS: Political Science and Politics 28:664-683. Putnam, Robert. 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster. Putnam, Robert. 2007. "E Pluribus Unum: Diversity and Community in the Twenty- first Century -- The 2006 Johan Skytte Prize Lecture." Scandinavian Political Studies 30:137-174. Quillian, Lincoln. 1995. "Prejudice as a Response to Perceived Group Threat: Population Composition and Anti-Immigrant and Racial Prejudice in Europe." American Sociological Review 60:586-611. Reagans, Ray, Ezra Zuckerman, and Bill McEvily. 2004. "How to Make the Team: Social Networks vs. Demography as Criteria for Designing Effective Teams." Administrative Science Quarterly 49:101-133
25 Rice, Tom W. and Brent Steele. 2001. "White Ethnic Diversity and Community Attachment in Small Iowa Towns." Social Science Quarterly 82:397-407. Riordan, Christine M., and Lynn McFarlane Shore. 1997. "Demographic Diversity and Employee Attitudes: An Empirical Examination of Relational Demography Within Work Unites." Journal of Applied Psychology 82:342-358. Ross, Catherine E., John Mirowsky, and Shana Pribesh. 2001. "Powerlessness and the Amplification of Threat: Neighborhood Disadvantage, Disorder, and Mistrust." American Sociological Review 66:568-591. Ruef, Martin. 2002. “Strong ties, weak ties and islands: structural and cultural predictors of organizational innovation.” Industrial and Corporate Change 11:427-449. Saguaro Seminar. 2001. "Social Capital Community Benchmark Survey Executive Summary." Civic Engagement in America at the John F. Kennedy School of Government of Harvard University, Boston. Sampson, Robert J., and Patrick Sharkey. 2008. "Neighborhood Selection and The Social Reproduction of Concentrated Racial Inequality." Demography 45:1- 29. Sampson, Robert J., Stephen Raudenbush, and Felton Earls. 1997. "Neighborhoods and Violent Crime: A Multilevel Study of Collective Efficacy." Science 277:918-924. Simmel, Georg. 1903 [1950]. "The Metropolis and Mental Life." Pp. 409-424 in The Sociology of Georg Simmel vol. Free Press, edited by K. Wolff. New York. Simmel, Georg. 1922 [1955]. "The Web of Group Affiliations." Pp. 125-95 in Conflict and the Web of Group Affiliations, edited by K. Wolff. Glencoe, IL: Free Press. Simmel, George. 1908 [1971]. “The Stranger.” Pp. 143–149 in George Simmel on Individuality and Social Forms, edited by Donald .N. Levine. Chicago, IL: University of Chicago Press. Soroka, Stuart N., John F. Helliwell, and Richard Johnston. 2007. "Measuring and Modelling Interpersonal Trust." Pp. 95-112 in Social Capital, Diversity, and the Welfare State, edited by F. M. Kay and R. Johnston. Vancouver: UBC Press. Stein, Robert M., Stephanie Shirley Post, and Allison L. Rinden. 2000. "Reconciling Context and Contact Effects on Racial Attitudes." Political Research Quarterly 53:285-303.
26 Stolle, Dietlind, Stuart Soroka, and Richard Johnston. 2008. "When does Diversity Erode Trust? Neighborhood Diversity, Interpersonal Trust and the Mediating Effect of Social Interactions." Political Studies 56:57-75. Stouffer, Samuel A. 1949. The American soldier. Princeton: Princeton University Press. Sturgis, Patrick, R. Patulny, and . N. Allum. 2007. "Re-evaluating the Individual level causes of trust: a panel data analysis." Paper presented at the European Survey Research Association conference, Prague, June 2007. Tajfel, Henri. 1981. Human Groups and Social Categories: Studies in Social Psychology. Cambridge: Cambridge University Press. Uslaner, Eric M. 2002. The Moral Foundations of Trust. Cambridge, UK: Cambridge University Press. Wellman, Barry. 2002. “Little Boxes, Glocalization, and Networked Individualism.” Pp. 10-25 in Digital Cities II: Computational and Sociological Approaches, edited by Makoto Tanabe, Peter van den Besselaar, and Toru Ishida. Berlin: Springer. Wirth, Louis. 1938. "Urbanism as a Way of Life." The American Journal of Sociology 44:1-24. World Value Study Group. 1994. World value survey 1981-1984 and 1990-1993. Ann Arbor: MI: ICPSR. Yamagishi, Toshio and Midori Yamagishi. 1994. "Trust and commitment in the United States and Japan." Motivation and Emotion 18:129-166
27 Table 1: Descriptive Statistics Variables N Mean Std. Dev. Min Max
Generalized trust 3000 2.92 0.71 1.00 4.00 Ethnic diversity 2931 0.29 0.19 0.01 0.75 Ethnic diversity squared 2931 0.12 0.12 0.00 0.56 Ethnic enclave 3000 0.06 0.23 0.00 1.00 Mixed neighborhood 3000 0.11 0.31 0.00 1.00 Percent of recent immigrants 2931 5.01 7.81 0.00 61.68 Percent of unemployed 2931 6.38 5.43 0.00 81.37 Percent of people below poverty 2931 13.16 10.62 0.00 70.87 Percent of renter 2931 34.09 22.72 0.00 100.00 Percent of people with BA degree 2931 23.43 16.40 0.69 87.36 Percent of people 65 years and older 2931 11.92 5.92 0.00 58.78 Average commuting time 2931 26.09 7.40 8.89 56.26 Gender (male=1) 3000 0.46 0.50 0.00 1.00 Age 3000 41.48 10.57 21.00 64.00 Homeownership 3000 0.73 0.44 0.00 1.00 Education 2995 6.05 1.28 1.00 8.00 Speaking English at home 3000 0.89 0.31 0.00 1.00 Nativity (native born) 3000 0.86 0.35 0.00 1.00 Household income 3000 16.66 7.82 1.00 28.00 Residential tenure 3000 1.71 1.48 0.00 6.50 White 3000 0.69 0.46 0.00 1.00 Black 3000 0.12 0.32 0.00 1.00 Latino 3000 0.14 0.34 0.00 1.00 Others 3000 0.05 0.22 0.00 1.00 Upper class 3000 0.19 0.40 0.00 1.00 Middle class 3000 0.56 0.50 0.00 1.00 Low class 3000 0.24 0.43 0.00 1.00 Big cities 3000 0.29 0.46 0.00 1.00 Suburbs 3000 0.24 0.43 0.00 1.00 Small towns 3000 0.36 0.48 0.00 1.00 Rural area 3000 0.11 0.31 0.00 1.00 Quota 3000 0.44 0.50 0.00 1.00
28 Table 2: OLS Regression on Generalized Trust Model 1 Model 2 Model 3 Model 4 Coef. SE t Coef. SE t Coef. SE t Coef. SE t Ethnic diversity -0.729 0.260 -2.81 ** -0.718 0.259 -2.77 ** -0.646 0.256 -2.52 * -0.663 0.256 -2.59 ** Ethnic diversity squared 1.138 0.391 2.91 ** 1.110 0.391 2.84 ** 1.013 0.386 2.62 ** 1.039 0.386 2.69 ** Ethnic enclave -0.188 0.062 -3.01 ** -0.156 0.062 -2.51 * -0.150 0.062 -2.42 * Mixed neighborhood -0.014 0.044 -0.31 0.001 0.044 0.02 0.003 0.044 0.08 Percent of recent immigrants -0.003 0.002 -1.44 -0.002 0.002 -0.76 -0.003 0.002 -1.18 -0.003 0.002 -1.25 Percent of unemployed -0.007 0.003 -2.02 * -0.007 0.003 -2.04 * -0.007 0.003 -2.11 * -0.007 0.003 -2.05 * Percent of people below poverty -0.002 0.002 -1.01 -0.002 0.002 -0.92 0.000 0.002 0.10 0.000 0.002 0.14 Percent of renter 0.000 0.001 -0.20 0.000 0.001 -0.33 0.000 0.001 -0.42 0.000 0.001 -0.37 Percent of people with BA degree 0.002 0.001 2.65 ** 0.002 0.001 2.49 * 0.002 0.001 1.67 † 0.001 0.001 1.58 Percent of people 65 years and older 0.000 0.002 -0.18 -0.001 0.002 -0.30 -0.002 0.002 -1.03 -0.002 0.002 -1.03 Average commuting time -0.005 0.002 -2.48 * -0.005 0.002 -2.59 ** -0.003 0.002 -1.50 -0.003 0.002 -1.51 Gender (male=1) 0.008 0.025 0.32 0.006 0.025 0.26 0.001 0.024 0.04 0.004 0.024 0.14 Age 0.004 0.001 3.04 ** 0.004 0.001 3.16 ** 0.004 0.001 2.83 ** 0.004 0.001 2.83 ** Homeownership 0.030 0.032 0.95 0.028 0.032 0.87 0.015 0.031 0.47 0.007 0.032 0.23 Education 0.058 0.011 5.38 *** 0.057 0.011 5.26 *** 0.045 0.011 4.22 *** 0.043 0.011 3.92 *** Speaking English at home 0.342 0.063 5.42 *** 0.300 0.065 4.63 *** 0.187 0.070 2.68 ** 0.187 0.070 2.68 ** Nativity 0.088 0.053 1.66 † 0.090 0.053 1.70 † 0.045 0.053 0.85 0.043 0.053 0.82 Household income 0.001 0.002 0.81 0.001 0.002 0.86 0.002 0.002 1.08 0.001 0.002 0.87 Residential tenure 0.013 0.009 1.37 0.012 0.009 1.36 0.017 0.009 1.90 † 0.017 0.009 1.89 † White -0.093 0.058 -1.62 -0.091 0.058 -1.58 Black -0.369 0.067 -5.48 *** -0.364 0.067 -5.40 *** Latino -0.446 0.070 -6.36 *** -0.440 0.070 -6.28 *** Upper class 0.065 0.039 1.66 † Middle class 0.067 0.030 2.20 * Big cities -0.030 0.050 -0.60 -0.027 0.050 -0.54 0.054 0.050 1.07 0.050 0.050 1.00 Suburbs -0.020 0.048 -0.43 -0.018 0.047 -0.38 0.024 0.047 0.50 0.017 0.047 0.35 Small towns 0.005 0.044 0.11 0.009 0.044 0.20 0.042 0.044 0.97 0.039 0.044 0.89 quota 0.036 0.028 1.29 0.037 0.028 1.32 0.060 0.028 2.18 * 0.059 0.028 2.15 * _cons 2.194 0.119 18.42 *** 2.248 0.121 18.60 *** 2.539 0.135 18.75 *** 2.520 0.136 18.48 *** Adjusted R2 0.11 0.11 0.14 0.14 N=2926, *** P<0.001; ** P<0.01; * P <0.05; † P<0.1
29 Table 3: OLS Regression on Generalized Trust by Race Whites Blacks Latinos Coef. Std. t P> Coef. Std. t P>|t Coef. Std. t P>|t Err. t Err. | Err. | Ethnic diversity - 0.278 -2.73 ** - 0.870 - 0.013 1.237 0.01 0.760 0.599 0.69 Ethnic diversity squared 1.147 0.446 2.57 ** 1.368 1.271 1.08 -1.626 - 0.642 0.39 Ethnic enclave 0.032 0.103 0.31 0.024 0.245 0.10 -0.137 - † 0.244 1.78 Mixed neighborhood 0.037 0.055 0.66 -0.141 - -0.128 - 0.114 0.81 0.006 0.05 Percent of recent immigrants 0.002 0.003 0.61 -0.009 - -0.006 - † 0.004 0.42 0.012 1.86 Percent of unemployed - 0.005 -0.39 0.005 0.010 0.51 -0.013 - * 0.002 0.026 2.05 Percent of people below poverty - 0.003 -0.22 -0.006 - 0.008 0.007 1.15 0.001 0.006 0.99 Percent of renter 0.000 0.001 -0.16 -0.002 - 0.001 0.003 0.24 0.001 0.45 Percent of people with BA degree 0.001 0.001 1.15 0.000 0.004 0.09 0.005 0.005 1.10 Percent of people 65 years and - 0.002 -0.90 0.010 0.008 1.23 -0.011 - † older 0.002 0.018 1.71 Average commuting time - 0.002 -1.65 † 0.001 0.005 0.19 -0.007 - 0.004 0.007 1.03 Gender (male=1) - 0.026 -1.51 0.072 0.079 0.91 0.066 0.104 0.63 0.039 Age 0.005 0.001 3.37 *** 0.008 0.004 1.95 † 0.000 0.005 0.04 Homeownership - 0.036 -0.15 -0.089 - 0.175 0.113 1.55 0.005 0.042 0.47 Education 0.025 0.012 2.00 * 0.052 0.036 1.46 0.070 0.036 1.93 † Speaking English at home 0.357 0.163 2.20 * 0.253 0.389 0.65 0.160 0.137 1.17 Nativity - 0.079 -0.77 0.137 0.154 0.89 -0.145 - 0.061 0.031 0.21
30 Household income - 0.002 -0.60 0.010 0.005 1.80 † 0.010 0.008 1.14 0.001 Residential tenure 0.015 0.010 1.56 -0.027 - 0.053 0.045 1.17 0.005 0.19 Upper class 0.154 0.043 3.62 *** - 0.123 - -0.180 - * 0.128 1.03 0.402 2.24 Middle class 0.115 0.034 3.33 *** - 0.087 - 0.043 0.103 0.42 0.103 1.19 Big cities - 0.053 -0.75 0.256 0.224 1.14 0.371 0.409 0.91 0.039 Suburbs 0.012 0.046 0.26 0.215 0.234 0.92 0.063 0.422 0.15 Small towns 0.022 0.042 0.53 0.133 0.225 0.59 0.258 0.412 0.63 Quota 0.099 0.029 3.45 *** - 0.096 - -0.147 - 0.013 0.14 0.149 1.01 _cons 2.467 0.202 12.2 *** 1.428 0.597 2.39 * 2.162 0.587 3.68 *** 3 Adjusted R2 0.03 0.001 0.11 N 2054 353 364 *** P<0.001; ** P<0.01; * P <0.05; † P<0.1
31 Table 4: OLS Regression on Generalized Trust by Class Upper class Middle class Low class Coef. Std. t P> Coef. Std. t P>|t Coef. Std. t P>|t Err. t Err. | Err. | Ethnic diversity - 0.583 - -0.333 -2.97 ** - 0.564 - 0.162 0.28 0.988 0.132 0.23 Ethnic diversity squared 0.473 0.920 0.51 1.602 0.502 3.19 *** 0.112 0.832 0.13 Ethnic enclave - 0.179 - † - 0.085 -2.74 ** 0.031 0.114 0.27 0.314 1.75 0.232 Mixed neighborhood - 0.118 - 0.003 0.057 0.05 0.007 0.092 0.08 0.013 0.11 Percent of recent immigrants 0.001 0.006 0.19 -0.003 -1.16 -0.005 - 0.004 0.001 0.27 Percent of unemployed - 0.007 - * - 0.005 -1.01 -0.006 - 0.019 2.54 0.005 0.004 0.58 Percent of people below poverty 0.005 0.005 0.98 -0.003 -0.63 0.003 0.004 0.72 0.002 Percent of renter 0.000 0.002 - -0.001 -0.71 0.000 0.002 - 0.15 0.001 0.20 Percent of people with BA degree 0.000 0.002 0.20 0.002 0.001 1.20 0.004 0.002 1.68 † Percent of people 65 years and 0.002 0.005 0.39 -0.003 -1.26 -0.005 - older 0.004 0.001 0.10 Average commuting time - 0.005 - -0.002 -1.14 -0.004 - 0.005 1.02 0.003 0.002 0.53 Gender (male=1) - 0.055 - 0.039 0.032 1.24 -0.055 - 0.072 1.32 0.035 0.63 Age 0.008 0.003 2.78 ** 0.001 0.002 0.61 0.007 0.003 2.45 * Homeownership 0.196 0.087 2.26 * 0.007 0.042 0.16 -0.062 - 0.062 1.01 Education 0.026 0.026 1.02 0.037 0.014 2.60 ** 0.058 0.023 2.50 * Speaking English at home - 0.177 - 0.178 0.090 1.99 * 0.364 0.157 2.32 * 0.009 0.05 Nativity 0.100 0.118 0.84 -0.071 -0.02 0.157 0.116 1.35 0.002
32 Household income 0.000 0.003 - 0.001 0.002 0.49 0.004 0.004 0.97 0.07 Residential tenure 0.037 0.021 1.78 † 0.018 0.012 1.56 -0.020 - 0.009 0.44 White - 0.128 - -0.074 -1.01 -0.137 - 0.161 1.26 0.074 0.062 0.45 Black - 0.158 - ** - 0.087 -4.62 *** - 0.151 - 0.469 2.96 0.404 0.190 1.26 Latino - 0.161 - *** - 0.089 -4.57 *** - 0.165 - 0.777 4.84 0.409 0.229 1.39 Big cities 0.040 0.117 0.34 0.109 0.065 1.67 † - 0.107 - 0.114 1.07 Suburbs 0.039 0.108 0.36 0.077 0.060 1.27 -0.111 - † 0.198 1.79 Small towns 0.044 0.101 0.43 0.090 0.057 1.57 -0.092 - 0.127 1.37 Quota - 0.059 - 0.085 0.036 2.38 * 0.077 0.065 1.18 0.005 0.09 _cons 2.568 0.332 7.74 *** 2.774 0.179 15.4 *** 1.965 0.295 6.67 *** 9 Adjusted R2 0.14 0.12 0.13 575 1647 704 *** P<0.001; ** P<0.01; * P <0.05; † P<0.1
33 Table 5: OLS Regression on Generalized Trust by Community Type
Big Cities Suburbs Small Towns Rural Area Coef. SE t Coef. SE t Coef. SE t Coef. SE t Ethnic diversity 0.121 0.612 0.20 -0.439 0.499 -0.88 -1.431 0.425 -3.37 *** 0.727 0.821 0.89 Ethnic diversity squared 0.088 0.825 0.11 0.816 0.738 1.11 2.330 0.691 3.37 *** -1.505 1.549 -0.97 Ethnic enclave -0.022 0.104 -0.21 -0.405 0.140 -2.90 ** -0.169 0.111 -1.52 -0.414 0.335 -1.24 Mixed neighborhood 0.052 0.079 0.66 -0.023 0.088 -0.26 0.007 0.079 0.08 0.094 0.155 0.61 Percent of recent immigrants -0.001 0.004 -0.29 -0.015 0.005 -2.72 ** -0.007 0.005 -1.23 0.043 0.016 2.76 ** Percent of unemployed -0.007 0.005 -1.25 -0.001 0.008 -0.18 -0.015 0.007 -2.12 * 0.007 0.018 0.38 Percent of people below poverty 0.003 0.004 0.63 -0.003 0.004 -0.61 0.004 0.003 1.08 -0.009 0.008 -1.20 Percent of renter -0.002 0.002 -1.25 0.001 0.002 0.71 -0.001 0.002 -0.42 -0.003 0.005 -0.57 Percent of people with BA degree 0.001 0.002 0.69 0.001 0.002 0.92 0.003 0.002 1.77 † -0.002 0.004 -0.37 Percent of people 65 years and older -0.006 0.005 -1.21 0.005 0.004 1.24 -0.004 0.004 -1.20 0.000 0.010 0.04 Average commuting time -0.002 0.004 -0.42 0.001 0.004 0.20 -0.007 0.003 -2.02 * -0.006 0.007 -0.95 Gender (male=1) 0.041 0.053 0.78 -0.017 0.045 -0.38 -0.017 0.039 -0.43 -0.004 0.068 -0.06 Age 0.006 0.003 2.01 * 0.006 0.002 2.49 * 0.002 0.002 1.25 -0.003 0.004 -0.69 Homeownership 0.012 0.063 0.20 -0.067 0.062 -1.09 -0.008 0.053 -0.15 0.134 0.098 1.37 Education 0.047 0.023 2.05 * 0.062 0.020 3.10 ** 0.030 0.018 1.65 † 0.022 0.034 0.65 Speaking English at home 0.317 0.112 2.82 ** -0.017 0.147 -0.12 0.138 0.144 0.96 0.557 0.354 1.57 Nativity 0.088 0.092 0.96 -0.007 0.107 -0.07 0.070 0.102 0.69 -0.150 0.184 -0.82 Household income 0.001 0.003 0.35 -0.001 0.003 -0.31 0.001 0.003 0.45 0.006 0.005 1.21 Residential tenure 0.019 0.020 0.94 -0.015 0.019 -0.80 0.018 0.014 1.26 0.043 0.025 1.75 † White -0.213 0.114 -1.86 † 0.040 0.104 0.38 -0.133 0.106 -1.25 0.006 0.168 0.04 Black -0.409 0.127 -3.23 *** -0.207 0.127 -1.63 -0.477 0.124 -3.84 *** -0.461 0.237 -1.95 † Latino -0.449 0.125 -3.59 *** -0.432 0.142 -3.04 ** -0.445 0.131 -3.41 *** -0.559 0.355 -1.57 Upper class -0.015 0.085 -0.18 0.110 0.080 1.37 0.138 0.062 2.23 * -0.045 0.112 -0.40 Middle class 0.058 0.061 0.94 0.137 0.067 2.04 * 0.096 0.048 2.00 * -0.116 0.083 -1.40 Quota 0.055 0.065 0.86 0.090 0.050 1.79 † 0.020 0.044 0.45 0.137 0.074 1.84 † _cons 2.272 0.274 8.29 *** 2.322 0.271 8.56 *** 2.944 0.232 12.71 *** 2.555 0.514 4.97 *** Adjusted R2 0.13 0.15 0.12 0.06 N 847 711 1059 309
*** P<0.001; ** P<0.01; * P <0.05; † P<0.1
34 The social capital of the rural migrants in Shanghai
Danching Ruan and Gina Lai Hong Kong Baptist University
This research was substantially supported by a grant from the Research Grants Council of Hong Kong Special Administrative Region (HKBU 2142/03H). The support from the Research Committee (FRG/02-03/I-39) of Hong Kong Baptist University is also gratefully acknowledged.
1 I. Background
1. The socialist legacy: Institutionalized rural-urban divide
2. Internal migration in China: from “still water” to 120 Million (from 1985-present)
3.Changing social landscapes in Chinese cities— locals vs. outsiders (mostly rural migrants)
2 II. Method • Data from a survey conducted in the summer of 2005 in Shanghai.
• Three stage cluster sampling: 7 districts from 18 sub- districts; 16 wards and 43 neighborhood committees.
• Data were collected through face-to face interviews from a sample of 1835 local Shanghai residents, aging from 16 to 60, and a sample of 2817 migrants, also aging from 16 to 60 (students were excluded from both samples).
3 III. Social contacts and social capital
4 Locals 住所邻居 nieghbor (%)
11% 3% mostly shanghai 50/50 mostly non local 86%
你平常接触的人大多是上海人还 是外地人 (%) 你的朋友大多是上海人还 是外地人 (%) daily contact friends
7% 1% all or mostly shanghai 4% 2% all or mostly shanghai
50/50 50/50
all or mostly not all or mostly not 92% shanghai 94% shanghai 5 Non-locals 住所邻 居 (%) neighbor
37% 40% mostly shanghai 50/50 mostly not shanghai 23%
你平常接触的人大多是上海人还是外地人 (%)People 你的朋友大多是上海人还是外地人 (%)Your friends in you meet everyday are mostly Shanghai are mostly
all or mostly shanghai all or mostly shanghai 37% 38% 19% 50/50 50/50
all or mostly not 58% 23% all or mostly not 25% shanghai shanghai 6 Non-locals Locals Occupation Occupational Prestige
.008 .1041 是否有科学研究人员 scientist 95 .0344 .1526 是否有大学教师 professor 91 .0317 .1700 是否有法律工作人员 lawyers 86 .1195 .2278 是否有工程技术人员 engineer/technician 86 .0933 .3471 是否有医生 doctor 86 .0407 .1984 是否有政府机关负责人 Government official 80 .1426 .3979 是否有中小学教师 school teacher 77 .1679 .3499 是否有企事业负责人 boss of an enterprise 71 .0702 .1635 是否有经济业务人员 persons in business field 64 (e.g. accountant) .0679 .2817 是否有行政办事人员 clerical worker 53 .1046 .3144 是否有民警 police 52 .0530 .1907 是否有护士 nurse 48 .3219 .4975 是否有司机 driver 25 .2997 .2365 是否有厨师、炊事员 cook 24 .4749 .4469 是否有产业工人 worker 20 .2105 .2289 是否有营销人员 sales people 15 .3083 .1597 是否有饭店餐馆服务员 waiter 11 .1182 .0469 是否有家庭保姆计时工 maid 6 7 N=2209* N=1835 Position Mean/ Stddev Mean/ Stddev Generator Percentage Percentage Variables (Locals) (migrants) Range 46.5 29.89 27.29 29.67 Upper 70.5935 29.02042 44.2933 32.54227 reachability Number of 4.5139 3.41323 2.6673 2.111 occupations Percent having 19.84 39.888 4.07 19.774 ties to officials Percent having 34.99 47.706 16.79 37.391 ties to bosses of enterprises Percent having 70.95 45.410 33.45 ties to people in 47.194 education and culture fields 8 If any from If any from Occupation Shanghai (%) Shanghai (%)
0.4 8.28 是否有科学研究人员 scientist 2.1 11.12 是否有大学教师 professor 2.0 13.19 是否有法律工作人员 lawyers 4.3 17.82 是否有工程技术人员 engineer/technician 5.1 28.12 是否有医生 doctor 2.9 16.78 是否有政府机关负责人 Government official 6.8 33.13 是否有中小学教师 school teacher 8.7 27.68 是否有企事业负责人 boss of an enterprise 2.5 12.70 是否有经济业务人员 persons in business field (e.g. accountant) 3.1 22.62 是否有行政办事人员 clerical worker 7.4 26.87 是否有民警 police 2.6 16.20 是否有护士 nurse 10.6 40.93 是否有司机 driver 0 18.20 是否有厨师、炊事员 cook 14.4 37.33 是否有产业工人 worker 6.7 17.55 是否有营销人员 sales people 4.4 11.01 是否有饭店餐馆服务员 waiter 9 1.0 2.07 是否有家庭保姆计时工 maid • IV. What helps?
10 Summary of Regression Results Having Range N of cadre boss good Occ. Upper. local friends Reachability District (ref=Luwan)
Xuhui (17) *
Yangpu (14) * * * * *
Baoshan (31) * * * * *
Pudong (79) *
Jiading (45) *
Qingpu (27) * * * * *
Men/Age
Education (ref=primary or below)
Junior high school * * * *
Senior high school * * * *
College (non-degree) * * *
University or above
Occupation (ref=worker in commerce)
Agricultural worker
Service worker * *
Clerical worker * *
Self-employed / owner of private * * * 11 business Manufacturing worker * * * Summary of Regression Results Having good Range N of Upper. cadre boss (cont.) local friends Occ. Reachability Time spent in Shanghai * * * *
First time leaving home -* -* -* -* -*
Understand SH dialect a bit -* -*
Not understand at all -* -* -* -*
Speak only a little bit of -* -* -* Shanghai dialect Cannot speak Shanghai Dialect -* -* -* -*
Origin (ref=Anhui)
Jiangsu .
Zhejiang * * * * *
Other provinces
Neighbors (ref= All or mostly non- * * locals) Daily contacts All or mostly * * * locals 50/50 * * *
12 13 社会网络变量 平均值/百分 标准差 *平均值/百 *标准差 比 分比 Locals migrants 网络顶端(最高声 70.5935 29.02042 44.2933 32.54227 望) Upper reachability 网络差异(职业个 4.5139 3.41323 2.6673 2.111 数) N of occupations 与领导层纽带关系 19.84 39.888 4.07 19.774 Tie to leaders 与经理层纽带关系 34.99 47.706 16.79 37.391 Tie to bosses 与知识层纽带关系 70.95 45.410 33.45 Tie to the people in education and culture 47.194 fields
14 Migrants std max min N
社会网络变量 平均值/百分 标准差 最大值 最小值 样本数 比
网络顶端(最高声 44.2933 32.54227 95 0 2209 望) Upper reachability 网络差异(职业个 2.6673 2.111 17 0* 2209 数) N of occupations
与领导层纽带关系 4.07 19.77 1 0 2209 Tie to leaders
与经理层纽带关系 16.79 37.39 1 0 2209 Tie to bosses 与知识层纽带关系 1 0 2209 Tie to the people in 33.45 47.19 education and culture fields
15 Locals std max min N 社会网络变量 平均值/ 标准 最大 最小 样本 百分比 差 值 值 数 网络顶端(最高 70.5935 29.02 95 0 1835 声望) 042 Upper reachability 网络差异(职业 4.5139 3.413 18 0 1835 个数) N of occu. 23 与领导层纽带关 19.84 39.88 1 0 1835 系 Tie to leaders 8 与经理层纽带关 34.99 47.70 1 0 1835 系Tie to bosses 与知识层纽带关 70.95 45.41 1 0 1835 系Tie to the people in education and culture fields 16 社会网络变量 平均值/百分 标准差 *平均值/百 *标准差 比 分比 Locals migrants 网络顶端(最高声 70.5935 29.02042 44.2933 32.54227 望) Upper reachability 网络差异(职业个 4.5139 3.41323 2.6673 2.111 数) N of occupations 与领导层纽带关系 19.84 39.888 4.07 19.774 Tie to leaders 与经理层纽带关系 34.99 47.706 16.79 37.391 Tie to bosses 与知识层纽带关系 70.95 45.410 33.45 Tie to the people in education and culture 47.194 fields
17 Neighbors of the locals
100 80 60 40 20 0 mostly Shanghai 50/50 mostly non-Shanghai
Daily contacts Friends
60 60 50 50 40 40 30 30 20 20 10 10 0 0 all shanghai mostly 50/50 mostly not all not all shanghai mostly 50/50 mostly not 18all not shanghai shanghai shanghai shanghai shanghai shanghai Neighbors of the migrants
50 40 30
20 10 0 mostly Shanghai 50/50 mostly non-Shanghai Friends Daily contacts
60 40 50 30 40 30 20 20 10 10 0 0 all shanghai mostly 50/50 mostly not all not all shanghai mostly 50/50 mostly not all not shanghai shanghai shanghai shanghai shanghai shanghai
19 农 民和城里人应该有平等的就业权利 (%) equal rights for employment
60 50 40 30 20 10 0 strong agree neutral disagree strong don’t know agree disagre
政府给 社 会 底 层 的人一些帮 助是公平的 (%) 农 民和城里人子女应该享有相同的受教育机会 (%) government should help the poor equal educational opportunities for all children
60 80 50 40 60 30 40 20 10 20 0 0 strong agree neutral disagree strong don’t know strong agree neutral disagree strong don’t know agree disagre agree disagre 20 外 来 人口为 上海经济发展做出了巨大的贡献 (%) major contribution to the Shanghai economy
50 40 30 20 10 0 strong agree neutral disagree strong don’t know agree disagre
外 来 人口大量进 城，给 市民生活带来了方便 (%) making life more convenient
40 30 20 10 0 strong agree neutral disagree strong don’t know agree disagre 21 外 来 人口大量进 城，增加了城市就业压力 (%) 外 来 人口大量进 城，城市环 境受到了损 害 (%) increase unemployment damaging living environment
50 50.0 40 40.0 30 30.0 20 20.0 10 10.0 0 0.0 strong agree neutral disagree strong don’t know strong agree neutral disagree strong don’t agree disagre agree disagre know
外 来 人口大量进 城，扰乱 了城市的治安 (%) incre ase crime
50 40 30 20 10 0 strong agree neutral disagree strong don’t know agree disagre 22 Table 3 Regression of Negative Attitudes
Independent Variables Unstandardized Coefficients Standardized Coefficients
Sex (1=men) -.01 -.003 Age .003 .01 Education (reference=University) Junior high -.29 -.07 Senior high -.34 -.08 College (non-degree) -.004 -.001
Neighbors (reference=all or mostly Shanghainese)* -.36 -.06 Daily contacts (reference=all or mostly Shanghainese) -.01 -.002 Friends (reference=all or mostly Shanghainese)*** 1.64 .18 Having non-Shanghainese as good friends*** .72 .11
District (reference=Luwan) Xuhui* -.66 -.11 Yangpu .09 .02 Baoshan .49 .08 Pudong -.27 -.06 Jiading* -.69 -.09 Qingpu -.38 -.06
Constant 5.36*** R-squared .10 Adjusted R-squared .09 N 1764 * p<.05 ** p<.01 23
*** p<.001 Table 4 Regression of Positive Attitudes Independent Variables Unstandardized Coefficients Standardized Coefficients
Sex (1=men) .002 .001 Age* .008 .05 Education (reference=University) Junior high .17 .05 Senior high* .36 .10 College (non-degree) .20 .04
Neighbors (reference=all or mostly Shanghainese) .05 .01 Daily contacts (reference=all or mostly Shanghainese)* -.39 -.06 Friends (reference=all or mostly Shanghainese) -.19 -.03 Having non-Shanghainese as good friends -.03 -.01
District (reference=Luwan) Xuhui -.12 -.03 Yangpu .26 .06 Baoshan** -.54 -.12 Pudong -.32 -.09 Jiading*** -1.01 -.16 Qingpu* -.46 -.09
Constant 4.53*** R-squared .06 Adjusted R-squared .05 N 1787
* p<.05 24 ** p<.01 *** p<.001 Internal and external measures of Philippine slum dwellers’ social capital and their relevance
Petr Matous, PhD Assistant Professor, Department of Civil Engineering, University of Tokyo
This presentation describes development of individual and aggregate social capital measures for the Philippine urban poor and tests the relation of these constructs to their access to basic services.
We have measured individual social capital of all 93 adults in a selected slum community in Metro Manila by a locally adapted Position Generator (18 positions). The set of positions was split by divisive clustering into two clusters (9 positions in each), so that the positions within one cluster tend to be accessed by the same respondents. Examination has shown that one cluster is composed of positions common in the slum area, or “internal contacts ”, and the other cluster is composed of positions that are not present in the area, or “external contacts ”. High access to the cluster of “internal contacts ” indicates good access to resources common in the slums. High access to the cluster of “external positions ” indicates extensive connections with the external society. In aggregate terms, these two measures can be interpreted as bonding and bridging and can be used for a two- dimensional numerical and visual expression of slum communities' aggregate social capital.
The measured individual social capital is strongly related to slum dwellers' access to water and electricity supply. Individuals with the most extensive external networks tend to gain exclusive access to sources of water and electricity outside the community and sell these resources at a profit to other community members. Members with the largest internal networks tend to obtain the resources with the least effort at the lowest price.
Keywords: Social capital Measurement Ego-centered networks Community
The materials for this presentation consist of two parts. Description of the case and the method development, data gathering process, the sample and data analysis can be found in the attached paper “Measuring social capital in a Philippine slum” (under review by the Journal of Field Methods).
Description of the actual findings, i.e. the relation of the developed measures with slum dwellers access to waters and electricity and created social capital by community-based infrastructure development, follows. A. Measuring Social Capital in a Philippine Slum 1
Petr Matous, University of Tokyo Kazumasa Ozawa, University of Tokyo Introduction Almost one third of the world’s urban population, or one billion people, live in slums. The majority of them are in the poor regions of the world. Moreover, a further ‘urbanization of poverty’ is observed, as the mass of world’s poor migrate to the metropolitan areas (UN-Habitat, 2003b). In slums, networks of personal relationships play a crucial role for survival. The provision of infrastructure lags behind the speedy urbanization, and peri-urban slum areas often have no formal utilities (Davis, 2004). Many people can satisfy their basic human needs only via informal personal connections. In the absence of access to official institutions, and services, necessary resources and information for survival are obtained mainly through unofficial channels.
FIGURE 1 Slum Dwellers’ Social Networks
Slum dwellers’ “Regular society” social network
Hi SC
Individual Social tie Low SC
Slums are not homogenous. A large part of inequalities between various social groups or genders can be explained by the difference in quality of their social networks (Figure 1). Individuals without reciprocal ties outside of their households, such as recent migrants or discriminated minorities, are the most vulnerable. Those, without a supportive network within the slum area can hardly cope during difficult times; those without connections outside of the slum area have low chances of getting a job in the formal sector, or participation in the local decision making (UN-Habitat, 2003a).
1 This article is drawn from Matous’s PhD dissertation titled “Relation of Slum Dwellers’ Social Capital and Their Gains from Community-Based Infrastructure Development: The Case of Water Supply in Manila, Philippines” (2007), University of Tokyo, Tokyo. The supervisors of the research were Kazumasa Ozawa, Tsuneaki Yoshida, Hideyuki Horii, Masahide Horita, and Keisuke Hanaki. It further benefited from the comments of Jude Esguerra, Taka Ueda, Ken’Ichiro Ikeda, Mike Handford, Naofumi Suzuki, and Shunsaku Komatsuzaki. I am grateful to James Esguerra, Bing Camacho, Jem Lapitan, Nat Dawn, Han Jarin, Takaki Tsuchiya and the people living in the studied area for their assistance during the survey. This study was supported by the Ministry of Education, Culture, Sports, Science and Technology, Japan.
The amount of resources accessible through one’s personal network can be expressed and measured as individual social capital (Lin 2001). Using this concept, we can explore the existing inequalities within slum communities. However, since social capital has a special meaning for slum dwellers and because of various obstacles to administration, the existing methods need to be adjusted before application in slum areas.
In this article, we describe how we developed an individual social-capital survey instrument for slum areas, and how it was verified through an application in a Philippine slum. After a brief review of the existing theoretical approaches to the measurement of social capital, we describe our field work which included the development of the tool by a series of tests and the actual data gathering process. This paper also covers an analytical development of slum dwellers’ social capital measures, which are based on the gathered data. Consequently, a way of aggregation and visualization of groups’ social capital is proposed. Moreover, a simple way of measurement of social capital creation during community activities is developed and tested. Definition and Measurability of Social Capital The amount of literature on social capital has been rapidly increasing but in many cases only vague definitions of social capital have been used, which allowed only fuzzy measurement. In this section the main conceptual approaches to social capital are briefly reviewed.
The social-capital construct is most often used as a collective-level attribute of communities (see Hayami 2006). The term has become widely popular through the works of Robert Putnam (for example Putnam 1993; Putnam 1995). His ‘social capital’ represents collective attributes such as norms, trust, and networks. However, the mechanism of the concept defined by multiple attributes is unclear. It is difficult to determine the causal relationships among the factors (Cook 2005; Durlauf and Fafchamps 2005). The broadness of this definition allowed to use the term as a new substitute for cohesion, social solidarity, or capacity for collective action or any other ethically valuable community attribute (Briggs 1998). Moreover, it obscures what is going on at the micro-level (ibid). Putnam measures social capital by assessment of civic participation, specifically organization membership. Since not all organizations produce social benefits, Isham and Kahkonen (1999) used coefficients for different types of organizations; however, which organization deserves which coefficient is debatable.
Another popular approach is that of James Coleman. Coleman (1990) placed the outcomes of social capital explicitly within its definition. According to his concept, social capital is anything related to social structure that facilitates individual or collective action. However, such a functional definition is theoretically questionable. Defining the construct by its outcomes does not distinguish causes and effects, and thus make independent measurement of social capital impossible (Foley and Edwards 1997).
Social Capital Definition Used in this Study In this study, Bourdieu’s approach to social capital is utilized (Bourdieu 1986). Bourdieu defines social capital as ‘the aggregate of the actual or potential resources which are linked to possession of durable network of more or less institutionalized relationships of mutual recognition’ (ibid, pp. 248). Bourdieu moreover clearly describes the meaning of social capital at the individual level. ‘The volume of social capital possessed by a given agent thus depends on the size of the network of connections he can effectively mobilize and on the volume of capital (economic, cultural, or symbolic) possessed in his own right by each of those to whom he is connected (ibid, pp. 249). Such individual social capital ‘is what we draw on when we get others, whether acquaintances, friends, or kin, to help us solve problems, seize opportunities, and accomplish other aims that matter to us’ (Briggs 1998, pp. 178).
This definition is not dependent on idealistic assumptions about community - pressure that social ties exert on community members is not always positive (Portes and Landolt 1996; Gargiulo and Benassi 2000). It takes into account that individuals can use their social capital, like any kind of resource, with positive as well as negative consequences for the society (see for example Durlauf and Fafchamps 2005), which the Putnam approach ignores (Schulman and Anderson 2001).
With this concept, we can quantitatively assess disparities within slums among individuals, social capital created during specific events, and aggregate social capital derived from the micro-level relations.
Types of Social Capital Social capital is a multi-dimensional concept. The main two (overlapping) forms that can be distinguished in the literature are (1) social capital to ‘get by’ or social support, and (2) social capital to ‘get ahead’ or social leverage (Briggs 1998).
Expressive Social Capital as the Social Support to ‘Get by’ Expressive social capital is directly contributes to an individual’s wellbeing. For example, it includes keeping a good mood by chatting with close friends. This type of support can be gained only through social networks (Bruggen 2001). Strong relations, which according to the homophily principle are more likely to be created with people who are similar to ego 2, are expected to provide a higher chance for the success of expressive action (Lin 2001). Also in case of slum dwellers, social ties with others of similar socio-economic status provide them the best opportunities to access expressive social capital.
Strong relations within a group, or network closure or a clique, are related to Putnam’s ‘bonding social capital’ (1995). Densely knit networks of strong ties can shield mental
2 Individual from whom social relations are referenced and physical stressors (especially in lower economies) and are therefore good for wellbeing (Haines and Hurlbert 1992). Moreover they are useful during emergency situations. In terms of social order, if everyone in a community knows each other, normative systems can be shared, a more effective sanctioning system is enabled and free-riding disabled. This can lead to trust, and enable actual mobilization of potential support. (called ‘bayanihan ’in the Philippines).
Instrumental Social Capital as the Social Leverage to ‘Get Ahead’ Instrumental social capital increases individuals’ chances for gaining more resources that are instrumentally useful for achieving well-being, for example, getting a job due to information obtained from an acquaintance (Lin 2001). This form of social capital helps one to ‘get ahead’ or change one’s opportunity set (Briggs, 1998). Diverse resources can be mobilized via relations to diverse social groups, or links across so-called structural holes (Burt 1992). Heterophilous 3 relations generally tend to access better instrumental social capital and tend to be weak (Granovetter 1973). In the case of slum dwellers, ties to the external society are likely to have the instrumental functions described above. This type of slum dwellers’ social capital needs to be increased to foster their inclusion in the whole society.
Links across various social groups are related to Putnam’s ‘bridging social capital’ (1995) are a precondition for development in terms of economics, pluralism and tolerance(Cook, Rice and Gerbasi 2002).
Existing Methods to Measure Social Capital and their Applicability in Slum Areas A number of tools for social capital measurement have been developed. This section briefly reviews existing survey instruments that meet the theoretical requirements expressed above and their cross-cultural applicability.
Name, Resource, and Position Generators At the individual level, social capital has been measured by name, resource, and position generators.
Eliciting which individuals are in ego’s network by name generators is probably the most popular network-based social capital measurement method. The personal network itself, however, is not social capital. Name generator does not measure the resources that are embedded in the network (Gaag 2005).
Gaag and Snijders (2005) have developed a direct method to measure resources embedded in social networks of Dutch respondents. Their resource generator is a set of
3 Relations with dissimilar people questions asking whether they know someone who posses some of the listed material and non-material resources. Unfortunately, this collection of resources would not be relevant in other cultures.
The position generator method (Lin and Dumil 1986), on the other hand, is more suitable for application in various cultures. This instrument usually consists of a list of occupations, and the respondents are asked whether they know anyone in each occupation. Since the general division of occupations across all complex societies in the world is similar (Treiman 1977), this method is most suitable for social capital measurement in various cultures. The principle of this method is based on the assumption that division of labour is the main factor behind social inequality, with one’s occupation giving the best estimate of his or her control of resources (Treiman 1977). Therefore, the diversity of accessible occupations can be used as an indicator of the diversity of social resources one can access, which is a good measure of social capital (Lin 1982).
Even this method, however, needs to be adjusted so that the set of specific occupations represent the most valuable resources in a non-Western slum setting.
Position Generator Measures Occupational prestige is strongly linked to socioeconomic status and social class. Based on occupational prestige, several measures of social capital derived from the position generator appear in the literature (Gaag 2005). The most popular measures and their meaning for general population are listed below.
The highest accessed prestige : indicates the access to people who have generally higher control over valued resources (wealth, power, status).
The lowest accessed prestige : indicates access to people who are generally helpful for carrying out manual work or unqualified tasks.
The range of accessed positions defined as the highest accessed position minus the lowest accessed position: indicates the reach of one’s ties in the hierarchical social structure
The number of accessed positions: indicates the diversity of accessed resources and the extensiveness of one’s personal network.
These measures have been developed for the study of a general population. All of them may not be equally suitable for people from the bottom of the socio-economic hierarchy.
Measuring Social Capital Creation To improve their lives, the slum dwellers need to strengthen their social support networks, which can protect them during difficult times. To increase their chances of obtaining regular jobs, they need better contacts with the society outside. However the creation of social capital in low-income communities is not yet well researched.
In, slums, change in levels of individuals’ social capital cannot be feasibly measured by usual panel surveys, because it is practically impossible to trace the respondents without any addresses. A simple theory-based method, which sidesteps this difficulty, is needed.
Slum Characteristics that make Special Measurement Approach necessary
The Area of the Study The studied area is on the outskirts of Metro Manila, the capital of the Philippines. Around the whole city, 2.5 million people squat on vacant private or public lands in 526 slum communities. Half of them have no work or work in the informal sector (UN- Habitat 2003). Slum dwellers are officially represented by their ‘neighbourhood organizations’ or ‘people’s organizations’ at the local authorities. Members of each organization come from the same neighbourhood delimited by distinct geographical features and tend to form coherent social units in terms of religion, years spent in the area, and shared infrastructure. In this paper, a ‘community’ is for practical purposes uniquely defined as a group of people who belong to the same organization.
First we carried out semi-structured interviews in twelve selected communities. The criterion for selection was to maximize diversity in terms of social characteristics within the sample. We interviewed all leaders and a total of thirty selected diverse members from the twelve organizations.
Subsequently, we selected one community of approximately 50 households for an in- depth social capital survey. Until the late nineties it had been occupied only by a few families living in tents. Most current inhabitants came to the area after 1999. In the following section the main characteristics of the studied social group and their implications for measurement are outlined in brief. Generally, these features are typical for slums around the world (as described in UN-Habitat 2003).
Slum Characteristics that Affect the Form of a Social Capital Survey A special approach needed to be taken to ensure that the voices of the poorest of the poor, who tend to be socially excluded, are heard. The following characteristics had to be taken into consideration when designing the survey in order to get reliable information from the maximum number of respondents, which is necessary for any evaluation of internal inequalities in a slum.
Education Literacy level of the respondents was lower than among the general population. Some of the respondents had never been exposed to formal surveys. A question format that is standard elsewhere was incomprehensible for some respondents. The respondents with less exposure to formal education generally concentrated for a shorter time. However, the willingness of this group to participate in the survey was crucial for its success. Interviews had to be sufficiently short and therefore the number of the positions in the position generator had to be relatively low.
Safety The area was relatively dangerous especially for outsiders. We could not approach the respondents directly without permission of the ‘gatekeepers’.
Language The respondents come from various parts of the Philippines and speak various languages. The instrument had to be prepared in the most relevant languages. The optimum wording had to be found in several languages.
Slum Characteristics that Affect the Content of a Social Capital Survey
Income The target population is the lowest income group in a low-income country. The slum dwellers have to use their social capital daily to deal with issues that are normally tackled by official institutions (obtaining drinking water, for example). On the other hand, some types of social capital which are valuable in a high-income setting would not be useful here (e.g. having a friend who can repair a computer, lend a yacht or provide advice about stock markets). Moreover, the social ties of the socio- economically underprivileged group with others may not be reciprocal. We decided to adapt the set of positions in the position generator for the respondents from the lowest socio-economical class and assess whether the elicited relationships are mutual.
Culture Firstly, terms indicating strengths of relationships have different meanings in different cultures. Secondly, the society in the area where the studied community belongs seemed to be strongly divided into communities of different religions. Therefore, we had to find meaningful classification of relationships. We also measured the bridging capital across communities of different religions.
Survey Strategy
Pilot Test We tested the instrument in a socially similar community nearby. It was obvious that self-administration of a pen-and-pencil questionnaire would not produce reliable results and sufficient response rate, especially among the poorest community members, and thus interviews had to be carried out face-to-face. All of the inhabitants could understand either Pilipino or Visayan; we tested the questionnaire in these two languages. The interviewers were four Philippine graduate students who could speak Pilipino and Visayan as well as English. Each time, while one of the interviewers was asking the questions, the others together with the supervisors were observing the procedure including the non-verbal reactions of the respondents.
During this process, for example, it seemed to us that some respondents felt it was appropriate to “agree” with all agree-disagree type of questions. Therefore, this format of questions was avoided in the final version. Furthermore, the poor respondents seemed frustrated when we asked them about their income and we did not find their answers to these questions reliable. Hence, we decided to ask only about what they did for a living, which did not produce such negative feelings and provided a reliable estimate whether or not the person earned at least a minimal wage. In the end of the interviews the interviewers asked the respondent for comments and opinions on each part of the questionnaire. After each interview, the team discussed and hand-altered the questionnaire on the spot and recommended changes in the interviewing protocol.
We printed out a new version of questionnaires every morning. After seven days, the respondents stopped providing negative comments and the questionnaires as well as the interviewing protocol were finalized. Respondents’ feedback after each interview during the actual survey verified the clarity of the final version.
Getting Access and Assuring Maximum Respondent Rate Research suggests that even in the middle class Western conditions, there are significant differences in answers between willing respondents and the others (Stoop 2004). Listening only to the voice of the loudest in a slum community would really miss the point of this survey. Therefore, our ideal target was to interview all men and women over 18 years of age who reside in the community for more than 3 months, plus all heads of local households regardless their age. Recruiting participation from residents of a small geographic area is not easy (see for example Marin and Hampton 2007); getting access to slum dwellers to research their private lives poses even more challenges (see Rashid 2007).
The survey was conducted in November and December 2006. We were introduced to the community leaders by an NGO who was trusted by them and at the same time knew well some members of our team. After explaining the purpose (for which we spent one whole day) we got their permission to conduct the survey. They became supportive of the research possibly for the following reasons. Firstly, they appreciated the interest of international academicians in their community and hoped that the research findings would highlight the positive role of their leadership. Secondly, they were curious about and enjoyed socializing with people from outside the slum, including a Czech and a Japanese. As a result the team members were treated as the leaders’ guests and could move freely and safely around the community which would be otherwise impossible for outsiders. Moreover, they provided the researchers precious information about the community. With their guidance, we could draw a schematic map of the slum with all households.
From the beginning of the survey, we took all meals in the community together with the leaders and their friends. In the evenings after finishing the surveys, the interviewers usually stayed in the area for dinner. Having a meal with someone has a special social importance particularly in the Philippines. This helped the researchers to get closer to the people and gradually gain their trust. When the pilot survey was finished, we bought some food and drinks and, with the help of the community leaders, organized an informal party in the slum. During this event the team could meet and discuss with many people from the community, which was not only very informative and enjoyable but it was also crucial to get accepted in the area. During breaks, we were playing games with children or chatting with someone from the community. The continuing socialization with the people from the area helped to gain their trust and increased their willingness to get interviewed and open themselves to the interviewer.
The interviewed people usually could not understand why we were interested in who individual community members knew. It was important to spend the necessary time to clarify the rationale of the study. After the first few days of the survey, news had spread among the community, in that there was nothing difficult about the survey, that it was not long and not so boring. When about half of the eligible respondents were interviewed, others started approaching the team to remind that they had not been interviewed yet. The not-yet-interviewed minority were interested in getting interviewed (some for the first time in their life). Some who came but were not eligible (for example, those who stay only temporarily) in fact seemed disappointed.
In a slum, it is almost impossible to conduct interviews in privacy. Usually many people share every room. When one person is interviewed the others gather around and interact with the interviewer, with many respondents also tending to ask others for advice. We witnessed that during the test surveys especially during the position generator module, the respondents tended to ask people around whether they know someone in the position. It is usually impossible to interview anyone alone. To prevent others from influencing the interviewee’s answers and also to make sure that other potential respondents are not influenced, it was decided to approach every group of people all together and interview all of them simultaneously. Up to four interviewers were conducting four separate interviews in one group at the same time, which effectively disabled interaction among the respondents. Moreover, some interviewers preferred this approach also because they still would not have felt completely safe alone in someone’s house surrounded by a group of people. The leaders were asked not to attend interviews of others, to avoid further increasing the possible bias caused by approaching the community via them (which was necessary). Naturally, we also assured the informants that their answers are confidential.
Finally, everyone who was present in the area during the survey period was interviewed; no one directly declined. In total 93 people were interviewed in seven days. There were only several individuals who lived in the area but, because of temporary absence, could not be accessed. In the end of this three-week survey (including the pilot phase), the interviewers remembered the names of most community-members and gained a lot of qualitative insights to supplement the quantitative analysis of the data.
The Social Capital Survey Instrument
Position Generator for Philippine Slum Dwellers Since our main goal was to explore inequality in slums, it was not necessary to use the same set of positions as in previous social capital studies of national populations in other countries. Instead, we chose a different set of positions, which should be more meaningful for the Philippine slum dwellers.
Based on the previous cross-sectional semi-structured interview survey, combined with a literature review of social capital survey studies and the Philippine labor statistics, we developed a pilot version of the position generator. We chose the specific occupations from each major occupational group based on our gained understanding of the slum population. During the testing, we further omitted the positions that led only to few positive answers or seemed to have little relevance for the respondents. Some of the respondents looked embarrassed if they were repeatedly asked about high-prestige occupations and they still did not know anyone. The final selection, although it covers the whole range of occupational prestige (Ganzeboom and Treiman 1996), includes only those high-prestige occupations that are most relevant to the life of the squatters. Since contacts with local officials are very valuable for slum dwellers (Appadurai 2001), a local government official, a barangay 4 leader, and also a police officer are in the list. Two non-occupational positions related to religious affiliation were also included to investigate ties bridging the strong local religious divides. We assume that the individuals that have access to various distinct social groups can also access higher variety of social resources.
It is recommended to randomize the order of positions during interviews (Erickson 2004). During the pilot test, however, the order from the lowest prestige occupation to the highest ones worked the best. Among the squatters, the lowest prestige occupations are most likely to produce positive answers, which are preferable for the start up of this process. Starting, for example, with a lawyer may be intimidating for some respondents. The religious positions are the most sensitive ones. Asking about them reinforced some respondents’ suspicion that the whole inquiry is for the purpose of religious recruitment (the only type of interview some of them have ever experienced).Therefore, the questions related to religion were placed in the very end of the interview to prevent any negative influences on the inquiry process.
A high number of positions (thirty or more) has been recommended for position generators (Erickson 2004). The number of positions in the existing position generators varies from twelve to forty. (Gaag ’s Social Capital Measurement Homepage). Since some of the respondents in the slum community lose their concentration and interest relatively quickly, no more than 18 positions could be used. The final set of positions is listed in TABLE 1 in the same order as we asked it.
4 The smallest administration unit of the Philippine government TABLE 1 The List of Positions in the Position Generator
1. Domestic Helper 2. Construction Worker 3. Driver 4. Police Officer 5. Farmer 6. Mechanic 7. Nurse 8. High-School Teacher 9. Priest 10. Accountant 11. Barangay Captain 12. Local Government Official 13. Soldier 14. Technician 15. Lawyer 16. Doctor 17. Person from another Christian denomination 18. Non-Christian
Like the respondents themselves, their alters tend not to have a single formal occupation but typically carry out various informal activities for a living. During the testing, the interviewers had to establish a consensus among themselves which typical activities can be considered as a part of the occupations in the position generator. For example, someone whose main income comes from driving a tricycle (a motorized three-wheeled vehicle) was considered to be a driver but persons who can drive only a pedicab (a human-powered three-wheeled vehicle) was not considered to be a driver.
Name Interpretation After going through all positions, we asked about the type of the generated relationships, characteristics of the alters (BOX 1). (This process is called name interpretation.) In the pilot, we started with four categories of strength of ties: professional, acquaintance, friend, and a family member. However, it was found that the respondents do not distinguish between a ‘professional’ relationship (propesyonal ) and an ‘acquaintance’ (kalikala ). Therefore, we merged these two categories, and the final classification had become (1) professional or acquaintance; (2) friend; (3) kin.
Finally, we asked about the origin of the relationships to estimate which activities catalyze creation of which type of social capital and to check the mutuality of the relationship. For example, one informant claimed to be acquainted with the barangay leader. However, the name interpretation revealed that she had just seen him once when he visited the community. Such alters 5, who are apparently not connected by relationships of ‘mutual recognition’ (Bourdieu 1986) were not included in the social capital computation.
5 Individuals in ego’s personal network BOX 1 Questions in the Final Version of the Position Generator
1. Please tell me whether you happen to know anyone in these positions? (Ask for each from the list of positions. (Record “Yes” or “No”.) If “Yes”: 2. Is this person (1) male or (2) female? 3. Is this person your (1) acquaintance (or professional relation); (2) friend; or (3) family member? For answers (1) and (2): 4. How have you become acquainted/a friend with this person? (Record the verbatim answer.) 5. Does this person live in (1) in [ name of this community ]; (2) other part of [ name of the whole slum area ]; or (3) outside of [ name of the whole slum area ]
Measuring Relative Gains in Social Capital Social capital is dynamic, as are networks of personal relations. Their evolution over time depends on processes of socialization (empirically explored in Bidart and Lavenu 2005, for example). During the previous semi-structured interview survey we identified community activities, which may provide opportunities for such socialization. All selected activities were community-based development projects carried out approximately five years before the survey.
Our goal was to measure which activities lead to social capital creation and which community members had gained most from them.
We asked the respondents whether they had made some new relationships (friends or acquaintances) during each listed activity. If they did, we asked them to recall those who come up first to their mind (maximum three of them) and also what the new alters did for a living. Since the new alters and their occupations could be named five years after the projects, the measured gains in social capital were not only temporary. The recalled alters are more likely to be those who are who had recently been in frequent contact with ego (Freeman, Romney et al. 1987). Despite this assumption, it is still likely that some of those elicited were not in touch with the respondent anymore. However, even such relationships tend to be perceived as potentially supportive and thus contribute to well-being (Cohen and Wills 1985).
Slum dwellers need ties both within the community for support and in everyday life, and outside to find a job in the formal sector, for example. Therefore, we also asked for every generated alter where he or she lived. In addition, we let the informants concretely identify alters who were from the same community (BOX 2).
We used a hand-drawn schematic map of all households in the studied community to (1) help the respondents recall the alters; (2) clearly identify the alters to be able to connect later all the personal networks in a complete directed network of the area; (3) verify that the respondents really knows the alter. It would have been impossible to uniquely identify the alters without using the map because the respondents tend not to use official names for each other but usually various nicknames.
BOX 2 Questions about Created Social Capital
1. Have you participated in this activity/project? (Ask for each from the list of identified community based activities. Record “Yes” or “No”.) If “Yes”: 2. Have you made some new acquaintances (including professional relationship) or friends during this activity/project? (Record “Yes” or “No”.) If “Yes”: 3. Can you recall specific persons you have become an acquaintance or a friend with? (Record a nickname, first name, or initials.) For each identified person ask: 1. Do you consider this person to be your (1) acquaintance (including professional relationship) or (2) friend? 2. Is this person (1) male or (2) female? 3. What does this person mainly do for a living? (Record the verbatim answer) 4. Does this person live in (1) in [ name of this community ]; (2) other part of [name of the whole slum area ]; or (3) outside of [ name of the whole slum area ] Identify on the map houses of the alters from this community.
Analysis After the survey was finished, it took the team two days to encode the answers and to decide meaningful categories for classification of the verbatim answers to the open- ended questions.
Suitability of the Selected Positions The results enabled us to distinguish respondents’ social capital in fine grades. The least popular position (construction worker) was accessed only by 22 respondents and the most popular one (accountant) by 84 people out of 93. The rest of the positions are distributed between these two at similar intervals.
FIGURE 2 shows total results of the whole community. Most respondents knew around half of the positions in the position generator but there are also three respondents who knew all eighteen positions and one respondent who knew only one position.
FIGURE 2 Distribution of Access to the Positions in the Position Generator
s t 12 n e d 10 n o p 8 s e R 6 f o r 4 e b 2 m u N 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Number of Known Positions
Meaning of the Usual Position Generator Measures for the Slum Dwellers We did not encounter any slum dweller s who did not know anyone in a low-prestige occupation. ‘The lowest accessed prestige’ is clearly not a relevant measure in a slum setting. Consequently, ‘the range of accessed positions’ means the same as ‘the highest accessed prestige’ because everyone has access to the lowest prestige occupations. ‘The highest accessed prestige’ and ‘the number of accessed positions’ are correlated (TABLE 2). However, the latter is a better measure of general social capital because it reflects the diversity of accessed resources and, being counted from multiple questions, it is more reliable than just the solitary highest accessed position. Moreover, using a measure that is not solely dependent on the occupational prestige enables us to include important non-occupational positions (e.g. religious affiliation-based position) which indicate access to other social groups and other types of resources.
TABLE 2 Correlations among Deductive Measures
Highest Lowest Range Sum Average Number 0.76 -0.39 0.81 0.98 0.60 Highest -0.12 0.95 0.78 0.86 Lowest -0.41 -0.32 0.22 Range 0.81 0.72 Sum 0.67
If the ‘lowest accessed prestige’ measure is excluded, the ‘number of positions’ and ‘average accessed prestige’ are the most independent couple of measures. The ‘number of positions’ may be supplemented with the ‘average accessed prestige’ to provide a simple idea about the prevailing type of accessed resources.
A measure of slum dwellers’ social capital based on a universal one-dimensional occupational prestige, however, has its drawbacks. Even if the assumption that the whole variety of occupational structure can be well represented by a unidimensional scale itself is true, the fact that it is calculated from averages over general populations makes its application in slum communities problematic. A universal prestige scale may not correspond well to the control of resources which are valuable for the slum dwellers.
From the viewpoint of our informants, the occupation as a priest seemed to be the most respectable but according to the prestige scale it is only in the middle of our list. Most notably nurses were thought of very highly, probably because this occupation gives valued opportunity to work abroad for a relatively high income. Having a nurse among relatives is considered very lucky. Their prestige is apparently much higher than, for example, that of soldiers. That is opposite to their relative ranking on the universal prestige scale. In the next section we present the analytically developed position generator measures which are based only on the ‘number of accessed positions’ independently on the prestige scale. We believe these reflect more accurately the reality that we observed in the slum community.
Construction of Inductive Measures that Reflect the Slum Dwellers’ Social Capital Dimensions We consider the positions in the position generator that tend to be accessed by the same people as similar to each other. Based on this consideration, we computed the dissimilarity of each pair of positions in the position generator. For each position a vector of all respondents’ answers concerning whether or not they know anyone in the position was composed. The dissimilarity of each pair of positions is computed as a Binary Euclidean distance between each pair of the position vectors divided by the number of positions. This equals the percentage of the respondents who know someone in one of the positions but do not know anyone in the other. Then, we split the list of the positions into two groups by divisive clustering, so that the positions in one cluster tend to be closer to each other than to positions in the other cluster. Specifically, the pair of the two most dissimilar occupations was found and the other occupations were assigned to the same cluster with the more similar of these two occupations (TABLE 3).
TABLE 3 Position Clusters
A – local positions B – non-local positions Domestic Helper Policeman Farmer Mechanic Driver Nurse High-School Teacher Priest Barangay Captain Accountant Local Government Official Technician Soldier Lawyer Construction Worker Doctor Person from Another Christian Denomination Non-Christian
Our examination of the positions in each cluster s shows that the first cluster (A) is composed of low-income occupations which are common in the studied community and in slums more generally (Domestic Helper, Farmer, Driver, Soldier, and Construction Worker). The remaining occupations in Cluster A are locally-based occupations and their representatives work with the local people as a part of their job (High-School Teacher, Barangay Captain, and Local Government Official). Finally, the last position which was assigned by the analysis of the similarity among respondents’ answers in Cluster A is a person from some Christian denomination that is different than the respondent’s denomination. (All respondents were Christians.) All positions in cluster A can be accessed locally.
The second cluster consists of higher-income and higher-prestige occupations and social positions which are not present in this community and are rare in slum areas generally.
Since each occupational group has some given population, individuals with access to more occupations are more likely to have larger social networks. Consequentially, individuals with relatively high access to occupations in the local and non-local cluster are likely to have large networks within and outside this slum respectively. Moreover, since access to Cluster B occupations indicates heterophilous ties to positions from different parts of social structure outside of the slum, it can be used as an estimate of the respondents’ instrumental social capital, or the so-called ‘social leverage’. Access to Cluster A occupations, on the other hand, which indicates connectedness within the slum area might be used for estimation of expressive social capital, or the social support.
Note that the division of the positions into two clusters corresponds with the two main social capital dimensions distinguished in the literature, although the division was based purely on respondents’ answers.
Expressing Aggregate Social Capital of a Slum Community The aggregation of the community-members’ social capital in FIGURE 2 does not distinguish different types of social capital. It does not take into account that a slum dweller who knows only doctors and lawyers has a different type of social capital than another one who knows only drivers and construction workers. Their social capital is qualitatively different even if their total number of alters is the same. Moreover, a community needs ties both inside and outside (Woolcock 1998) and an aggregate social capital measure should reflect it.
A better aggregate measure and visualization can be produced by using the inductive measures developed above. In FIGURE 3, the horizontal axis represents an individual’s access to positions in Cluster A, which indicates the size of the respondent’s local personal network. From the viewpoint of the whole community, this is a part of its bonding social capital. The vertical axis represents an individual’s access to positions in Cluster B, which indicates the size of the respondent’s non-local personal network. At the community-level, it corresponds to the community’s bridging social capital to the external society.
FIGURE 3 Respondents’ Internal and External Network 6 - an Aggregate Picture of the Community
9
8
7
Bridging 6 ～ 5 B 4
3
2
1 External Contacts Contacts External 0 0 1 2 3 4 5 6 7 8 9 ～ Internal ContactsA Bonding
Concluding Remarks The conditions in slums require a special approach to social capital surveys. Therefore, we had to adapt existing methods for application in a slum. For the measurement of social capital, we used a position generator consisting of 16 occupational positions and 2 religious positions, which were selected according to their relevance for the slum dwellers. The order from the lowest prestige positions to the highest prestige positions, with religious positions in the end was best accepted by the respondents. Asking about the following three types of relationships seemed to be most understandable for the interviewed Filipinos: an acquaintance or a professional; a friend; and a family member. We also decided to ask about the origin of each elicited relationship to verify its mutuality and to learn about the ways of social capital creation. Furthermore, we developed a new way for estimating the created individual social capital during specific activities by name generating the new friends and acquaintances, and their occupations. The feasibility of administration of the measurement tool in slums was tested by an actual survey and the comprehensibility of the whole combined instrument has been verified by respondents’ feedback.
It is useful to distinguish different types of slum dwellers’ social capital which are usable for different purposes. Instead of using a unidimensional prestige scale which does not seem to fit the reality of a slum, we developed two inductive measures of slums dwellers’ social capital based on the gathered data. These can be interpreted as measures of access to resources within and outside the slum area. They can be used as two dimensions to express aggregate social capital of a slum community. The results (which are not presented here) confirm that the constructs measured by the developed instrument are strongly related to the slum dwellers’ chances to fulfil their basic needs.
6 The coordinates of the centre of each circle are determined by respondents’ access to Clusters A and B. The circle size represents the number of respondents with the particular combination of A and B alters. The smallest circle represents one respondent; the largest circle represents five respondents. The presented method meets the criteria of a suitable social capital measurement tool (as stated by Erickson 2004). It has a clear theoretical foundation, known relationships to important causes and consequences, and desirable measurement characteristics such as easy administration and reliability. It can be used also in future studies in other Philippine slum communities and, with some adjustments, also in slums of other countries.
Finally, short social surveys are inherently less reliable than full-blown long-term anthropological observations, but are often the only feasible option. We have learnt the following lessons to increase the trustworthiness of the gathered information.
1. Extensive testing of the instrument by all interviewers together (one is interviewing, the others observe) serves also as a good training and helps to establish internal consensus on interpretation of the terms and categories used in the survey which leads to consistent data coding.
2. Simplicity of the measurement tool is necessary for smooth administration in slum community. It has to be brief and enjoyable to assure a high response rate.
3. Spending some extra time within the community is worthwhile. For example, sharing small meals and drinks with participants increases informants’ willingness to participate in the survey and open themselves to the research. Getting support of the community leaders was also crucial for the survey administration. As a result, not one of the available community members declined to be interviewed.
4. Interviewing individual household members simultaneously can prevent their unwanted interaction.
5. Using a map for alters’ identification in combination with a name generator increases reliability of the produced data.
References Appadurai, A. (2001). "Deep democracy: urban governmentality and the horizon of politics." Environment and Urbanization 13 (23). Bidart, C. and D. Lavenu (2005). "Evolutions of Personal Networks and Life Events." Bourdieu, P., Ed. (1986). Forms of capital Handbook of Theory and Research for the Sociology of Education. New York, Greenwood Press. Briggs, X. d. S. (1998). "Brown kids in white suburbs: Housing mobility and the many faces of social capital." Housing Policy Debate 9(1): 177-221. Bruggen, A. C. v. (2001). Individual production of social well-being: An exploratory study. Psychologische, Pedagogische en Sociologische Wetenschappen . Groningen, Rijksuniversiteit Groningen. Doctoral Dissertation . Burt, R. S. (1992). Structural Holes . Cambridge, MA: Harvard University Press. Cohen, S. and T. A. Wills (1985). "Stress, social support, and the buffering hypothesis." Psychological Bulletin 98 : 310-357. Coleman, J. S. (1990). Foundations of social theory. Cambridge, Harvard University Press. Cook, S. K. (2005). "Networks, Norms, and Trust: The Social Psychology of Social Capital." Social Psychology Quaterly 68 (1): 4-14. Durlauf, S. N. and M. Fafchamps (2005). Social capital. Handbook of Economic Growth . P. Aghion and S. Durlauf, Elsevier : 1639-1699. Erickson, B. (2004). A report on measuring the social capital in weak ties. Toronto, The Policy Research Initiative. Foley, M. and B. Edwards (1997). "Social capital and the political economy of our discontent." American Behavioral Scientist 40 (5): 669-678. Freeman, L. C., A. K. Romney, et al. (1987). "Cognitive structure and informant accuracy." American Anthropologist 89 (2): 310-325. Gaag, M. P. J. v. d. "Measurement of individual social capital webpages " Retrieved 6 August, 2007, from http://www.xs4all.nl/~gaag/work/ . Gaag, M. P. J. v. d. (2005). Measurement of individual social capital. Doctoral Dissertation . Groningen, Rijksuniversiteit Groningen. Doctoral Disseration . Gaag, M. V. D. and T. A. B. Snijders (2005). "The resource generator: Social capital quantification with concrete items " Social Networks 27 (1): 1-29. Ganzeboom, H. B. G. and D. J. Treiman (1996). "Internationally comparable measures of occupational status for the 1988 International Standard Classification of Occupations." Social Science Research 25 (3): 201-239. Gargiulo, M. and M. Benassi (2000). "Trapped in Your Own Net? Network Cohesion, Structural Holes, and the Adaptation of Social Capital." Organization Science 11 (2): 183-196. Granovetter, M. (1973). "The strength of weak ties." American Journal of Sociology 78 (6): 1360-1380. Haines, V. A. and J. S. Hurlbert (1992). "Network range and health." Journal of Health and Social Behavior 33 (3): 254-266. Hayami, Y. (2006). Social capital, human capital and the community mechanism: Toward a conceptual framework for economists. Discussion Paper Series on International Development Strategies . Tokyo, FASID. Isham, J. and S. Kahkonen (1999). What determines the effectiveness of community- based water projects? Evidence from central Java, Indonesia on demand responsiveness, service rules, and social capital. Social Capital Initiative , World Bank. Lin, N. (1982). Social resources and instrumental action. Social structure and network analysis . P. V. Marsden and N. Lin. Beverly Hills, Sage : 131-145. Lin, N. (2001). Social capital: A theory of social structure and action . Cambridge, Cambridge University Press. Lin, N. and M. Dumil (1986). "Access to occupations through social ties." Social Networks 8: 365-385. Marin, A. and K. N. Hampton (2007). "Simplifying the personal network Name Generator: Alternatives to traditional multiple and single Name Generators." Field Methods 19 (2): 163-193. Portes, A. and P. Landolt (1996). "The downside of social capital." The American Prospect 94 (25): 18-21. Putnam, R. D. (1993). Making democracy work: Civic traditions in modern Italy . Princeton, University Press. Putnam, R. D. (1995). "Bowling alone: American’s declining social capital." Journal of Democracy 6(1): 65-78. Rashid, S. F. (2007). "Accessing married adolescent women: The realities of ethnographic research in an urban slum environment in Dhaka, Bangladesh." Field Methods 19 : 369 - 383. Stoop, I. A. L. (2004). "Surveying nonrespondents." Field Methods 16 : 24-54. Treiman, D. J. (1977). Occupational prestige in comparative perspective . New York, Academic Press. UN-Habitat (2003a). Slums of the world: The face of urban poverty in the new millennium? Nairobi, United Nations Human Settlements Programme. UN-Habitat (2003b). The Challenge of slums: Global report on human settlements , United Nations Human Settlements Programme. Woolcock, M. (1998). "Social capital and economic development: Toward a theoretical synthesis and policy framework." Theory and Society 27 (2): 151-208.
B. Relations Between gains from Community-Based projects and Social Capital
In this part, by using the developed methods, the distribution of individual social capital in the selected community is shown with regards to individuals’ occupation and gender. Then, the relationships between the cost of water and electricity individuals pay and their social capital is examined. Finally, creation of social capital in community-based water and electricity supply, footbridge and road construction projects is explored. The background conditions in the area, the sample, the method development, data gathering process and data analysis is described in the attached paper “Measuring social capital in a Philippine slum” (under review in the Journal of Field Methods).
The data were gathered using (1) semi-structured interviews with leaders of 12 community-based water systems, 30 consumers supplied served by these systems, the representatives of the central water utility; and (2) fixed-form interviews in one of the communities. In this community, 93 adults were interviewed in total. This sample consists of all 66 available adults belonging to the community organization that took the responsibility for water supply and other infrastructure development and public service delivery in the area and a random sample of 27 individuals who live in the neighborhood but belong to other organizations. The findings about the social capital distribution within the community are based on data from all 93 individuals. The findings related to infrastructure development are derived from analysis of the all the members of the community organization.
Individual Social Capital Distribution
An adapted position generator was used to measure individual social capital. After the survey, the list of positions was split by divisive clustering into two groups. Positions in first group are common in the community. The number of connections to this group reflects the size of the ego’s local or internal slum social network.
Positions in the second group, on the other hand, are higher prestige positions with more control over resources. These positions are rare in the area. The number of connections to this group reflects the size of ego’s external social network. Assuming that the relations with these individuals are mainly heterophilous, the more positions from this group the respondent knows someone in, the better instrumental social capital is he or she likely to have. 9
8
7
6
5
4
3
2
1
0 0 1 2 3 4 5 6 7 8 9 Internal contacts
Figure 4 – Number of Alters in the Position Generator Accessed by Ego 7
Next, the respondents were split into two categories: those with occupation that is estimated to earn at least minimum income and those without. It was found that while the average access to the internal positions is the same for respondents from both groups, those with relatively stable income have on average more external contacts. All of the respondents from this group knew someone in at least to of the external positions.
9 8
7 6
5 Low er income 4 Higher income
3 External contacts External 2
1 0 0 5 Internal contacts
Figure 5 – Social Capital of Those with Relatively Stable Income and Those Without
7 The area of each disk is proportional to the number of respondents with the particular combination of internal and external contacts. The smallest disk represents one respondent. Respondents’ social capital is not related to their gender. Although interviewed separately, married couples tend to name the same alters in the Position Generator. Friends of one of them are also known by the other one. That may be because of the slum dwellers’ economical situation, and properties of their housing. Friends meet each other around their homes and the whole family is involved. It should also be noted that the Philippines belongs to the countries with the highest gender equality in the world. It ranked the 6 th in The Global Gender Gap Report 2007 (Hausmann, Tyson et al. 2007)
100% 90% 80% 70% 60% Male 50% Female 40% 30% 20% 10% 0%
r l H W e ce n ia D iv ian iest ic C r ldier oli ic r ff D o hanic Nurse n P o S P Farmer c h aptai eacher t Lawyer Doctor C t v ec o on person Me T Hs g i ion person Accountantl ig Brgy. rel t- oca n L e ame-religr S iffe D
Figure 6 – Popularity of Occupations 8 among Men and Women
450.0 400.0 350.0 300.0 250.0 Male 200.0 Female 150.0 100.0 50.0 0.0 # Alters TOP Bottom Range Sum Average
Figure 7 – Position Generator Measures of Social Capital of Men and Women 9
8 Percentage of respondents knowing someone in the given occupation 9 The meaning of the measures is explained in the attached methodology paper 9 8
7 6
5 Female 4 Male
3 External contacts External 2
1 0 0 1 2 3 4 5 6 7 8 9 Internal contacts
Figure 8 – Social Capital of Men and Women
Organization of Community-Based Water Supplies
The community leader buys water from the central water supply utility and distributes it through a local pipe network to the community members, who pay to her an initial connection fee and the regular fee based on their consumption.
Analyzing the semi-structures interviews of the leaders of the 12 community-based systems and the representatives of the central utility shows that only individuals with the highest social capital become leaders of bulk water systems (Table 4).
Only people, who know from other communities about the possibility to form a partnership with the central utility, approach the utility with the offer to start a community-based system in their area. Alternatively, when the utility looks in some area for capable people to run local water systems, only the people who learn about the offer through informal channels apply. In either case, having access to information through personal networks is necessary.
The utility decides who to cooperate with based on references. Only people with good credentials in the area are considered.
Connections to local government officials are almost inevitable condition to obtain permissions for water systems implementation. Finally, having access to financial credit is necessary to implement the systems. The leaders have to be able to mobilize the necessary starting capital from pre-payments from potential customers, and loans from friends or possibly their cooperatives if they belong to any.
Phase Conditions Social Capital? Know the systems elsewhere Approach Offered via their Access to info contacts Selection Good references Credentials Connection to Obtain permissions officials Implementation Capital for initial investment Access to credit
Table 4 – Necessary Characteristics of the Community-based Water Supply Leaders
In the studied communities, not all community members have a water connection; some buy water from their neighbours in buckets. Analyzing the semi-structured interviews, shows reasons why some respondents did not obtain a pipe connection (Table 4Table 5).
Claimed Reason Failed to SCSocial Domain Capital? 1."Couldn't afford Borrow from friends Access to the connection Share the connection credit fee" Get an exception to pay later 2. "Didn't know Access to Get informed (how)" info
Table 5 – Reasons for not Having a Connection
The respondents without a water connection mostly claimed that they could not afford the connection fee. Slum dwellers generally do not have enough cash for high lump sum payments. Official loans are usually not available. Those who could not afford the connection fee are those who (1) failed to borrow money from friends, (2) failed to share the connection and its cost with other households, and (3) failed to get exception from the leader to pay later. All these reasons are related to their lack of credit.
The second most common reason for not having a connection was a lack of information. Some consumers would have liked to buy a connection but they were not told about the possibility, or did not know who to approach about the matter. Apart from that, some consumers opted not to get a connection because they were misinformed about the real costs and benefits. All these reasons are related to lack of accurate information available through their personal networks.
Moreover, analyzing correlations with supplementary questions showed that the individuals that sell or resell water to others (7 respondents) are significantly more socially active in the area than others. They were more active in helping new inhabitants to move into this slum area, all of them attend birthdays of other people in the area, and have relatively more godparents in the area.
The individuals that share their household water connection with others (12 respondents) also tend to share electricity. When necessary, all of them borrow money for water and electricity expenditures non-commercially from friends and kins.
Measured Relation of Water Cost and Individual Social Capital
The community-based water supply leader buys the water relatively cheaply (22 Php/m 3) and sells it to the community (Figure 9). Those who could gather the necessary funds to get a connection from the organization in the beginning pay the second lowest rate (35Php/m 3). Those who could not mobilize all the necessary funds, or did not have the information in the beginning, obtained their connection later for a partial payment but their rate is higher (40Php/m 3). Those who could neither borrow enough money nor have anyone with whom they could share the connection and its cost, remained disconnected. Some people were not even invited to join. These people without any connection pay the highest price for the water they buy in buckets from others . When these small volumes are converted in cubic meters, the approximate unit price they pay is the highest (around 100Php/m 3).
Community Leader
A shared Originally Connected connection Consumers
Later Connected Consumers
Unconnected Consumers
Figure 9: Water Distribution within the Community
Relation of respondent’s unit price of water with his or her social capital (measured by the number of positions accessed in the Position Generator) is shown in Figure 7. Clearly, those with the highest social capital are in the best position in the bulk water supply structure.
120 16 Water Price 14 100 SC 12 80 10 60 8 6 40 4 Water Price [P/m3] Water Price 20 2 SC [#PositionsSC Accessed] 0 0 The Leader Originally Later No connection connected Connected
Figure 10: Individual Social Capital and Cost of Water
The next diagram shows the average number of internal and external contacts in the Position Generator for the groups of respondents who pay 22, 35, 40 and 100 Php/m 3 of water respectively. The area of each disk represents the number of valid respondents in each group. Only the leader pays 22 Php/m 3.
9
8
7 6 22 5 35 4 40
3 100 External contacts External 2
1 0 0 1 2 3 4 5 6 7 8 9 Internal contacts
Figure 11 – Relation of Internal and External Social Capital to the Cost of Water
Although those who do not buy a water connection (on average, lower social capital individuals) have no initial expenditures, they lose most in the long term (Figure 12). ] 100 The leader Water connection Others owners (HI-SC) (MID-SC) (LOW-SC) 0 1.7 -23.5
-100
-200
-235
-300 NPV [2 yrs, 10% discount 10% [2 yrs, unit: rate, NPV thousands Php
Figure 12 - Net Present Value for Different Consumer Groups at the Start of Bulk Water Supply Operation
Measured Relation of Electricity Cost and Individual Social Capital
Some similarities can be observed in the electricity supply. Several individuals from the community have obtained a direct connection from the electricity provider. The others buy electricity either from them or from people from neighboring areas.
Electricity provider
Directly connected
Connected to other consumers
Figure 13 – Electricity Supply Structure
To obtain official electricity connection directly from the provider is more complicated than just connecting a wire to a neighbor. Applicants for a direct connection have to pay a deposit, which is determined by the provider on a case by case basis. The company may refuse to provide an electricity connection to “unsuitable” applicants.
The respondents who have a direct electricity connection have on average higher general social capital than others, who buy electricity from them with a surcharge (4- 7P/kWh) (Figure 14).
16 14
12 10
8 Electricity Price 6 SC 4
2 [P/kWh, # accessed positions] 0 Direct To others Figure 14 – Type of Electricity Connection, Electricity Price, and Social Capital
Four individuals from the community who have an official direct connection from the electricity provider have higher both internal and external social capital than the rest of the measured sample (Figure 15).
9
8
7
4 6
5 Others Direct 4
89 3
2
1
0 0 1 2 3 4 5 6 7 8 9 Internal contacts
Figure 15 - Relation of Internal and External Social Capital and Access to Electricity Supply
Potential of Community-Based Projects for Social Capital Creation
Water supply and electricity supply projects, although carried out entirely by people from the community, did not induce socialization and social capital creation, because in each case only several individuals could do all the necessary work to implement and operate the projects and keep the profits in return.
Extensive socialization was induced only in community-based projects that met two conditions. First, they were labour-intensive, which required widespread participation to achieve the projects’ stated ends. Second, there was not sufficient revenue potential from operating the projects and no commercial motivation for any single member to do all the work to keep all the profits (Table 6). Such projects provided opportunity for people to get together and create new relationships. The construction of footbridges and a road met these conditions (Table 7)
Revenue potential Labour-intensive SC created Water Yes No No Electricity Yes No No Footbridges No Yes Yes Road No Yes Yes
Table 6 – Relation of Participation in the Community-Based Projects to their Revenue Potential and Demand on Labor
Recalling footbridge and road construction combined: Categorized Verbatim Open-Ended Answers Before Answers related to the times during the construction After necessary (3) it was difficult/exhausting (20) new relationships (7) thinking of returning to province everyone cooperated (13) improvement (2) scared of the water giving contributions (6) well done (2) hopes to get a title for the seemed impossible working at night (5) land technical information (3) cooking with others (2) kindness (2) absent at work these were happy times it was ok nostalgic memories it was well organized worked on the part in front of own house only some husbands not around, women worked hard started the construction
Table 7 - What comes up to your mind when you recall the footbridge/road construction? 10
Creation of Social Capital during Footbridge and Road Construction
The method of measurement of social capital creation in the community projects is described in the attached methodology paper.
Out of the 66 interviewed community members, 41 (62%) said they created new relationships during the construction of the footbridges and a road. The respondents who claimed to have made new relationships were asked to name up to three of them. 22% of new relationships (average 0.53 per participant) are with a people who earn at least the minimum wage. These new relationships are likely increase the participant’s instrumental social capital.
It is illuminating to split further the new relationships according to the occupation of the respondent. It can be seen that those who are better of in terms of occupation (those who were estimated to earn at least the minimum wage) tend to form new relationships with other such participants. The average number of their new contacts is close to the maximum number allowed during the interview (2.6 out of 3). Almost half (46%) of these new relations is with other people with stable income. Apparently, the individuals in this group gained better social capital. The tendency maybe explained by the homophily principle.
3.0
2.5
2.0
Higher Income Alter 1.5 Low er Income Alter
1.0
Average number ofalters new number Average 0.5
0.0 Higher Income Ego Low er Income Ego
Figure 16: New Ties Created During the Projects
10 Numbers in brackets indicates how many people provided the particular answer. Women gained better social capital probably because they are more actively involved in organization of the projects. Men, on the other hand, spend more of their time at their jobs.
Higher Lower Lower Income Income Income % Male ’s new relations total 31 5 86% New relations per one male 1.94 0.31 Female ’s new relations total 29 12 71% New relations per one female 1.71 0.71 Table 8 – Social Capital Gained by Males and Females
Conclusions
The most of the respondents’ social capital appears to be embedded within their internal slum dwellers’ networks. The respondents without any contacts outside the slum area tend to have lower-prestige occupations. Men and women are equal in terms of social capital.
The slum dwellers depend on their social capital to obtain information and mobilize cash for lump-sum payments necessary to get access to affordable water and electricity. The individuals with the highest measured social capital earn profit from the operation of the community-based water and electricity supply. The individuals with lower measured social capital tend to pay higher prices for these resources and are more like to be excluded from the water supply network.
The leaders of the community-based water supply have the highest individual social capital (access to 15 positions from the Position Generator, in the surveyed community). Thanks to their access to information, connections to people in power, ability to mobilize funds, and influence in the area, they were able to initiate these water systems. Subsequently, the same qualities helped them to successfully operate them. These people get the water directly from the central water utility for the minimum payment and sell the water with a profit via direct connections to core members of the community. Those who were connected earlier and were given a slightly lower tariff have on average higher individual social capital (12 positions) then those who connected later and have to pay (average of 10 positions). Those who could neither borrow enough money nor have anyone with whom they could share the connection and its cost, remained disconnected. Some people were not even invited to join. These people have on average the lowest measured social capital (around 8 positions). They are left without any connection and pay the highest price for the water they buy in buckets from others . The individuals with direct connections (who tend to have higher social capital) can recover part of their expenditures for water from resale to those with no connection.
Analogically, the highest-social capital individuals use their superior access to information, informal loans, and local authorities to obtain official direct electricity connections from the electricity provider and sell it at a profit to the rest who lack these social resources.
Comparing various community-based projects in the area, widespread socialization occurred only during construction of footbridges and road. New lasting relationships were formed. Everyone’s cooperation was required to carry out these labour-demanding projects as opposed to the water and electricity supply which can be operated by a single individual motivated by the profits it enables.
Individuals with occupations that pay at least minimum wage, increased their instrumental advantage by creating new relations with other such people. The rest, who can be considered poor, created relations mainly with other poor people. Females who tended to be more involved in the organization of the community-based projects gained slightly better social capital. At the community level, the projects increased bonding but did not significantly increase bridging to other communities. For example, no one had developed a new relationship with a person from another religion. Thus, participation in community-based project development does not necessarily contribute to social capital creation that would lead to development in terms of economics, pluralism and tolerance.
The findings are in accordance with the Bourdieu’s claim that social capital has ‘a multiplier effect’ on the capital individuals already posses (Bourdieu 1986). Without a regulatory framework for the infrastructure development the individuals obtain very unequal returns for equivalent investment of economic capital depending on their social capital. The findings support the claim that poor communities cannot well develop by themselves unless measures are taken to deal with individual structural inequalities (Berner and Philips 2005).
Based on this research practical policy implications for infrastructure development in slums have been made. An ongoing research is testing the found relationships in other countries and other types of infrastructure in the region.
References:
Berner, E. and B. Philips (2005). "Left to their Own Devices? Community Self-Help between Alternative Development and Neoliberalism." Community Development Journal 40 (1): 17-29. Bourdieu, P., Ed. (1986). Forms Of Capital Handbook of Theory and Research for the Sociology of Education. New York, Greenwood Press. Cook, K. S., E. R. W. Rice, et al. (2002). Commitment and Exchange: The Emregence of Trust Networks under Uncertainity. Hausmann, R., L. D. Tyson, et al. (2007). The Global Gender Gap Report 2007. Geneva, World Economic Forum. Pacione, M. (2001). Urban Geography: A Global Perspective . London, Routledge.
Diverse Ties, Diverse Effects: Looking to Networks to Help Explain Tolerance towards Ethnic Minorities
Rochelle R. Côté, Bonnie H. Erickson and Robert Andersen Department of Sociology, University of Toronto August 2007
ABSTRACT
Research in the "contact hypothesis" tradition shows that some types of contacts can lead to more positive attitudes towards minorities. This paper compares the effects of friends, networks, and voluntary associations on positive attitudes towards immigrants and ethnic minorities in Canada, using the 2004 Canadian federal election survey. We assess the impact of a person's network as a whole through a gendered position generator. This allows us to measure the variety of contacts people have with middle class occupations and with working class occupations, as well as the number of occupations in which they know men and the number in which they know women. Consistent with a social influence argument, we find that (1) the occupational variety of ties to men or women make no difference (because men and women do not differ in their level of tolerance in Canada today). Consistent with both a social influence argument and a competitive threat argument, (2) varied contacts with middle class people increase tolerance while varied working class contacts reduce tolerance (because higher status people are both more educated, which increases tolerance, and less in competition with minorities for lower status jobs). We compare these effects to the possible effects of close ties (are people with more ethnically diverse friends more tolerant?) to further assess contradictory arguments about the relative importance of strong ties, weak ties, and contacts with local strangers.
1
DRAFT PAPER ONLY
PLEASE DO NOT CITE WITHOUT WRITTEN CONSENT FROM AUTHORS
INTRODUCTION
As recent waves of immigrants to Canada and the United States increasingly include non-whites, the categories of minority, non-white, and immigrant have become blurred. The ensuing climate is a constant renegotiation to increase tolerance towards changing minority groups. Research increasingly looks at factors that impact tolerance as a way of contributing to a solution and in part tells us that our friends and family have the capacity to exert varying levels of influence in our lives. Social tolerance or the capacity to accept people who are different from us is in part explained this way. Inter-group friendships in the form of network ties play an important role in facilitating increased tolerance and decreased prejudice of ethnic minorities and immigrants. As an example, people with diversified networks, including many different kinds of contacts, gain positive experience of others not like themselves. They share interests with people outside their own groups, interact with them, come to like them, learn about their contributions and the problems they face, and become more supportive of policies helpful to them. That is, they become more tolerant in our sense. This argument has been developed for personal networks and extended to activity in voluntary associations. Association activity often brings people into contact with others who are very different from themselves yet share the common interests on which associations are based. Earlier research does not do full justice to the complexities of the connections between social networks and tolerance. One kind of complexity concerns the variable nature of social networks ties and their effects. We argue that diversity is not unitary. There are different kinds of networks and network diversity, which for good theoretical reasons have different results. For example, do relationships with close friends and family members provide the education and close contact necessary to impact tolerance, or is it the diversity and range of weak ties that provides a broad range of experiences and types of contact with minorities to impact tolerance? With this study, we will show that while one form of network diversity increases tolerance, others reduce tolerance or have no effect at all. We will further show that while strong and weak ties have an effect on tolerance towards ethnic minorities, our data show that this is not necessarily the case. This variability is not confined to our outcome, tolerance. It is well known that many aspects of social networks vary from positive to neutral to negative, depending on the outcome of interest, the context, and the particular form that the network variable takes. We begin with a brief review of our theoretical engines. These include contact theory, as revised by Putnam and further revised by us; the threat or competition hypothesis, as originated by Bonacich (1972, 1976); learning; and social network influence. We then use these theoretical resources to develop a series of alternative hypotheses connecting networks and tolerance in various ways. We examine these arguments using the 2004 Canadian federal election survey, a large national survey with the range of variables needed to look at the diverse nature of factors leading to tolerance.
2
EXPLAINING TOLERANCE: GENERAL THEORIES The Contact Hypothesis The contact hypothesis asserts that positive orientations towards a group grow as contact with group members increases, provided that the contact is of the right kind. For example, economic competition (discussed below) can lead to negative orientations toward a competing group. Positive orientations flow from contact that is personal, positive, equal status, voluntary, and includes shared goals (Dixon and Rosenbaum 2004; Erickson and Nosanchuk 1998; Sigelman, Welch, Combs and Bledsoe 1996). This kind of contact encourages people to interact, to cooperate, to share valued experiences, to come to like each other, and to learn relatively accurate and positive things about each other. If people have the right kind of contact with minorities, their orientations toward minorities should become more positive. When people study possible effects of contact on tolerance, they often focus on contact with close ties such as friends and families (Pettigrew 1997; 1998). But, as Putnam (2000) and others point out, contact with a small number of special people is too specific to lead to changes in one’s generalized orientations towards whole social groups. If one has a best friend from a minority group, one can easily define that friend as unique, and esteem the friend while still derogating the group. As such, researchers alternatively argue that tolerance flows from extensive contact with many people, from experiences in one’s entire network including acquaintances as well as the near and dear (Dovidio et al. 2003; Putnam 2000). Here, we address a more recent generalization of the contact hypothesis. People who have positive contacts with a variety of others do not just become more tolerant of the particular kinds of people they interact with. They generalize their positive experiences with many kinds of people different from themselves, and develop positive views of people of all kinds (Huckfeldt et al. 1995). This generalized form of the contact hypothesis implies that we should find greater tolerance among people with diversified social networks, and among those who participate in more diversified social settings such as voluntary associations and metropolitan areas. But we argue that diversity is not unitary. There are many kinds of diversity in social networks, which bring different kinds of contact experiences and hence have different effects on tolerance, as we elaborate below. The Competition Hypothesis The competition hypothesis is the negative flip side of the contact hypothesis. Where cooperative contact fosters tolerance, competition and lack of personal contact fosters intolerance. A negative consequence of ethnic groups’ competing occurs when they perceive each other as threats and positive interpersonal contacts become fewer. Tensions increase, leading to negative views (Allport 1954, Bonacich 1972, Forbes 1997, Kunovich 2004, Olzak 1992). Competition and threat can take many forms, but the most discussed are threats to social status, such as economic competition for jobs or business (Bonacich 1972, 1976; Boswell 1986; Brown 2000; Kunovich 2004; Olzak 1992). Competition with minorities is greatest where minorities are most numerous, in some but not all economic sectors in metropolitan areas. Minorities are highly concentrated in large cities in Canada, and too few to be much threat to anyone in smaller communities. Within cities, minorities are further concentrated in some kinds of work. Employers often discriminate against minorities 3
and often undervalue immigrant human capital when it was gained outside of Canada. Minorities often cannot compete effectively for middle class jobs, even when qualified for them, and they are forced to compete for working class jobs instead. As a result, working class people may see minorities as a threat more often than middle class people. We discuss an extension of this hypothesis below. The Learning Hypothesis People become more tolerant of minorities through personal experience, but also because they have learned more about minorities and about issues of tolerance in general. This is the most popular explanation for the well-documented link between education and tolerance: education enlightens. The learning hypothesis is also part of the contact hypothesis; positive kinds of contact teach people positive things about others. Yet it is also part of the competition hypothesis. Groups in competition who see their rivals as threats or negative outgroups, pay more attention to negative ideas about them, and Alearn@ how intolerable they are. People do not only learn about the people they have contact with, they also learn and are influenced by them (Pettigrew 1998). By extension, contact diversity only promotes tolerance when it consists of richly varied contacts with tolerant people (Kawakami et al. 2000). If people have varied contact with the intolerant, diversity increases intolerance. And if people have varied contacts with social groupings that are average in tolerance, their own tolerance levels will neither rise nor fall. Below we discuss different kinds of diversity in social networks, arguing that this diversity contributes to differential effects on tolerance, from positive, to negative, to neutral.
EXPLAINING TOLERANCE: THE DIVERSE ROLES OF NETWORKS AND OTHER FACTORS Social Networks The contact hypothesis predicts that close, personal and frequent contact with ethnic minorities will enhance tolerance. This would suggest that the most important ties in a network are family and friends that provide many opportunities to get to know individuals from a minority group. Ideally, the more close friends that are a part of a minority group, the more likely they are to be tolerant of that group. To evaluate the role of close friends and family, a measure of the racial or ethnic diversity of strong ties is required. Such a measure exists within the Canadian Elections Survey that allow us to look at several types of diversity within respondents close ties: educational, racial, occupational, age and gender. Since we are primarily interested in the effect of racial diversity of close ties on tolerance, we use this as a test of the classic contact hypothesis as well as a measure of the importance of strong ties. A second type of network influence on tolerance stems from the generalized contact hypothesis. It predicts that widely diversified networks, especially diversified weaker ties, enhance tolerance. Our revision of this view calls for more refined attention to various kinds of diversity, which have varying effects on tolerance. We see it as essential for measuring and theorizing multiple forms of diversity in a person=s wider network, including acquaintances as well as the closer ties which previous studies of tolerance have emphasized. Measuring such diversities may well seem a daunting, even impossible task, since people do not know the distribution of important social groupings in their networks. Fortunately, social network researchers have developed a simple, effective, and theoretically well grounded way 4
around this obstacle. The first and still core strategy is to sample occupations which are fairly well represented and range from high to low in prestige in the society one is studying. Respondents are asked whether they know anyone in each of the occupations, and the researcher counts the number in which a respondent knows at least one person. The result is an effective measure of the occupational diversity in a person=s network as a whole. Since occupation plays a very powerful role in modern societies, it is strongly related to a host of other important variables, so the occupational diversity measure is the best single proxy for network diversity overall. Lin and Dumin (1986) launched this approach, and Lin et al. (2001) provide a recent update. Since occupational diversity is not the only important kind, Erickson (2004) developed an extension to include gender. This version selects occupations that vary in gender composition at each level of prestige, and asks respondents whether they know a man and whether they know a woman in each occupation; not just whether they know anyone of indeterminate gender. Since Erickson=s new measure was included in the 2004 federal election study in Canada, we can assess network diversity of two kinds in this paper: middle class diversity (do respondents know people in many or few middle class occupations?), working class diversity (do people know others in many or few working class occupations?), and male and female diversity (do respondents know men in many or few occupations? or women in many or few occupations?). These measures are guides to the kinds of people in a network, as well as the kinds of influences network members will exert on tolerance. Middle class people are likely to be more tolerant for two reasons. They are better educated, and more educated people are more tolerant. They are also less often in competition with minorities, because minority group members do not have equal access to middle class positions (e.g. Li 1992). Contacts with many middle class people should bring influences that increase tolerance. People in working class positions are less well educated, and more often in competition with minorities. They are more likely to see minorities as threats, so extensive contacts with working class people bring influences that reduce tolerance. Gender (as we shall see) was not overly correlated with tolerance towards minorities in Canada in 2004. Thus diversity of ties to men or to women should not have any impact on tolerance either, if networks shape tolerance through social influence. This set of strong and weak tie network diversity measures provides a neat opportunity to explore our claim that network diversity can have all three possible effects on tolerance (positive, negative, or null) depending on the type of diversity. We can also compare the relative effect of strong versus weak tie diversity on tolerance. Individual Attributes Networks are made-up of individuals that ultimately impact the nature of diversity within them. Given the focus of this paper considers the impact of diversity within types of networks on tolerance, we will briefly review the relative impact of personal attributes on tolerance and also show how they relate to networks. Education is one of the strongest correlates of tolerance (Bobo and Licari 1989; Jackman and Muha 1984; Kingston et al. 2003; Schumann et al. 1997). This may be because the learning hypothesis applies: education opens people=s minds to critical debunking of received stereotypes and enhances people=s knowledge of the real positive aspects of many different groups. But the correlation of education with tolerance may not be a direct product of education itself. More educated people have more diversified social networks (Erickson 2004) and are more active in 5
voluntary associations (Curtis and Grabb 1992): two forms of experience thought to develop tolerance (see below). If so, the oft-reported Aeffect@ of education on tolerance will vanish or at least weaken after controlling for network diversity and/or association activity. Occupation is an indicator of class status and is also correlated with tolerance (Grabb 1979). In comparing people in working class occupations to middle class occupations, it is the working class who are least tolerant of ethnic minorities. The competition hypothesis provides details as to why: given the number of immigrants that are forced into working class jobs, the perceived threat to the working class is tangible. Where full time jobs are also scarce, this may intensify the feeling of having employment security threatened. Membership in a minority group is associated with more positive feelings towards minorities, in part because people tend to see their own groups in a more positive light compared to other groups (Judd and Park 1993), and in part because people have better knowledge of their own groups and are less reliant on stereotypes for ideas and judgments. Unlike the effect of education, there is little reason to think that the apparent effect of ethnic or immigrant status is really a masked effect of networks, since minorities have less network diversity (e.g. Erickson 2004). Residents of urban areas have many opportunities for positive learning about minority groups, from extensive and explicit diversity education in schools to discussions of local issues in the media (Nunn, Crockett, and Williams 1978; Stouffer 1955). Cities provide many settings in which members of different groups participate, interact, and feel some kind of common cause such as the vibrancy of ethnically diversified neighborhoods. Since most immigrants flock to the metropolitan areas, and avoid rural ones, such processes are especially relevant to urban dwellers. Cities provide a rich array of subcultures and voluntary associations, and an equally rich range of opportunities to meet many diversified people, so one might expect part of the apparent effect of cities to operate through the more diversified networks and greater associational activity of city dwellers B except that network diversity is actually greatest in rural areas and lowest in the metropolis (Erickson 2004). Youth is often associated with tolerance (Cutler and Kaufman 1975; Stouffer 1955), since younger people are more flexible in their views. Researchers also suggest the presence of a cohort effect (Davis 1975; Wilson 1994) perhaps because the younger people are, the more diversified and tolerant society was during their formative period. Younger people may also be more tolerant because they are more highly educated, more often live in urban areas, and are more often themselves minorities. We find no effect of age in 2004, and will leave this discussion alone for now. Women are often thought to be more nurturing and social than men, and women in Canada today are more liberal than men on many issues (Gidengil, Blais and Nadeau 2003). Thus women might be more tolerant. However, both men and women are found in all ethno-racial groups and in both immigrant and native born populations. The tolerance-related experiences of men and women are probably much the same, and they probably do not experience tolerance issues as gender issues. Thus men and women may not differ in tolerance, as we will indeed show is partially the case. Voluntary Associations Participation in voluntary organizational life is another important source of tolerance (Cigler and Joslyn 2002; Paxton 2004; Pickering 2006; Stolle and Rochon 2001; Uslaner 1999; 6
Warren 2001). Voluntary associations at their best are settings meeting the key conditions of the contact hypothesis. People who differ in many ways, except for the special interest that their association serves, meet as equals on a fully voluntary basis and interact in pursuit of their common concerns. In this way, the opportunity to expand and diversify one’s network increases with participation in voluntary associations. This is the main reason that Putnam (2000) and others stress the importance of voluntary associations as sources of civilized social views. Voluntary association activity may also have indirect effects on tolerance. Putnam (2000) argues that associations contribute to tolerance precisely because associations bring diverse people together, and Erickson (2004) shows that the extent of activity in voluntary associations is the strongest single predictor of network diversity. Closely examined, such arguments imply that the effects of voluntary associations should vary, because associations vary in their levels of the factors thought to produce tolerance. Putnam (2000) acknowledges that some associations are toxic for tolerance, not supportive, although he argues strongly that associations have a good effect on average. Cigler and Joslyn (2002) report that tolerance is relatively low for members of veterans= and ethnic associations, and relatively high for members of political and cultural associations. They speculate that these variations stem from different levels of membership diversity, but cannot directly test this mechanism. Some associations are settings favoring the competition hypothesis. Occupational groups, including labor unions, business groups, and professional associations, bring together people experiencing similar job pressures and tensions. Where market competition with minority groups is high, such associations can increase awareness of and discussion of such competition, heighten perception of group threats, and lead to relatively low levels of tolerance. With so many different kinds of associations, we expect that activity in associations will have all possible effects on tolerance: activity will increase tolerance, reduce it, or have no effect, depending on the nature of the association.
DATA AND MEASURES The data source is the study of the May, 2004 federal election in Canada. This study selected a national representative sample and administered three surveys: a telephone survey during the election (N = 4,323, response rate 53%), another telephone survey just after the election (N = 3,138), and a mailed-out survey shortly afterwards (N = 1,674). The demographic composition of the three waves is similar, except for a slight under-representation of young respondents in the last wave. This paper uses variables from all three waves, with a total N of 1409. The measure of tolerance. Our tolerance scale combines ten items concerning views of immigrants, non-whites, and minorities (see Figure 1). Since immigrants to Canada were once overwhelmingly white, but have become predominantly non-white in recent decades, the three categories (immigrants, non-whites, minorities) overlap as real populations and in popular perceptions. Thus the scale has good reliability (Cronbach=s alpha = .79). The scale combines awareness of the real problems of discrimination in Canada (AIs it more difficult for non-whites to succeed in Canada?@), positive evaluation of immigrants in Canada (AImmigrants make an important contribution to society,@ “On a scale of 0 to 100, how do you feel about racial minorities?” where 100 is a positive evaluation of attitudes, and AToo many immigrants just don=t want to fit into Canadian society@), willingness to welcome new immigrants (ACanada should 7
admit more immigrants@), and approval of social support for immigrants and minorities (AWe should look after Canadians born in this country first and others second@, AHow much more should be done for minorities in Canada?@ coded so that doing much more ranks first, “We have gone too far in pushing equal rights in this country”, “Political parties spend too much time catering to minorities”, and “Minority groups need special rights”). In a final step, the scale values were reversed so that higher values represent more positive attitudes toward immigrants, non-whites and minorities. Measures of attributes. Gender is coded 1 for female, 0 for male. Age is years of age, square rooted to correct for skew. Level of education is an ordinal variable with five levels from less than high school to a post-graduate degree. Place of residence is three dummy variables (for rural areas, mid-size areas, and metropolitan areas) with rural areas as the omitted category in multivariate tables. Employment status was recoded as a dummy variable: “1” included all respondents who were self-employed, working full and part time, “0” was all other categories of work status such as student, volunteer, unemployed and/or currently looking and retired. For this paper, we were able to use two indicators to approximate non-white, minority or immigrant group status; Canadian Born and Non-European Heritage. Canadian Born began with the question “In what country were you born?” and where we can distinguish between immigrants and native born respondents. We created a dummy variable where “1” was Canadian Born and “0” included all other birth countries. Non-European uses the question ATo what ethnic or cultural group do you belong?@ All those reporting non-European ethnic identity were classified as Non-European (coded 1, with all others coded 0). 25 respondents to this first question declared themselves Canadians, so their responses were replaced by their answers to the follow-up question Ain addition to being Canadian, to what ethnic or cultural group did you, or your ancestors belong on first coming to this continent?@ 3 respondents still insisted they were just Canadians, but the rest reported non-European ancestry and we added these to the Non-European group. We also added First Nations, Métis, and Inuit peoples. Most or all Non-Europeans are non-white, though there may well be a few cases of (for example) white immigrants from the Caribbean... For a measure of respondent’s occupation, we chose to use an occupational prestige measure instead of occupation only, consistent with our measure of weak tie network diversity (please see below for details). Initially, Blishen (1981) values were assigned to respondent occupations in determining prestige scores. When using only these scores however, we encountered a significant reduction in sample size. While many respondents were unable or refused to provide an occupation, others were simply left uncoded when there was not enough detail or information to distinguish what type of occupation the respondent held. We follow the method set-out by Ross and Mirowsky (1992)1 in order to include all possible responses and increase the sample size. This technique allows us to compare people with jobs to people without jobs by testing the effect of having an occupation or not , as well as the effect of occupational prestige for those with work. We create the following equation: Working (Prestige Score – Mean of Prestige), whereby “Working” is our employment status dummy variable and “Prestige” is the original occupational prestige variable. By following this equation, we are able to include those who are not working as part of the analysis, by assigning a value of “0”, while those with jobs and
1 For a detailed description, please see Ross and Mirowsky (1992). 8
prestige scores are assigned values ranging from -27.96 to 31.66. Measures of network diversity. For the election study, Erickson developed a new kind of measure of network diversity (Figure 2), a variant on the original developed by Lin and Dumin (1986); see Erickson (2004) for a more detailed explanation. Respondents were asked whether they knew a man, and whether they knew a woman, in each of 15 occupations ranging from high to low in prestige. Figure 2 reports the prestige scores, taken from Ganzeboom and Treimann (1996). To clarify the division into middle class (higher prestige) and working class (lower prestige) occupations, Figure 2 orders occupations by prestige, but the actual survey item randomized the order to prevent order effects on responses. From this question we created four kinds of network diversity measures. Middle class diversity is the number of middle class occupations in which the respondent reports knowing someone, with middle class occupations including lawyer, pharmacist, human relations or sales manager, social worker, and computer programmer. Working class diversity is the number of working class occupations in which the respondent reports knowing someone, including all remaining occupations in Figure 2 from Atailor, furrier or dressmaker@ to Aserver.@ Male diversity is the number of occupations in which the respondent knows a man, and female diversity is the number of occupations in which the respondent knows a woman. The diversity of close ties within a respondent’s network is another important kind of network diversity that can account for tolerance toward ethnic minorities. The truest test of the contact hypothesis is to look at the diversity of racial groups represented within the respondent’s close ties network. The 2004 CES includes such a question, asking respondents to first indicate how many close friends they have, not including family members or relatives. Respondents then indicated whether most, some or none of these friends were of the same race as them (“How many of these friends are of the same race as you?”). For our purposes, this variable was reverse coded to reflect racial diversity as opposed to similarity in the network and then recoded as a dummy variable (0 = most; 1 = some/none). Measures of association activity. Respondents reported whether or not they were active in several kinds of associations during the past five years. Note that these measures do not refer to mere memberships in associations, since many members take no active part in association activities and hence do not experience any of the processes by which associations may foster tolerance. Respondents reported whether or not they were active in associations of these kinds: parents association, religious association, sports association, ethnic association, labour union, women’s group, environmental association, professional association, business association, community association or other. Since we expected these types to have different effects on tolerance, we created dummy variables for each kind (1 for those active and 0 for those not active in a given type of group). Since the Aother@ category is a mysterious mix of things, we used it as the omitted category for multivariate analyses.
RESULTS We begin with a series of multivariate regressions, showing the unstandardized coefficients for attributes, networks, and associations, alone and in combination in Table 1. We start by examining each set of possible predictors separately. Given that attributes such as education and are generally acknowledged to be strong predictors of tolerance and also lend 9
support to the various theoretical models we are developing, we begin by briefly looking at the relative impact of personal attributes on tolerance before focusing on forms of network diversity. Attributes and Tolerance Are attributes correlated with tolerance as expected? The answer is in Model 1 of Table 1. As predicted, we find that tolerance is greater for those with more education. Some researchers argue that part or all of the apparent effect of education consists of response bias. More educated respondents often react negatively to questions that are too simplistically worded, that use qualified language, or that use categorical responses (Campbell et al. 1960, Peabody 1961, Jackman 1973, Jackman and Senter 1980). Yet other research argues that education has real positive effects (Jacob 1957). A study designed to assess real effects beyond response bias found that education is positively related to racial attitudes (Jackman and Muha 1984: 758). Since our tolerance questions (Figure 1) are clear, concrete, and varied in orientation so that a negative response is sometimes tolerant and sometimes intolerant, we argue that the education effect is more reality than response set. This supports the learning hypothesis. We also find the expected results for ethnicity, Canadian born and size of community. Tolerance is greater for those who are themselves ethnically Non-European and hence members of some (probably non-white) minority group. Tolerance is also greatest for those who were not born in Canada, in other words, who are themselves immigrants and perhaps more sympathetic to the plight of other immigrant groups. Tolerance is lowest for those in rural areas, higher in mid-size areas and significantly highest in metropolitan areas. In comparing only attributes, our data initially confirm women=s general reputation for tolerance. There is a significant correlation between gender and tolerance for minorities in Canada 2004. This may be because men and women differ in other attributes we find related to tolerance: age, Non-European status, education, or residence in metropolitan areas. Thus men and women may well have life experiences that affect their tolerance levels in very different ways. We find however, that this effect does not hold-up when compared with other factors. The role of age in increasing tolerant behavior was not confirmed. As results turn out, age is not an important predictor of tolerance towards ethnic minorities or immigrants in Canada in 2004. This result provides evidence contrary to the learning hypothesis, perhaps suggesting that other factors are more important than number and frequency of contact with minorities over time. Voluntary Associations and Tolerance As anticipated, model 2 of Table 1 shows an intriguing mix of correlations between association activities and tolerance. Depending on the type of association, those active are sometimes more tolerant, sometimes less, and sometimes just average. These effects may result from associational activity itself, or may operate through the strong role associations play in network diversity, or may be spurious associations stemming from the kinds of people who join these kinds of associations. Network Diversity and Tolerance If social network relationships with different people are a form of positive contact, and if people generalize such positive experiences to different others in general, then all kinds of network diversity should build tolerance for minorities. With this data, we have the opportunity to compare the effect of strong ties versus weak ties on tolerance. Arguments have suggested that close ties are a more important form of contact, and that it is this type of network diversity that can positively 10
impact tolerance towards minorities. Alternatively, the equally important role of weak ties has been researched and shown to be plausible. Given we can look at both forms of network ties on tolerance we are able to see if there is a difference in the relative effect of each on tolerance. We also suggest that not all forms of network diversity are positive influences on tolerance. If network relationships affect tolerance through social influence processes, then different kinds of network diversity will have different kinds of impacts on tolerance. This is what we find in our data and what we explore below. As we see in our final model, gender is not related to tolerance. Consistent with a social influence argument, tolerance is also not related to network diversity of contacts with men nor to network diversity of contacts with women. Since these null results are important but not significant, we do not report the details. More lively are the results for class, shown in Model 3 of Table 1. Those with more richly diversified ties to middle class people are significantly more tolerant, while those with more diversified ties to working class people are significantly less tolerant. These results are consistent with a social influence argument, in which middle class people provide more tolerant influences because of their higher education and there relative insulation from economic competition with minorities. Since it is only network diversity with respect to class that is related to tolerance, we will only consider these forms of diversity in our models. We next consider the role of strong ties on tolerance. Model 4 of Table 1 shows that overall racial diversity of close ties has no effect on whether or not respondents are more tolerant towards ethnic minorities. However, the more interesting question is whether this effect is different for nonwhites as opposed to whites. For example, do racially diverse networks increase tolerance of whites over nonwhites? The contact hypothesis would suggest this to be the case. Our data, however, show the opposite effect. Diverse close tie networks only matter for respondents of non-European background: in other words, who are already themselves most likely ethnic or racial minorities. For respondents of European background, racially diverse networks have little to no effect on whether or not this group is tolerant of ethnic minorities. According to this data, it would seem as though weak ties matter more for tolerance than do close tie networks, supporting the idea that influences lie perhaps outside the boundaries of close friends and family members and that for contact to be effective, it must also be diverse. Mixed Effects of Attributes It is possible that the apparent effects of attributes are not direct effects of those attributes themselves, but of the kinds of networks or the kinds of association activity with which the attributes are associated. We explore the possible mediating role of network diversity by comparing Model 5 of Table 1 (attributes plus network diversity) to Model 1 (attributes alone). The regression coefficients for attributes are essentially the same before and after controlling for networks. This is somewhat surprising for education at first, since we know that educated people tend to have more diversified networks. However, we also know that more educated and higher status people tend to have more diversity of ties up and down the ladder of stratification, with more ties that are both high and low status. It is lower status people who tend to have ties limited to just part of the range (their own, lower status part) (e.g. Lin 1982). More educated people have both more diverse ties to middle class people with their influence towards greater tolerance, and more diverse ties to 11
working class people with their influences toward less tolerance, and the two opposite kinds of network influence cancel each other out. Other attributes associated with tolerance also remain associated once network diversity is controlled, with the exception of the effect of living in a metropolitan center. Mixed Effects of Network Diversity Perhaps our network diversity measures go with tolerance because a wide range of contacts brings many invitations to join voluntary associations, and people learn tolerance in associations rather than in networks themselves. However, comparing Model 7 of Table 1 to Model 3 shows that the coefficients for our network measures are about the same whether we control for association activity or not. There may perhaps be a slight decline in the strength of the effect of middle class diversity, because middle class people are more active in associations, and likely to recruit new members into associations full of tolerant middle class people. However, the apparent effect of middle class diversity is still strong after controlling all sorts of association activity, so associations do little if anything to account for possible network effects. Network diversity may also be related to tolerance in spurious ways, because the kinds of people who differ in tolerance are also the kinds of people who differ in their networks. Education, for example, is a source of both network diversity and tolerance. Network effects on tolerance are in part, but only part, spurious with respect to attributes. Comparing Model 5 to Model 3 in Table 1, we find that network effects are indeed reduced after controls for attributes, but still clearly significant. When we control for both voluntary association activity and attributes in Model 8, we see no further change in the network effects compared to just controlling attributes in Model 5. Looking at our final model (Model 8) and the relative effect of racial diversity of strong ties on tolerance, we see its effect significantly diminished in contrast to that of weak tie diversity. This suggests that in Canada, in 2004, strong ties were not an important predictor of tolerance. The bottom line however, is that a healthy amount of network effects for weak ties remains whatever we control. Interactions with middle class people, likely to be tolerant, influence people to be more tolerant themselves. Interactions with working class people, likely to be less tolerant, reduce tolerance. These findings provide strong evidence for the social influence model of network diversity effects, but none for the view that all forms of network diversity teach or equally influence tolerant attitudes. Mixed Effects of Voluntary Associations Putnam and others have argued that associations produce tolerance by bringing people into contact with a healthy variety of other people. If so, any apparent effects of associations should shrink or vanish when we control for social network diversity. This could be true both for associations for which activity is positively related to tolerance (because such associations have usefully diverse members, especially perhaps diverse middle class members) and for associations for which activity is negatively related to tolerance (because such associations lack member diversity, or have diverse working class members). But, when we compare Model 7 to Model 2, we find that the coefficients for all types of associations are much the same whether we control for network diversity or not. If we could look closely at more detailed types of associations, or better yet individual associations, we would see more variation in the extent and type of diversity in associations. This might uncover more of a role for social network formation within specific 12
associations. But we cannot do that here. It is also possible that some associations do not correlate with tolerance because of association activities, but because of the kinds of people who choose to engage in these activities. If the apparent effects of associations are spurious with respect to attributes, they should weaken or vanish when attributes are controlled. Comparing Model 6 to Model 2, we do indeed see some changes of this kind. The coefficient for professional association activity goes from quite strong and significant to nothing at all, quite likely because professionals are highly educated. The effect for parents associations, community associations and ethnic groups also weaken. When we control for network diversity as well as attributes, in Model 8 of Table 1, we see no further change in the effects of associations as compared with Model 6, in which only attributes are controlled.
CONCLUSIONS Our research shows that different types of networks, attributes and associations all contribute to tolerance B but not in all of the ways that have been argued in past work. The contact hypothesis, most generally, asserts that network diversity leads to greater tolerance. People generalize their positive experiences with some forms of diversity to positive views of all forms of diversity. But, as we expected, diversities differ in their effect. Contact with others consists of channels for social influence, and the effect of a given kind of diversity depends on the kind of influence that those known are likely to exert. We found much stronger support for this view, since different forms of network diversity had very different relationships with tolerance. The most conservative interpretation of the contact hypothesis suggests that frequent and close contact with individuals from minority groups increases tolerance. In other words, it is the strong ties in one’s network that have the greatest impact on tolerance. By extension, racial diversity of strong ties should lead to tolerant attitudes towards the racial groups represented. We find however, that this is not necessarily the case. While racially diverse strong ties matter for non-European respondents, this is not true for European respondents. Further, when compared to other factors that impact tolerance, like diverse weak tie network measures, racially diverse strong tie measures become nonsignificant. Why these types of ties are more important for tolerant attitudes of ethnic minorities is an interesting question. Further study needs to be done in order to determine why this is the case. Diversity of ties to middle class people goes with greater tolerance, consistent with the higher education levels of middle class people and their lower degree of economic competition with minorities, both of which contribute to higher tolerance. Diversity of ties to working class people goes with lower tolerance, consistent with their lower education levels and greater economic competition with minorities. These theoretically pivotal effects could not be explained away as disguised effects of attributes or voluntary associations, although attributes do help to account for a modest portion of the initial network effects. We hasten to point out that these findings are not in any way a reflection of the essential nature of the working class or the middle class, as was once argued for the Aworking class authoritarianism@ that we now know to be mythical. If working class people are more often in competition with minorities, this is not their doing, but the result of employer discrimination and government failure to intervene with policy levers like affirmative action or quick accreditation of 13
foreign-trained professionals. The ultimate sources of the conditions that promote intolerance lie in the upper and middle classes, not the working class. Given equally intense competition, middle class people will also be likely to see minorities as threats and to become less tolerant. Indeed, our strongest support for the competition hypothesis comes from the significant tendency for those active in professional associations to be less tolerant when in metropolitan areas where competition with minorities is greatest. For labor unions, which are more often working class, we found no such trend. Associations also vary greatly in how they are associated with tolerance. Depending on the kind of association, activity in the association goes with greater tolerance, with lower tolerance, or with just average tolerance. Clearly, associations are not uniformly wonderful; they are not even mostly wonderful (as Putnam often argues). Further, some of the apparent effects of association life seem to be, in fact, membership composition effects while other apparent effects seem to be, in fact, real effects of participation in certain kinds of associations. Associations with well educated members, or a relatively high proportion of minority members, look good for tolerance because they include kinds of people disposed toward tolerance. Associations with poorly educated memberships look bad. But often the associations as such are doing nothing in particular beyond attracting people with different levels of education or different numbers of minorities. However, we found that some kinds of association activities have real effects that do not vanish when we control attributes or various kinds of network diversity. We suspect that one important form of such activity is intense, direct discussion of related issues in an engaged or even emotional way. We think environmental groups may have positive forms of such discussion and action, while sports groups have more negative forms. Clearly, future research needs to examine more detailed kinds of associations and to make more refined and direct measurement of association characteristics. How do ethnic groups for minority ethnic groups differ from others? How does the ethno-racial composition of an association affect the ethno-racial diversity of the networks of members, and how do both of these affect tolerance? How democratic is an association? What is the extent, and the intensity of discussion of issues connected to minorities? Social networks are a major factor in tolerance, but not at all uniform. Social networks are diverse and come in many forms. These differ greatly as experiences of contact with minority groups, competition with the, learning about them, and influence concerning them. Social networks promote tolerance, erode it, leave it unaffected, and matters for some and not others, depending on the form it takes.
14 REFERENCES
Allport, G.W. 1954. The Nature of Prejudice. Reading, MA: Addison-Wesley. Bobo, Lawrence and Frederick C. Licari 1989. Education and Political Tolerance. Public Opinion Quarterly, 53: 285-307. Bonacich, Edna 1972. A Theory of Ethnic Antagonism: The Split Labor Market. American Sociological Review, 37: 547-559. — 1976. Advanced Capitalism and Black/White Race Relations in the United States: A Split-Market Interpretation. American Sociological Review, 41: 34-51. Boswell, T. 1986. A Split Labor Market Interpretation of Discrimination against Chinese Immigrants, 1850-1882. American Sociological Review, 51: 352-371. Brown, C. 2000. The Role of Employers in Split Labor Market: An Event -Structure Analysis of Racial Conflict and AFL Organizing. Social Forces, 79: 653-682. Campbell, Angus, Philip E. Converse, Warren E. Miller and Donald E. Stokes 1960. The American Voter. New York, NY: Wiley and Sons. Curtis, J., and Grabb, E. 1992. Voluntary Association Activity in English Canada, Fench Canada and the United States: A Multivariate Analysis. Canadian Journal of Sociology, 17(4): 371-388. Cutler, Stephen J., and Robert L. Kaufman 1975. Cohort Changes in Political Attitudes: Tolerance of Ideological Nonconformity. Public Opinion Quarterly 39(Spring): 249-274. Davis, James A. 1975. Communism, Conformity, Cohorts and Categories: American Tolerance in 1954 and 1972-73. American Journal of Sociology, 81: 491-513. Dixon, Jeffrey C. and Michael S. Rosenbaum 2004. Nice To Know You? Testing Contact, Cultural and Group Threat Theories of Anti-Black and Anti-Hispanic Stereotypes. Social Science Quarterly, 85(2): 257-280. Dovidio, John F., Samuel L. Gaertner, and Kerry Kawakami 2003. Intergroup Contact: The Past, Present and Future. Group Processes and Intergroup Relations, 6: 5-21. Erickson, Bonnie H. 2004. The Distribution of Gendered Social Capital in Canada. Pp. 27-50 in Henk Flap and Beate Volker (eds.), Creation and Returns of Social Capital: A New Research Program. London and New York: Routledge. —. 2006. Persuasion and perception: new models of network effects on gendered issues. Pp. 293-322 in Brenda O'Neill and Elisabeth Gidengil (eds.), Gender and Social Capital. New York, NY: Routledge. Erickson, Bonnie H. and T.A. Nosanchuk 1998. Contact and Stereotyping in a Voluntary Association. Bulletin de Méthodologie Sociologique 60: 5-33. Forbes, H.D. 1997. Ethnic Conflict: Commerce, Culture and the Contact Hypothesis. New Haven, CT: Yale University Press. Ganzeboom, Harry B. and Donald J. Treiman 1996. Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations. Social Science Research, 25: 201-39. Gidengil, Elisabeth, Andre Blais, R. Nadeau, and Neil Nevitte 2003. Women to the Left? Gender Differences in Political Beliefs and policy preferences. Pp. 148-59 in M. Tremblay and L. Trimble (eds.), Women and Electoral Politics in Canada. Don Mills, ON: Oxford University Press. Grabb, Edward G. 1979. Working Class Authoritarianism and Tolerance of Outgroups: A Reassessment. Public Opinion Quarterly, 43(1): 36-47.
Huckfeldt, Robert, Paul Allen Beck, Russel J. Dalton, and Jeffrey Levine 1995. Political Environments, Cohesive Social Groups, and the Communication of Public Opinion. American Journal of Public Opinion, 39: 1025-54. Jacob, PE. 1957. Changing Values in College: An Exploratory Study of the Impact of College Teaching. New York, NY: Harper Collins. Jackman, Mary J. 1973. Education and Prejudice or Education and Response-Set? American Sociological Review, 38: 327-339. Jackman, Mary J., and Michael E. Muha 1984. Education and Intergroup Attitudes: Moral Enlightenment, Superficial Democratic Commitment, or Ideological Refinement? American Sociological Review, 49: 751-769. Jackman, Mary R. and Senter, Mary Scheuer 1980. Images of Social Groups? Categorical or Qualified? Public Opinion Quarterly, 44: 341-361. Judd, Charles, and Bernadette Park 1993. Definition and Assessment of Accuracy in Social Stereotypes. Psychological Review, 100(1): 109-128. Kawakami, K., J.F. Dovidio, J. Moll, S. Hermsen, and A. Russen 2000. Just Say No (To Sterotyping): Effects on Training in Trait Negation on Sterotype Activation. Journal of Personality and Social Psychology, 78: 871-888. Kingston, P. W., R. Hubbard, B. Lapp, P. Schroeder, and J. Wilson 2003. Why education matters. Sociology of Education, 76: 53-70. Kunovich, Robert M. 2004. "Social Structural Position and Prejudice: An Exploration of Cross-National Differences in Regression Slopes." Social Science Research 33:20-44. Li, Peter S. 1992. Race and Gender as Bases of Class Fractions and Earnings. Canadian Review of Sociology and Anthropology, 29: 488-510. Lin, Nan 1982. Social Resources and Instrumental Action. Pp. 131-45 in Peter V. Marsden and Nan Lin (eds.), Social Structure and Network Analysis. Beverly Hills, CA: Sage. Lin, Nan and Mary Dumin 1986. Access to Occupations Through Social Ties. Social Networks, 8: 365-85. Lin, Nan, Yang-Chih Fu, and Ray-May Hsung 2001. The Position Generator: Measurement Techniques for Investigations of Social Capital. Pp. 57-81 in Nan Lin, Karen Cook, and Ronald S. Burt (eds.), Social Capital: Theory and Research. New York, NY: Aldine de Gruyter. Nunn, Clyde Z., Harry J. Crockett, and J. Allen Williams 1978. Tolerance for Non-Conformity. San Francisco, CA: Josey-Bass. Olzak, Susan 1992. The Dynamics of Ethnic Competition and Conflict. Stanford, CA: Stanford University Press. Pattie, Charles and Ron Johnson 2000. >People Who Talk Together Vote Together=: An Exploration of Contextual Effects in Great Britain. Annals of the Association of American Geographers, 90: 41-66. Paxton, Pamela 2004. Who Do You Trust? A Multilevel Model of Generalized Trust in Thirty-One Countries. 99th American Sociological Association Annual Meeting. San Francisco, CA. Peabody, D. 1961. Attitude Content and Agreement Set In Scales of Authoritarian, Dogmatism, Anti-Semitism and Economic Conservatism. Journal of Abnormal Psychology, 63: 1-11. Perrin, Andrew J. 2005.Political Microcultures: Linking Civic life and Democratic Discourse. Social Forces, 84(2): 1049-1082. Pettigrew, T.F. 1998. Intergroup Contact Theory. Annual Review of Psychology, 49:65-85.
16
Pickering, Paula M. 2006. Generating Social Capital for Bridging Ethnic Divisions in the Balkans: Case Studies of Two Bosniac Cities. Ethnic and Racial Studies, 29(1): 79-103. Putnam, Robert 2000. Bowling Alone.. New York, NY: Simon & Schuster. Ross, Catherine E. and John Mirowsky. 1992. Households, Employment, and the Sense of Control. Social Psychology Quarterly, 55(3): 217-235). Schuman, Howard, Charlotte Steeh, Lawrence Bobo, and Maria Krysa. 1997. Racial Attitudes in America: Trends and Interpretation. Cambridge, MA: Harvard University Press. Sigelman, L., S. Welch, M.W. Combs and T. Bledsoe 1996. Making Contact? Black-White Social Interaction in an Urban Setting. American Journal of Sociology, 101: 1306-1332. Stolle, D., and T.R. Rochon 2001. Are all associations alike? Pp. 143-156 in Beyond Tocqueville: Civil Society and the Social Capital Debate in Comparative Perspective, edited by R. Edwards, M.W. Foley, and M. Diani. Hanover, NH: Tufts University Press. Stouffer, Samuel A. 1955. Communism, Conformity and Civil Liberties. Garden City, NY: Doubleday. Uslaner, E. M. and R.S. Conley 2003. Civic engagement and particularized trust - The ties that bind people to their ethnic communities. American Politics Research, 31: 331-360. Warren, M.E. 2001. Democracy and Associations. Princeton, NJ: Princeton University Press. Wilson, Thomas C. 1994. Trends in Tolerance Toward Rightist and Leftist Groups, 1976- 1988 - Effects of Attitude-Change and Cohort Succession. Public Opinion Quarterly, 58(4): 539-556.
17
FIGURE 1
Tolerance of Ethnic Minorities and Immigration
1. How much more should done for racial minorities in Canada? 1 = much more, 2 = somewhat more, 3 = about the same as now, 4 = somewhat less, 5 = much less
2. . Immigrants make an important contribution to society. 1 = strongly agree, 2 = agree, 4 = disagree, 5 = strongly disagree 3. We have gone too far in pushing equal rights in this country 1 = strongly agree, 2 = agree, 4 = disagree, 5 = strongly disagree
4. Too many recent immigrants just don't want to fit into Canadian society. 1 = strongly agree, 2 = agree, 4 = disagree, 5 = strongly disagree
5. We should look after Canadians born in this country first and others second. 1 = strongly agree, 2 = agree, 4 = disagree, 5 = strongly disagree
S (α = .699)
18
FIGURE 2
Occupation (Prestige)
Lawyer (73) Pharmacist (64) HR Manager (60) Middle Class Sales Manager (60) Occupations Social Worker (52) Computer Programmer (51) Tailor, Furrier, Dressmaker (40) Farmer (40) Carpenter (37) Cashier (34) Working Class Delivery Driver (31) Occupations Security Guard (30) Sewing Machine Operator (25) Janitor (25) Server (21)
19
TABLE 1 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 RESPONDENT ATTRIBUTES Female 0.30 ª 0.31 ª 0.15 0.81 Sqrt Age -0.07 -0.09 -0.16 ª -0.18 * Canadian Born -1.27 *** -1.24 *** -1.18 *** -1.00 *** Non-Europeen 0.78 * 0.32 0.80 * 0.61 -0.69 Level of Education 0.46 *** 0.43 *** 0.45 *** 0.44 *** Mid-Size 0.18 0.12 0.23 0.23 Metro 0.47 * 0.36 0.50 * 0.48 * Workstatus 0.26 0.20 0.32 ª 0.22 Occupational Prestige 0.03 ** 0.02 * 0.03 ** 0.02 *
TYPE OF ASSOCIATION Parents -0.47 ª -0.46 ª -0.36 -0.58 * Religious -0.05 0.15 -0.03 0.22 Sports -0.61 ** -0.76 *** -0.64 ** -0.77 *** Ethnic 1.63 *** 0.59 1.75 *** 0.60 Labour 0.01 0.17 0.12 0.24 Womens 0.18 0.22 0.19 0.09 Environmental 0.23 0.01 0.14 0.06 Professional 1.18 *** -0.04 0.90 *** -0.01 Business -0.55 * -0.61 * -0.69 ** -0.63 * Community 0.59 ** 0.60 ** 0.54 ** 0.64 **
NETWORK DIVERSITY Middle Class Contacts 0.55 *** 0.17 * 0.49 *** 0.16 * Working Class Contacts -0.35 *** -0.12 ** -0.34 *** -0.11 * Same Race as You -0.02 -0.10 Non-Europeen x Same Race as You 1.799 * 1.89 *
Constant 11.62 *** 13.44 *** 13.47 *** 13.69 *** 11.96 *** 12.22 *** 13.44 *** 12.34 *** Adjusted R 0.156 0.043 0.048 0.011 0.159 0.174 0.083 0.182 N 1375 1523 1526 1443 1375 1373 1523 1325 ?p<.10; * p< .05; ** p<.01; *** p<.001
20
Social Capital Across the Life Course: Age and Gendered Patterns of Network Resources*
Steve McDonald and Christine Mair
North Carolina State University
*Direct correspondence to Steve McDonald ([email protected]), Box 8107, Department of Sociology & Anthropology, North Carolina State University, Raleigh, NC 27695- 8107. This research is funded by Academic Sinica, Taiwan (through the Research Center for Humanities and Social Sciences and the Institute of Sociology). Social Capital Across the Life Course: Age and Gendered Patterns of Network Resources
Steve McDonald and Christine Mair
Abstract
This paper highlights the usefulness of applying a life course perspective to the study of social capital. To address the dearth of empirical evidence on social capital variation across the life course, we examine age variation in access to social capital in an analysis of nationally representative data on U.S. respondents age 22-65. Analyses identify a statistically significant relationship between age and various indicators of social capital. The results show that resources from occupational contacts tend to increase with age, but eventually level off among older respondents. Changes in voluntary memberships follow a similar pattern, while daily contact is negatively associated with age. The findings suggest that resources embedded in occupational networks tend to accumulate across the career, even in the face of a decline in sociability.
Moreover, the relationship between age and social capital is often conditioned by gender. Social Capital Across the Life Course: Age and Gendered Patterns of Network Resources
Introduction
Sociologists have recently focused much attention on identifying and explaining patterns of resource accumulation across the life cycle (for a review, see DiPrete and Eirich 2006).
Researchers have found, for example, substantial divergence in health (Dupre 2007; Willson,
Shuey, and Elder 2007) and economic resources (Alon and Haberfeld 2007; Maume 2004) over the course of careers. However, little is known about life course patterns of social capital—i.e., resources embedded within social relationships (Lin 2001). Insight into trajectories of social capital would aid in our understanding how people navigate the various spheres of their lives
(e.g., work, family, and educational spheres, to name just a few). Furthermore, life course trajectories of social capital may contribute to broader patterns of cumulative advantage/disadvantage as well as persistent race and gender inequities, as many have suggested
(e.g., see Bourdieu 1986).
Some have suggested that social resources tend to accumulate over time (Bridges and
Villemez 1986), while others have suggested that social resources tend to depreciate over time
(Coleman 1990). Few researchers have explored the nuances of network dynamism across the life course. Rather, most of the research on social networks has remained preoccupied with examining how static network configurations and features explain various outcomes (Smith-
Doerr and Powell 2005). The bulk of the evidence on network change comes from the social support literature (e.g., Cornwell, Laumann, and Schumm 2008; Kalmijn 2003). Yet these studies examine strong ties to network contacts which tend to be less salient than weak ties for understanding attainment processes (Granovetter 1973). The few studies that focus on change in work contacts are limited primarily to networks within work organizations rather than more
1 broadly representative populations. Moreover, analyses of network change are not always embedded within the context of individual’s biographical experiences. Such analyses are often focused on examining changes in networks over time (for example, Time 1 è Time 2 è Time
3) rather than linking changes in networks to aging processes. Consequently, such research lacks sensitivity to the ways that age structures human experiences.
The current investigation attempts to overcome some of the limitations of prior research by using a nationally representative sample of United States residents age 22-65 to examine age variation in social network characteristics. Our examination is focused on (but not limited to) occupational network characteristics drawn from the position generator approach to collecting network data (Lin and Dumin 1986; Lin, Fu, and Hsung 2001). The data are cross-sectional, which limits our ability to distinguish aging trajectories from period and cohort effects.
Nonetheless, the findings are consistent with prior research and robust across various indicators of social capital. The results suggest that highly valued work-related social resources tend to accumulate over the course of careers, with diminishing returns among the most senior respondents. Voluntary association membership follows and identical pattern, but non-work daily contact is negatively associated with age. The analyses also highlight many gender differences in the form of the relationships between age and social network characteristics, suggesting that gendered careers structure life course patterns of social capital accumulation.
The first half of the paper begins with a theoretical justification for our study. We link the study of social capital to a life course approach aimed at understanding changing lives within biographical and historical context. Empirical justification follows from a critical review of the relevant empirical literature on network dynamics. We then discuss a set of broad empirical expectations to guide our analyses.
2 Social Capital and Principles of the Life Course
Noting the static character of network studies, Smith-Doerr and Powell (2005) emphasize the need to incorporate principles of the life course into the study of social networks and economic life. The life course perspective is highly compatible with a relational approach to understanding human behavior (see Emirbayer 1997). A life course perspective helps to embed the network features and configurations possessed by individuals within their broader set of experiences, highlighting the cross-fertilization, intersecting fields, complex motivations, and cross-cutting network affiliations. Moreover, understanding the role that networks play in determining attainment outcomes requires greater attention to life course social processes, the history, timing, and duration of relationships. In this section, we summary the core principles of the life course perspective as they relate to the study of social networks.
Linked Lives – Life course theory explicitly notes that lives are lived interdependently.
An individual’s opportunities are linked to the fortunes of other people with whom they are associated. Status attainment research (e.g., Sewell, Haller, and Ohlendorf 1970) has long recognized that educational and occupational opportunities of individuals are affected by the role played by significant others in their lives: for example, parents (Kim and Schneider 2005), peers
(Crosnoe 2000), teachers (Crosnoe, Johnson, and Elder 2004), and mentors (McDonald,
Erickson, Johnson, and Elder 2007). Network configurations are therefore consequential for patterns of life course attainment. Because of homophilous network formation and maintenance
(McPherson, Smith-Lovin, and Cook 2001), social resources are unequally distributed throughout society (Lin 2000). Resource advantages and disadvantages are consolidated, helping to reproduce existing societal inequities within society (cf., McDonald, Lin, and Ao 2008).
Individuals rich in social capital tend to share their good fortune with the people who they know
3 and care about. Whereas other individuals may be forced to hoard the few available resources
(Smith 2005) lest other network members prove to be a drag on opportunities for advancement
(Stack 1974). In this way, social capital structures life chances.
Agency – The life course perspective also stresses the importance of agency. Individuals construct their life course through the choices and actions that they take. The formation and maintenance of social networks over time is an essential part of this process. The cultivation of contacts has often been viewed as an investment process (Bourdieu 1986; Burt 1992; Lin 2001), such that individuals spend considerable time and effort interacting with others. Some have considered these actions as instrumentally motivated in order to increase wealth, status, and power (Lin 2001), while others have left open the possibility of network actors to socialize for non-goal directed purposes. Both classical (Weber [1922]1968) and contemporary (Bourdieu
1986) scholars have acknowledged that much of social interaction is unconsciously motivated by habit and impulse. Friendships tend to form, at least in part, on the basis of mere spatial proximity (Blau 1963; Verbrugge 1977).
Consequently, Kilduff and Tsai (2003) have argued that interaction within organizational settings can be instrumentally motivated or serendipitous. This distinction, though extremely useful, fails to expose the further complexities of social capital investment. Granovetter (2003) has argued that engaging in social interaction can involve multiplex motivations. For example, graduate students might be inclined to attend a departmental holiday party not only for the requisite “face time” that they can get with faculty members, but also because of the free food.
Relationships are also dynamic: the usefulness of relationships is almost certain to change over time. Motivations for relationship formation are likely to differ from the motivations for relationship maintenance. Furthermore, relationships can have unintended consequences. A
4 friendship forged in youthful debauchery can lead to a job referral later in life. In brief, action is an essential element of the construction of social networks across the life course, though the motivations for engaging in social interaction are multiplex and dynamic.
Biography – Aging occurs through processes that are path dependent and cumulative in nature (Elder and Shanahan 2006; O'Rand 2006). Career patterns are constitutive of life course trajectories—that is, sequences of related events or experiences in individual lives. These life course trajectories set individuals on specific paths in life, helping individuals to garner experiences, skills, and resources that ultimately structure future opportunities and decision making. At the same time, individuals are capable of making subtle or even dramatic changes to the paths that they are on. Transitions refer to the distinct events in life that denote changes in the trajectories of social roles. Any given transition in a person’s life needs to be understood within the context of these broader life course trajectories.
Social relationships, too, follow their own unique trajectories over time. Most research on social networks has focused on the presence of social relationships and how relationships affect various outcomes (Smith-Doerr and Powell 2005), but few have examined the ways in which relationships are formed, maintained, erode, go into abeyance, and are rekindled (Burt 2000).
Even studies of change in network structures can treat such transitions as independent from larger life course experiences. For example, recent advances in the statistical modeling of network dynamics are derived from Markov chain models, which are based on the assumption that the probability of a future transition in state is dependent on a person’s current state, but independent of past experiences (Snijders 2001). Such an approach is insufficient for understanding divergence in mobility outcomes over time (Fuller 2008) and fails to account for
5 the ways that prior actions and experiences transform network structures. In this way, lifetime
experiences help shape networks and opportunities.
Timing – The life course perspective recognizes that the impact of any given event is
contingent on its timing. Individuals who fail to follow normative pathways can be labeled as
deviant and/or face substantial institutional impediments to accumulating resources. For
example, individuals who complete college “off-time” fail to receive the same wage advantages
of individuals who complete college “on-time” (Elman and O'Rand 2004). Similarly, the timing
of parenthood can have important implications for relationships with children (Heuvel 1988) as
well as long term economic consequences for parents (Hofferth 1984). The timing of these life
transitions is therefore consequential for life course trajectories, setting some on pathways of
advantage while disadvantaging others. Furthermore, the character of social relationships, the
resources embedded within those relationships, and the returns to those relationships all vary
across the life course. Network composition is linked to age and to specific life course transitions
(Bidart and Lavenu; Degenne and Lebeaux; Fischer and Oliker 1983; Kalmijn). The influence of
social relationships on life events is also contingent on timing (McDonald and Elder 2006;
McDonald, Erickson, Johnson, and Elder 2007; Moerbeek and Flap; Rhee).
History – Finally, the life course approach examines individual trajectories within the larger historical context. Individual biographies intersect with historical circumstances (Mills
1959). History makes its own unique mark on each cohort of individuals. People traverse life as part of a convoy of social relationships (Antonucci and Akiyama 1987), experiencing historical circumstances and developing collective meanings and memories (Schuman and Scott 1989).
Social interactions can only be understood within these historical contexts. The various historical patterns can have a profound impact on the character of social networks. For example, women’s
6 increased entry into the labor force, the rise of suburban communities and urban isolation, and
the rapid expansion of internet technology (among many other societal trends) have all had major
impacts on the makeup off social networks and their usefulness in society. Social scientists have
only begun to address historical change in network structures (see McPherson, Smith-Lovin, and
Brashears 2006; Putnam 2002).
Approaches to the Analysis of Network Change Over Time
A variety of different research fields have explored changes in social networks and social capital
over time. Yet few studies explicitly link changes in social capital to experiences across the life
course. The approaches to studying social network change are briefly outlined here.
Modeling Dynamic Network Structures
Recent research has addressed change over time in social networks through the statistical modeling of dynamic network structures. This branch of research examines the complexity of network structures via statistical simulations (Doreian 2002; Doreian 2006; Hummon 2000;
Hummon and Doreian 2003). Structural balance theory is used to assess whether or not a set of social ties (a triad, for example) is “balanced” based upon its configuration and ties with other groups. This balance reflects of a level of stability in that social configuration and is used to predict changes, or lack of changes, in social ties over time. These mathematically-based models of social networks are informative because they allow researchers to theorize about change in social tie arrangements over time. However, this line of research is problematic for a number of reasons. First, these studies are based on overly simplistic assumptions of rational behavior that fail to account for multiplex and dynamic motivations for action. Second, the models are based primarily on simulations and therefore lack an empirical basis. Third, time is de-contextualized in these studies. In other words, time is divorced from historical context and disembedded from
7 the life course experiences of individual actors. Current network configurations are presumed to affect future network change independent of individual biographies.
Changes in Friendship and Social Support Networks
Research on friendship and social support networks has focused much attention on life course patterns of network resources. Founded in family and community theory, research in this tradition has examined the ways in which age, family transitions, career transitions, and historical events affect network configurations. In general, the results show that life course transitions can result in major network disruption. Friendships are formed quickly and easily at young ages (Cairns and Leung 1995). Drastic transitions, such as those from grade school to junior high for example, result in major upheaval of social networks leaving some individuals vulnerable to a lack of emotional support (Cantin and Boivin 2004).
Prior research suggests that friend networks tend to shrink over time (Kalmijn 2003;
Wellman, Wong, Tindall, and Nazer 1997). Labor force participation results in more homogeneous networks (Bidart and Lavenu 2005) and greater network density leads to higher retention of contacts over time (Martin and Yeung 2006). Despite a general decline in contacts across the life course, older individuals tend to experience increases in neighborhood socializing, religious attendance and volunteering from age 57 to 85 (Cornwell, Laumann, and Schumm) and older individuals demonstrate a high level of flexibility and adaptability in the maintenance of ties over time (Wenger and Jerrome 1999).
Gender and family structure also affect friendship networks over time. Married and cohabiting individuals have smaller networks than non-married (Bidart and Lavenu 2005;
Fischer and Oliker 1983; Kalmijn 2003). The number of friends and contacts a couple shares increases over the life course, although women tend to have greater contact with friends than
8 men (Kalmijn 2003). Gendered patterns of labor force participation also influence network characteristics across work careers. Women tend to have more discontinuous work histories than men (Han and Moen 1999). Consequently, men socialize primarily coworkers while women have higher proportions of kin-based ties in their networks (Fischer and Oliker 1983). Women gained more opposite-sex friends in their networks with entrance into the working world while the number of opposite sex friends for men is unaffected by setting (Kalmijn 2002). Women’s social resources, though, are particularly vulnerable to changes in socioeconomic status (Ajrouch,
Blandon, and Antonucci 2005). Other structural changes such as retirement may actually benefit female networks as they gain more friendship connections but result in network shrinkage for males (Fischer and Oliker 1983).
This line of research has demonstrated that social networks (1) vary over time, (2) are sensitive to life course transitions, and (3) are structured by gender and family/work career trajectories. There are, however, a number of limitations of this research. First, these studies have also tended to rely on data from age-specific samples (e.g., Cairns and Leung 1995; Cornwell,
Laumann, and Schumm 2008), which are unable to fully assess age variation in social network characteristics. Second, the studies have questionable generalizability for the U.S. context, since the samples are typically not nationally representative. Third, few of these studies have directly addressed issues related to changes in occupational networks over time. Most of these studies focus exclusively on changes in close friendship ties rather than work-related ties.
Changes in Occupational Networks
An emerging line of research focuses on change in occupational ties over time. Drawing from a sample of investment bankers, Burt (2000) notes that social ties are not stable over time but rather, are subject to a high level of “decay” or disintegration of ties. If a tie is strongly
9 embedded in a given network then it is less likely to decay or more likely to decay at a slower rate than ties that are not as embedded in a social network. Further investigations reveal that newly formed ties are more vulnerable to decay and the risk of decay is likely to decrease over time as the tie persists (Burt 2002). Social resources tend to accumulate for those who start out with much social capital and decay is less frequent among high performance bankers (Burt
2002).
Researchers have also examined the importance of timing for conditioning the effects of social capital. For example, Rhee (2007) demonstrates that the impact of different types of network ties on promotions within a high technology firm is contingent on the duration of time that workers had known their contacts. Moerbeek and Flap (2008) show that, among Dutch household members, access to social resources through relatives has a greater impact on the prestige of first jobs, while access to resources through friends has a greater impact on current job prestige. Analyses of panel data from the NLSY survey show that the use of informal job matching methods (i.e., getting jobs through personal intermediaries) results in divergent outcomes depending on the timing of the job change in people’s lives (McDonald and Elder
2006). The authors find that certain types of informal job matches (i.e., non-searching) are more effective during the middle of the work career than early in the career, implying that workers tend to accumulate social capital across the life course. Then again, research on a sample of white collar workers finds that younger and older network members tended to have the least amount of diversity in their networks (Lambert, Eby, and Reeves 2006), suggesting a curvilinear relationship between social capital and age.
These studies are important because they point to patterns of the accumulation and erosion of work-related contacts over the life course and the contingency of social capital effects
10 on the timing of employment transitions and relationship formation. However, the evidence is far from conclusive. Like with the social support literature, the studies are subject to questionable generalizability due to non-representative and age-specific sampling. Furthermore, little attention is paid in these studies to gendered patterns of network change over time.
Expectations
The empirical investigation presented here attempts to overcome some of these problems by relying on a nationally representative sample of the United States to examine age variation in social capital and how these life course patterns vary by gender. We explore a number of operationalizations of social capital. To begin with, we focus on resources embedded in occupational networks because (1) relatively little research has examined change in occupational networks over time and (2) these resources are most salient for attainment processes. Social capital resources are examined by using a composite measure of the quantity and quality of occupational networks. We analyze a number of other features of occupational networks: gender, closeness, density, and trust. The age patterns for these occupational network characteristics are compared to changes in daily contact and voluntary association membership.
We anticipate finding that social capital resources vary across the life course. Based on evidence from prior empirical investigations, we expect to find a steady accumulation of resources embedded in occupational networks (McDonald and Elder 2006; Moerbeek and Flap
2008), though the extent of these resources should level out and perhaps even decline at more advanced ages (Lambert, Eby, and Reeves 2006). Research on friendship and social support networks has found the reverse: declining socialization with age (Kalmijn 2003; Wellman,
Wong, Tindall, and Nazer 1997). Therefore, we expect the results to show that age is negatively associated with daily contact and voluntary association memberships. Finally, prior research
11 suggests that men and women display unique patterns of socialization over the life course due to gendered careers (Fischer and Oliker 1983; Han and Moen 1999; Kalmijn 2002). Consequently, we expect to find gender differences in the relationship between social capital resources.
Data and Methodology
We use data from the Social Capital-USA survey. SC-USA is a national, random-digit dialed telephone survey of adults in the U.S. age 22-65 who were currently or previously employed.
The survey was specifically designed to examine social capital and was part of a three-society
(U.S.A., China and Taiwan) study, jointly sponsored by Academia Sinica, Taiwan, and Duke
University. The survey was administered from November of 2004 to March of 2005, averaged 35 minutes to complete, and resulted in 3,000 completed interviews. The response rate (43%) is comparable to other recent national RDD surveys (see Groves, Fowler, Couper, Lepkowski,
Singer, and Tourangeau 2004). To assess the potential bias due to non-response, parameter estimates from SC-USA were compared to those obtained from the March 2005 Current
Population Survey (limited to respondents age 22-65). Few significant differences were found, although SC-USA respondents are more highly educated than the CPS respondents. Therefore, sample weights were constructed (using the rake procedure in STATA) to match the gender, race, age, marital status, and education characteristics of SC-USA to the CPS estimates. The sample weights are employed in the analyses presented here to ensure generalizability. Five cases do not have valid weights due to missingness and are therefore excluded from the analyses.
Measurement
The weighted means for the variables used in this analysis are presented in Table 1. This study focuses on age and gendered patterns of social capital. Age is calculated based on the self- reported year of birth and is reported in years. The sample ranges from age 22-65, with a mean
12 age of about 42 years. In the analyses, we compute squared and cubic age terms to capture curvilinear variation in the dependent variables across age. Gender also serves as an independent variable in the analysis and is measured as a dummy variable (female=1, male=0).
The remaining 8 variables are measures of social network characteristics and serve as the dependent variables for the analyses. Most of the social network variables are drawn from the position generator methodology, which measures the characteristics of occupational networks
(Lin and Dumin 1986; Lin, Fu, and Hsung 2001). The position generator measures quantity and quality of embedded resources in occupational social networks by identifying access to contacts in a list of 22 occupations. Each respondent is asked, “Do you know anyone among your relatives, friends, or acquaintances that has one of the following jobs? (“Knowing” means that you and the person can recognize and greet each other. If you know several persons that have a particular job, please name the person that comes to mind first.)” The list of jobs is designed to capture the scope of positions in the occupational hierarchy: from janitors and hairdressers to lawyers and CEOs. Upon identifying a contact with a particular job, respondents are asked a series of follow-up questions about the characteristics of the position occupant.
We calculate a number of variables based on these questions. First, we calculate a measure of social capital based that taps into the quantity and quality of social resources available in occupational networks. Extensity is a measure of the total number of positions accessed by respondents. The quality of contacts is measured by upper reachability of network contacts, calculated by selecting the highest prestige score of the occupations to which respondents had access (using the Standard International Occupational Prestige Scale). This variable spans from 0 (for respondents who have access to zero positions in the generator) to 85
(for respondents who said they knew a Congressperson). Extensity and upper reachability are
13 highly correlated (r=.62) and are therefore combined into a single social capital index using principle component factor analysis with varimax rotation. The factor score was recoded such that values range from 0 to 5.96, with a mean value of 3.36.
We also include a measure of the proportion of contacts mentioned in the position generator. The variable ranges from 0 to 1 and is equal to the number of male contacts mentioned divided by extensity. Respondents were also asked about how close they are to each person they mentioned. The closeness variable was recoded (such that 1=not close at all, 2=not close, 3=so, so, 4=close, and 5=very close) and the average closeness of all contacts mentioned in the position generator was calculated. The resulting continuous variable ranges from 0 (for individuals with access to zero positions in the generator) to 5 (for people who reported only very close occupational contacts).
Density of occupational networks is gleaned from a question asking, “Among the above people you mentioned [in the position generator], how many know each other?” Higher values on the density variable indicate a greater degree closure among network contacts. This ordinal level variable ranges from a value of 0 (no contacts know each other) to 4 (all contacts know each other). Trust is measured in a similar way. Respondents were asked, “Among the above people you mentioned [in the position generator], how many can be trusted?” The values on the trust variable range from 0 (all can’t be trusted) to 4 (all can be trusted).
We also examine a question on daily contact, which identifies the number of people that each respondent interacts with in a typical day. The variable spans from a value of 1 (0-4 people) to 6 (more than 100) and has been shown to be related to expressive and instrumental network dimensions of other social capital indicators (Fu 2005). A follow up question asked, “Among those [daily contacts], how many of those do you make contact with because of work?”
14 Responses vary from “almost none” (0) to “almost all” (4). Finally, we include a measure of the number of memberships in voluntary organizations, which is a count of up to 10 different types of organizations.
In the subsequent analyses, the sample size depends on the missing values on the dependent variables, since the independent variables (age and gender) contain no missing values.
The only variable with substantial missing data (i.e., greater than 7%) is the work-related contact variable, which is explained by the fact that responses from non-employed individuals were coded as “not applicable.” Therefore, when interpreting the results, it is important to recognize that valid responses refer only to currently employed individuals.
Analysis
The analysis is divided into three steps. First, the relationship between age and social network characteristics is presented graphically to provide a visual representation of how social capital varies across the life course. Second, the form of the relationship between age and social capital is modeled using ordinary least squares regression (for the continuous dependent variables) and ordinal logistic regression (for ordinal dependent variables). Third, analyses are split to identify gender similarity/difference in the form of the relationships between age and social capital. The figures and tables are organized according to these three steps, but we describe all three steps for each dependent variable before moving to the next variable.
Results
Figures 1 and 2 present visual evidence of how social network characteristics vary across different age groups. The first figure (1a) graphs the relationship between age and the social capital factor score. Social capital increases steadily with age, though additional returns to age tend to diminish to the point where social capital peaks (age 46-50) and then drops off among the
15 50+ population. This pattern is generally consistent with our expectations. Interestingly, there is a slight uptick in social capital scores among the oldest age group. This finding might be explained by life course patterns of career renewal. Moen (Moen 2005) notes that careers often involve “second acts” in which people retire from high pressure jobs to take on more flexible enriching work roles. Alternatively, the higher levels of social capital could be reflective of generational differences, reflecting the greater overall civic involvement of the World War II
“greatest generation” (Mettler 2007).
The functional form of the relationship is modeled using ordinary least squares regression
(see Table 2). The best fitting model includes both a linear age term and a squared age term. The significant positive age effect captures the pattern of social capital accumulation with age, while the significant negative age-squared effect highlights the diminishing returns to age among the older respondents. The summary from Table 3 shows that this pattern holds for both males and females in the sample. This is illustrated by the positive linear age term, negative squared age term, and insignificant cubic age term.
The relationship between age and the proportion of male contacts is displayed in Figure
1b. Again, the results reveal a curvilinear relationship: initial steep decline in male contacts followed by a steady increase to the oldest age group. Overall, this evidence is consistent with what one might expect to find. Female dominated occupations (e.g., service and clerical) are common in the early part of careers, while increasing occupational attainment may lead individuals to male dominated workplaces, which offer greater wages and benefits (not to mention male contacts). However, the pattern could also reflect changing labor market conditions over time. For example, the decline in male dominated manufacturing jobs, the rise in the female-dominated service sector, and increasing female labor force participation could all
16 contribute to lower proportions of male contacts in occupational networks among younger cohorts. The relatively high proportion of male contacts among young adults age 22-26 is more difficult to explain. This may be related to gender differences in labor force participation or in networking behavior among this age group. The best fitting form of this relationship, though, is a positive linear relationship that captures the steady increase in male contacts with age. The results in Table 3, however, show that this relationship is significant for women only. For men, the relationship between age and proportion of male contacts is statistically insignificant.
The closeness of contacts tends to diminish over time (see Figure 1c). After an initial rise in the closeness of contacts from 22-26 to 27-30, closeness declines and tends to level off around age 46-50. This suggests a general weakening of occupational ties across the life course. As individuals gain greater work experience, their occupational networks expand and diversify, resulting in broader yet more weakly tied networks. The significant negative linear age effect best captures this pattern of declining closeness across age groups (see Table 2). The relationship between age and closeness, however, is conditioned by gender. Only men experience this significant decline in closeness across the life course.
Patterns of network density across age are similar to closeness. Figure 1d shows a general decline in density, with a particularly steep drop between ages 27 to 40. The negative relationship between network density and age provides further evidence of the accumulation of social capital resources across the life course. Dense networks lack structural holes (Burt 1992), which are useful for procuring non-redundant information and for brokering exchanges.
Furthermore, the similarity between the closeness and density graphs is consistent with what one might expect. Both factors are empirically associated and strength of ties has often been used as a proxy measure of network density (Burt 1992). Using ordinal logistic regression, we find that
17 the best fit for the form of the relationship between age and density uses both linear and squared age variables. The significant negative effect captures the rapid decline in density in the first half of the work career, whereas the positive squared term illustrates the leveling off of the effects during the second half of the work career. The gender-specific models (see Table 3) indicate that the curvilinear relationship is best for men only. For women, density declines across the career with less leveling at the end of the career than men.
Figure 2a shows that, in general, older respondents are more likely than younger respondents to perceive their occupational contacts as trustworthy. This pattern could reflect a social process by which individuals develop greater trust in their contacts over time or that individuals eventually rid their networks of untrustworthy contacts. The increase in network trust across age categories provides further support for the notion that social resources tend to accumulate across the work career. However, the increase in trust is not consistent across all age groups. There is an early rise in trust from age 27 to 45, followed by a sharp drop in trust for the
46-50 age group, and then a steady rise in trust again among the oldest age groups. The sharp drop in the levels of trust at age 46 might be reflective of a historical shift in trust, but the trend does not mesh with prior assertions that trust in others has declined in successive cohorts
(Putnam 2002), which would predict that the later cohorts would have substantially higher levels of trust in their network contacts than earlier contacts. The linear pattern of increasing trust across age is confirmed with ordinal logistic regression (see Table 3). The results show that, on average, each additional year is associated with higher levels of trust in occupational networks.
Splitting out the results by gender, however, shows that increasing trust with age is significant for women, but not for men.
18 The results also show a general decline in the amount of daily contact in which people engage, though again the pattern is not entirely linear (see Figure 2b). The highest levels of daily contact are among the youngest age group. Daily contact dips among 27 to 35 year olds, then rises among 36 to 40 year olds, and steadily declines thereafter. Using OLS regression, the age pattern of daily contact is best captured by a negative linear term for age (see Table 3). The overall decline in daily contact is consistent with prior research that shows a general decline in network contacts over time (Kalmijn 2003). Further analyses summarized in Table 4 reveal gender differences in the form of the relationship between age and daily contact. For men, the daily contact relationship is best described as linear. For women, the best fitting model contains a negative linear age term, a positive squared age term, and a negative cubic age term (all of which are statistically significant). This model predicts an S-curve characterized by an initial decline, then increase, and finally a decline in daily contact across age. This suggests greater fluctuation in the daily contact of women across the life course.
Age patterns of daily social interaction are distinct from changes in occupational network characteristics over time. Indeed, Figure 2c tracks the proportion of daily contact that is work- related and shows a pattern that is more consistent with what was found for the variables constructed from the position generator. That is, the proportion of work-related contact tends to increase with age and levels off around age 46. In spite of the apparent diminishing returns, the ordinal logistic regression model with a single linear age variable produced the only statistically significant coefficients. However, both men and women display similar linear increases in work- related contact across age.
Finally, we examine age variation in the number of voluntary organization memberships.
Here the pattern with steady increases in memberships early in the life course, diminishing
19 returns and a reversal of the trend during the middle years, followed by a slight uptick in memberships among the oldest age group. This increase in voluntary association membership is consistent with recent evidence that shows an increase in community participation and volunteering among older individuals (Cornwell, Laumann, and Schumm 2008). Furthermore, the pattern is strikingly similar to what was found for social capital. Results from OLS regression show that the relationship is best described with a positive linear age term and a negative squared age term. This captures the parabolic form of the relationship between age and organizational memberships. Further elaboration by gender in Table 4 shows that the relationship is similar for both men and women.
Discussion
Drawing from nationally representative U.S. data, we investigate age variation in social capital and the extent to which these patterns may be conditioned by gender. As such, this paper provides a number of important contributions to the research literature. First, we apply a life course perspective to the study of change in social capital over time. This study demonstrates the importance of embedding studies of network change within the context of individual biographies. The results show that social capital resources indeed vary by age, as each of the eight social resource indicators is significantly associated with age. Second, we employ a multidimensional approach to examining social capital. Third, this research fills a gap in knowledge about age-related patterns of occupational networks and compares these patterns to changes in other forms of socialization (daily contact and voluntary association memberships). In general, occupational network resources tend to accumulate as individuals age, but have a tendency to taper off (and sometimes diminish) at older ages. The same is true for voluntary association memberships. However, daily contact tends to decline across the life course. Finally,
20 the results explore the moderating influence of gender on age-based patterns of social capital.
There remains some consistency in the relationships between age and the social network measures by gender, but gender divergence in these relationships is more common.
The analysis presented here was designed to overcome some of the limitations of previous research literature on network dynamism, but retains a number of critical limitations.
Most importantly, this study relies on cross-sectional data, which does not allow for the separation of age, period, and cohort effects (Glenn 2003). Our own interpretation of the age- related patterns emphasizes aging processes by which resources tend to accumulate or dissolve, but alternative interpretations cannot be ruled out. The greatest impediment to understanding variation in social capital across the life course is a lack of longitudinal data on social networks.
The data presented comes from the first wave of a larger panel study. As these and other panel datasets become available, researchers should explore these empirical findings in greater detail to distinguish aging process from changes in historical circumstances and enhance understanding of these life course processes.
The findings presented here are primarily descriptive and therefore provide many additional opportunities for further elaboration. First, age-based patterns of social capital are more orderly and consistent during the middle years and are more erratic earlier and later in adult lives. This presumably reflects the transitional character of these times in people’s lives. The youth labor market involves considerable “churning” or “milling about” (Gardecki and Neumark
1998) as individuals complete education and attempt to settle on a career. The late career period also contains numerous transitions out of primary careers and into retirement, secondary careers, and volunteering (Moen 2005). Relatively little is known about the role that social capital plays in these particular transitions. Second, the findings presented here reveal a striking congruence
21 between age-related patterns of social capital and those of voluntary association memberships, which seems to suggest that participation in voluntary associations may be causally linked to the expansion of occupational networks. Recent empirical analyses have found little support for an overarching causal relationship between social capital and association membership (Song 2008), but it remains possible that the relationship is contingent on timing. In other words, voluntary associations may be critical for developing diverse occupational contacts for older individuals only, helping to explain the uptick in social capital among the 56-65 age group in the SC-USA sample. Third, the gender differences in age-related patterns of social capital identified in this paper were not examined in any real detail. Future research should examine how family and work events/histories contribute to gender-specific social capital trajectories. Finally, future studies should not only clarify trajectories of social capital accumulation across the life course, but also examine how these trajectories are associated with other forms of resource accumulation. Social capital likely serves as a social mechanism that reproduces social inequality
(cf., McDonald, Lin, and Ao 2008) by contributing to broader patterns of cumulative advantage and disadvantage across the life course.
22 References
Ajrouch, Kristine J., Alysia Y. Blandon, and Toni C. Antonucci. 2005. "Social Networks Among
Men and Women: The Effects of Age and Socioeconomic Status." Journal of
Gerontology 60B:S311-S317.
Alon, Sigal and Yitchak Haberfeld. 2007. "Labor Force Attachment and the Evolving Wage Gap
Between White, Black, and Hispanic Young Women." Work and Occupations 34:369-
398.
Antonucci, T. C. and H. Akiyama. 1987. "Social Networks In Adult Life And A Preliminary
Examination Of The Convoy Model." Journals Of Gerontology 42:519-527.
Bidart, C. and D. Lavenu. 2005. "Evolutions of personal networks and life events." Social
Networks 27:359-376.
Blau, Peter M. 1963. Exchange and Power in Social Life. New York: Wiley.
Bourdieu, Pierre. 1986. "The Forms of Capital." Pp. 241-58 in Handbook of Theory and
Research for the Sociology of Education, edited by J. G. Richardson. Westport:
Greenwood Press.
Bridges, William P. and Wayne J. Villemez. 1986. "Informal Hiring and Income in the Labor
Market." American Sociological Review 51:574-82.
Burt, R. S. 2000. "Decay functions." Social Networks 22:1-28.
—. 2002. "Bridge decay." Social Networks 24:333-363.
Burt, Ronald S. 1992. Structural Holes. Cambridge, MA: Harvard University Press.
Cairns, Robert B. and Man-Chi Leung. 1995. "Friendships and Social Networks in Childhood
and Adolescence: Fluidity, Reliability, and Interrelations." Child Development
Child Development J1 - Child Development 66:1330-1345.
23 Cantin, S. and M. Boivin. 2004. "Change and stability in children's social network and self-
perceptions during transition from elementary to junior high school." International
Journal Of Behavioral Development 28:561-570.
Coleman, James S. 1990. Foundations of Social Theory. Cambridge: Harvard University Press.
Cornwell, Benjamin, Edward O. Laumann, and L. Phillip Schumm. 2008. "The Social
Connectedness of Older Adults: A National Profile." American Sociological Review
73:185-203.
Crosnoe, R. 2000. "Friendships in childhood and adolescence: The life course and new
directions." Social Psychology Quarterly 63:377-391.
Crosnoe, R., M. K. Johnson, and G. H. Elder. 2004. "Intergenerational bonding in school: The
behavioral and contextual correlates of student-teacher relationships." Sociology Of
Education 77:60-81.
Degenne, A. and M. O. Lebeaux. 2005. "The dynamics of personal networks at the time of entry
into adult life." Social Networks 27:337-358.
DiPrete, Thomas A. and Gregory M. Eirich. 2006. "Cumulative Advantage as a Mechanism for
Inequality: A Review of Theoretical and Empirical Developments." Annual Review of
Sociology 32.
Doreian, P. 2002. "Event sequences as generators of social network evolution." Social Networks
24:93-119.
—. 2006. "Actor network utilities and network evolution." Social Networks 28:137-164.
Dupre, Matthew E. 2007. " Educational Differences in Age-Related Patterns of Disease:
Reconsidering the Cumulative Disadvantage and Age-As-Leveler Hypotheses." Journal
of Health and Social Behavior 48:1-15.
24 Elder, Glen H., Jr. and Michael J. Shanahan. 2006. "The Life Course and Human
Development." Pp. 665-715 in The Handbook of Child Psychology, vol. 1, edited by W.
Damon and R. M. Lerner. New York: Wiley.
Elman, Cheryl and Angela M. O'Rand. 2004. "The Race is to the Swift: Socioeconomic Origins,
Adult Education, and Wage Attainment." American Journal of Sociology 110:123-160.
Emirbayer, Mustafa. 1997. "Manifesto for Relational Sociology." American Journal of Sociology
103:281-317.
Fischer, Claude S. and Stacey J. Oliker. 1983. "A Research Note on Friendship, Gender, and the
Life Cycle." Social Forces 62:124-132.
Fu, Yang-chih. 2005. "Measuring Personal Networks with Daily Contacts: A Single-item Survey
Question and the Contact Diary." Social Networks 27:169-186.
Fuller, Sylvia. 2008. "Job Mobility and Wage Trajectories for Men and Women in the United
States." American Sociological Review 73:158-183.
Gardecki, Rosella and David Neumark. 1998. "Order from Chaos? The Effects of Early Labor
Market Experiences on Adult Labor Market Outcomes." Industrial and Labor Relations
Review 51:299-322.
Glenn, Norval D. 2003. "Distinguishing Age, Period, and Cohort Effects." Pp. 465-476 in
Handbook of the Life Course, edited by J. T. Mortimer and M. J. Shanahan. New York:
Kluwer Academic/Plenum Publishers.
Granovetter, Mark S. 1973. "The Strength of Weak Ties." American Journal of Sociology
78:1360-1380.
25 —. 2003. "A Theoretical Agenda for Economic Sociology." Pp. 35-78 in The New Economic
Sociology: Developments in an Emerging Field, edited by M. F. Guillen, R. Collins, P.
England, and M. Meyer. New York: Sage.
Groves, Robert M., Floyd Jr. Fowler, Mick Couper, James Lepkowski, Eleanor Singer, and
Roger Tourangeau. 2004. Survey Methodology. New York: Wiley & Sons.
Han, S. K. and P. Moen. 1999. "Work and family over time: A life course approach." Annals Of
The American Academy Of Political And Social Science 562:98-110.
Heuvel, Audrey Vanden. 1988. "The Timing of Parenthood and Intergenerational Relations."
Journal of Marriage and the Family 50:483-491.
Hofferth, Sandra L. 1984. "Long-Term Economic Consequences for Women of Delayed
Childbearing and Reduced Family Size." Demography 21:141-155.
Hummon, N. P. 2000. "Utility and dynamic social networks." Social Networks 22:221-249.
Hummon, N. P. and P. Doreian. 2003. "Some dynamics of social balance processes: bringing
Heider back into balance theory." Social Networks 25:17-49.
Kalmijn, M. 2003. "Shared friendship networks and the life course: an analysis of survey data on
married and cohabiting couples." Social Networks 25:231-249.
Kalmijn, Matthijs. 2002. "Sex Segregation of Friendship Networks: Individual and Structural
Determinants of Having Cross-Sex Friends." European Sociological Review 18:101-117.
Kilduff, Martin and Wenpin Tsai. 2003. Social Networks and Organizations. Thousand Oaks,
CA: Sage.
Kim, Doo Hwan and Barbara Schneider. 2005. "Social Capital in Action: Alignment of Parental
Support in Adolescents' Transition to Postsecondary Education." Social Forces 84:1181-
1206.
26 Lambert, T. A., L. T. Eby, and M. P. Reeves. 2006. "Predictors of networking intensity and
network quality among white-collar job seekers." Journal Of Career Development
32:351-365.
Lin, Nan. 2000. "Inequality in Social Capital." Contemporary Sociology 29:785-795.
—. 2001. Social Capital: A Theory of Social Structure and Action. Cambridge: Cambridge
University Press.
Lin, Nan and Mary Dumin. 1986. "Access to Occupations through Social Ties." Social Networks
8:365-385.
Lin, Nan, Yang-chih Fu, and Ray-May Hsung. 2001. "The Position Generator: Measurement
Techniques for Investigations of Social Capital." Pp. 57-81 in Social Capital: Theory and
Research, edited by N. Lin, K. Cook, and R. S. Burt. New York: De Gruyter.
Martin, John Levi and King-To Yeung. 2006. "Persistence of Close Personal Ties Over a 12-year
Period." Social Networks.
Maume, David J. Jr. 2004. "Wage Discrimination over the Life Course: A Comparison of
Explanations." Social Problems 51:505-527.
McDonald, Steve and Glen H. Elder, Jr. 2006. "When Does Social Capital Matter? Non-
Searching for Jobs Across the Life Course." Social Forces 85:521-550.
McDonald, Steve, Lance D. Erickson, Monica Kirkpatrick Johnson, and Glen H. Elder. 2007.
"Informal mentoring and young adult employment." Social Science Research 36:1328-
1347.
McDonald, Steve, Nan Lin, and Dan Ao. 2008. "Networks of Opportunity: Gender, Race, and
Job Leads." Unpublished manuscript, North Carolina State University.
27 McPherson, M., L. Smith-Lovin, and M. E. Brashears. 2006. "Social isolation in America:
Changes in core discussion networks over two decades." American Sociological Review
71:353-375.
McPherson, Miller, Lynn Smith-Lovin, and James M. Cook. 2001. "Birds of a Feather:
Homophily in Social Networks." Annual Review of Sociology 27:415-444.
Mettler, Suzanne. 2007. Soldiers to Citizens: The G.I. Bill and the Making of the Greatest
Generation: Oxford University Press.
Mills, C. Wright. 1959. The Sociological Imagination. London: Oxford University Press.
Moen, Phyllis. 2005. "Beyond the Career Mystique: "Time in," "Time Out," and "Second Acts"."
Sociological Forum 20:189-208.
Moerbeek, Hester and Henk Flap. 2008. "Social Resources and their effect on occupational
attainment through the life course." in Social Capital: An International Research
Program, edited by N. a. B. E. Lin. Oxford: Oxford University Press.
O'Rand, Angela M. 2006. "Stratification and the Life Course: Life Course Capital, Life Course
Risks, and Social Inequality." Pp. 146-162 in Handbook of Aging and the Social
Sciences, edited by R. H. Binstock, L. K. George, S. J. Cutler, J. Hendricks, and J. H.
Schultz. Burlington, MA: Academic Press.
Putnam, Robert. 2002. Bowling Alone: The Collapse and Revival of American Community. New
York: Simon & Schuster.
Rhee, Mooweon. 2007. "The Time Relevance of Social Capital." Rationality and Society 19:367-
389.
Schuman, Howard and Jacqueline Scott. 1989. "Generations and Collective Memories."
American Sociological Review 54:359.
28 Sewell, William H., Archibald O. Haller, and George W. Ohlendorf. 1970. "The Educational and
Early Occupational Status Attainment Process: Replication and Revision." American
Sociological Review 35:1014-1027.
Smith-Doerr, Laurel and Walter W. Powell. 2005. "Networks and Economic Life." Pp. 379-402
in Handbook of Economic Sociology, edited by N. J. Smelser and R. Swedberg. Thousand
Oaks, CA: Sage.
Smith, Sandra S. 2005. "'Don't Put My Name on It': Social Capital Activation and Job-finding
Assistance among the Black Urban Poor." American Journal of Sociology 111:1-57.
Snijders, Tom A. B. 2001. "The statistical evaluation of social network dynamics." Pp. 361-395
in Sociological Methodology 2001, Vol 31, vol. 31, Sociological Methodology.
Song, Lijun. 2008. "A Tale of Two Concepts of Social Capital: Network Resources and Civic
Networks." in Southern Sociological Society annual meeting. Richmond, VA.
Stack, Carol B. 1974. All Our Kin: Strategies for Survival in a Black Community. New York:
Harper & Row.
Verbrugge, Lois M. 1977. "The Structure of Adult Friendship Choices." Social Forces 56:576-
597.
Weber, Max. [1922]1968. "Economy and Society." vol. 1, edited by G. Roth and C. Wittich.
Berkeley, CA: University of California Press.
Wellman, Barry, Renita Yuk-lin Wong, David Tindall, and Nancy Nazer. 1997. "A Decade of
Network Change: Turnover, Persistence and Stability in Personal Communities." Social
Networks 19:27-50.
Wenger, G. C. and D. Jerrome. 1999. "Change and stability in confidant relationships: Findings
from the Bangor Longitudinal Study of Ageing." Journal Of Aging Studies 13:269-294.
29 Willson, Andrea E., Kim M. Shuey, and Glen H. Elder, Jr. 2007. "Cumulative Advantage
Processes as Mechanisms of Inequality in Life Course Health." American Journal of
Sociology 112:1886-1924.
30 Table 1. Descriptive Statistics Variable Description Mean Range N Age In years 42.38 22-65 2995 Gender Female=1, Male=0 .51 0-1 2995 Social capital Factor score combining number of contacts with 3.36 0-6 2995 highest prestige from position generator (PG) Male contacts Proportion of male contacts in PG .49 0-1 2995 Closeness Closeness of contacts in PG 3.43 0-5 2995 Density Proportion of contacts in PG that know each other 1.82 0-4 2812 0=None know each other 1=Only a few know each other 2=About half know each other 3=Most know each other 4=All know each other Trust Trust in contacts in PG 3.23 0-4 2904 0=All can’t be trusted 1=Most can’t be trusted 2=Some can be trusted 3=Most can be trusted 4=All can be trusted Daily contact Number of people R makes contact with each day 2.30 0-5 2995 0=0-4 1=5-9 2=10-19 3=20-49 4=50-99 5=100+ Work contact Proportion of daily contact that is work-related 2.74 0-4 2652 0=Almost none 1=Few 2=About half 3=Most 4=Almost all Organizational Number of organizational memberships 1.73 0-10 2995 memberships
31 Table 2. Multiple Regression of Social Capital, Male Contacts, Closeness, and Density on Age Social capital Male contacts Closeness Density OLS OLS OLS Ordinal logistic Age .078 *** .001 * -.006 *** -.080 ** (.016) (.001) (.002) (.029) Age squared -.0008 *** .0008 * (.0002) (.0003) Constant 1.579 *** .435 *** 3.694 *** (.332) (.023) (.086) Cut1 constant -3.772 *** (.626) Cut2 constant -1.998 ** (.618) Cut3 constant -1.161 (.616) Cut4 constant .018 (.618) N 2995 2995 2995 2812 R-squared .021 .003 .006 .004 * p<0.05, **p<.01, ***p<.001
32 Table 3. Multiple Regression of Trust, Daily Contact, Work Contact, and Organizational Memberships on Age Trust Daily contact Work contact Org. memberships Ordinal logistic OLS Ordinal logistic OLS Age .008 * -.008 ** .018 *** .123 *** (.004) (.003) (.004) (.023) Age squared -.0013 *** (.0003) Constant 2.648 *** -.963 * (.137) (.472) Cut1 constant -3.619 *** -1.494 *** (.223) (.194) Cut2 constant -2.673 *** -.600 ** (.188) (.183) Cut3 constant -1.165 *** .135 (.170) (.184) Cut4 constant .445 ** 1.177 *** (.168) (.187) N 2904 2995 2652 2995 R-squared .001 .004 .004 .014 * p<0.05, **p<.01, ***p<.001
33
Table 4. Summary of Gender-specific Regressions of Social Network Characteristics on Age Social capital Male contacts Closeness Density Male Female Male Female Male Female Male Female Age + + n.s. + – n.s. – – Age squared – – n.s. n.s. n.s. n.s. + n.s. Age cubed n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s.
Trust Daily contact Work contact Org. memberships Male Female Male Female Male Female Male Female Age n.s. + – – + + + + Age squared n.s. n.s. n.s. + n.s. n.s. – – Age cubed n.s. n.s. n.s. – n.s. n.s. n.s. n.s.
34
3.6 .54
l .52 s t a 3.4 t c i a p t
a .50 n c o
l 3.2 c a i
e .48 c l a o S
3.0 M .46
2.8 .44 22- 27- 31- 36- 41- 46- 51- 56- 22- 27- 31- 36- 41- 46- 51- 56- 1A 26 30 35 40 45 50 55 65 1B 26 30 35 40 45 50 55 65
3.7 2.2 2.1 3.6 s
s 2.0 y t e i n s e
3.5 n 1.9 s e o l D 1.8 C 3.4 1.7 3.3 1.6 22- 27- 31- 36- 41- 46- 51- 56- 22- 27- 31- 36- 41- 46- 51- 56- 1C 26 30 35 40 45 50 55 65 1D 26 30 35 40 45 50 55 65
Figure 1. Social capital, male contact, closeness, and density by age
35
3.4 2.5
2.4 t c a
3.3 t t
n 2.3 s o u c r
T y
l 2.2 i
3.2 a D 2.1
3.1 2.0 22- 27- 31- 36- 41- 46- 51- 56- 22- 27- 31- 36- 41- 46- 51- 56- 2A 26 30 35 40 45 50 55 65 2B 26 30 35 40 45 50 55 65
3.0 2
t
c s
a 2.9 p t
i 1.8 n h
o 2.8 s r c
e 1.6 d
2.7 b e t m a e l 2.6 1.4 e m r
- 2.5 . k g r r 1.2 o
2.4 O W 2.3 1 22- 27- 31- 36- 41- 46- 51- 56- 22- 27- 31- 36- 41- 46- 51- 56- 2C 26 30 35 40 45 50 55 65 2D 26 30 35 40 45 50 55 65
Figures 2. Trust, daily contact, work-related contact, and organizational membership by age
36 Access to and Use of Social Capital among Married Couples in Hong Kong
Gina Lai and Danching Ruan
Department of Sociology Hong Kong Baptist University
May 30, 2008
Presentation at the International Social Capital Conference, Taipei, Taiwan, May 29-30, 2008.
This research is supported by a grant from the Research Grants Council of Hong Kong Special Administrative Region (Project No. HKBU2032/02H). Research Background
1. Marriage and network change
a. Network integration
• Bridging function of marital partner and cross-network linkages
• Affective interdependence and mutual friendship
• Benefits • Marital identity • Marital satisfaction • Marital stability • Enhanced access to social capital
b. Network restructuring • Tendency towards gender homogeneity of social ties
2. Network structure and social capital
a. Network diversity (Erickson, Lin)
b. Importance of male ties (Erickson) Research Issues
1. How might the network changes induced by marital coupling influence the access to spouse’s social capital?
• Getting into the spouse’s social network + access?
• Change in the gender composition of social ties
2. Would the access to spouse’s social capital promote the use of it?
3. Any gender differences? Data
1. Sample
801 Chinese married couples residing in Hong Kong at the time of the survey
2. Sampling method: Two-stage random sampling
a. Households
b. Married couples
3. Data collection method
a. Separate face-to-face interviews with husbands and wives
b. July to September, 2005 Socioeconomic Characteristics of Couples.
Individual Characteristics Husbands Wives Age (mean years)*** 43.97 40.62 Born in Hong Kong (%)** 75.20 68.40 Education (%)** Junior high or below 44.30 45.70 Senior high or non-degree courses 44.30 47.60 University or above 11.40 6.70 Employed (%)*** 93.20 56.70 Religion (%) No religion 64.90 62.90 Chinese religion / Buddhism 28.20 27.80 Other religion 6.90 9.20
* p<.05; ** p<.01; *** p<0.001 Spousal Characteristics Age difference (mean years) 3.71 Birth place (%) Both born in Hong Kong 62.90 Both not born in Hong Kong 19.40 Only husband born in Hong Kong 39.10 Only wife born in Hong Kong 5.50 Education (%) Same 54.20 Husband > wife 27.60 Husband < wife 18.20 Employment status (%) Both employed 54.20 Both non-employed 4.20 Only husband employed 39.10 Only wife employed 2.50 Religion (%) Same religion 27.70 Both no religion 57.30 Different religions (including cases in which one partner has no religion) 15.00 Studied in the same school before 7.20 Worked in the same firm before 19.20 Marriage Characteristics Length of marriage (%) 5 years or less 16.90 6-10 years 17.90 11-15 years 16.40 16-20 years 17.50 21-25 years 18.00 More than 25 years 13.40 Duration of marriage (mean year) 15.74 Number of children (mean) 1.65 Children aged 12 or under (%) 44.19 Getting into Spouse’s Social Network: Kin Ties.
Kin Husbands Wives Acquaintance with spouse’s kin (%) Know almost all of them 28.00 28.40 Know most of them 44.90 45.90 Know about half of them 23.10 21.80 Know very few of them 3.60 3.60 Hardly know any of them 0.40 0.30 Contact with kin (%)*** Mainly with husband’s kin 22.30 9.40 Mainly with wife’s kin 13.80 29.60 50/50 63.90 61.10 Joint participation in activities with spouse’s kin (%) Often 17.60 20.00 Sometimes 57.50 54.00 Seldom 23.80 23.40 Never 1.10 2.60
* p<.05; ** p<.01; *** p<0.001 Getting into Spouse’s Social Network: Friendship Ties.
Friends Husbands Wives Acquaintance with spouse’s friends (%)** Know almost all of them 7.80 6.60 Know most of them 41.50 42.80 Know about half of them 37.50 31.40 Know very few of them 11.40 16.40 Hardly know any of them 1.80 2.90 Contact with spouse’s friends (%) Often 0.50 1.10 Sometimes 17.10 16.30 Seldom 38.50 38.40 Never 40.60 42.10 No acquaintance at all 3.30 2.10
* p<.05; ** p<.01; *** p<0.001 Changes in Personal Network.
Network Change Husbands Wives Network size (%) Increased a lot 4.40 4.50 Increased a bit 26.90 30.50 No change 56.70 52.10 Decreased a bit 10.00 11.10 Decreased a lot 2.00 1.80 Number of male friends (%)*** Increased a lot 1.60 1.00 Increased a bit 25.50 18.90 No change 63.00 61.10 Decreased a bit 8.50 16.10 Decreased a lot 1.40 2.90 Number of female friends (%)*** Increased a lot 0.50 2.50 Increased a bit 21.10 30.80 No change 63.60 58.10 Decreased a bit 12.20 7.40 Decreased a lot 2.60 1.30 Measurement of Social Capital
1. A modified version of Lin’s Position Generator suitable for the local context
2. A list of 15 occupations spanning across the structural hierarchy
a. Whether they know a kin, friend, and/or an acquaintance in the occupation
b. Whether they know a man and/or woman in the occupation
c. Whether they know the person(s) directly or indirectly through their spouse
3. Network diversity: Number of social positions reached Access to Social Capital.
Access to Social Positions (ISEI) (%) Husbands Wives Secondary school teacher (69)** 38.08 44.57 Electrician (40)*** 72.91 58.30 Medical doctor (88) 45.44 48.56 Domestic helper (16)*** 63.17 72.91 Functional manager (61)*** 58.18 49.44 Accountant (69) 30.34 29.46 Waiter/waitress (34) 64.17 67.54 Salesperson (43)*** 66.17 74.91 Construction worker (30)*** 61.92 41.20 Government administrator (77) 13.61 14.86 Police (50) 37.08 33.08 Engineer (73)* 26.47 21.47 Street vendor (29) 48.69 51.44 Lawyer (85) 20.10 16.35 Clerk (45) 83.02 82.52
* p<.05; ** p<.01; *** p<0.001 Access to Social Capital.
Access to Social Positions Husbands Wives Number of positions reached (mean) 7.29 7.07 Number of positions reached through spouse (mean) 0.83 0.91 Percent of positions reached through spouse (%) 11.51 13.18 Number of positions reached through male ties (mean)*** 5.83 4.98 Number of positions reached through female ties (mean)*** 3.91 4.53 Percent of positions reached through male ties (%)*** 79.64 67.29 Percent of positions reached through female ties (%)*** 52.37 66.08 Highest occupational prestige (ISEI) reached (mean) 73.49 74.57 Highest occupational prestige (ISEI) reached through male ties (mean) 67.73 66.24 Highest occupational prestige (ISEI) reached through female ties (mean) 52.84 55.20 Highest occupational prestige (ISEI) reached through male and 55.84 58.31 female ties (mean)* Highest occupational prestige (ISEI) reached directly (mean) 72.27 72.90 Highest occupational prestige (ISEI) reached through spouse (mean) 56.41 55.19
* p<.05; ** p<.01; *** p<0.001 Predicting the Access to Social Capital through Spouse
1. Duration of marriage
2. Parental status
3. Relative status a. Education b. Employment status c. Birthplace d. Religion e. Age gap
4. Getting into spouse’s social network a. Acquaintance with spouse’s kin b. Contact with kin c. Joint activities with spouse’s kin d. Acquaintance with spouse’s friends e. Contact with spouse’s friends
5. Change in the numbers of male and female friends Regression of Number of Social Positions Reached through Spouse.
Independent Variables Husbands Wives Education (ref=both senior high) Both junior high or below -.09 (-.03) .05 (.02) Both university .94 (.15)*** 1.03 (.16)*** Respondent > spouse .33 (.09)* -.002 (.000) Respondent < spouse .15 (.03) .44 (.12)** Contact with kin (ref=mainly with own kin) Mainly with spouse’s kin -.04 (-.01) -.14 (-.03) 50/50 -.33 (-.12)** -.19 (-.07) Birthplace (ref=both born in Hong Kong) Only respondent born in Hong Kong -.31 (-.07)* -.06 (-.01) Only spouse born in Hong Kong -.18 (-.03) -.15 (-.03) Both not born in Hong Kong -.30 (-.08)* -.18 (-.05) Acquaintance with spouse’s kin .10 (.06) .19 (.11)** Joint participation in activities with spouse’s kin .28 (.14)*** -.01 (-.01) Acquaintance with spouse’s friends .22 (.14)*** .26 (.17)*** Increase in the number of male friends -.16 (-.07) .26 (.14)*** Increase in the number of female friends .23 (.11)** -.10 (-.05) Constant -1.34** -1.31** R2 .15 .13 Adjusted R2 .12 .10 N 790 794 Regression of Proportion of Social Positions Reached through Spouse.
Independent Variables Husbands Wives Education (ref=both senior high) Both junior high or below .02 (.04) .03 (.08) Both university .11 (.13)*** .14 (.15)*** Respondent > spouse .04 (.07) .02 (.04) Respondent < spouse .04 (.07) .06 (.11)** Employment status (ref=both employed) Only respondent employed -.03 (-.08)* .01 (.01) Only spouse employed .01 (.01) .01 (.01) Both non-employed .04 (.04) .03 (.03) Birthplace (ref=both born in Hong Kong) Only respondent born in Hong Kong -.04 (-.07) .001 (.001) Only spouse born in Hong Kong -.04 (-.05) .003 (.004) Both not born in Hong Kong -.04 (-.09)* -.01 (-.03) Acquaintance with spouse’s kin .02 (.09)* .03 (.13)*** Joint participation in activities with spouse’s kin .03 (.12)** -.004 (-.01) Acquaintance with spouse’s friends .02 (.09)* .04 (.16)*** Increase in the number of male friends -.02 (-.08)* .03 (.12)** Increase in the number of female friends .02 (.07) -.03 (-.09)* Constant -.14** -.19*** R2 .11 .12 Adjusted R2 .09 .09 N 789 787 Getting Help from Spouse’s Social Ties in the Past Year
Sources of Help Men Women
Family and relatives .18 .19
Friends .13 .12 Predicting the Likelihood of Getting Help from Spouse’s Social Ties
1. Duration of marriage
2. Parental status
3. Relative status a. Education b. Employment status c. Birthplace d. Religion e. Age gap
4. Getting into spouse’s social network a. Acquaintance with spouse’s kin b. Contact with kin c. Joint activities with spouse’s kin d. Acquaintance with spouse’s friends e. Contact with spouse’s friends
5. Change in the numbers of male and female friends
6. Number and proportion of social positions reached through spouse Logistic Regression of Getting Help from Spouse’s Kin Ties in the Past Year
Independent Variables Husbands Wives Education (ref=both senior high) Both junior high or below -.60* -.44 Both university -.24 -.004 Respondent > spouse -.58 .52 Respondent < spouse .18 .35 Religion (ref=both no religion) Both Chinese traditional Chinese religion/Buddhism .58* .26 Both other religion -.49 .25 Different religions -.38 -.20 Joint participation in activities with spouse’s kin .28 .44** Contact with kin (ref=mainly with own kin) Mainly with spouse’s kin -.42 .89** 50/50 .10 .28 Contact with spouse’s friends .45** .11 Logistic Regression of Getting Help from Spouse’s Friends in the Past Year
Independent Variables Husbands Wives Employment status (ref=both employed) Only respondent employed -.18 -.90 Only spouse employed -.46 -.89* Both non-employed .33 -.10 Birthplace (ref=both born in Hong Kong) Only respondent born in Hong Kong .07 -.57 Only spouse born in Hong Kong -.82 .08 Both not born in Hong Kong -.41 -.81* Religion (ref=both no religion) Both Chinese traditional Chinese religion/Buddhism .49 1.03** Both other religion .25 -.01 Different religions -.17 .01 Contact with kin (ref=mainly with own kin) Mainly with spouse’s kin -.37 .63 50/50 .39 .69* Acquaintance with spouse’s friends .44** .32 Contact with spouse’s friends .35* 1.25*** Increase in the number of male friends -.07 .61** Increase in the number of female friends .03 -.60* Summary of Findings
1. Both men and women are able to get into their spouse’s social network.
a. Men tend to know more of their spouses’ friends than do women.
a. But respondents are more likely to have contact with their own kin ties than their spouses’ kin.
b. Respondents also tend to have limited contact with their spouses’ friends.
2. There is no gender difference in the access to social capital, directly or indirectly through spouse.
3. Getting into the spouse’s social network would facilitate the reach of social capital through spouse.
4. Change in the gender composition of social ties due to marriage brings differential impacts on the access to social capital through spouse for men and women.
5. Network integration would facilitate the use of resources embedded in spouse’s social network, but the access to social capital through spouse does not encourage the use of it. Concluding Remarks
1. Marriage tends to promote the sharing of social capital between husband and wife, particularly among better-educated couples.
2. Due to the gendered change in network composition associated with marriage, women may have a greater reliance on marriage for access to social capital than do men.
3. Access to social capital through spouse does not seem to generate a tendency to use it.
a. Personal networks resourceful enough
b. Difficulties in mobilizing spouses’ social ties
c. Proximity or availability of helper is more important
4. A more refined classification of social relationships is needed to ascertain the social bridging role of spouse. Dan Ao Department of Sociology The Chinese University of Hong Kong May 30th , 2008 Social capital is defined as resources embedded in one’s social network (Lin, 2001). From the perspective of resource composition, two types of social capital can be identified: ◦ Homophilous (homophily) ◦ heterophilous (heterophily) Lazarsfeld & Merton 1954
2 There is voluminous evidence in the literature that the principle of homophily exists in social networks (e.g., Fischer 1982; Homans 1950; Laumann 1966, 1973, 1976; Lazarsfeld & Merton 1954; Lin 1982; Marsden 1981, 1988, 1990; McPherson et al. 2001). The extent of homophily varies for each respondent, which may be one of the sources that differentiations in a variety of outcomes come out.
3 Less attention has been paid to the heterophily principle. How are the ties with groups other than that to which a respondent belongs formed? ◦ Are the heterophilous ties formed randomly? ◦ Are the heterophilous ties formed with certain patterns? Social distance (Bogardus (1933), Laumann (1966), or McFarland and Brown (1973))
4 Marsden (1988) examined the patterns of inbreeding and social distance on respondent-alter dyads discussing important matters using data collected with the name generator in the 1985 GSS. The stratifying variables include: ◦ Age ◦ Education ◦ Race/ethnicity ◦ Religion ◦ Sex
5 His main conclusions are: ◦ Discussion relations are most constrained by race/ethnicity, and least by sex and education; ◦ Inbreeding effects are present for all five stratifying variables, and account for virtually all structure in dyads classified by race/ethnicity and religion; and ◦ Appreciable social distance biases in the formation of these strong ties are found for age and education, but not for other stratifying variables.
6 Smith-Lovin et al. (2007) replicated the name generator questions in the 2004 GSS. 1985 2004? They found that: ◦ (1) racial homophily is still the most salient dimension among these five statuses and has remained relatively constant over the two decades; ◦ (2) the tendencies of homophily over other statuses increase, with the exception of sex.
7 To date, no studies have directly examined the patterns of homophily and heterophily using position-generated network data. Studying network data from the position generator can yield fruitful results. The differences between the name-generator and the position generator.
8 The name generator The position generator technique: technique: ◦ Stronger ties, stronger A more diverse social ties in terms of tie strength, role relations, or role relations, and geographically limited ties number of ties (Campbell & Lee 1991) ◦ Composed of both Consisting mostly of the homophilous (friends and homophilous ties (friends relatives) and and relatives) heterophilous ties Core network (friends of friends, acquaintances) ◦ Extended network
9 This study tries to follow the methods used in Marsden (1988) to model the patterns of homophily and heterophily using the position- generated network data in SC-USA 2004-5. Two types of special ties: ◦ (1) very close ties ◦ (2) non-kin ties
10 11 12 13 14 As Marsden (1988) mentioned in his study, respondent-alter pairs are cluster sampled within respondents, therefore, there may be a design effect due to the clustering. It is important to examine the effect of survey design on the distribution of X2 and propose simple adjustments of this statistic. Holt et al. (1980) propose a one moment adjustment of the X2 : a weighted sum of cell design effects.
15 16 17 18 19 20 21 22 23 24 25 The inbreeding bias exists in all of the three statuses, although the extent varies. Compared with core discussion networks, the results show that in the extended networks, the degree of racial/ethnic homophily is still tremendously strong. It helps explain why a certain racial/ethnic group (Latinos) is disadvantaged in the access to social capital and the eventual utility of social capital in status attainment.
26 Measurement of Social Capital: Recall Errors 1 and Bias Estimations Kuo-hsien Su, National Taiwan University Nan Lin, Academia Sinica and Duke University Change in number of positions accessed from wave I to wave II (N=2,707 respondents)
2 15
No change : 12% 10
Decrease : 52.9% Increase : 35% Percent 5 0 -20 -10 0 10 20 Differences in number of positions accessed (wave II - wave I) Differences between the sets of accessed positions during two interviews may reflect… 3
Genuine changes in network Observed change of accessed positions
Measurement Error Motivations
4
Measurement instability poses a serious challenge to the study of network changes. Need a clear measurement or better understanding of the possible sources of error. The two periods panel survey provided an opportunity (1) to model factors associated with changes in accessed position (2) to detect whether the respondent forgot a subsequently/previously named contact . Prior research
5
Forgetting is a pervasive phenomenon in the elicitation of network contacts. Research on forgetfulness has been disproportionately based on name generator instrument. Little research on the reliability of position generator. Tasks
6
Identify the sources of potential bias
Analyze the factors associated with forgetting
Examine effects of forgetting on network resource indices Data
7
Social Capital Project: the Taiwan Survey, conducted in late 2004 and 2006 Consists of 1,695 men and 1,585 women aged 20- 65. Problem of Non-response
8
Wave I Wave II 2004 2006 N = 3,280 N = 2,710 Re-interview = 82.6% Non-response = 17.4% Table 1. Characteristics of the follow-up and non-response sub- sample Non-response Full sample Follow-up sample (N=3280) (N=2710) (N=570) Mean % Mean % Mean % Gender Male 51.7% 51.5% 52.5% Female 48.3% 48.5% 47.5% Age 41.3 41.7 39.5 Years of schooling 11.7 11.7 11.8 Marital Status Single 23.9% 22.8% 29.1% Married/cohab 70.2% 71.4% 64.4% Widow/divorced 6.0% 5.8% 6.5% Network resource indices Extensity 8.5 8.5 8.2 Upper reachability 62.4 62.8 60.4 9 Range of prestige 36.7 37.0 35.1 Three types of research designs (Brewer, 2000).
10 A B C • Comparisons • Comparison Comparison between between between two recall and recall data recall data recognition and elicited in two data objective records of separate interaction interviews within a short period of time Limitations of our data
11
Our survey was not designed to examine forgetting specifically. No recognition data or objective records to compare with. Two years interval is too long: Test-retest design is usually within a very short time interval. Revised method C: Comparison of accessed positions elicited in two separate interviews
12
How many years have you known this person？
2004 2005 2006 Wave I Wave II Whether the respondent forgot a subsequently named contact？ Forgetting = (Contact mentioned in wave II but not mentioned in wave I) AND (duration >= 3 years) Assumption: durations reported in wave II are more or less accurate. Coding scheme for tie changes
13 Wave II (2006) NO YES (1) Consistent “NO” (2)New contacts (less than 3 years) NO (3)Forgetting at wave I (more Wave I (2004) than 3 years) (4)Lost contact (5) Consistent “YES” /Forgetting at YES wave II The distribution of length of relationship of forgotten ties (N=4,332 dyads, 7.3%) .1 .08 The average duration of ties forgotten is 13 years .06 Density .04 .02 0 0 20 40 60 Length of relationship (in years) How much does the respondent forget？
15
know more Wave I Wave II Categories N Percent than 3 years? Consistent YES YES 14,330 49.9% "YES" NO YES NO New contact 1,240 4.3% Forgotten at YES 4,332 15.1% wave I Unique Contact lost/ = 51.1% YES NO Forgotten at 8,794 30.7% wave II Total 28,696 100% approximately 15% of forgetting Distribution of respondents by number of ties forgotten (N=2707 respondents)
16
40 35.6% of the respondents did not forget any ties
30 64.4 %of the respondents failed to mention at least one contact, with an average of 1.6 forgotten ties per
20 respondent. Percent These numbers suggest that forgetting a contact was not a rare 10 occurrence. 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Number of forgotten ties in wave 1 Analytical Strategies
17
Analysis for the effect of forgetting on Model predicting “forgetting” estimates of accessibility
What factors are Whether “forgetting” associated with affects estimates of forgetting？ network resources ？ Unit of analysis: Unit of analysis: person-contacts dyads person Model : Multilevel logit Sample
18
A multi-level logit approach The models estimate the odds of “forgetting” versus “not forgetting”; the reference population consisted of all contacts mentioned in the first interview (2004). LEVEL 2 LEVEL 1
The multi-level approach nurse requires us to transform the individual-based data to lawyer person-contacts observations. Respondent A doctor Positions professor nested within Data structure individuals CEO The final sample consists of 2,682 respondents janitor and 28,343 person- Respondent contact dyads. B Taxi driver
Security guard Variables
20
Level 2 (respondent level): Age Years of schooling Marital status (married) Employment status (employee) Occupational prestige score Size of daily contact Variables
21
Level 1 (ties level): Type of relationships Group into six categories: kin, neighbor, school tie, work- related ties, friends, indirect tie Length of relations (in years) Closeness Gender homophily Status difference Status distance = absolute difference between respondent’s prestige score and contact’s prestige scores Status disparity = respondent’s prestige score – contact’s prestige score Descriptive statistics (individual level) Level-2 Total Male Female (N=2676) (N= 1383) (N=1293) Age(in years) 41.62 41.46 41.79 (11.66) (11.62) (11.70) Years of education 11.77 12.30 11.19 (4.23) (3.75) (4.63) Marital status single 0.23 0.25 0.21 divorced/widowed 0.06 0.03 0.09 married 0.71 0.72 0.71 Employment status employee 0.72 0.68 0.76 self-employed/employer 0.18 0.23 0.12 part-timer 0.03 0.03 0.03 family worker 0.08 0.05 0.10 Occupation prestige 39.88 41.26 38.39 score (12.91) (13.13) (12.50) Size of daily contacts 3.42 3.52 3.31 (1.36) (1.31) (1.41) Descriptive statistics (dyad level) Level-1 Total Forgetting Not forgetting (N=27,103) (N=4,315 ) (N=22,788) Type of relationship kin 0.21 0.24 0.21 neighbor 0.07 0.09 0.07 school tie 0.07 0.07 0.08 work-related ties 0.35 0.42 0.33 friends 0.24 0.12 0.26 indirect tie 0.05 0.07 0.05 Same sex 0.61 0.60 0.61 Length of relationship 12.89 12.95 12.88 (11.92) (11.98) (11.91) Closeness 3.46 3.34 3.49 (0.99) (0.99) (0.99) Status Distance 15.79 16.74 15.61 (11.66) (12.21) (11.54) Status Disparity -3.56 -5.89 -3.12 (19.30) (19.87) (19.16) Multi-level model predicting “forgetting” (level-2 model)
24 MODEL (1) Level-2 Model Intercept -1.197*** Female (male) -.146*** Age (in years) .000 Years of schooling -.054*** Marital status (married) Single .105+ Divorced/widowed -.168+ Employment status (employee) Self-employed/employer -.080 Part-timer -.076 Family worker -.048 Occupation prestige scores -.007*** Size of daily contacts -.125*** Multi-level model predicting “forgetting” (Level-1 model)
25 MODEL (1) MODEL (2) Level-1 Model Type of relationship (work-related ties)
Kin .100* .100 * Neighbor .015 .011 School ties -.196*** -.197 *** Friends -.807*** -.804 *** Indrect ties .096 .104 + Same sex -.039+ -.106 *** (same-sex)×female .122 ** Length of relationship -.008*** -.007 *** Closeness -.173*** -.174 *** Status Distance .007*** .011 *** (status distance)×female -.007 *** Multi-level model predicting “forgetting” (Level-1 model)
26 MODEL (3) MODEL (4) Level-1 Model Type of relationship (work-related ties) Kin .108* .107* Neighbor .022 .020 School ties -.197*** -.200*** Friends -.814*** -.812*** Indrect ties .098+ .104+ Same sex -.040+ -.106*** (same-sex)×female .119 * Length of relationship -.008*** -.008*** Closeness -.179*** -.181*** Status disparity -.002** -.004*** (status disparity)×female .004** Findings
27
Recall error may not be random. Forgetting is more likely among weak ties. How does recall error affect the estimation of network-driven indices？ Table 4. Discrepancy between “true” (corrected) and “observed” (raw) network resources indices 28 Corrected Raw Differences t-test score score Extensity Mean 9.9 8.5 1.38 39.2 SD 5.5 5.5 Range Mean 40.6 36.7 3.92 25.8 SD 16.8 18.6 Upper reachability Mean 65.2 62.4 2.83 19.3 SD 15.2 17.6 Because forgetting is more likely among weak ties, position- generator underestimate embedded network resources. Table 5. Correlations between “true” (corrected) and observed (raw) network resources indices at wave I (N=3,272)
29 Raw indices at Corrected indices wave I Extensity Range Reachability Extensity Range
Corrected Extensity -- indices Range .817 Reachability .692 .886 Raw Extensity .934 .745 .632 indices at wave I Range .792 .884 .776 .832 Reachability .674 .804 .880 .694 .865 Conclusions
30
Forgetting a contact was not a rare occurrence; Recall error is largely nonrandom. Status difference appears to govern the recall process. Position generator systematically underestimates network-driven resource indices. The Spatial Dimension of Social Capital: An Exploration
Zong-Rong Lee 李宗榮 Institute of Sociology Academia Sinica Taipei, Taiwan 1 Spatially Bounded Social Interactions
Earlier writers have found that people are localized in their social contacts and that their interactions are mediated by conditions of urbanization and locals of neighborhood where they dwell in (Park, Fisher 1982, Wellman…)
Geographical and social space structure the likelihood of social interactions; proximate actors form ties more frequently and have stronger influence on each other. (Festinger 1950; W. White1956; Blau 1977)
Other similar studies on mate selection, political attitude, social trips between city areas etc.
Spatially proximate companies are more likely to share directors. Mints and Schwartz (1985); Kono et al, (1998) Burt (2006)
The likelihood that a venture capitalist invests in a new target declines with the distance between venture capitalist and its target. Sorensen and Stuart (2000)
2 Spatial Dimension of Social Capital? Social Capital: resources embedded in a social networks that can be accessed and mobilized in instrumental actions (Lin 2001)
> Is there a spatial dimension behind the generation of social capital? And if so, what’s its pattern?
1. As social capital represents the network embedded resources and is mostly unequally distributed, is there any spatial association of this distribution? Do people endowed with different levels of social capital also show differences in geographic reach?
2. What social economic factors account for such differences? For example, do people of a higher status (e.g, job prestige, education, wealth) tend to have a social capital of a higher geographic reach, or vice versa?
3. Studies show consistent impact of social capital on instrumental actions. How may this effect be conditioned by the factor of proximity. What kind of roles does the factor of spatial distance play in the process of instrumental actions where the effect of social capital is at work?
3 Distance to Accessed Position
Sample: Social Capital Survey, USA 2007 Cross-Sectional Data. (n=1443 )
Distance measurement in Position-Generator module
Is there anyone you know who is a NURSE? [Yes, No.]
Typically, how long does it take you to travel to meet this person?
(1) Less than 15 minutes (2) 15-30 minutes (3) 30-60 minutes (4) 1-2 hours (5) 2-3 hours (6) 3-5 hours (7) 5-12 hours (8) more than 12 hours
4 Social Relations Tend to be LOCAL
76.7% of all reported ties fall within a 1 hour distance range between egos and reached alters.
(1433 respondents report 10807 ties for 22 positions) 40 30 20 % of Total N Total %of 10 0 <15min. 15-30min.30-60min. 1-2hrs 2-3hrs 3-5hrs 5-12hrs >12hrs 5 Some Descriptive Statistics ( 22 positions)
N=1433 % of Average Average Position (Prestige Score) Respondent Respondent Distance
Accessing Prestige (1-8) Professor (78) 40.27 50.80 3.29 Lawyer (73) 59.39 48.67 3.12 CEO (70) 22.12 49.32 3.21 Congressman (64) 16.61 48.58 3.21 Production manager (60) 17.24 46.00 2.53 Middle school teacher (60) 33.43 47.18 2.30 Personnel manager (60) 45.64 48.57 2.26 Writer (57) 25.33 50.61 3.36 Nurse (54) 72.09 47.64 2.88 Administrative assistant (53) 34.40 49.19 2.51 Computer programmer (51) 44.45 49.10 2.91 Bookkeeper (49) 29.59 48.42 2.60 Farmer (47) 41.10 47.07 3.46 Policeman (40) 50.59 47.05 2.39 Receptionist (38) 38.38 48.43 2.09 Operator in a factory (34) 23.10 44.62 2.53 Hair dresser (32) 60.78 47.05 2.10 Taxi driver (31) 9.49 43.35 2.26 Security guard (30) 25.75 45.00 2.27 Janitor (25) 34.12 47.97 1.95 Babysitter (23) 26.80 46.63 1.81 Hotel bell boy (22) 3.49 46.48 3.06
6 Correlation
N=22 2 3 4 5 6 7 8 9 10 1. Ave. Distance .61* .47* .45* .06 .26 .20 .61* .24 .29 2. Position Prestige .64* .70* .51* .25 -.18 .42 .51* .43* 3. Ego Prestige .96* .31 .16 -.19 .32 .58* .11 4. Ego Education .36 .24 -.28 .31 .63* .19 5. Ego Income .19 -.37 .21 .19 .47* 6. Weak Tie (%) -.65* .33 .03 .60* 7. Kin (%) -.28 .03 -.42 8. Executive (%) .08 .29 9. White (%) .18 10 .Age * Significance at 5%
7 “Social Capital in the Creation of Human Capital”
80 prfssr_78
lwyr_73 ceo_70
cngrss_64 prmgr_60prdmgr_60 tchr_60 60 wtr_57 nrs_54 adasst_53 prgmmr_51 bkkp_49 fmr_47
plcman_40 40 rcpnist_38 oprtr_34 hrdrssr_32 Prestige of Accessed Position Accessed Prestigeof taxid_31 secur_30
jntr_25 bllb_22 bbstr_23 20 14 14.5 15 15.5 16 16.5 ave. ego-edu
Fitted values prestige
8 Prestige Homophily and Social Networking
80 prfssr_78
lwyr_73 ceo_70
cngrss_64 prdmgr_60 prmgr_60 tchr_60 60 wtr_57 nrs_54 adasst_53 prgmmr_51 bkkp_49 fmr_47
plcman_40 40 rcpnist_38 oprtr_34 hrdrssr_32 Prestigeof Accessed Position taxid_31 secur_30
jntr_25 bllb_22bbstr_23 20 44 46 48 50 52 Ave. Ego-Prestige
Fitted values prestige
People of a higher status tend to connect with others also of a higher status. 9 Distance is NOT a Function of Income Level
3.5 fmr_47 wtr_57 prfssr_78 cngrss_64 ceo_70 lwyr_73 bllb_22 3 nrs_54 prgmmr_51
bkkp_49 oprtr_34 prdmgr_60 adasst_53 2.5 plcman_40 tchr_60 secur_30 prmgr_60 taxid_31
Average Distance hrdrssr_32 rcpnist_38 2 jntr_25 bbstr_23 1.5 17.5 18 18.5 19 19.5 20 Ave. Ego's Income
Fitted values distance
10 Ego’s Prestige vs. Distance to Accessed Position
3.5 fmr_47 wtr_57 prfssr_78 cngrss_64ceo_70 lwyr_73 bllb_22 3 nrs_54 prgmmr_51
bkkp_49 oprtr_34 prdmgr_60 adasst_53 2.5 plcman_40 tchr_60 taxid_31 secur_30 prmgr_60 ave.distance
hrdrssr_32 rcpnist_38 2 jntr_25 bbstr_23 1.5 44 46 48 50 52 ave. ego-prestige
Fitted values distance
Respondents of a higher job prestige tend to have spatially distant networks (that are also more likely to be high in job prestige) 11 Spatial Distance vs. Positions Prestige Percentage Distribution of Respondents’ Distances to 22 Positions
100%
80%
>12hr 5~12hr 60% 3~5hr 2~3hr 1~2hr 30~60m. 40% 15~30m. <15min.
20%
0% nrs_54 ceo_70 jntr_25 wtr_57 fmr_47 tchr_60 bllb_22 lwyr_73 oprtr_34 bbstr_23 taxid_31 bkkp_49 secur_30 prfssr_78 adasst_53 prmgr_60 cngrss_64 rcpnist_38 hrdrssr_32 prdmgr_60 plcman_40 prgmmr_51 12 Spatial Distance vs. Position Prestige
80 prfssr_78
lwyr_73 ceo_70
cngrss_64 prmgr_60tchr_60 prdmgr_60 60 wtr_57 adasst_53 nrs_54 prgmmr_51 bkkp_49 fmr_47
plcman_40 40 rcpnist_38 oprtr_34 hrdrssr_32 Prestige of Accessed Position Accessed Prestigeof taxid_31secur_30
jntr_25 bbstr_23 bllb_22 20 1.5 2 2.5 3 3.5 Average Distance
Fitted values prestige
The distance b/w egos and alters is highly associated with alter job prestige. Valuable networks are geographically distant! 13 A Prestige-Spatial Network Structure
1.Respondents with higher education level, job prestige are more likely to access valuable networks. (human capital principle).
2.They are also more likely to extend the reach of their networks beyond their surrounding geographic neighborhood.
3.The prestige of accessed positions is highly associated with the geographical distances between egos and reached alters.
A Three Way Interaction! Individuals of a higher social status tend to have spatially distant networks that are also more likely to be of a higher status.
When the principle of homophily is working so that two persons both demonstrating high job prestige will be more inclined to maintain their friendship, the underlying fact is that their geographic distance from each other more likely will be greater than that of others.
14 Executives Have Greater Geographic Reach Distance vs. Percentage of Executive Respondents
3.5 fmr_47 wtr_57 prfssr_78 cngrss_64 ceo_70 lwyr_73 bllb_22 3 nrs_54 prgmmr_51
bkkp_49 oprtr_34 prdmgr_60 adasst_53 2.5 plcman_40 tchr_60 taxid_31secur_30prmgr_60 ave.distance
rcpnist_38hrdrssr_32 2 jntr_25 bbstr_23 1.5 .08 .1 .12 .14 .16 .18 executive (%)
Fitted values distance
15 The Strength of Distant Ties
Remember that valuable networks tend to be geographically distant. Do they give STRENGTH?
Empirical test on acquisition of job information
J1. Now I would like you to think of the last months, did someone mention job possibilities, openings or opportunities to you, without your asking, in casual conversation? (1) Yes (0) No
Would social capital at different geographical distances deliver differing effects? Would distant ties fare better?
Previous Studies: Weak Ties (Granovetter) Social Capital: extensity, range, upper prestige (Lin) Structural Holes: brokerage position (Burt)
16 Variables
Extensity: the number of positions accessed (0~22) Highest Prestige: prestige score of highest position accessed Range: range of the prestige scores of positioned accessed (difference b/w highest and lowest scores)
<30min. 30min~2hrs 2hrs~5hrs 5hrs~12hrs Social Capital Extensity ExtensityD1 ExtensityD2 ExtensityD3 ExtensityD4 Highest Prestige SC_HighD1 SC_HighD2 SC_HighD3 SC_HighD4 Range SC_RangeD1 SC_RangeD2 SC_RangeD3 SC_RangeD4
Sum of Distance (1~8) for 22 accessed positions Mean of Distance for 22 accessed positions
In general, we expect social capital that is geographical distant to deliver a better effect on an individual’s acquisition of job information.
17 >The further the network extensity, the greater its effect! >Network resources bounded locally get discounted!
Determination of Receipt of Routine Job Information [exp (B): odds ratio] Variable Model 1 Model 2 Agea 1.000 (.994) 1.003 (.954) Age (squared)a 1.000 (.672) 1.000 (.617) Male 1.143 (.311) 1.127 (.366) Native born 1.338 (.202) 1.371 (.170) Race/ethnicity (base: non-Latino white) African-American 0.544*** (.005) 0.566*** (.009) Latino 1.947*** (.001) 1.975*** (.001) Other 1.167 (.613) 1.149 (.650) Marrieda 0.931 (.608) 0.935 (.635) Year of education 1.030 (.124) 1.020 (.296) Social Capital Extensity 1.078*** (.001) Extensity_D1 1.046** (.016) Extensity_D2 1.134*** (.001) Extensity_D3 1.148** (.017) Extensity_D4 1.182*** (.001) Observations 1,411 1,411
* Significance at 10%; ** Significance at 5%; ***significance at 1% 18 >Network range further away has greater effect! >The greatest effect is at third distance (D3) level.
Determination of Receipt of Routine Job Information [exp (B): odds ratio] Variable Model 3 Model 4 Agea 1.009 (.853) 1.014 (.771) Age (squared)a 1.000 (.557) 1.000 (.488) Male 1.160 (.259) 1.139 (.327) Native born 1.410 (.131) 1.400 (.141) Race/ethnicity (base: non-Latino white) African-American 0.577*** (.010) 0.578** (.011) Latino 1.937*** (.001) 1.939*** (.001) Other 1.196 (.557) 1.133 (.684) Marrieda 0.939 (.654) 0.954 (.736) Year of education 1.031 (.108) 1.023 (.233) Social Capital SC_Range 1.018*** (.001) SC_Range_D1 1.008** (.025) SC_Range_D2 1.013*** (.003) SC_Range_D3 ---> 1.019*** (.007) SC_Range_D4 1.016*** (.005) Observations 1411 1411 * Significance at 10%; ** Significance at 5%; ***significance at 1% 19 >Upper network prestige that is further away has stronger effect.
Determination of receipt of routine job information [exp (B): odds ratio] Variable Model 5 Model 6 Agea 1.014 (.765) 1.014 (.777) Age (squared)a 1.000 (.482) 1.000 (.486) Male 1.132 (.346) 1.140 (.324) Native born 1.406 (.134) 1.429 (.122) Race/ethnicity (base: non-Latino white) African-American 0.586** (.013) 0.602** (.019) Latino 1.905*** (.002) 1.995*** (.001) Other 1.161 (.624) 1.096 (.765) Marrieda 0.946 (.689) 0.944 (.683) Year of education 1.029 (.140) 1.018 (.372) Social Capital SC_High 1.020*** (.001) SC_High_D1 1.005 (.102) SC_High_D2 1.007*** (.001) SC_High_D3 1.006*** (.008) SC_High_D4 1.008*** (.001) Observations 1411 1411 * Significance at 10%; ** Significance at 5%; ***significance at 1% 20 Predicted Probability of Job Information Acquisition for Network Extensity at Different Levels of Distance 1 .8 .6 .4 Pr(Ln Job Information) Pr(LnJob .2
0 5 10 15 20 Network Extensity
Distance Within 30 Min. Distance 30min.~2hrs. Distance 2hrs.~5hrs. Distance 5hrs.~>12hrs.
21 Predicted Probability of Job Information Acquisition for Network Range at Different Levels of Distance .4 .35 .3 .25 Pr(Ln Job Information) Pr(LnJob .2 .15 0 20 40 60 Social Capital (range)
SC_Range Within 30 Min. SC_Range 30min.~2hrs. SC_Range 2hrs.~5hrs. SC_Range 5hrs.~>12hrs.
22 Predicted Probability of Job Information Acquisition for Highest Network Prestige at Different Levels of Distance .3 .25 .2 Pr(Ln Job Information) Pr(LnJob .15 0 20 40 60 80 Highest Network Prestige
Distance Within 30 Min. Distance 30min.~2hrs. Distance 2hrs.~5hrs. Distance 5hrs.~>12hrs.
23 Summary
1. A spatial dimension of social capital was identified, and social capital varies by distance: The prestige of accessed positions is highly associated with the geographical distances between egos and reached alters. Higher-status networks tend to be geographically distant.
2. Prestige Homophily principle works Over the Proximity principle: People of a higher prestige form networks with each other, despite the likely long distance separating them.
3. “The Strength of Distant Ties”: A geographically distant social capital delivers a stronger effect on the acquisition of job information than a geographically closer one. Beneficial effect of social capital gets discounted when network bounded locally.
4. The function of spatial influence on network utility may not be linear. (curvelinear ?)
24 Discussion
Spatial Extensity is congruent with the concept of Social Capital Individuals with a wide variety of networks (i.e., extensity, diversity) might also have networks of wider geographic reach.
Such spatial extensity can probably prevent them from being constrained by local networks that mostly deliver redundant information, as weak-ties and structural holes arguments suggest.
Instead of asking “Why do weak ties have strength?”, we should ask
Why do ties with strength tend to be WEAK?
This study suggests that it’s probably because valuable networks tend to be geographically distant; as the result, the relationships may tend to be weaker as well).
Is SPACE an endogenous dimension of social capital? remember the prestige-spatial network structure that we identified
25 Is “Structural Holes” A Spatial Concept?
Ave. Distance of Robert’s Ties > Ave. Distance of James’ Ties
26 Thank You!
27 — page 1 —
Commentary Academia Sinica Conference on Social Capital: Its Origins and Consequences
These slides draw from a course, “Strategic Leadership,” in the Chicago GSB Executive M.B.A. Program, research papers, and a recent book, Brokerage and Closure (2005, Oxford University Press). All rights are reserved (© Ronald S. Burt, 2008). A course syllabus, book overview, and related research papers can be downloaded from the author’s university website (http://ChicagoGSB.edu/fac/ronald.burt). — page 2 —
Presentation note: Jiminy-Cricket Chinese Thank you Nan Lin and Yang-chih Fu for invitation intellectual stimulation, renew old acquaintances
What can I add in a few minutes to what is already well-done? Factors working against us finding value together end of conference, end of day, end of week, jet lag (more clever in AM)
No interest in review of papers or sugar-buzz congratulations work was good and interesting, but you are tired, there is insufficient time for good review, and many papers are early analyses of new data
I decided to play the role of network broker. Nan means a great deal to me. I am a keenly sympathetic outsider to the use of positional generators. I too have data from the 3-country survey and plan to look at correlates of positional versus name-generator networks.
The task for me is to look for external frames of reference that people could find useful as they move forward with the interesting work presented. We can address any topic when Nan opens discussion, but I limit my initial remarks to three issues: method, mechanism, and endogeneity. — page 3 —
Self congratulations on social capital theory or method is premature. In an otherwise positive review of Putnam’s Bowling Alone, Claude Fischer (Social Networks, 2005:157) had the following to say about the concept of social capital:
“Social capital” is also unnecessary, because clearer and simpler terms – such as membership, family, sociability and trust – serve perfectly well. Putnam implicitly recognizes the problem when he switches to other metaphors to describe types of “social capital”: “bridging” and “bonding.” These are both terms much more suited to modify the metaphor of “ties” than that of capital. “Social capitalism” has expanded in all directions like a swamp in wet weather. Glaeser et al. (2000, p. 3), for example write that “individual social capital (is) a person’s social characteristics – including social skills, charisma, and the size of his Rolodex – which enables him to reap market and non-market returns from interactions with others.” This is not much different from saying that social capital is everything psychological and sociological about a person.7 “Social capital” has become one of the many species of “capitals” that have recently infested sociologists’ prose.
7These complaints are some of the reasons that, as past-editor of the American Sociological Association’s popular sociology magazine, Contexts, I put “social capital” on the do-not-use style-sheet (unless put into ironic quotation marks).
When you are feeling arrogant, check out the Web of Science to see what research in your area is finding an audience. — page 4 —
1. METHOD: Do the data capture the mechanism? Name & position generators both weak. Data-concept distance below highlights risk of theory-detachment. short: Aral & van Alstyne (2007) -- info content of emails (variety within ties & between contacts) longer (ask respondent to interpret meaning of relationship): Festinger et al (1950), Coleman et al (1966) -- name generators with saturation sample survey longer: Berelson et al (1944), Laumann (1966), Wellman (1979) -- name generators in interviews with random sample of community residents (no links between contacts cited by different respondents) longer: Marsden (1987) -- name generator in interviews with national probability sample (GSS) longer (replace contact as person with status/group): Lin & Dumin (1986) -- positional generators define contact with statuses/groups longer (replace relations with correlates; What mechanism changes Y if X is increased?): Putnam (1993) -- counts of group memberships is proxy for network density in region Kasarda (1970s) -- appearance of wig store indicates decay in community network Liebert (1976) -- city incorporation year indicates reform ethic dominating civic politics Florida (2002) -- gays in city indicates creativity by indicating tolerance for heterophily really long (network measure is reduced to little more than an opinion-poll item): Ask respondent to evaluate network (e.g., GSS item, item in Structural Holes; "Everything considered, my network is about as effective as others here." 5 absolutely yes to 1 absolutely not) — page 5 —
1. METHOD (cont.): Do positional analyses display concern about method or mechanism? For example, how would you compare alternative network measures of social capital? If you focus on similarity/differences between the measures, you are writing a methods paper. Social capital is about network associations with advantage (Lazarsfeld’s “point substitutability”).
Measure A advantage in explaining differences not predicted by B
Measure A variance
Methodological aside on SNA Criterion variance indicating Contested criterion advantage variance allocated to method with larger advantage Measure B variance
Measure B advantage In explaining differences not predicted by A — page 6 —
2. MECHANISM: It is difficult to focus data on the mechanism when the mechanism is unclear. What exactly is the mechanism by which having or mobilizing ties to status groups constitutes social capital?
I have heard explanations claiming that advantage results from the diversity of statuses reached and claims that advantage results from the prestige of the statuses reached. Both are reasonable, but what is the mechanism we can use to motivate students or we can cite in response to criticisms such as Claude Fischer’s that social capital is an empty abstraction better replaced with more concrete network terms like bridging and bonding?
Let me illustrate with my own efforts to figure out the mechanism responsible for diversity creating advantage. My efforts have led me away from information and resource flows to focus explicitly on cognitive processes -- which have been mentioned from time to time during this conference.
I don’t propose that what I have done is what everyone should do. I just want to highlight the issue of digging below network correlations with advantage to get at the underlying network mechanisms responsible. Strategic Leadership Creating Value: The Social Capital of Growth and Innovation (page 12)
Robert Z-Score Relative Performance (compensation, evaluation, promotion) -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
1 5 (C) withineachof eightstudypopulations.Dashed linegoesthroughmeanvalues ofZfor E J Circles areaveragez-scoreperformance (Z)forafive-pointintervalofnetworkconstraint E E E E E J E 2 5 E E intervals ofC.Bold lineisperformancepredicted bythenaturallogofC. E E E J E E E E E E E E E E J E E E 3 5 many ———Structural Holes———few E E E E E E E E J E E E E E E constraint (35points) median network E E E J E E E E E Network Constraint (C) 4 5 E E E E E E E J E E E E E E E E E E 5 5 E E E J E E Social CapitalofBrokerage Note: E E E E J E Manifest asbetterideas,more-positiveevaluations,highercompensation, 6 5 E J E E E Brokerageisalargepercentageofexplainedperformancedifferences. E J E E E 7 5 — page7 E J E E E E J E E E 8 5 E J E E earlier promotion,andfasterteams. E J E E 9 5 Z =2.78-.82ln(C) E J E E E E J E 5 James E J E E study-population graphs in Appendix IIon measuringnetwork constraint. r =-.53 From Figure1.8 in 17% 28% 2% 33% Brokerage andClosure. 5% 10% 64% 55% 85% Data pooledacross eight electronics rank inalarge to seniorjob early promotion Third pieis bureaucracies. in corporate compensation HR manager supply-chain and Second pieis Research Team. All-America election tothe and analyst compensation banker is investment (blue). Firstpie other factors rank (red),and (white), job constraint Network Managers: between Differences Performance Variance in Predicted Much ofthe Contributes Brokerage
firm. Strategic Leadership Creating Value: The Social Capital of Growth and Innovation (page 32) Social NetworksCouldConstitute Capital Three InformationProcessesbywhich From Figure 2.8inbook tentativelytitled by whichSocialNetworksCouldConstitute Capital Appendix VI:ProcessCluesfromSpillover Figure 2.8:NeighborNetworkCluesto Neighbor Networks — page8 social interfaceegomanages). strong independent ofsocialcapitalinneighbornetworksdespite immediate network emotional skillsdevelopfrommanaginginformationin the performance exceptasperformance-enhancingcognitive and Personal Processes Austrian marketmetaphor) the local.(Affiliateterms:tacitinformation,localinstitutions, covariance decreasesrapidlyasneighbornetworks as itdoeswithsocialcapitalinego'sownnetwork,but performance relationships: so socialcapitalisafunctionofmorecompleteaccesstolocal relations butnotthroughindirectconnectionsbeyondthelocal, Local Processes neoclassical marketmetaphor) explicit information,"Metcalfe'sLaw,"maturecapitalmarket, with socialcapitalinego capital inneighbornetworks(definesego known farandwide. social capitalisafunctionofaccesstomorerelations: average probabilitywhenitmovesthroughanyrelationship,so Global Processes covariation covariation covaries covaries She istheleadingpersoninarea. She with socialcapitalinego'sownnetwork (the with – : – Information retainsmeaningthroughlocal She iswise. She Information retainsmeaning withsome Information Egoperformance – with socialcapitalinneighbornetworks with Information flowisirrelevant to ’ s immediatenetwork. s Processes forcing functionforhuman capital In short,s In Egoperformanceis covaries covaries ’ s reach)morethan s ocial capital ocial (Affiliate terms: (Affiliate with social with go beyond go is a is Ego She is . Strategic Leadership Creating Value: The Social Capital of Growth and Innovation (page 33)
association across455supply-chain managers. ego's contacts).Parentheses containt-teststatisticsfor performance atlevels of indirectconstraint(constrainton own network).Hollow dotsanddashedlineindicate at levelsofdirectconstraint (constraintwithinego's horizontal axis.Soliddotsandlineindicate performance Constraint scoresarepooledfor5-pointintervals onthe brokers isanadvantage it willseemthataffiliation with If youlookatcolleagues, Relative Salary (z-score) -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 From Figure3.8 inbooktentativelytitled 10 20 (-9.1) many ———StructuralHolesfew (-14.9) 30 Network Constraint(C) 40 50 median direct (60 points) network constraint 60 Neighbor Networks 70 80 90 A (cf.,Burt,"Secondhand brokerage,"2007, 100 — page9 10 (-3.4) 20 (-7.4) (-6.7) 30 (-2.5) 40 Academy ofManagement Journal 50 60 70 80 90 B C ) 100 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Relative Value of Best Idea Performance Evaluation (z-score) (z-score) Strategic Leadership Creating Value: The Social Capital of Growth and Innovation (page 34) Holes among Performance brokerage," 2007 of Structural of ManagementJournal (cf. Burt,"Secondhand Correlates Direct and Contacts Neighbor Networks control variablesin from tableslisting Test statisticsare Indirect Academy
). in financialorganization(n=469) Investment bankerworkrelations Pacific productlaunch(n=258) supply-chain managersinlarge financial organization(n=351) Buying andsellingrelations Discussion networkin Asia- Discussion networkamong Discussion networkamong defining supplier-customer Analyst workrelationsina networks around four-digit HR employeesinfinancial electronics firm(n=455) American manufacturing organization (n=283) industries (n=632) — page10 Structural Holes Direct Contacts implies personal processeson page8. Negligible benefitofneighbor networks among 3.79 3.38 3.78 4.29 4.17 2.70 Indirect Contacts Structural Holes among 3.86 0.24 1.33 0.23 0.92 1.00
Center-Periphery Balkanized Networks Networks — page 11 —
3. ENDOGENEITY: Manski (1993, Review of Economic Studies; see Mouw, 2006, Annual Review of Sociology) “reflection” problem. When a person watching herself in a mirror moves, person and reflection move simultaneously. How do you know that the reflection did not make the person move? Answer: Know the mechanism by which reflection occurs.
This issue of criterion variable being used to predict itself can be used to call almost any kind of analysis into question, but it is not equally a problem for all forms of network analysis. The issue is interestingly entwined in positional measures because alter SES is often used to predict ego’s SES. For example, mean occupational prestige of contacts is often used to predict ego’s occupational prestige.
In the extreme, you get a network autocorrelation model or the eigenvector model of network centrality/status (Pi = Σj zjiPj). In fact, my initial interpretation of Nan’s positional work (1981 ASR) when he was moving from Blau-Duncan models was that he described a network contagion effect in which ego achieved prestige comparable to alters, as Berelson et al. (1944) showed in Elmira that ego voting resembled the voting preferences of friends.
When the criterion advantage variable also defines social capital, then endogeneity warrants consideration when thinking about knowledge claims, gathering longitudinal data, looking for counterfactuals, instrumental variables, or opportunities for field experiments.