THE DISTRIBUTION AND RETURN OF SOCIAL CAPITAL: EVIDENCE FROM SWEDEN

This paper studies the role of social capital in the status attainment process and examines the link between the hiring process and the potential pool of social capital embedded in a person's network. The analysis is based on a sample of people newly employed by the municipal services in Malmö, the third largest city in Sweden. Jobs in this sector of labour market are mainly low-paid, and are dominated by women and immigrants. The position generator method is used to measure social capital, understood as assets captured by individuals in social networks. The findings demonstrate that access to social capital is positively related to work experience, a higher educational level, having a partner, and active membership of voluntary associations. It is also apparent that being an immigrant is associated with a substantial social capital deficit. Regarding the return on capital, the results show that both human capital and social capital were rewarded with higher wages and more adequate jobs. Furthermore, we found that social capital is associated with better labour market outcomes, whether or not respondents reported that they obtained their current jobs using informal job-search methods. Results also show similar returns on access to social capital for natives and immigrants.

Social capital consists of the resources embedded in an individual's social networks that are potentially available to her or him through these contacts (Flap and Völker 2004). Good networks are valuable to job-seekers because, for example, they increase the chances of obtaining better information about vacancies, influential contacts are good references in job-seekers resume and give more confidence in job interview situations (Lin 2001a). The purpose of this study is to examine the effects of access to social capital on status attainment in the labour market: in short, how social capital improves the likelihood of getting better jobs.

The subject of this article is related to the general research field known as the status attainment process. In this field the focus is on the relationship between social origin (parents’ educational, economical and social resources), educational achievements and class destinations (Jonsson 1988). This study, however, investigates only the connections between education and work experience (human capital) on the one hand, and social capital on the other, in so far as they affect status attainment in the labour market. As Erickson (2001: 128) suggests, the effect of human capital variables – that is, education and labour market experience – have been overestimated because greater amounts of human capital lead to greater social capital, and it is possible that ‘individuals with better social capital get better jobs, but the effect is spurious, because people with better education and work experience get both better jobs and have greater social capital’. The empirical results of this paper confirm the positive effect of education on access to better social capital, and show that social capital is linked to better jobs even when we control for education and work experience.

The following research questions are posed in this study:

• 1.

  What characteristics (such as educational level, work experience, gender, birthplace and marital status) enhance or hinder access to social capital? Or, who has better or worse access to social capital?

• 2.

  How well is social capital rewarded in the labour market compared with education and work experience?

Due to the large proportion of immigrants in our data, we can examine the impact of being an immigrant rather than a native as a determinant of access to and the return on social capital. Immigrants in this study are defined as individuals born outside Sweden. Here we single out two different regions of origin: north-west countries (NW), which comprise north-west Europe and North America, and countries outside north-west Europe and North America (ONW), which account for rest of the world.¹ As Erickson (2004: 40) puts it, immigrants from ONW countries (which she names ‘non-whites’) ‘face a double handicap in forming social capital’. They are in many cases residents of stigmatized neighbourhoods or are concentrated in inferior parts of the labour market; consequently ‘prejudiced people do not find them attractive as potential acquaintances’. Earlier Swedish empirical studies confirm that individuals from ONW countries have an inferior status (in terms such as employment level or wage) to individuals from NW countries, and run a higher risk of suffering from discrimination (Arai and Villhelmson 2001; Le Grand and Szulkin 2002; Nekby 2003).

As earlier studies have shown, hiring processes vary among different lines of business (Erickson 2001: 129). Consequently, any investigation of the hiring process must be located within a specific industry or organisation. Granovetter (1995) also traces variations in how people get jobs, among other things, to the wider institutional and historical context within which people seek jobs and employers seek labour. Thus, he argues, it is important to locate research within a specific setting. Considering this argument, we investigate one specific workplace, the substantial public sector municipal services in Malmö, Sweden's third-largest city. Malmö has a relatively high proportion of immigrant inhabitants (nearly 24 per cent). Of the total employed in the municipality in August 2002, almost 17 per cent were immigrants. Day care, elderly care, primary education, adult education and the fire brigade are administered by local authorities in Sweden. Public sector occupations in the Swedish labour market are mainly low-paid, and dominated by women and immigrants. This is the case with Malmö's municipal services as well as in our sample.

Since our survey is limited to those already working in the Malmö municipality, we cannot compare those who got jobs with those who did not, but only job outcomes within the organisation. It is plausible to assume that immigrants (particularly those from ONW countries) are severely overrepresented in the group who did not obtain any job (see for example Arai and Villhelmson 2001; Nekby 2003).

Social capital consists of the resources that are embedded in one's social networks and accessible through one's direct and indirect ties (Lin and Dumin 1986). Bourdieu (1986) defines the volume of social capital as a function of the size of the network and the volume of capital (economic and cultural) possessed by networked individuals. Social capital is not possessed by individuals but consists of (second-order) resources which belong to one's relatives, friends, or acquaintances. When looking for a job, persons may mobilize their social resources by reaching out to relatives, friends, or acquaintances in their personal networks in order to access their information and influence, which may improve the chances of obtaining a job. Therefore, access and use of social capital would lead to a better job-search outcome.

Regarding access to social capital, the theory suggests that the initial position of an individual plays an important role. For example, parental socioeconomic resources ‘promote the likelihood of reaching better social resources’ (Lai et al. 1998: 160). At the same time, the composition of an individual's network is largely understandable in terms of the ‘homophily’ principle (Laumann 1966), which means that individuals in a network tend to resemble each other in several ways. Many studies which support this hypothesis for several dimensions (see, for example, Campbell et al. 1986, for the network structure of people with more advantaged socioeconomic origins, Green et al. 1995 for the social networks of poor people, and Erickson 2004 about gender and social capital). Further, inequality in different types of capital – which Lin (2000) calls capital deficit – brings about social inequality, such as in socioeconomic status. For Lin (2000), capital deficit is the consequence of a process whereby differential opportunities result in a relative shortage of capital for one group as compared with another. Some social groups (for example, ‘racial’/'ethnic’ groups or poor people) may be embedded in social networks that constrain their members’ ability to obtain valuable resources.

Lin (2001: 100) suggests that capital inequality may result from two distinct processes. In addition to capital deficit, he mentions ‘return deficit’, which is ‘the consequence of a process by which a given quality or quantity of capital generates a differential return or outcome for different social groups’, as, for example, when natives and immigrants with the same quality and quantity of social capital receive different wages in a workplace. In this case, if we know already that immigrants suffer from a social capital deficit, inequality in labour market outcomes between natives and immigrants may be due to capital deficit, return deficit, or both. In the case of return deficit, either immigrants have not been able to mobilize and use their social capital or they have received different responses from employers. As Lin (2001: 102) says: ‘Return deficit may or may not occur independent of capital deficit’.

Regarding the return on social capital, the theory suggests that success in, for example, finding an adequate job is positively associated with access to and use of better social capital (Lin 2001a: 60). Access to better social capital is in turn associated with being connected to better-placed contacts. This is because, as Lin (2001: 60) suggests,

• 1.

  the better-placed individuals in one's social network, have control over more resources, and the greater their influence, the more the benefit to the job-seeker;

• 2.

  the better-placed individuals in one's social network, given their advantageous view of the structure, and can provide better information to the job-seeker;

• 3.

  the better-placed individuals in one's social network have better social credentials, as, for example, when they are cited as referees in the job-seeker's applications;

• 4.

  access to better-placed individuals in one's social network itself enhances the job-seeker's confidence and self-esteem and improves his or her conduct in job interviews; and

• 5.

  being socialized with better-placed social groups elevates the applicant's ‘acceptability’.² This means that the job-seeker learns more about their speech style, manner, aesthetic preferences and generally about their ‘taste’. As Bourdieu (1984: 174) suggests about ‘tastes’ and social classes; ‘The system of matching properties, which include people … is organized by taste … Taste is the practical operator of the transmitation of things into distinct and distinctive signs, of continuous distributions into discontinuous oppositions; it raises the differences inscribed in the physical order of bodies to the symbolic order of significant distinctions’.

What measures can be suggested as indicators of social capital? When we define social capital as resources embedded in one's social networks, measurements focus on the valued resources of others accessed by individuals in their networks and ties. Lin (2001b: 14) suggests: ‘measurements focus on the valued resources (e.g., wealth, power, and status) of others accessed by individuals in their networks and ties’. The Following indicators are suggested as measurement of resources in the networks:

• i.

  extensity or the number of positions that are reachable, reflects the diversity of positions, and their embedded resources;

• ii.

  upper reachability: the best possible resources in the social network or among ties;

• iii.

  heterogeneity: the range of positions (distinction between the highest and lowest reachable positions) whose resources are reachable through social ties,³ and

• iv.

  the composition of resources or average or typical resources (Lin 2001b: 15).

Extensity indicates the quantity of ties that give access to occupations, while the other three characteristics indicate the quality of access or the volume of capital possessed by networked individuals. According to earlier research, correlations among these various measures of social capital tend to be high, and they can form a single factor (Lin 2001b).

A survey of individuals employed by local authorities in Malmö municipality during 2001–2003 was conducted in April 2004 in collaboration with Statistics Sweden (SCB). All individuals who obtained a job and continued working in Malmö municipality during this period received a copy of a questionnaire. Alongside with the questionnaire, we obtained further information (such as age, educational level and birthplace), about recently employed people from the longitudinal register databases of Statistics Sweden. Of those who received our questionnaire (a total of 3,702 individuals), 2,504 respondents, or about 67 per cent, cooperated with the research project and completed the survey. In 65 cases values were missing in some variables, and these cases were dropped. The final sample of 2,439 employees is reasonably representative. Appendix Table 1 compares the sample to the total population. The comparison has been undertaken with the help of information provided by Statistics Sweden about the total population. A summary of respondent characteristics is shown in Table 1.

There are several methods for counting resources in a social network, such as the ‘name generator’, ‘position generator’ and ‘resource generator’ methods (M. Van Der Gaag and T. Snijders 2004). This study employed the position-generator method for the measurement of social capital (Lin and Dumin 1986, Lai et al.1998). This method is based on particular theoretical determinations: ‘It chooses to sample positions in a hierarchical structure, rather than sampling ego-centred interpersonal ties. To the extent that social capital reflects embedded resources in the structure, then this approach should yield meaningful information regarding job-seeker's access to such structurally embedded resources’ (Lin et al. 2001: 63, emphasis added).

To use this method, we developed a list of occupations that range from high to low in the hierarchical structure of the Swedish society, and asked respondents whether or not they know anyone in each of these occupations. Under this method, occupations are assumed to reflect important aspects of contact persons’ social location such as their power, class background and education status (Lin 2001b). The measure of social capital includes 15 occupational categories. For prestige scores of accessed positions, we followed the Standard International Occupational Prestige Scale (SIOPS), a prestige rating constructed by Harry B. G. Ganzeboom and Donald J. Treiman (1996). The occupations in this study were: medical doctor (with SIOPS score of 78), civil engineer (70), business manager (70), senior local government official (63), local politician (63), secondary teacher (60), primary teacher (57), registered nurse (54), local government official (52), foreman (46), machine worker (40), hospital orderly (42), office clerk (37), child-care worker (23), and cleaner (21). The categories are quite varied in general ways, so the measures of social capital in this study resembled those used in previous studies (Lin and Dumin 1986, Lai et al. 1998). On the other hand, these varieties were given specific indicators for access to information and resources which were valuable in the context of our study, that is, a public sector organisation delivering services. Thus, this is a measure of social capital, both in the general sense of access to a wide range of useful resources, and in the sense of access to resources useful in Malmö's municipal services.⁴ The position generator questions were: ‘Of your relatives, friends, or acquaintances, is there anyone who has the jobs listed in the following table?’ If the response was ‘yes’, the respondent was asked if she or he knew this person at the time when she or he was looking for the current job. If the response was again in the affirmative, the respondent was asked what the relation was between the respondent and the occupier of the position.

From the data, four variables were constructed: (i) extensity; (ii) upper reachability; (iii) heterogeneity; and (iv) the average or typical resources.⁵ It must be emphasized that only those positions were considered as accessed positions that respondents knew at the time when they were looking for their current job. Previous research indicates that these four measures are highly correlated and tend to form a single factor (Lin 2001b: 15). Measures of position data were also highly correlated in our data. In line with earlier studies, we assumed that variables’ extensity, upper reachability, range and average were observable measures of the unobserved variable ‘social capital’⁶ (Lin et al. 2001), and that the indicators of social capital were ‘effect indicators’, which means that the latent variable social capital caused the observed variables (Bollen 1989: 65). We performed a factor analysis on these four variables. A factor analysis, as presented in Table 2 (with principal component methodology, varimax rotation, and a criterion of an eigenvalue equal to or greater than 1), yielded a single factor solution. A factor score was constructed with differential weights assigned to the four variables.

Education, measured as the educational level (primary school, secondary school, university undergraduate and university graduate) completed by the respondent at the time of the survey, was included in the analysis of access to social capital because higher education brings the potential to meet other individuals who have better initial positions (for example, parental socioeconomic resources) and better access to social capital (Erickson 2004). In analyses of the return on social capital, educational attainments measured in education years were assumed to be an important control for individual ‘human capital’ (Becker 1964). We used the level of education as an indicator of access to social capital because we were able to compare different levels of education; but for return on social capital we were more interested in comparing the effect of a single education variable with the other main variables, that is, work experience and social capital.

In analysing access to social capital, we used the individual's total work experience (years of the participation in the labour force), and assumed that work experiences enrich individuals’ social contact. This variable and education have been conventional variables which indicate individual's human capital (Mincer 1974). In the ‘Mincer earnings function’, in addition to the number of years of labour market experience there is also a quadratic on experience that captures the concavity of the earning profile. Hence, when estimating the return on social capital we have both the number of years of labour market experience and its quadrate.

Marital status was included because social capital is assumed to be greater for those with a partner, since one meets some of the one's partner's contacts (Erickson 2004). Having children under 18 is included because caring for children potentially brings one into contact with other adults involved in care for children. On the other hand, Erickson (2004: 42) suggests that adults involved with children younger than 18 years ‘are often women (one's children's friend's mothers, the children's teachers or daycare workers, and so on). These women add little to network diversity for women respondents’. Another control variable for access to social capital is active membership in voluntary associations. Association activity is assumed to be an important pathway to more contacts (Stoloffe et al. 1999). Moreover, membership of trade unions was added because it is a conventional variable in estimating labour market outcomes in the public sector (Borjas 1996: 392). We included this variable, too, in studying access to social capital in order to examine its possible effects. Two main control variables in this study were gender and birthplace, because, as Lin (2001: 95) puts it, ‘inequality of social capital offers fewer opportunities for women and minorities to mobilize better social resources to attain and promote careers’.

The research task of this section is to study differential access to social capital: who has better or worse access to social capital? Or, what characteristics would enhance or hinder access to social capital? To this end, the variable social capital is regressed to the control variables described below. Educational level (primary school as reference, secondary school, university undergraduate and university graduate), work experience (total years of the participation in the labour force), active in voluntary associations (dummy variable), marital status (single or married), gender, and birthplace (native or immigrant).

As presented in Table 3, access to social capital is positively associated with education (with university qualification, both undergraduate and graduate level, as compared with primary school qualification),⁷ and with work experience. Having a partner and active membership in voluntary associations, as expected, are also positively and significantly associated with social capital acquisition. Having a birthplace outside Sweden has a considerable negative effect on access to social capital, which implies a substantial social capital deficit, particularly for ONW immigrants. Whether or not such unequal access to social capital for immigrants translates into disadvantage in labour market outcomes will be examined in the following sections.

The non-significant coefficient of being male rather than female in this estimation runs counter to what earlier studies have shown (see, for example, Campbell and Rosenfeld 1985). But it is probably due to the specification of our sample, which is from a section of labour market with mainly low-paid jobs, and dominated by women and immigrants. It is plausible that newly employed males who try to find a job in such a sector are not among those with more valuable social resources, since otherwise they would search for work in more attractive parts of the labour market.

In addition, with the same procedure as before, we estimated social capital and access to social capital only for immigrants, including duration of residence as a new control variable (see Appendix Table 3). Duration of residence measures the number of years an individual has lived in Sweden since migration and is coded into five, 5-year categorical variables. This variable is generated from information on years since immigration, which was available from register databases of Statistics Sweden. Similar to the previous estimation (for whole sample), access to social capital is positively affected by education and having a partner. But immigrants’ work experience and their active membership in voluntary associations (unlike for total sample) have no significant association with access to more social capital. Work experience is a non-significant variable for immigrants presumably because some work experience has been gained before immigration.⁸ Membership of voluntary associations is a non-significant variable for immigrants possibly because they have tended to be involved in associations that are not mainstream and standard Swedish organisations but have a more immigrant-oriented character. Regarding association between duration of residence and access to more social capital for immigrants, results indicate that a significant positive relationship first appears only after more than 15 years since immigration. This is presumably due to obstacles to regular employment for immigrants in the Swedish labour market, particularly during the economic downturn of the early 1990s (see Nekby 2003 for the effect of duration of residence on the employment probabilities of immigrants in Sweden).

The initial hypothesis is that, following Granovetter's argument (1974), weaker ties (that is, non-kin ties) tend to access better general social capital. In line with earlier research, we divided all contacts of individuals in our sample into three distinct categories; relatives, friends and acquaintances (Lin and Dumin 1986, Lin et al. 2001), and respondents were asked whether the persons they knew in the jobs listed in our questionnaire were relatives, friends or an acquaintances. Following Völker and Flap (1999: 22), if a respondent knew two or more persons in a particular occupation, the instruction was to sign on the strongest tie. The result shows that, acquaintances (the weakest of the three kinds of tie) not only provided the best access to the number of positions that are reachable (extensity), they also provided access to greater resources in the social network (upper reachability), access to the broadest range of positions (heterogeneity) as well as the highest average of typical resources.

In addition, we calculated the percentage of the access of our three groups (natives, NW immigrants and ONW immigrants) to various positions mediated merely by kin ties. Table 5 shows that immigrants are less likely than natives to use kin ties to access positions, presumably either because immigrants have fewer relatives in Sweden than natives or because their relatives do not have such desirable positions in the labour market as their friends and acquaintances.

The research task of this section is to assess the effects of social capital on status attainment. Two labour market outcome variables are used: (I) being in the high wage group and (II) getting an adequate job. For independent variables, we include: years of formal education, total years of working experience, gender (male = 1), marital status (have a partner = 1), having children under 18 years old (dummy variable), membership of a trade union (dummy variable), the job-search method that helped them to secure their current job (informal search method = 1, others = 0), social capital, and birthplace (with Sweden as reference group). These control variables are included in estimations in line with earlier research on the labour market, and are assumed to have associations with labour market outcomes.⁹

At the first stage we examined the effects of the control variables on the wage level of employees. We compared those who were in the high-wage group, that is, those who had a monthly wage more than 20,000 SEK, with those who were not. Table 6 shows logistic regressions predicting the employees who were in the high-wage group.

As Model 1 of Table 6 shows, work experience, years of formal schooling, membership of a union and having children younger than 18 years are generally associated with a greater chance of being in the high-wage group. Men in our sample are almost twice as likely to be in the highest wage groups as are women.

Model 2 includes the use of informal channels: whether or not respondents reported getting their jobs with the help of friends and relatives or from contact with previous co-workers and employers. But the effect of informal job-search methods in our sample and in this case is not significant (and in some other estimations is negative), which means that using informal recruitment channels does not increase (and sometimes decreases) the probability of producing superior labour market outcomes. When we estimate a logit equation predicting a dichotomous variable indicating the use of informal methods (with the same control variables as in Table 3), we find a negative significant effect for university education and a positive significant effect for active membership of voluntary associations. As Lin (2004: 166) puts it, a possible explanation for the lesser likelihood of highly educated individuals using informal job-search methods is that individuals in advantaged socioeconomic positions with access to greater resources in their social networks routinely receive helpful job information in their social circles and are in fact ‘“passive” recipients of job information and offers rendered by social ties in the networks through routine social interactions’. Moreover, in this study, as many other earlier studies, we defined the use of informal job-search methods as using of contacts, either as job-seeker–contact–employer or as job-seeker–employer (‘direct offer from the employer’). But recent research in this field on the ‘job-search network chain’ extends the notion of informal job-search beyond these straightforward and one-to-one connections between job-seeker and employer, and tries to measure the entire chain of contacts between job-seeker and employer (Lin 2004).

The exponentiated coefficient of social capital in Model 3 indicates that social capital plays an important role in improving the chance of being in the high-wage group, and indicates that the impact of social capital is significant and substantial, even after controls for work experience and education.¹⁰

In Model 4 we add the birthplace variables, which demonstrate that natives have a much better chance of being in the high-wage group than do immigrants, given the same levels of human capital and social capital. One conceivable explanation is that the education and work experience of immigrants (possibly not obtained in Sweden) are not valued as highly as those of natives. But when we run the same logistic regression separately for natives and immigrants, the coefficients for education and work experience in our estimations are roughly the same for both groups – which refutes such an explanation in this estimation. From descriptive data presented in Table 1 it is clear that immigrants are overrepresented in the groups with involuntary part-time jobs and non-permanent jobs, which could explain their smaller chances of being in the high-wage group. This in turn means that they successfully obtained a job, but not an adequate one (this is the research task for the next section). Obviously, the social capital deficit of immigrants is important for their lesser chances of being in the highest-wage group, because when we estimate the effect first of birthplace variables and then of social capital, the exponentiated coefficient of immigrant variables increases, which indicates that including the social capital variable improves the chances of being in the high-wage group.

At last, in Model 5 we add the interaction between social capital and ONW immigrant, to test the suggestion that the social capital of ONW immigrants and that of natives have different effects on the probability of being in the highest-wage group. The non-significance of the interaction term in this model shows that there are no different effects for the amounts of social capital that these two groups have access to.¹¹

In this section we examine the quality of employment obtained. To find employment is no doubt the primary purpose of job-seekers, but it is still important to consider the quality of the employment gained. One can join the labour market by getting a job which is marginal in nature (a job with less employment security, an involuntary part-time job, or a job with a low level of remuneration). Consequently, individuals may find themselves working in jobs that which do not match their qualifications. Thus, it is important to examine the quality and adequacy of the new job. Clogg and Sullivan (1983), in their Labour Utilization Framework (LUF), offer a schema which attempts to capture the quality of employment and to distinguish between adequate employment and underemployment. The LUF is composed of several labour force states, which indicate different hierarchical levels of employment hardship. The categories used in Clogg and Sullivan (1983), differentiate between: adequate employment, educational mismatch, involuntary part-time employment, low-paid job, unemployment, and discouragement. Inspired by this framework, we have constructed an index for several states of job quality (job adequacy). Variables entering in this index are;

• 1.

  Wage level: less than 10,000 SEK = 1; 10,000–15,000 SEK = 2; 15,001–20,000 SEK = 3; 20,001–25,000 SEK = 4; 25,001–30,000 SEK = 5; and more than 30,000 = 6.

• 2.

  Involuntary part-time jobs: 1–19 h/week, involuntary = 1; 20–34 h/week, involuntary = 2; and all others (voluntary part-time jobs and full-time jobs),=3.

• 3.

  Employment security: permanent job = 4; probationary employment = 3 project-employment or temporary job = 2; labour market policy programmes = 1.

The variable ‘adequate job’ is constructed by performing a factor analysis (principal component) on the aforementioned three variables. The variable adequate job is then regressed to the same control variables as in the preceding section. Table 7 shows OLS regressions predicting adequate job.

Model 1 of Table 7 shows that work experience, years of education, being male, having a partner and trade union membership¹² are positively associated with the probability of having an adequate job. Model 2 adds whether or not respondents reported getting their jobs with the help of informal recruitment channels. The significant negative coefficient for informal recruitment channels shows that in this sample those who used these types of recruitment channel have a lower probability of having an adequate job. We have already mentioned that disadvantaged groups in our sample tend to use these methods more often.

Introducing social capital in Model 3 shows that social capital is also here associated with a higher probability of having an adequate job.¹³ Model 4 adds the interaction between social capital and using an informal recruitment channel. The non-significant interaction shows that social capital goes with a greater probability of having an adequate job whether or not an informal recruitment channel was used.

Model 5 introduce birthplace variables. Being an immigrant from an ONW country has a significant and substantial negative association with the outcome variable of having an adequate job.¹⁴ This indicates that being born in an ONW country reduces the probability of having an adequate job as compared with being born in Sweden, after control for human capital variables, social capital and other control variables. When we run separate regression for natives and immigrants with the same outcome and control variables, the results demonstrate that immigrants are remunerated less than natives for their education in this estimation.¹⁵ The lesser likelihood of ONW immigrants obtaining adequate jobs seems to be partly explained by Swedish employers lacking the ability to evaluate formal education acquired in ONW countries or attributing lower value to foreign degrees. But, as earlier research shows, a relatively large part of labour-market outcome differences between ONW immigrants and natives remain after grades from Swedish education have been accounted for (Le Grand and Szulkin 2002). Apparently, the social capital deficit of immigrants is here also another important explanation for their lower probability of having an adequate job.

Finally, Model 6 adds the interaction between social capital and ONW immigrant to examine whether social capital has a different effect on immigrants’ and natives’ probabilities of having an adequate job. The non-significance of the interaction term in this model shows that social capital has the same effect on having an adequate job irrespective of the birthplace of individuals in our sample.

Social capital is unevenly distributed, and tends to be higher for those with higher amounts of other kinds of capital. Since it is assumed that social inequality results from inequality of access to various forms of capital (including social capital), it becomes important to clarify the mechanisms whereby inequality of capital arises for different social groups. The results of this study demonstrate that access to social capital was positively associated with more work experience, higher educational levels, active membership of voluntary associations and having a partner. An important limitation in our data was that we had no information about the social origins (parents’ educational, economic and social resources) of the individuals in our sample. Another finding of this study is that being an immigrant was associated with a substantial social capital deficit, which arises because immigrants are embedded in social networks that constrain their ability to gain valuable social resources. The inferior position of immigrants in the Swedish labour market is in part due to this social capital deficit, since such a capital deficit offers fewer opportunities for immigrants to mobilize better social resources and improve their labour market outcomes. More research is needed to shed light on the processes that leads to such a differential access to social capital.

Regarding the return on capital, the results show that both human capital and social capital were rewarded with higher wages and more adequate jobs. Further, we found that social capital goes with better outcomes whether or not a person reported getting the current job with someone's help. But the definition of the use of informal job-seeking methods in this study was not extended beyond the one-to-one connection between job-seeker and employer, since our data do not contain information that allows measuring the entire chains between job-seekers and employers.

As earlier research suggests, return deficit may or may not occur independently of capital deficit. The results of this study show that, despite a substantial social capital deficit for immigrants, there are some indications that the returns on accessed social capital are similar for natives and immigrants. An important limitation that restricted our findings was that we analysed the labour market outcomes only of individuals who were hired. It is plausible that immigrants in our sample are a positively selected group as compared with the entire group of immigrants in the Swedish labour force.

Social capital is so far a theory under development. We need to refine our ideas about how to conceptualize and measure it. Given that social capital plays an important role in hiring and remuneration in labour market, more research is needed to establish how social capital works in the process of recruitment and in different parts of the labour market. Importantly, replication of this study in other part of the Swedish labour market would make it possible to assess whether the associations observed in this study hold in different samples and with different types of data. In addition, more longitudinal research is needed to know how those who get a job actually use the resources embedded in their social network to improve their wage and position within workplaces.

Many thanks to the anonymous reviewers of European Societies for their helpful comments and suggestions. Thanks also to Carl-Ulrik Scheirup, Mahmood Arai and Anders Neergaard for valuable comments.

Alireza Behtoui is a research fellow at the Department of Social and Welfare Studies (SIV), Linköping University, Campus Norrköping/ITUF, SE-601 74 Norrköping, Sweden.