Jcl22(4).book(jcl001.fm)

VOLUME 22
SEPTEMBER 2006
353–368
DOI:10.1093/esr/jcl001, available onliOnline publication 28 April 2006 Social Networks and Labour
Market Outcomes:
The Non-Monetary Benefits
of Social Capital
We contrast Granovetter’s hypothesis (Granovetter, M. (1973). American Journal of Sociol-
ogy, 78, 1360–1380; Granovetter, M. (1974). Getting a Job: A Study of Contacts and
Careers. University of Chicago Press; Granovetter, M. (1995). Getting a Job: A Study of Con-
tacts and Careers. University of Chicago Press) that social networks help individuals to find
better-paid jobs with a new model, which predicts that networks are helpful with respect
to non-pecuniary job characteristics but not concerning the monetary pay-offs. Following
Montgomery (Montgomery, J. D. (1992) American Sociological Review, 57, 586–596), our
model is a combination of classical job-search theory and the network hypothesis. First,
concerning the monetary consequences, we test our hypotheses empirically by analysing
the 2001 International Social Survey Programme on social relations and support systems.
We show that using social ties is a common job-search strategy in all countries. However,
using social networks does not increase the monetary pay-off. Second, we use a sample of
8,000 Swiss university graduates who recently entered the labour market to show that
informal job-search channels are beneficial with respect to important non-monetary job
characteristics. Thus, graduates who received their jobs through social contacts tended to
get jobs that are linked to their educational degree and offer better career perspectives.
Furthermore, using personal networks is related to lower search costs. Therefore, the
results suggest overall that networks improve the non-pecuniary characteristics but not the
monetary pay-offs.
Introduction
First, he proposed that many employees find their jobsthrough social contacts and not only through formal Since Granovetter’s (1974) Getting a Job, the question of channels such as direct applications, employment how individuals find jobs and what effect social contacts agencies, or job advertisements. Second, according to have on the labour market has emerged to be one of the Granovetter, the use of social networks allows job seek- most interesting and controversial research questions in ers to gather better information about the availability labour market research. Granovetter’s (1973, 1974) of jobs as well as job characteristics. This informational central ideas can be summarized as three hypotheses.
advantage should enable job seekers to select better The Author 2006. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journal FRANZEN AND HANGARTNER
jobs. Hence a job found through the network should Are Granovetter’s Hypotheses
result in a better match, that is, higher wages and Refuted?
higher job satisfaction. Third, information about thelabour market can best be generated through weak ties.
Granovetter’s (1974) original study is based on a sam- The advantage of weak ties as opposed to strong ties ple of 282 professionals, technical, and managerial lies in the fact that the information in close-friendship workers1 living in Newton, Massachusetts, who were circles is rather redundant and similar and that more interviewed by him and partly surveyed by written new information is generated by networks whose questionnaires in 1969. Fifty-six per cent reported that members are dispersed and dissimilar.
they found their jobs through social contacts. This res- Granovetter’s first proposition has been confirmed in ult has been reconfirmed repeatedly in many studies in many studies. Most empirical research shows that a sub- the United States (U.S. Department of Labor, 1975; stantial proportion of individuals find their jobs via their Corcoran et al., 1980; Marsden and Campbell, 1990; contacts with friends, relatives, colleagues, or acquaint- Staiger, 1990) as well as in Great Britain, Japan, and ances. However, hypotheses two and three are contro- The Netherlands (see Afterword in Granovetter, 1995).
versial. Extensive reviews by Granovetter (1995) himself Some differences between studies appear from the fact as well as others (Lin, 1999) suggest that most empirical that sometimes only active job seekers are taken into studies were not able to confirm the wage bonus. This account. However, there is evidence that social con- conclusion is nurtured by recent results presented by tacts also play an important part in those matches in Mouw (2003) who concludes that contacts have no which respondents received an offer from an employer causal effect on labour market outcomes.
without prior search. These respondents are often These results raise the question whether the social excluded from the analyses, a procedure that results in embeddedness of individuals has any consequences at all some biases as Granovetter convincingly argues. Stud- on labour market outcomes. We argue that they have ies that also pay attention to the ‘non-seekers’ show consequences and will present theoretical arguments as that in about 80 per cent of these cases, a friend or rela- well as empirical evidence, which support the notion tive was involved. Hence, Granovetter’s proposal that that contacts matter. Jobs found through social contacts networks are involved in about half of all job matches have non-monetary benefits, particularly a better match between employees’ education and the job require- Granovetter’s second and third hypotheses, that jobs ments. Furthermore, social networks reduce the search found through social contacts are better paid and more satisfying for employees and that weak ties are better The remainder of the article is organized into four than strong ties, are very controversial. Granovetter sections. Are Granovetter’s Hypotheses Refuted? starts (1974) found that 54 per cent of those who found their with a brief review of existing findings, namely that most jobs through contacts reported to be very satisfied with studies show that networks do not matter with respect to their work compared to 30 per cent who found their earnings. We then refer to job-search theory and to work through formal methods. Similarly, a larger pro- Montgomery (1992) in order to explain these non- portion (ten percentage points) of the former is found in findings. Moreover, we formulate an extension of his the higher income group. However, these findings were model and propose that job offers obtained through only replicated by a few studies (Corcoran et al., 1980; social networks are superior with respect to non-monetary Staiger, 1990; Wegener 1991; Coverdill, 1994; Jann, characteristics. The ISSP 2001 describes our first data 2003), while many others could not detect a wage differ- source and the results obtained. We analyse the ISSP ential (Lin et al., 1981; Bridges and Villemez, 1986; 2001 data and show that jobs found through social con- Marsden and Hurlbert, 1988; Preisendörfer and Voss, tacts are not superior with respect to payment. In The 1988; Lin, 1999; Mau and Kopischke, 2001). Some stud- Swiss Graduate Survey, we refer to another data source, a ies (De Graaf and Flap, 1988; Flap and Boxmann, 2001) survey of university graduates who entered the labour even find a negative wage effect for social contacts. Fur- market in 2001 to show that search strategies are related ther evidence against Granovetter’s hypothesis has also to the non-monetary job characteristics. In particular, been presented recently by Lin (1999), and Mouw (2003) we evaluate the outcomes of different job-search strate- who concludes ‘I believe the weight of anecdotal evidence gies with respect to earnings, educational adequacy, and und intuition suggests that being “well connected” is search costs. Finally, the last section concludes and dis- an advantage in the labor market (.). At the moment, cusses our findings and shortcomings.
SOCIAL NETWORKS AND LABOUR MARKET OUTCOMES
intuition und anecdote aside, we have little empirical shows that the expected wage from weak-tie offers may be, counter-intuitively, lower than the wage expected However, this conclusion holds only with respect to from strong-tie offers. We extend Montgomery’s argu- the direct effect, that is, for wages of those jobs that were ment by applying them to the difference between formal offered with the help of the social network. The litera- and informal search channels instead of the difference ture on social resources (see Lin, 1999) also demon- between weak and strong ties. First, we assume that most strated that individuals in high job positions are found individuals use both formal (direct applications, answer- to have a large social network as well. Hence, the two ing job advertisements, placing an advertisement, using findings that on the one hand high levels of social capital a labour office) and informal (social contacts) search are correlated with high-income jobs, but that using the channels. Second, we assume that the wage distributions network on the other hand does not affect the wage level of both channels are identical. Thus, at least in principle, constitute a paradox. Mouw (2003) suspects that high most available jobs can be found by various search chan- job positions and size (and quality) of social networks nels and are not exclusively restricted to one specific are merely associated and that both depend on unob- search method. Third, we assume that the offer rate of served individual characteristics. Hence, he speaks of informal channels is higher than the one from formal spurious social capital. However, an alternative interpre- search methods. This assumption basically follows tation is given by Montgomery (1992) who offers an Granovetter who asserts that information about new job interesting combination of economic job-search theory opportunities is particularly efficiently (fast) transported and Granovetter’s network hypothesis.
through network ties as compared with formal search The difficulty of choosing a job is that job offers do modes. In order to receive a formal offer, a job seeker not arrive simultaneously but sequentially in time (e. g.
first has to find a job advertisement and has to issue a Lippman and McCall, 1976). Thus, a job seeker is con- formal application. This procedure takes more effort fronted with the following decision problem: either to and is more costly than receiving the information from a accept an offer and stop searching or to reject the offer network tie and applying with the help of the tie. Thus, and continue searching. Since searching is costly (direct we assume that almost every person receives one or costs and opportunity costs), a worker who maximizes more offers through the social network channel and only lifetime earnings will accept an offer of wage wR (or fewer offers through the formal channel. The problem higher) if this offer exceeds his value of leisure and if he now is that most of the time the number and quality of does not expect to find a higher wage offer that compen- job offers are unobserved. Instead what researchers (and sates for the continued search costs (see Mortensen, we) observe is only the accepted job and the search 1986). wR is called the reservation wage. Obviously, the channel through which it was found. Thus, it could be higher an individual’s reservation wage the longer his or the case that an individual who accepted an offer her search time until he or she finds a wage offer that received through the formal channel passed on other matches the reservation wage. Moreover, the reservation offers from the social network. Hence, we infer that a wage depends (among other things) on the arrival rate of seeker who accepted a formal offer had on average more job offers. The more offers an individual expects the offers to choose from and was therefore better able to higher is his or her reservation wage and the higher the select the best offer. Those who accepted an offer from probability of finding a better-paid job. Following Mont- the social network are on average likely to have had gomery (1992), one way to interpret the effect of social fewer offers to choose from, which results in a lower capital on wages is via the reservation wage. Individuals with larger networks (or alternatively a higher propor- Wages are of course not the only characteristics of tion of weak ties) may expect to receive more job offers, jobs. This notion is well known and accepted by many which increases the reservation wage. This indirect effect researchers. Particularly, we assume that in addition to of social networks on earnings is in line with empirical wages, workers consider how well they fit into a job in findings reported by Lin (1999) and Mouw (2003).2 terms of their interests and qualifications. One indicator Networks not only can affect the reservation wage but concerning the quality of the match is how well a can also have direct implication via the job-search strat- worker’s education and qualifications fit the require- egies. This is Granovetter’s crucial hypothesis who does ments of the job. Let us call this the educational ade- not consider the indirect effect via reservation wages.
quacy (a) of a job. We propose that among the jobs Montgomery (1992) interprets Granovetter in the way offered that are above the reservation wage, workers that weak ties elicit more job offers than strong ties and choose the one that best meets their qualifications.
FRANZEN AND HANGARTNER
While we assume that the wage distributions of the However, they differ in educational adequacy in such formal and the informal offer distributions are identical, a way that workers who found their job through the net- the job adequacy distributions of both search channels work should have jobs that match their qualification are not. Our fourth and new assumption is that the ade- better. Note that our argument that jobs found through quacy distribution of jobs offered through social net- the network have higher adequacy but no wage advant- works should be stochastically superior concerning the age implies that adequacy and wages are not positively first and third moment of the distribution (see Figure 1).
correlated (ceteris paribus, particularly human capital).
More specifically, jobs from the formal offer distribution There is often a trade-off between initial wages when might be right skewed with respect to adequacy, because entering the labour market and adequacy. However, this on average more jobs have low than high adequacy. How- counter-intuitive argument is in line with human capital ever, the offer distribution from the social network should theory (Becker, 1964; see also Acemoglu and Pischke, be skewed to the left side, since networks offer more ade- 1999). Thus, jobs that offer general on-the-job training quate than inadequate jobs. The rationale behind this should have relative low initial payment and steeper proposition is that the network is rather well informed earning profiles since employees have to compensate about the job seekers’ interest and qualifications and employers for their training. Hence, the degree of gen- selects jobs with higher adequacy. Alternatively, it could eral training and job adequacy should be positively cor- be argued that networks are usually homogenous, which related. As will be shown below, the first implication might also result in more adequate job offers. Of course, (the non-positive correlation between wages and ade- information about the educational requirements of jobs is quacy) can be tested with our data. However, we have no usually also transferred via formal channels. However, it data to test the second implication, that is, the positive seems reasonable to assume that the information available correlation between adequacy and general training.
through networks is more detailed and more specific thaninformation received through formal channels.
With respect to wages, jobs that are offered through The ISSP 2001
the network should not be superior to the formal channel-offer distribution. When members of a network transmit The ISSP 2001 was conducted on social relations and information about jobs to a seeker, they are probably support systems in 28 countries.3 Next to some socio- not very well informed about the wage of the job nor do demographic information (earnings, education, and they know a worker’s reservation wage. Wages are often work experience), the surveys contain questions about the result of negotiations between employer and the number of respondents’ friends and how they found employee. However, the network is usually very well their present jobs. Thus, participants were asked to informed about the qualification and education of a report the number of close friends at the work place and worker, and it filters jobs in such a way that it offers in their neighbourhood, and other close friends. Fur- what it believes to be a good match. Hence, our expecta- thermore, they were asked ‘Please indicate how you first tion is that jobs do not differ in wages depending found out about work at your present employer’. We on whether they were found with the help of the net- grouped the answers into strong ties if participants named family members, other relatives, or close friendsas contacts. Answers were grouped into the categoryweak ties if respondents said acquaintance. Table 1 dis-plays the percentages of strong- and weak-network con-tacts as well as the number of valid cases for theparticipating nations of the ISSP 2001. Overall, we canobserve a substantial degree of variance. Proportions ofnetwork contacts are comparatively high in the southernEuropean countries (Italy, Hungary, and Cyprus) as wellin some developing countries such as the Philippinesand Brazil. Relatively low proportions are observable inthe Scandinavian countries. The United States, Japan,and Germany are in the middle. With few exceptions Figure 1 Job offer distributions from networks and formal
(most eastern European countries) and contrary to channels. Note: a refers to job adequacy.
expectation, the proportion of strong-tie contacts is SOCIAL NETWORKS AND LABOUR MARKET OUTCOMES
Table 1 Job placement via social networks
Strong ties (%)
Weak ties (%)
N
NA, not available.
Data source: ISSP 2001, own calculations. larger than that of weak-tie contacts. However, this We had to drop both countries from our analysis so might be partly due to our coding of other relatives into that we end up with 15 remaining nations. The esti- the strong-tie category. Nonetheless, the descriptive mated coefficient for education in model 1 of Table 2 impression presented in Table 1 confirms Granovetter’s tells us that every additional year in education is notion that on average a substantial amount of jobs are rewarded on average by 7.9 per cent ((exp(0.076)–1) × 100) increase in hourly wages. Also, the other results, the pos- Next we analyse whether network size and job place- itive but concave effect for work experience, confirm the ment via social contacts are related to wages. Such analy- well-known results of the standard income regressions.
ses are presented in Table 2. The first model is a standard Next, in model 2, we introduce the network indicators Mincer income OLS-regression model controlling for and two dummy variables if respondents found their country-specific differences by country dummies. In jobs through strong or weak ties as compared with order to compare the hourly incomes, we transformed formal methods (reference group). First, the more the national wages into purchasing power parity (PPP) friends respondents have at work and the more other units. Since the transition from the original currencies friends they report to have, the higher their hourly wage.
into PPP units can be done much more reliably for Thus, these results are in line with the hypothesis of OECD members, we restrict our analysis to these. Fur- Montgomery (1992) that those with more contacts on thermore, some countries (e.g. Austria, Norway) have average expect more job offers, which raises their reserva- missing data with respect to some central variables tion wage and finally also their realized wage. Counter- (income, education, or network indicators).
intuitive is the negative effect of the number of friends in FRANZEN AND HANGARTNER
Table 2 Extended Mincer-type wage regressions
P-values in parentheses, computed with Huber–White-corrected standard errors.
*Significant at 0.05; **significant at 0.01.
~Under control of country dummies.
Models 1 and 2 are estimated by ordinary least squares, model 3 is estimated by maximum likelihood. In all three models, the following OECD countries are included: Australia, Canada, Czech Republic, Denmark, Finland, France, Germany, Great Britain, Hungary, Japan, New Zealand, Poland, Spain, Switzerland, and the United States. the neighbourhood. However, more wealthy people confirm former empirical findings (Mouw, 2003) that job might live further apart from each other in suburban placement through social contacts is not positively associ- neighbourhoods, which may reduce their neighbourhood ated with higher wages. Furthermore, our analysis of the contacts. The data also confirm the second part of Mont- data confirms a hypothesis by Montgomery (1992) that gomery’s (1992) argument, namely that respondents who the size of the social network is positively related to wages.
accepted a job offered via a strong tie have on average alower wage. This negative effect can also be observed forweak ties confirming our extension of Montgomery’s The Swiss Graduate Survey
(1992) model. Thus, respondents who accepted an offerthrough the network (weak or strong) either had a lower Our second data source is a survey of all Swiss university search time or did not receive as many formal job offers.
graduates, which has been conducted biannually by the Both causes lead to a lower number of total job offers, Swiss statistical office since 1977. This data is collected from which has the consequence of reducing the realized wage.
university graduates one year after graduation via written Finally, model 3 in Table 2 is a multilevel model in which questionnaires and is concerned with respondents’ the country dummies are replaced by country-specific entrance (first job) into the labour market. We analyse the covariates that should affect wages, such as GDP per cap- newest available data of respondents who graduated in ita, GDP growth, and the labour force participation of 2000.4 In this year, 12,447 graduates left the universities.
women (LFB). All three coefficients are positive and sta- They were contacted about 12 months later via a written tistically significantly related to wages. We also tested two questionnaire. A total of 8,151 graduates returned the ques- cross-level effects between GDP growth and placements tionnaire, constituting a response rate of 65 per cent.5 via strong and weak ties. Both interaction effects are not The data have some advantages that make them par- significantly related to wages. Overall, the ISSP 2001 data ticularly suitable for an analysis of our propositions.
SOCIAL NETWORKS AND LABOUR MARKET OUTCOMES
Since all respondents are labour market entrances, they are The results we present here concern the sample that at the start of their career, and the sample is homogeneous entered the labour market in the second half of 2000 with respect to their working biography. Many other (which was a prosperous economic year in Switzerland).
investigations into the effects of network contacts on job However, we also analysed the interviews from 1995, 1997, characteristics, such as our own of the ISSP, use cross- and 1999 of graduates entering the labour market during sectional data from the entire working population and more recessive periods. We did not find substantial differ- have to control for career-specific effects, such as the last ences, which suggests that our results do not seem to position before the present job was entered, the segment depend on the specific economic situation in 2000.
of the labour market, the level of on-the-job training, The survey contains information on respondents’ whether respondents changed employer, and so forth.
search strategies, search costs, labour market outcome, Such heterogeneity of individual working biographies and educational adequacy. Each of these variables is makes it more difficult to isolate the effect of network measured by several indicators. The labour market out- contacts on the job in question. Also, Granovetter (1995: come is measured by wages and additionally by the occu- 154) discusses that past positions may have been found pational position (with managing responsibilities as by network contacts but not necessarily the present one.
opposed to without). Search costs are measured by the However, since the past position influences the charac- search time, the number of job applications, and the teristics of the present job, networks can have an indirect number of job interviews, and by a subjective measure of effect on present positions. This indirect effect of net- how difficult respondents perceived the job search to works is usually not taken into account so that the net- have been. Educational adequacy is measured by four indicators, most importantly by the specificity of the Another advantage of our data is that the survey took degree that the employer required. Additionally, the place 12 months or less (the median is 3 months) after questionnaire contains a few subjective measures, that is, respondents entered the labour market. Hence, biases whether respondents believe that the job is a long-term due to memory problems that are usually present in ret- engagement, which offers career perspectives, the extent rospective questioning should be less of a problem in to which respondents can use their ability, how they per- this data. Information about job searching can be inac- ceive the possibility to exert influence, and how well the curate in representative surveys that contain a cross- wages correspond to their qualifications. Hence, the data section of the entire labour force for those respondents allow for an analysis of the effects of search strategies on who have not experienced a job shift for a longer time the labour market outcome, the search costs, and the A further difficulty of analysing the effects of social For the analysis of wages, we restrict the analysis to capital is the dependence of the labour market on eco- those graduates who received their first university degree nomic cycles. Granovetter (1995) supposes that strong in 2000 and had entered the labour market by the time ties are more important during economic recessions of the interview.6 The questionnaire distinguishes 12 dif- than weak ties because strong ties feel more obliged to ferent job-search strategies: for example, graduates may help their friends or relatives in difficult times. Another have applied directly, asked different employment agencies possibility mentioned in Granovetter (1995) is that (official employment office or one from the university), employers have more bargaining power during economic responded to job advertisements in the media, or placed recessions than workers and may determine the job an advertisement themselves. Graduates may also have match weakening the influence of networks. Theories of contacted friends, relatives, or colleagues or have looked labour market segmentation suggest that ‘social closure’ for jobs by asking professors and former employers they is stronger during recessive cycles, which would increase the importance of personal contacts (Preisendörfer and In 2001, 25.2 per cent of the graduates reported that Voss, 1988). Some evidence of the dependence between they received a job offer without prior search. The most the economic situation and the shape of the labour mar- common job-search strategy among the graduates in ket was presented by Osberg (1993) with Canadian data.
Switzerland is direct application (50.2 per cent) followed He found that more unemployed used social networks by formal search strategies (46.9 per cent) and the help during times of higher unemployment. However, at the of personal networks (40.5 per cent).7 More important same time, the proportion who found a job through the than the question which strategies were employed is the network decreased, possibly because a larger proportion question which ones were successful. Figure 2 shows that about a quarter of the graduates found a job through FRANZEN AND HANGARTNER
Notice: The question wording was: „Which of the strategies you used was decisive in finding the job?“. Figure 2 Proportion of successful search strategies, 1995–2001.
each of the search channels, that is, formal, informal, business administration or economics are higher than and direct applications. Moreover, these proportions the ones from social sciences, history, or language. Fur- remain fairly constant over time. Comparing 1995 with thermore, the highest wages are observed among gradu- 1999, the importance of social contacts decreased a little.
ates from universities in the German-speaking part as However, in 2001, the search via social networks was the compared with the French and Italian part, which decisive route into the labour market for 19.6 per cent of depends on the shape of the regional labour markets and the graduates. Calculating the success ratio by dividing has little to do with the quality of the universities.8 the number of individuals who found a job by the First of all, the results reveal that searching is number who used a given strategy reveals no substantial rewarded. Respondents who searched for a job receive a differences (direct application 44.8 per cent, formal 4 per cent wage bonus as compared with those who search 49.5 per cent, and social contacts 45.3 per cent).
accepted an offer by an employer without prior search Summarizing the first part of our descriptive analysis (see Table 3, model 1). However, wages do not increase again confirms Granovetter’s (1974, 1995) first hypothe- with increasing search time. This finding is not consist- sis. A substantial proportion of individuals find their ent with job-search theory, which assumes that individ- jobs due to the help of their personal networks. This res- uals with a higher reservation wage should search longer ult replicates that of studies conducted in the United and realize a higher wage. However, job-search theory States (Young, 1974; Sagen et al., 1999).
makes the (highly unrealistic) assumption that job The interesting question is whether job matches through searchers know the offer distribution (see Mortensen, social networks are beneficial as compared with other 1986). The zero effect could be the result of the mixture strategies. Table 3 shows the results of an OLS regression of two types of individuals in our sample: those who of the logarithm of the hourly wages. Presented are the search and find the better-paid jobs and those whose effects of different search strategies controlling for other reservation wage is higher than what the market is will- mostly socio-demographic influences such as respond- ing to offer and who, therefore, have difficulties finding ents’ age, sex, or nationality. The analysis also controls for a job that meets their reservation wage. More important the effects of different universities and subjects of study.
with respect to our hypothesis is the negative effect on However, we do not show the latter effects to keep the wages if the job was found with the help of social net- table readable. As is usual, wages for subjects such as works. Jobs that were found through social contacts pay SOCIAL NETWORKS AND LABOUR MARKET OUTCOMES
Table 3 The influence of social networks on hourly wages, occupational position, and educational adequacy
Income regression
Managing position
Educational adequacy
*Significant at the 5% level; **significant at the 1% level.
Depicted in model 1 are the unstandardized coefficients from OLS regression. Numbers in parenthesis denote the standard errors of the estimates.
Model 1 is an OLS regression with the logarithm of hourly wages as the dependent variable. The model controls for university dummies and for subject dum-mies, which are not displayed due to place restrictions. The university dummies consist of Basel, Berne, Fribourg, Geneva, Lausanne, Neuchâtel, St. Gallen, Ticino, ETH Zurich, EPF Lausanne with the University of Zurich as the reference. Subjects are distinguished into Theology, Language, History, Social Sciences, Law, Natural Science, Medicine, and technical subjects. Economics is used as the reference category.
Model 2 is a logistic regression. The dependent variable is coded as 1, if graduates received a job with management responsibilities, and 0 otherwise.
Model 3 is an ordered-probit model. The dependent variable is the educational adequacy coded as 1 if employer did not require any university degree, 2 with only a general university degree, 3 if a university degree from similar subjects were also accepted, and 4 if the employer required a specific degree. on average 5 per cent less as compared with jobs found jobs for which the employer demands a specific degree through formal search channels. Thus, also our results as compared with more general university degrees are on show that searching via social contacts has no monetary average paid 5 per cent less. The results also show that advantage (Mouw, 2003) and, moreover, might even women have a 4 per cent wage disadvantage in the have negative effects (see De Graaf and Flap, 1988; Flap labour market.10 Children increase the wage by 7 per cent, which can be explained with social benefit pay- Also, the number of different strategies people use to ments employers have to make. A small positive effect of find a job does not affect wages. This finding is also 2 per cent can also be observed for the graduates’ age inconsistent with job-search theory. A more intensive and a 3 per cent income advantage for work experience search should increase the number of job offers, which acquired during university enrolment. The education of in turn should increase the chance of finding a better- a respondent’s father or mother does not affect a gradu- paid job. However, this again might be due to the mix- ate’s wage level at labour market entrance. Hourly wages ture of two groups of individuals, namely, those who are also not affected if respondents work only part-time find well-paid jobs by searching and those whose chances are worse to begin with and who are therefore In addition to wages, we also analysed the occupa- tional position at which graduates entered the labour In addition to the effects of search strategies, model 1 market. The questionnaire contains a dichotomous variable also controls for educational adequacy and certain that indicates whether individuals received a position socio-demographic effects. First, as hypothesized, job with or without management responsibilities. Assuming adequacy is negatively correlated with wages. Hence, that management positions have more occupational FRANZEN AND HANGARTNER
prestige, it is expected that offers from the network 0.05 and 0.10 significance level), indicating that jobs should lead to management positions more often. Since found via the network are perceived to offer more we deal with a dichotomous variable, model 2 in Table 3 opportunities for personal influence and ability.12 shows the logistic-regression coefficients. However, only Network jobs, however, are negatively associated with direct applications lead to jobs that start in a managing the perceived adequacy of payment (model 4 in Table 4).
position significantly less often. Apart from this, two Thus, this finding corresponds rather well to the com- further significant effects emerge from the model: man- paratively lower hourly wages reported in Table 3. Also, aging positions are more often obtained by older gradu- noteworthy are the positive effects of the work experi- ates and less often open for part-time employment.
ence graduates acquired during their study. Graduates Thus, model 2 in Table 3 echoes the results obtained for who worked while still enrolled at the university have a better chance of finding a job that is related to their sub- Finally, in model 3 in Table 3, we attend to the ques- ject of study. Obviously, this work experience also tion whether network contacts increase the probability increases graduates’ knowledge of where to find ade- of receiving a job with higher educational adequacy.
Graduates were asked whether their current employer Finally, we will take a look at the search costs. If required no university degree at all, only a general degree, Granovetter’s (1974, 1995) and our models are correct, one from a related subject, or a specific university degree.
graduates who use networks for job searching should We assume that jobs that do not require a university receive job offers more often and sooner. Thus, the degree or only an unspecific one are less adequate for search time should be reduced for all those who use their graduates and less preferred by them. Since the depend- networks. This hypothesis is supported by our analysis ent variable has four categories that can be ordered, of the search time. We analysed the search time until model 3 in Table 3 presents the results of an ordered- respondents found a job by event-history analysis (more probit model. The results suggest that search strategies particularly we use a Weibull model), which takes right- matter. Respondents who searched (coefficient of 0.16), censored cases (respondents who were still looking for a received a job through social networks (coefficient of job at the time of the interview) into account. Model 1 in 0.20), or applied directly (coefficient of 0.14) report to Table 5 shows the effects on the hazard rate of leaving have jobs that more often required an adequate degree the stage of search and entering employment. Thus, as compared with respondents who found jobs through graduates using the network have an increased hazard formal search channels.11 This result confirms our rate of 17 per cent ((exp(0.16)–1) × 100) as opposed to hypothesis that friends, relatives, and colleagues seem to those who use formal job-search strategies. Moreover, pay attention to a graduate’s educational qualifications we analysed two more indicators of the search costs, when informing about job opportunities.
namely, the number of applications and the number of The adequacy of jobs that were found through social job interviews individuals went through before accept- networks is also better in the respondents’ own percep- ing a job. Since these variables are count data, we analyse tion. The participants of the survey were asked whether them using negative-binomial models (see Cameron and they view their current job as a temporary means to earn Trivedi, 1998).13 The estimation results suggest that money or as a long-term career investment, as well as graduates who used the network wrote 11 per cent fewer how well they are able to exert influence and apply their applications and went through 11 per cent fewer job abilities on the job. Clearly, jobs that are more adequate interviews. Thus, models 2 and 3 confirm the results we to individuals’ interest and education should be viewed obtained through our analysis of the search time.
more often as long-term investments and should offer Our fourth model (in Table 5) contains the analysis of better opportunities for personal influence and ability.
respondents’ perception whether they encountered diffi- The results of the analysis are displayed in Table 4. The culty during the job search. This indicator also reflects logistic regression (model 1 in Table 4) shows that grad- the results we already obtained from models 1 through 3 uates who found their job through the network (as well of Table 5. Respondents who used the network have a as direct applicants) have a higher chance of finding jobs lower chance (the odds are reduced by 0.66) to report with a long-term career perspective (the odds increase difficulties. Models 1 through 4 in Table 5 also control by exp(0.43) = 1.54). Models 2 and 3 contain the results for a number of socio-demographic effects that are pos- of exerting influence and using their abilities at the work sibly associated with the search costs. A few systematic place. In both models, the non-standardized OLS-regression and crucial results emerge from the control variables. Thus, coefficients are significantly positive (0.17 and 0.07 on a women and academics who are looking for part-time SOCIAL NETWORKS AND LABOUR MARKET OUTCOMES
Table 4 Social networks and subjective indicators of job adequacy
Career investment
Apply ability
Exert influence
Perceived payment
†Significant at the 10% level; *significant at the 5% level; **significant at the 1% level.
Numbers in parenthesis denote the standard errors. The regressions contain but do not show dummies for universities and dummies for the subject of study.
Model 1 is a logistic regression. The dependent variable is coded with 1 = long-term employment intention with possibility of upward mobility and 0 = short-term employment with no upward mobility.
Model 2 is an OLS regression. The dependent variable measures respondents’ rating of the adequacy of job concerning the possibility of using their knowledge and ability.
Model 3 is an OLS regression. The dependent variable is the perceived adequacy concerning the possibility of having an impact on the job.
Model 4 is an OLS regression; the dependent variable contains the rating of the perceived adequacy of payment. Same results were obtained for models 2 through 4 if an ordered-probit was applied instead of an OLS regression. employment seem to encounter higher search costs. Obvi- which may at first seem counter-intuitive, might stem ously, the number of search strategies respondents used from the fact that those who accepted an offer through increases the number of applications and job interviews.
network contacts could have overall received less offersor, alternatively, had a shorter search time, which conse-quentially results in a lower realized wage.
Summary and Discussion
Furthermore, we extended Montgomery’s (1992) model by assuming that the distribution of job offers This article tries to demonstrate that social networks from networks is superior to the distribution of job matter for finding a job. First, we test some implications offers due to formal channels with respect to educational of Granovetter’s (1974, 1995) and Montgomery’s (1992) adequacy. We tested our hypotheses by using the Swiss hypotheses by analysing the ISSP 2001 data. Three basic Graduate Survey. Overall, four results emerge from this findings emerge from this analysis. First, a substantial analysis: First, also in this survey, a substantial propor- proportion of individuals report that they found their tion of individuals report that they found their first job job through network contacts. Second, individuals with by network contacts. Second, we analysed the hourly a larger number of friends (particularly friends at the wages and were not able to discover a wage bonus for work place) indeed report to have a higher income. The individuals who had accepted an offer through the net- effect can be explained by Montgomery’s (1992) work. More specifically, controlling for the search time assumption that a larger network increases respondents’ those who accepted an offer through the network had reservation wage and consequentially their income.
even a monetary disadvantage. Thus, our analysis con- Third, however, jobs that are directly found with the firms other findings (De Graaf and Flap, 1988; Flap and help of a network tie are not better paid. This result, FRANZEN AND HANGARTNER
Table 5 Search strategies and the cost of job search
Duration of
Number of
Number of job
Difficulties in
applications
interviews
job search
†Significant at the 10% level; *significant at the 5% level; **significant at the 1% level.
Numbers in brackets denote the standard errors. The regressions contain but do not show dummies for universities and dummies for the subject of study.
Model 1 is a Weibull model of the hazard rate to enter employment.
Models 2 and 3 are negative-binomial models with the number of applications and the number of job interviews as dependent variables.
Model 4 is a logistic regression analysis, the dependent variable indicating whether respondents report difficulties in finding a job. Third, our results suggest that jobs found with the entrances. Other results suggest that the influence of net- help of friends, colleagues, or relatives have a higher works should decrease, since during recessions a larger educational adequacy. Thus, employers more often proportion of an individual’s network should be unem- require a specific university degree for jobs that are ployed as well. In order to exclude the possibility that found by network contacts. Moreover, respondents our results depend on the good health of the economy in more often view jobs found over the network as long- 2000, we also analysed the data of graduates who entered term engagements compatible with their career plans the job market during recessive times in 1995, 1997, and versus short-term employment that has little or no rela- 1999. However, we obtained almost identical results tion to career plans. The notion that educational job from the other three data sets as well.
adequacy is higher is also supported by respondents’ To summarize, the acceptance of a job offer through evaluation of job characteristics. Thus, network jobs are the network seems to have non-monetary advantages for more often perceived as offering the opportunity to labour market entrances. The help of the network increases the chance of an appropriate match concern- Fourth, the analyses show that searching via the net- ing respondents’ education and the type of work. At the work saves search costs. Hence, respondents who found same time, our results replicate that finding a job their job through the network did so faster, applied less through the network has no monetary benefit. Thus, the often, and went through a lower number of job inter- results support Montgomery’s (1992) model as well as views. Therefore, searching via the network has some our extension of it that the distribution of job offers monetary benefits regarding individual’s lifetime earn- from the network is stochastically dominant with respect ings. However, these benefits are small (on average to educational adequacy. Our results also suggest that search time is about two weeks shorter) given an indi- graduates face a trade-off between adequacy and wages.
Higher adequacy is associated with lower entrance As discussed above, labour market outcomes may wages. Employers who look for specific university depend on the business cycle. Some research suggests degrees seem to provide general on-the-job training that networks become more important during reces- more often, which is associated with a steeper wage–age sions since the market is more closed towards new SOCIAL NETWORKS AND LABOUR MARKET OUTCOMES
Both of our data sources have particular advantages 5. The data are available from the Swiss Statistical
but also some disadvantages. Thus, our analysis of the Graduate Survey does not allow for a test of the weak-tie versus strong-tie hypothesis. The Graduate Survey has 6. This excludes graduates who received a second
no information concerning what type of contacts are best degree (e.g. a dissertation) and those 4.1 per cent with respect to finding adequate jobs. Furthermore, it graduates who were still looking for a job at the time contains no information on the size or other features of individuals’ networks. We can also not exclude the pos- 7. These proportions correspond closely to those
sibility that individuals using the network for their job reported by Young (1974), which are based on a sur- search have some unobserved characteristics that deter- vey of 750,000 university graduates in the United mine the outcome of the search instead of the used States. For a similar result concerning direct applica- search strategy. However, the data at hand do not support this possibility. Thus, none of the graduates reported to 8. The detailed results of universities and subjects’
have only used a single strategy, and estimating one’s hourly wages can be obtained from the authors.
chance to use the network was only significantly related 9. Note that the estimated effect of the use of social
to the number of strategies used, as well as to gender networks on wage might be biased if those who (men use networks more often than women). However, found their jobs via social networks are different in overall the McFadden R2 of a probit model was too low some characteristics not controlled for in the regres- (0.04) to corroborate the obtained estimate into a treat- sion equation from individuals who did not use ment-effect model (Greene, 2000: 933). Thus, clarification social contacts. One possible procedure to correct of the problem of possible unobserved heterogeneity as for this possible bias is the estimation of a switchingregression model (Wooldridge, 2002). However, in well as the question to which extent the results can be our case, the estimation of the selection equation generalized to a larger proportion of the labour market (probit model) did not show any fundamental dif- has to be left to further research. However, we believe that ferences between the two groups. The detailed the analyses presented here draw attention to the non- results can be obtained from the authors.
monetary benefit of social networks on the labour market.
10. If women have a lower labour market participation
than men, the analysis needs a correction (Heckit corrections; Heckman, 1979) to obtain an unbiasedestimator. However, female participation in our 1. Most of the people interviewed did have a university
sample is about 90 per cent, so that such a correction 2. Note that Mouw (2003) calls this ‘spurious social
11. The calculation of the exact change in probabilities
capital’. This term implies that it cannot be inter- would require further transformations of the coeffi- preted causally. However, the argument via the res- ervation wage implies a causal explanation. It does 12. The extent to which respondents believe that they
not need to be either the one or the other but may can apply their ability or exert influence on the job well be a mixture of a spurious relation and a causal was measured on a four-point-rating scale varying effect via wage expectations. Networks could also from very much to not at all. Our OLS regression increase the reservation wage by increasing the value assumes interval measurement. However, the results remain robust even if we apply ordered-probit mod- 3. The data are available from the Swiss Information
and Data Archive Services for the Social Sciences 13. Usually count data require the analysis via Poisson
regressions. In our case, the so-called assumption of 4. Switzerland has 12 universities, 6 in the German-
equidispersion is not fulfilled so that we apply the speaking part (Universities of Basel, Berne, St. Gallen, negative-binomial model. The assumption of equid- Lucerne, Zurich, and the Swiss Federal Institute of ispersion is fulfilled if a = Var(y|x)/E(y|x) does not Technology in Zurich), 5 in the French-speaking significantly deviate from 1. In model 2 a = 1.5 and part (Universities of Fribourg, Geneva, Lausanne, in model 3 a = 1.2 indicating over-dispersion.
Neuchâtel, and the Swiss Federal Institute of Tech- However, the estimation results do not differ sub- nology in Lausanne), and 1 in the Italian-speaking stantially between the Poisson model and the FRANZEN AND HANGARTNER
Acknowledgements
Lin, N. (1999). Social Networks and Status Attainment.
Annual Review of Sociology, 25, 467–487.
We are indebted to Norman Braun, Cedric El-Idrissi, Lin, N., Vaughn, J. C. and Ensel, W. M. (1981). Social Josef Hartmann, and Peter Preisendörfer for their help- Resources and Occupational Status Attainment.
Social Forces, 59, 1163–1181.
Lippman, S. and McCall, J. (1976). The Economics of Job Search: A Survey. Economic Inquiry, 14, 155–189,
347–368.
References
Marsden, P. and Campbell, K. (1990). Recruitment and Selection Processes: The Organization Side of Job Acemoglu, D. and Pischke, J. S. (1999). The Structure of Searches. In Breiger, R. (Ed.), Social Mobility and Wages and Investment in General Training. Journal Social Structure. New York: Cambridge University of Political Economy, 107, 539–572.
Becker, G. S. (1964). Human Capital. A Theoretical and Marsden, P. and Hurlbert, J. (1988). Social Resources Empirical Analysis with Special Reference to Educa- and Mobility Outcomes: A Replication and Exten- tion. New York: Columbia University Press.
sion. Social Forces, 66, 1038–1059.
Bridges, W. P. and Villemez, W. J. (1986). Informal Mau, W.-Ch. and Kopischke, A. (2001). Job Search Meth- Hiring and Income in the Labor Market. American ods, Job Search Outcomes, and Job Satisfaction of Col- Sociological Review, 51, 574–582.
lege Graduates: A Comparison of Race and Sex. Journal Cameron, C. A. and Trivedi, P. K. (1998). Regression of Employment Counselling, 38, 141–149.
Analysis of Count Data. Cambridge, UK: Cambridge Montgomery, J. D. (1992). Job Search and Network Com- position: Implications of the Strength-of Weak-Ties Corcoran, M., Datcher, L. and Duncan, G. (1980). Infor- Hypothesis. American Sociological Review, 57, 586–596.
mation and Influence Networks in Labor Markets.
Mortensen, D. T. (1986). Job Search and Labor Market In Duncan, G. J. and Morgan, J. N. (Eds), Five Thou- Analysis. In Ashenfelter, O. and Layard, R. (Eds), sand American Families: Patterns of Economic Handbook of Labor Economics. New York: Elsevier Progress. Ann Arbor, MI: Institute for Social Mouw, T. (2003). Social Capital and Finding a Job: Do Con- Coverdill, J. (1994). Personal Contacts and Post-Hire Job tacts Matter? American Sociological Review, 68, 868–898.
Outcomes. Manuscript, Department of Sociology, Osberg, L. (1993). Fishing in Different Pools: Job-Search and Job-finding Success in Canada in the Early De Graaf, N. and Flap, H. (1988). With a Little Help 1980s. Journal of Labor Economics, 11, 348–386.
from My Friends: Social Resources as an Explana- Ports, M. (1993). Trends in Job Search Methods, 1970–92.
tion of Occupational Status and Income in West Monthly Labor Review, 116, 63–37.
Germany, The Netherlands, and the United States.
Preisendörfer, P. and Voss, Th. (1988). Arbeitsmarkt Social Forces, 67, 452–472.
und soziale Netzwerke. Die Bedeutung sozialer Kon- Flap, H. and Boxmann, E. (2001). Getting Started: The takte beim Zugang zu Arbeitsplätzen. Soziale Welt, Influence of Social Capital on the Start of 39, 104–119.
the Occupational Career. In Lin, N., Cook, K. S. and Sagen, B., Dallam, J. and Laverty, J. (1999). Job Search Burt, R. S. (Eds), Social Capital: Theory and Research.
Techniques as Employment Channels: Differential New York: Aldine de Gruyter, pp. 159–181.
Effects on the Initial Employment Success of College Granovetter, M. (1973). The Strength of Weak Ties.
Graduates. The Career Development Quarterly, 84,
American Journal of Sociology, 78, 1360–1380.
Granovetter, M. (1974). Getting a Job: A Study of Con- Staiger, D. (1990). The Effect of Connections on the Wages tacts and Careers. Chicago and London: University and Mobility of Young Workers. Cambridge, MA: Granovetter, M. (1995). Getting a Job: A Study of Con- U.S. Department of Labor. (1975). Job Seeking Methods tacts and Careers, 2nd edn. Chicago and London: Used by American Workers. Bureau of Labor Statis- Greene, W. H. (2000). Econometric Analysis, 4th edn.
Wegener, B. (1991). Job Mobility and Social Ties: Social London: Prentice Hall International.
Resources, Prior Job, and Status Attainment. Ameri- Heckman, J. (1979). Sample Selection Bias as a Specifi- can Sociological Review, 56, 60–71.
cation Error. Econometrica, 71, 153–167.
Wooldridge, J. M. (2002). Econometric Analysis of Cross Jann, B. (2003). Old Boy Network. Militärdienst und Section and Panel Data. Cambridge, MA: MIT Press.
ziviler Berufserfolg in der Schweiz. Zeitschrift für Young, A. (1974). Labor Market Experience of Recent Soziologie, 32, 139–155.
College Graduates. Monthly Labor Review, 97, 33–40.
SOCIAL NETWORKS AND LABOUR MARKET OUTCOMES
Authors’ Addresses
Dominik Hangartner, Institute of Sociology, Univer- sity of Bern, Lerchenweg 36, CH-3012 Bern, Axel Franzen (to whom correspondence should be addressed), Institute of Sociology, RWTH Aachen,Eilfschornsteinstr. 7, D-52062 Aachen, Germany.
Emai Appendix 1
Table A1 Measurement of variables, means, and proportions of the Swiss Graduate Survey
Swiss Institute of Technology in Lausanne Adequacy of university degree concerning job FRANZEN AND HANGARTNER
Table A1 (continued)
Perception of difficulties during the search The category History also includes Philosophy, Archaeology, History of Art, Ethnology, Music, Theatre, and Film. Social Sci-ences include Psychology, Pedagogic, Sociology, Social Work, Political Science, and Media Science. Other subjects include Ecol-ogy, Sport, and Military Science. Natural Sciences include Mathematics, Astronomy, Physics, Computer Science, Geography, Chemistry, and Biology. Economics also includes Business Administration.

Source: http://www.sociologia.unimib.it/DATA/Insegnamenti/3_2293/materiale/franzen.pdf