Referral-based Job Search Networks
Christian Dustmann
a,b
Albrecht Glitz
c
Uta Sch¨onberg
a,b,e
Herbert Br¨ucker
d,e
May 2015
Abstract
This paper derives novel testable implications of referral-based job search networks in
which employees provide employers with information about potential new hires that they
otherwise would not have. Using comprehensive matched employer-employee data cov-
ering the entire workforce in one large metropolitan labor market combined with unique
survey data linked to administrative records, we provide evidence that workers earn higher
wages and are less inclined to leave their firms if they have obtained their job through a
referral. These effects are particularly strong at the beginning of the employment relation-
ship and decline with tenure in the firm, suggesting that firms and workers learn about
workers’ productivity over time. Overall, our findings imply that job search networks
help to reduce informational deficiencies in the labor market and lead to productivity
gains for workers and firms.
Key Words: Networks, Referrals, Uncertainty
JEL Classification: J61, J63, J31
a
University College London (UCL);
b
Centre for Research and Analysis of Migration (CREAM);
c
Universitat Pompeu Fabra and Barcelona GSE;
d
University of Bamberg;
e
Institute for Employment
Research (IAB). We thank Marco Hafner and the IAB for their support with the data, and the NOR-
FACE research programme for financial support. Albrecht Glitz further acknowledges the support of the
Barcelona GSE Research Network, the Government of Catalonia, and the Spanish Ministerio de Econom´ıa
y Competitividad (Project No. ECO2008-06395-C05-01, ECO2011-30323-C03-02, and ECO2014-52238-
R).
1
1 Introduction
Several studies show that at least one third of employees have obtained their current
job through family members or friends, pointing towards the importance of informal
social networks in the job search process.
1
Such networks may serve as an information
transmission mechanism and therefore have the potential to enhance the efficiency of
the labor market by reducing informational uncertainties and search frictions. So far,
however, little is known about how job search networks actually operate, and whether
they indeed lead to efficiency gains.
In this paper, we focus on an information transmission mechanism in which employees
refer network members to their employers and thereby provide them with information
about potential job market candidates that they otherwise would not have, as in the
referral models by Montgomery (1991), Simon and Warner (1992) and, more recently,
Galenianos (2013).
2
Similar to Borjas (1992, 1995), Bertrand et al. (2000), and Bandiera
et al. (2009), we define networks in terms of ethnicity. Based on a search model that
encompasses both uncertainty in the labor market and the possibility of hiring through
either formal channels or through the network, we propose novel empirical implications
of referral-based job search networks. We test these implications using both large-scale
matched employer-employee social security data covering all workers and firms in one large
German metropolitan area over a 20 year period, and unique survey data linked to social
security records. Our most conservative estimates show that referrals lead to around 2.5
percent higher initial wages and a 1.9 percentage point lower initial probability of leaving
one’s firm. Consistent with our theoretical framework, both of these initial gains from
referrals decline over time spent in the firm.
Our model builds on the learning-matching model by Jovanovic (1979, 1984). We
extend his analysis by distinguishing between recruitment through networks and through
1
See, for instance, Granovetter (1974, 1995), Corcoran et al. (1980), Holzer (1988), Gregg and
Wadsworth (1996), and Addison and Portugal (2002). For recent surveys of the literature, see also
Ioannides and Loury (2004) or Topa (2011).
2
An alternative way of how information can be exchanged within networks is among potential em-
ployees by informing each other about job opportunities, as in Topa (2001) and Calv´o-Armengol and
Jackson (2004, 2007).
2
the external market, and by endogenizing the probability of obtaining a job through
a referral and relating it to the workforce composition of the firm. In our model, the
worker’s match-specific productivity is more uncertain in the external than in the referral
market. This larger uncertainty implies a larger opportunity for future wage growth, as
workers are partially insured against low realizations of their productivity by quitting their
job (see Jovanovic 1979, 1984). Consequently, referral hires turn down wage offers that
otherwise identical external hires would accept and are therefore initially better matched
than external hires. But since low realizations of the match-specific productivity lead,
over time, to separations of the least suitable workers from their firms, the difference
in match quality, and hence in wages and turnover probabilities, between external and
referral hires declines with tenure in the firm.
We confirm the key predictions of our model based on analysis of two complementary
data sources, using alternative estimation approaches. First, motivated by our theoretical
model and confirmed by empirical evidence from two novel survey data sets linked to social
security records, we use the share of workers from the same ethnic group in the firm at
the time of the hire as a proxy for a referral hire. Second, we use direct information on
referrals obtained from linked survey-social security data. Both approaches show that
referral hires earn higher wages, but experience slower wage growth, than external hires
once we account for the non-randomness of workers’ sorting into firms and use of referrals.
Furthermore, referral hires are initially less likely to leave their firms than external hires,
but this effect also declines with tenure in the firm.
According to our most conservative estimate, we compute that uncertainty in the
referral market is 41.4 percent lower than in the external market, and that referrals,
through the provision of additional information to employers, increase total welfare in the
economy by 0.62 percent. Overall, the combined evidence from the large-scale matched
employer-employee data and the smaller linked survey-social security data provide strong
evidence for the hypothesis that, through referrals, job search networks help to reduce
informational deficiencies in the labor market and lead to productivity gains for workers
and firms.
3
Our paper is related to the literature on job search networks. Most of the existing
evidence on such networks comes from surveys where workers are asked how they found
their current job (see Ioannides and Loury, 2004, for an excellent overview). While the
widespread use of social contacts in the job search process is a consistent finding, the
evidence on its effect on wages is mixed. For instance, while Marmaros and Sacerdote
(2002) report that individuals who received help from fraternity/sorority contacts were
more likely to obtain high-paying jobs, Bentolila et al. (2010) find significant wage
discounts for jobs found through family and friends.
3
One concern in this literature
is that both employees and employers who rely more on networks in their job search
process may not be randomly selected. An important contribution of our paper is that
the longitudinal nature of our data allows us to eliminate any potential bias due to the
fact that low productivity workers, or low productivity firms, may use networks in their
job search process more or less intensively than high productivity workers and firms. Our
results illustrate that addressing selection is indeed important, and that the different ways
to deal with it may be one reason for some of the contradictory findings in the literature.
While most studies rely on worker surveys to analyze job search networks, two recent
studies by Brown et al. (2014) and Burks et al. (2015) use data from a single (or a set
of) firms that include explicit information about whether or not a new hire was referred
by a current employee, and find evidence in line with ours. Since these papers compare
the wage and turnover behavior of referred and non-referred workers within the same
firm, they are, like us, able to account for the possible non-random selection of firms
into the recruiting method. We add to these studies by additionally accounting for the
possible non-random selection of workers, by investigating the effects of a referral on wage
and turnover trajectories for a representative set of firms, and by providing a theoretical
framework that allows us to interpret our findings in a concise manner.
Other recent research does not use direct information on the job search method used,
but instead provides indirect, yet compelling, evidence on the existence of job search
3
Other papers with varying findings include Holzer (1987), Kugler (2003), Loury (2006) and Pa-
tel and Vella (2013). Pellizzari (2010) provides an overview of wage differentials between jobs found
through informal and formal methods in a number of European countries, and Topa (2011) provides a
comprehensive survey of this literature.
4
methods. For instance, while Bayer et al. (2008) and Hellerstein et al. (2011) show that
network members cluster together in the same work-location or firm, Kramarz and Skans
(2013) find that firms are more likely to hire children of current employees than otherwise
comparable job candidates. In a similar spirit, Oyer and Schaefer (2009) present evidence
that partners hire graduates from their own law school with a much higher probability
than randomization would predict. We complement these studies by analyzing ethnicity-
based (as opposed to location-, family- or education-based) networks and go beyond these
papers by presenting novel evidence on the productivity of networks.
4
A number of recent papers provide, like our paper, both a theoretical and empirical
analysis on the use of networks in the labor market, but focus on different mechanisms
than us. For instance, while Hensvik and Skans (2013) systematically test and provide
support for Montgomery’s (1991) referral model whereby referrals allow firms to attract
workers with high unobserved (to the market) productivity, Heath (2013) provides evi-
dence that firms use referrals in order to mitigate limited liability (moral hazard) prob-
lems. Schmutte (2014) develops a search model in which workers who are connected to
workers earning high wages are assumed to draw from a better wage offer distribution
than workers who are connected to workers earning low wages, while Goel and Lang
(2009) focus on the effects of networks on wages which arise through the number of job
offers strongly and weakly connected workers may receive. Bandiera et al. (2009) and
Beaman and Magruder (2012) study how favoritism or the type of referral changes in
response to different incentive schemes.
The structure of the paper is as follows. In the next section, we set up a referral
model that forms the basis of our empirical analysis. After describing the data and main
minority groups, we provide comprehensive survey evidence of the relevance of ethnicity-
based networks in the job search process in Germany in Section 3. We then explain
our empirical methods in Section 4 and report the corresponding results as well as their
4
In a field experiment for a specific online market, Pallais and Sands (2014) show that referred workers
perform better on the job than all other workers who applied for the job. Although consistent with
ours, Pallais and Sands’s findings suggest that referred workers are more productive than the average job
applicant, while our findings (and those of Brown et al., 2014, and Burks et al., 2015) suggest that referred
workers are (initially) more productive than workers who obtained the job through other channels.
5
