Optimal policy and the role of social contacts in a search model with heterogeneous workers

• Date: 01 October 2017
• Author: Anna Zaharieva,Yuliia Stupnytska
• Published date: 01 October 2017
• DOI: http://doi.org/10.1111/jpet.12249

Received: 16 January 2017 Accepted: 18 March 2017

DOI: 10.1111/jpet.12249

ARTICLE

Optimal policy and the role of social contacts

in a search model with heterogeneous workers

Yuliia Stupnytska1AnnaZaharieva2

1BielefeldUniversity and University of Paris 1

2BielefeldUniversity

YuliiaStupnytska, Center for Mathematical Eco-

nomics,Bielefeld University, 33501 Bielefeld,

Germany,and CES, University of Paris 1, 75013,

Paris,France (ystupnytska@uni-bielefeld.de).

AnnaZaharieva, Center for Mathematical Eco-

nomics,Bielefeld University, 33501 Bielefeld,

Germany(azaharieva@wiwi.uni-bielefeld.de).

Weare grateful to Herbert Dawid, Jean-Olivier

Hairault,Barbara Petrongolo, and the audiences

atEBIM-EDEEM Jamboree (2013), the annual

conferenceof the Italian Association of Labour

Economists(2013), the 8th Workshop in Eco-

nomicTheory in Bielefeld University (2013),

theannual congress of the European Economic

Association(2014), and the annual meeting of

theEuropean Association of Labor Economists

(2014)for useful comments and suggestions.

This paper develops a search model with heterogeneous workers

and social networks. High-ability workers are more productive and

have a larger number of professional contacts. Firms can choose

between a vacancy in the regular market and a job opening in the

referral market. The model predicts that a larger number of social

contacts is associated with a larger wage gap between high- and

low-ability workers and a larger difference in the unemployment

rates. The net welfare gain of referrals is estimated at 1.2%.There

are three reasons for the inefficiency of the decentralized equilib-

rium. First is the traditional search externality described by Hosios.

Second, firms share their profits with workers, whereas the social

optimum implies that full surplus should be given to firms in the

referral market. This inefficiency can be internalized by means of

referral subsidies. Third, there is the “pooling inefficiency” in the reg-

ular market. If high-ability workers are sufficiently more productive

they impose a positive externalityon low-ability workers and should

be rewarded by positive transfers. On the contrary, if productiv-

ity differences are small, high-ability workers achieve unreasonably

high wages by referring each other and reduce employment chances

of low-ability workers.In this latter case, high-ability workers should

be penalized. If optimal policy is implemented the net welfare gain of

referrals rises up to 1.8%.

1INTRODUCTION

Workers are born with heterogeneous abilities. Later this leads to different productivities when they enter the labor

marketand start searching for jobs. This creates uncertainty for firms since they open vacancies prior to meeting work-

ers.In a standard search and matching model (Pissarides, 2000) job creation depends on the average productivity of the

laborforce, which means that a larger fraction of high-productivity workers leads to more open vacancies and improves

employment chances of workers, including those with low productivities. In this paper, we show that this result may

be violated if high-productivity workers form professional networks and refer each other for open positions, which is

known as informal job search (Montgomery, 1991). Intuitively,better outside opportunities due to professional con-

nections raisewages of productive workers and reduce expected profits of firms. We provide a formal characterization

Journal of Public Economic Theory.2017;19:957–985. wileyonlinelibrary.com/journal/jpet c

2017 Wiley Periodicals,Inc. 957

958 STUPNYTSKA AND ZAHARIEVA

of this negative externality,which has been neglected by other studies, and analyze conditions under which low types

gain/lose welfare from the presence of high types in a search and matching context. More formally,we investigate the

impact of social networks and referrals on social welfare and wage inequality in a frictional labor market with hetero-

geneous workers.

Thereis strong empirical evidencethat 30%–60% of new hires find jobs through personal contacts (see, for example,

Bentolila, Michelacci, & Suarez, 2010; Granovetter,1995; Kugler, 2003; Pelizarri, 2010; Pistaferri, 1999; Staiger,1990,

for different countries). In addition, Hensvik and Skans (2013) report that incumbent workerswith a high test score are

more likely to be linked to the new hires than low-ability employees.In particular, in their data firms rely on referrals

from high-ability workers in order to attract applicants with higher unobservedability. To incorporate these empirical

findings into the model, we assume that high-ability workers are linked in a network and have an identical number of

professional contacts who can give a reference for the job. In contrast, low-ability workers do not have professional

contacts and are restricted to search for jobs in the regular labor market. Even though this approach is highly stylized,

it captures the intuitive idea that more able workersare more likely to be referred for the job.

The choice of search methods by firms is endogenous. When entering the labor market, firms decide between a

high-cost vacancy in the regular job market and a low-cost informal job opening in the referral market. The pool of

job applicants in the regular labor market is mixed, because both types of unemployed workersapply for the publicly

advertised positions. On the contrary, the pool of applicants in the referralmarket is limited to unemployed workers

with high ability as only these workers are connected in a network. This assumption is in line with the original idea of

Montgomery (1991) that workers hired through social networks are on average more productivethan job applicants

hired through the formal channel of search.

Tokeep the model tractable, we assume that the worker type is immediately observed by the firm upon the match.

Thus there is no asymmetric information in the model and wages are negotiated ex post between the firm and the appli-

cant by means of the individual Nash bargaining. Depending on the parameters,there are two types of equilibria. If the

number of social contacts is low it is not optimal for firms to rely solely on referrals as the probability of hiring in the

referral marketis relatively low. In this situation, there exists a unique equilibrium without referrals where both types

of workers are mixedin the regular labor market. In contrast, if the number of social contacts is sufficiently large, then

some firms prefer to use referrals in the hiring process, so both search channels are active in the equilibrium. In the

numerical part of the paper,we estimate the net welfare gain of referrals at 1.2%.

Focusingon the equilibrium with referrals, we find that it is inefficient for any value of the bargaining power. Specif-

ically,there are three sources of inefficiency. First, inefficiency stems from a classical search externality in the regular

market;that is, the number of vacancies in this market is distorted if worker’s bargaining power deviates from the tradi-

tional Hosios value (see Hosios, 1990; Pissarides, 2000). Intuitively,every new open vacancy in this market changes the

search behavior of unemployed workersand imposes a negative externality on competing vacancies. This inefficiency

is also present in the equilibrium without referrals and it is neutralized for the Hosios value of the bargaining power.

The second source of inefficiency is specific to the referral market.From a social perspective, vacancies in this mar-

ketshould be created starting from the point where the expected cost is equal to the expected social surplus of the job.

In contrast, in the decentralized equilibrium firms create positions from the point where the expectedcost is equal to

the expectedprofit. We show that expected profits of firms in the decentralized economy are lower than the social sur-

plus if productivity differences between workersare small or the worker’s bargaining power is larger than 0.5. In these

cases, the threshold number of contacts, which is necessary to open the referral market, is higher in the decentralized

economy compared to the social optimum. Thus referral potential is underutilized by the decentralized marketif the

number of social contacts is in between the two thresholds. In the paper, we show that this inefficiency may be miti-

gated by means of employment subsidies in the referral market.In reality, such subsidies can take the form of referral

bonuses which are reimbursed by the state.

Then, there is the third source of inefficiency (pooling inefficiency), stemming from the fact that heterogeneous work-

ers are pooled together in the regular market. On the one hand, more productive workers are linkedby their profes-

sional networks, which improvestheir hiring chances, raising their outside options and hence their wages. This discour-

ages firms from posting vacancies and can be seen as a negative externality onl ow-ability workers.In such a situation,