You are on page 1of 8

The Journal of Socio-Economics 38 (2009) 456463

Contents lists available at ScienceDirect


The Journal of Socio-Economics
j our nal homepage: www. el sevi er . com/ l ocat e/ soceco
Low wages and high unemployment rates: The role of social interactions
in hiring discrimination

Jean-Franc ois Jacques


a,
, Emmanuelle Walkowiak
b
a
LEDA, Paris-Dauphine, Place du Marechal de Lattre de Tassigny, 75775 Paris Cedex 16, France
b
ERUDITE
1
, Universit Paris-Est and Centre dEtudes de lemploi (CEE
2
), TEPP (FR n

3126, CNRS), France


a r t i c l e i n f o
Article history:
Received 29 May 2007
Received in revised form
14 November 2008
Accepted 12 December 2008
JEL classication:
J41
J82
D23
Z13
Keywords:
Discrimination
Social interactions
Unemployment rate
a b s t r a c t
The purpose of this paper is to explain why low-wage workers with identical qualications to higher-
wage workers are more exposed to unemployment. Each worker is considered to belong to a social group
(dened according to his/her gender, age, and nationality). We assume that workers experience both pro-
ductive interdependencies andsocial interactions withinthe rm. Alsointer- andintra-groupinteractions
determine worker productivity, and frictions on the labor market limit the hiring of the most productive
workers. Consequently, externalities acting both within the rm and in the labor market can lead to a
higher rate of unemployment for low-wage workers.
2008 Elsevier Inc. All rights reserved.
1. Introduction
Despite notable improvements in their qualication levels
and/or professional experience, workers who are socially catego-
rizedas young, old, female, or foreigncontinue toexperience higher
rates of unemployment and receive lower wages (OECD, 2003).
These gaps can clearly be seen in trends in participation rates,
unemployment rates, and wage differentials from the 1970s to the
present.
Numerous laws against sexual discrimination notwithstanding,
unemployment rates for women are generally higher than those

We are particularly indebted to J. Scheinkman for his constructive remarks dur-


ing the construction of our model. We would like to thank C. Aubert, A. Clark, N.
Greenan, M. Gurgand, J. Gauti, R. Kierzenkowski, B. Kukla, E. Lavery, T. Weitzen-
blum, B. Burgess and the participants of the REMMI seminar of the Centre dEtudes
de lEmploi and EURiSCO-Dauphine. We also thank C. Sofer for valuable comments
on an earlier version of this paper which was presented at the LIIIth Congrs of the
AFSE (2004). Finally we are grateful to the anonymous referees for their stimulating
comments. We retain full responsibility for any remaining errors.

Corresponding author. Tel.: +33 1 44 05 44 60.


E-mail addresses: jacques@dauphine.fr (J.-F. Jacques),
emmanuelle.walkowiak@univ-paris12.fr (E. Walkowiak).
1
UniversitParis Est, 61, avenueduGnral deGaulle, 94010Crteil Cedex, France.
2
Le Descartes I - 29, promenade Michel Simon, 93166 Noisy-le-Grand Cedex,
France. Tl.: +33 1 45 92 68 00.
for men, and their average wages are lower. For example, in Spain,
the European country with the greatest unemployment gender gap
after Greece (4.5 points), the mean wage for women is 82 percent of
that for men in 2007. Similarly, Portuguese female unemployment
rates are 3 points higher than those for men, and their wages are 8
percent lower. In 2007, these kinds of gaps were typical in the EU
as well as for most other developed countries. As shown in Fig. 1
below, there are only a few exceptions to this general trend, with
a gender employment gap favorable for women (Romania, Latvia,
Ireland and Estonia) or unemployment rates almost at parity for
men and women (Japan, United States, United Kingdom, Germany,
Lithuania).
Foreign workers are also under-represented in the labor mar-
ket (OECD, 2003). In OECD countries, from 1999 to 2000, the
employment-rate gapbetweennon-citizens andcitizens was 4 per-
cent for men and 8 percent for women. According to the European
Labor Force Survey in 1998, this gap was the largest in Belgium,
Sweden and France, where the unemployment rate of foreigners
(non-EU members) was more than 20 points higher than that for
the local population.
These observations suggest substantial inequalities in employ-
ment outcomes between social groups. Any cumulation of these
different social characteristics reinforces the negative effects on
outcomes (OECD, 2003). For example, the French Labor Force Sur-
vey from1998 calculates an unemployment rate for non-EUforeign
1053-5357/$ see front matter 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.socec.2008.12.008
J.-F. Jacques, E. Walkowiak / The Journal of Socio-Economics 38 (2009) 456463 457
Fig. 1. Gap in the unemployment rate between females and males in 2006.
women between the ages of 15 and 24 of 55 percent. Different the-
ories have been developed to explain these inequalities amongst
individuals with the same qualications.
Mortensen (2003) focuses on wage inequalities, and suggests
that these result fromjob-searchfrictions, whichare the direct con-
sequence of asymmetric information regarding wages. Mortensen
concludes that these inequalities do not, therefore, result from
discrimination. In our paper, we use this standard assumption of
labor-market frictions as part of our explanation why similar work-
ers have unequal access to employment.
In contrast to Mortensen, economic theories of discrimina-
tion (Becker, 1971; Phelps, 1972; Arrow, 1973) explain inequalities
in employment access for equally qualied workers. However,
these theories do not explain why discrimination persists in the
long run. To simplify, we can classify theories of discrimination
into two groups.
3
The rst brings together papers in the Becker
(1971) tradition, where discrimination is considered to result from
discriminatory preferences. As a result of this preference for dis-
crimination among employers, employees and/or consumers, rms
that discriminate will not employ the most productive workers.
Therefore, in the long run, discriminatory rms are less efcient
than non-discriminatory rms. In theory, competition should elim-
inate these rms fromthe market and hence discrimination should
progressively disappear. However, this is not the case. Phelps (1972)
and Arrow (1973) developed a second approach known as statis-
tical discrimination. Employers statistically discriminate because
they lack information regarding the productivity and turnover of
their applicants. While the individual productivity of an applicant
is unknown, the productivity of his/her reference group is com-
mon knowledge. Due to the inaccuracy of imprecise observations
of individual productivity, employers tend to assess an applicants
productivity according to their reference group. When reference
groups differ in average productivity, statistical discrimination
3
Athird approachconsiders that institutions shape behaviors or keep some social
groups out of certain professions. As this paper focus on rmbehavior, this approach
will not be developed here. See Altonji and Blank (1999) for a survey of theoretical
and empirical papers using this approach.
implies that some categories of workers, for example women, are
paidless thantheir marginal productivity. During the 1970s women
were generally less educated than men, so that such statistical
discrimination could be explained. Today employers are proba-
bly aware that in developed countries men and women receive
the same level of education, and therefore that this kind of dis-
crimination is groundless. Our paper offers an alternative way of
understanding bothkinds of discrimination. Incontrast to the Beck-
erian view, we show the persistence of discrimination in the long
run; and in contrast to the statistical discrimination model, we con-
sider the productivity of each individual applicant to be perfectly
known, but endogenous as it depends on work organization.
Explaining the persistence of long-run discriminatory behavior
requires an understanding of how employers assess the quali-
cations of the workers they choose to hire. The nature of the
discriminationproblemsuggests that theabilities of certainminori-
ties are systematically underestimated given their qualications or
diplomas. Why are qualications or diplomas assessed differently
for different groups of workers? When considering the criteria that
determine worker selection, it is crucial to understand why worker
qualications are differentiated from their abilities.
We appeal to the socio-economic status of the worker to explain
the persistence of discrimination. The main objective is to illus-
trate how the workers individual socio-economic characteristics
combined with the social composition of the workforce can affect
individual productivity via social interactions. That is, by com-
municating with other workers (who are similar to or different
fromthemselves), workers improve their problem-solving abilities.
The workers qualications and their ability to resolve problems
through worker interactions jointly determine their productivity.
This assumption may be particularly relevant in innovative rms
or rms facing complex and unstable environments. Mintzberg
(1981) has shown that informal communication is an efcient
coordination mechanism that enhances productivity in such envi-
ronments. Empirically, Gant et al. (2003) analyze differences in
information sharing and connections between workers across the
workplace with traditional and innovative HRM systems in sev-
eral production lines in the Steel industry. They show that rms
favor workplace communication between employees in innovative
458 J.-F. Jacques, E. Walkowiak / The Journal of Socio-Economics 38 (2009) 456463
production lines. All workers communicate more extensively to
solve operating problems on these production lines, which achieve
greater productivity levels than do lines where there is less com-
munication.
Our approach follows recent research, which has investigated
the effects of worker diversity on productivity. Work on diversity
within teams has however produced contrasting results. The case
studies in Kochan et al. (2003) on diversity within teams showthat
context is crucial in determining the impact of diversity on perfor-
mance. In Hamilton et al. (2006) teams with more age diversity
were less productive, while those composed of only one ethnic
group were more productive. However, these ndings depend on
the model specication. Greenan and Walkowiak (2005), using
French data, show that a workers ability to communicate depends
strongly on diversity in qualications, age, gender and citizenship
in the rm.
By analyzing the role of social interactions in the production
process, three dimensions of problem-solving abilities
4
are iden-
tied: individual, collective and social. Abilities are deemed to
be individual when based on the qualications of the worker,
whereas they are collective or social when they rely on inter- and
intra-group social interactions. From an efciency point of view,
as inter- and intra-group communications will not be the same
for all workers, both the social characteristics of workers and the
socio-demographic composition of the rms workforce affect their
abilities to solve problems and therefore their individual productiv-
ity, as in Gant et al. (2003).
The paper is organized as follows. Section 2 introduces the con-
cepts used here, and puts them into perspective by comparing
them to discrimination theories. Section 3 outlines the model, and
Section 4 analyses the equilibrium of the model. Last, Section 5
concludes.
2. Social interactions and productive activity: some
theoretical concepts
We focus on one aspect of the labor process to understand the
way in which employers assess abilities: the interdependencies
between workers. In economics, interdependencies at work are
formalized as productive complementarities. Two tasks are com-
plementaryina productionfunctionif the effort involvedinthe rst
positively affects the marginal productivity of the second (Alchian
andDemsetz, 1972). As a result, the productioncontributionof each
individual task cannot be fully identied.
One of the basic assumptions of our model is that workers are in
a situationof bothproductive interdependencies andsocial interac-
tions. The quality of the social interactions canaffect the production
process. Also, productive activity may contribute to the develop-
ment of social interactions. We now consider the consequences of
taking social interactions into account on the economic analysis of
production.
Social interactions rst lead to a process of categorization
among workers. Reference groups are categories resulting from
a social process of categorization, much like those described by
Fryer and Jackson (2003). This process explains how the struc-
ture of social interaction draws distinctions between employees
due to their socio-demographic characteristics. When interactions
occur, employees class others into categories based on observable
characteristics. This type of behavior occurs for efciency rea-
sons: workers cannot remember every detail about every person
4
We do not cover all of the relevant debates, particularly in sociology, over the
concept of ability, but consider this concept in the light of human-capital theory.
In the latter, abilities are attached to individuals, whatever their nature (general or
specic, transferable or not transferable).
they meet, and this kind of broad classication allows individu-
als to store and retrieve salient features quickly and efciently.
Fryer and Jackson show that for a xed number of categories,
an individual looks for cognitive efciency by minimizing the
variance of socio-demographic characteristics within categories.
Consequently, individuals in the same category tend to have similar
socio-demographic characteristics. We consider the characteristics
on which these categories are based (age, gender, citizenship, etc.)
to be socio-demographic, as they are observable by all employees.
These characteristics could also be cultural (for example, language
or religion).
The survey in Williams and OReilly (1998) reveals the effects of
socio-economic diversity on labor efciency and rmperformance.
The work surveyed there has shown that the socio-economic com-
position of the workforce plays a role in both social interactions
and the process of group formation. This composition modies
communication, cohesion and conict within the rm, as well
as decision-making style, which affects overall performance. As
such, the socio-demographic composition of employees can play
an important role in worker productivity. In our model, the refer-
ence groupis translatedinto vectors of norms where values provide
a common language.
When the employees run into problems at work, they interact
by communicating; this improves their problem-solving abilities.
Models formalizing social interactions generally consider a utility
function that depends on the agents actions and also on the mean
action of the reference group to which the agent belongs (Glaeser
and Scheinkman, 2002, and Becker and Murphy, 2000). We fol-
low this tradition, and assume that the ability to solve problems
depends on both the individuals own qualications and the mean
qualication level among other workers in the rm.
When a worker interacts with other workers, he/she draws on
the human capital of the other workers in the rm to solve the
problem at hand, and this will increase his/her own social capital
(i.e. the human capital associated with the interactions). This social
capital in turn, improves the collective problem-solving ability of
all of the workers. Finally, as in Glaeser et al. (2000), we consider
that social interactions lead to the accumulation of social capital,
and that social capital is a component of human capital. Whereas
these authors introduce social capital into the utility function, we
here introduce it into the production function.
Furthermore, following Leana and Van Buren (1999), we sup-
pose that social capital has twocomponents: one emotional andthe
other based on qualications. The emotional component of social
capital reects the fact that the worker is not just required to carry
out certaintasks, but also to be a certaintype of person. We propose
todistinguishthesociabilityof workers fromdifferent communities
as a function of their reference group.
We also assume that the efciency of communication depends
on each workers reference group. In this respect, we formalize two
hypotheses discussed in the social capital literature. For Bourdieu
(1980) and Coleman (1990), closed social networks, where indi-
viduals are in close proximity, favor the development of social
capital because the ows of information and trust are stronger.
In this view, workers in the same reference group may com-
municate more efciently than workers in different reference
groups. But the opposite may also hold. Proximity between work-
ers may lead to an excess of information and to lower efciency
(Burt, 1992). Here the interaction between different workers may
favor the accumulation of social capital in a close proximity set-
ting.
The integration of social interactions into a production function
highlights the collective and social nature of abilities. The external-
ities created by the interactions between workers inside the rm
are added to the externalities generated by competition between
the unemployed in the labor market.
J.-F. Jacques, E. Walkowiak / The Journal of Socio-Economics 38 (2009) 456463 459
3. The model
In this section, we propose a new production function that
includes social interactions. This production function renders the
organizationof labor withinthe rmendogenous due tothe search-
based labor market.
We assume that the economy is populatedby Nagents denotedi.
We consider two exogenous sources of worker heterogeneity: qual-
ications denoted by a, and the reference group denoted by g. To
simplify, we assume that the human capital of an agent is given by
his/her qualications and we assume that there are two reference
groups, g =1, 2. Each agent belongs to only one reference group,
which indicates his/her type. Belonging to this group denes social
proximity between individuals. This social proximity provides a
common language and conveys norms and collective values. These
reference groups imply a similar language, which is not necessar-
ily efcient. We do not assume, a priori, any ranking for the relative
efciencyof intra- andinter-groupcommunication. We dohowever
assume that all of the workers interact.
3.1. The production function when workers interact
Let N
1
and N
2
denote the number of workers of type-1 and 2
within the rm. The total number of workers is denoted by N (with
N=N
1
+N
2
). Total production Y
tot
is the sum of the individual out-
puts. Letting y
g
i
denotes the output of the worker i of type g, total
production is equal to:
Y
tot
=
N
1

i=1
y
1
i
+
N
2

j=1
y
2
j
(1)
Individual output depends on both an externality denoted E and
individual problem-solving ability denoted m as follows:
y
g
i
= m
g
i
E for i = 1, . . . , N and g = 1, 2 (2)
As in Rosen (1982), E is a positive organizational externality that
has a symmetrical effect on the production of all workers. This
relates to the indivisibility of managerial choices. Whereas in the
model of Rosen (1982) the positive scale effect associated with the
hierarchical organization of labor comes from the manager, here
this effect arises from the horizontal interdependencies between
posts formalized via an O-Ring function (Kremer, 1993). This scale
effect does not allow for substitution between quantities and
quality in the chain of production. We have:
E =

N
1

i=1
m
1
i

N
2

j=1
m
2
j

1/N
(3)
In the models of Rosen and Kremer, the match between workers
andrms relies onthe workers abilities, whichare exogenous. Here
abilities are endogenous, depending on the workers qualications
(a) and his/her level of social capital (b). For the worker i of type g,
this ability is:
m
g
i
= a
g
i
b
g
i
for g = 1, 2 (4)
In this paper we focus on inequalities which are not explained
by qualication differentials. Wage residual inequality increased
steadily throughout the 1970s and 1980s, and continued to increase
sharply in the 1990s (Acemoglu, 2002 and Lemieux, 2006). For this
reason, we assume that qualications are exogenous and identical
among workers, a
g
i
= a whatever their type, in order to focus on
inequalities in the access to employment that do not result from
differences in qualications.
5
According to Eq. (4), differences in
problem-solving ability do not only come from worker qualica-
tions. The social capital of eachworker is endogenous. As mentioned
above, we consider two hypotheses about social-capital accumula-
tion: a closed group of homogenous individuals, or an open group
of heterogeneous individuals. We formulate these via a peer effect,
denoted S, which is assumed to be the same for the two reference
groups. Denoting the proportion of workers of type-1 in the rmas
p, and that of workers of type-2 as (1p), the social capital levels b
of the workers i of type-1, and of the workers j of type-2 are written
as
b
1
i
= b
1
= a(Sp +(1 p)) for workers of type 1
b
2
j
= b
2
= a((1 p)S +p) for workers of type 2
(5)
The parameter S can be greater or less than one. In the rst case
(S >1), at a given qualication level, the exchange of knowledge
from social interactions is greater between workers of the same
type (the BourdieuColeman hypothesis). The communication in a
group of homogenous workers in terms of their type (intra-group
interactions) is more efcient than the communication between
heterogeneous individuals (inter-group interactions). In the second
case (S <1), inter-group interactions are more efcient than intra-
group interactions (the Burt hypothesis). Inter- and intra-group
communication is equally efcient when there is no peer effect
(S =1). We exclude the extreme cases where S =0 or S =+, which
correspond, respectively to no effect of communication between
members of thesamegroupor members of different groups. Implic-
itly, everybody communicates.
In Eq. (5), all workers belonging to group 1 (respectively group
2) ownthe same level of social capital. Individual indices couldthen
be omitted. However, workers of type-1 and 2 differ in their level of
social capital. The variable b then depends on the peer effect, and
consequently on the socio-demographic composition of the work-
ers (p) in addition to the inuence of the qualications of the other
workers in the rm. We obtain m
g
i
= a
g
i
b
g
by combining Eqs. (4)
and (5), and we directly see the individual, collective and social
dimensions of the ability to solve problems.
Regarding the individual dimension, ability (m
g
i
) increases with
the individual human capital of the worker (a
g
i
). The collec-
tive dimension of abilities raises the question of the effect of
the socio-demographic diversity/homogeneity of the workforce
within organizations. Social capital (b
g
) depends on the socio-
demographic composition of the workforce (since p and 1p
represent the proportions of each group). Communication will be
more efcient with higher qualication levels, given the socio-
demographic composition of the workforce. Finally, the third
dimension is social, as the value of b
g
depends on the reference
group to which the worker belongs. This means that the social cap-
ital of a worker depends on both his/her reference group and the
percentage of the workforce who are of the same type as her. This
view of abilities ts the empirical evidence (Hamilton et al., 2006;
Greenan and Walkowiak, 2005).
The production function dened by Eqs. (1)(5) therefore takes
into account both productive interdependencies and social interac-
5
This assumption makes the question of worker segregation by skill irrelevant,
since there is no skill variance in our model. Some papers have shown that during
the 1990s, the correlation between the wages (and also seniority) of workers within
rms has increasedinthe US, UKandFrance (for instance Kremer andMaskin, 1996).
The explanation of skill segregation is beyond the scope of our paper. We focus on
other forms of segregationlike gender, ethnicity, and age, as highlighted ina number
of papers (see for instance Bayard et al., 2003).
460 J.-F. Jacques, E. Walkowiak / The Journal of Socio-Economics 38 (2009) 456463
Fig. 2. The production function.
tions between workers. This can be written as follows
6
:
Y
tot
= E

N
1

i
a( a(Sp +(1 p)) +
N
2

j
a( a(S(1 p) +p))

(6)
Fig. 2 below shows total output as the proportion of workers
of type-1 varies, with the same number of workers. When intra-
group communication is the more efcient (S >1), and when there
are fewer workers of type-1 (p<1/2), it is not efcient to increase
their proportion. Onthecontrary, if type-1workers areinthemajor-
ity (p>1/2), productivity rises with their group size. So, when S >1,
the output curve is U-shaped. This representation of aggregate out-
put is fairly similar to the empirical nding of Goldin (2002), who
considered gender segregation in the US. The rm thus always has
an interest in increasing the proportion of the majority group. On
thecontrary, wheninter-groupcommunicationis themoreefcient
(S <1), the curve is hump-shaped. The rm would be better off in
encouraging a mix of workers, i.e. in increasing the proportion of
the minority group.
3.2. Labor-market equilibrium
To determine the market equilibrium, we assume that all the
rms have the same production function given by (6) and focus on
the symmetric equilibrium. Each group has its own labor market
out of which a dual market emerges, as in Saint-Paul (1996). Even if
discriminationis illegal inhiring announcements, the pre-selection
round (in which costs for both workers and the rm are negligible,
i.e. submitting a CV) leads to a segmentation of the labor market.
After this rst selection, only workers of type g continue to compete
for the job catering to group g. Both the vacancies and the workers
are directedtowards a specic labor market. The number of workers
hired H
g
in segment g of the labor market depends positively on the
number of unemployed U
g
and on the number of vacancies open
to that category of workers V
g
. The matching function (Pissarides,
2000) of workers with vacancies is H
g
=m(U
g
, V
g
).
If there is no friction in the labor market, and if S >1, the rm
should hire only workers of type-1 or 2, depending on their respec-
tive productivity. However, frictions prevent such corner equilibria.
On the contrary, if S <1, the rmhires just as many workers of type-
6
As E depends on the proportion of each category of worker in the rm, we only
analyze the symmetric equilibrium.
1 as workers of type-2. As in Saint-Pauls model, the equilibrium
depends on the conditions in the labor market, which are exoge-
nous: wages (w
1
, w
2
),
7
the labor supply of each category (L
1
, L
2
),
andthe efciency of the matching process. Incontrast to Saint-Paul,
the gains the rm obtains for each post are endogenously deter-
mined, and depend on the socio-demographic composition of the
rm (Eqs. (1)(5)). Individual productivity y
g
depends on the orga-
nizational externality E, which is linked to the socio-demographic
organization of the workers inside the rm. As a consequence, the
gain from hiring a worker of type g is endogenously determined by
the matching outside the rm and the interactions inside the rm.
Mathematically, at the market level, two equations determine
the stationary equilibrium. The rst relates to the externali-
ties inside the rm resulting from productive interdependencies
and social interactions between workers. The second relates to
labor-market externalities, which stem from the bilateral search
processes of job-hunting and employee-searching.
The rst equation reects the trade-off of the rm when it
decides to post a vacancy of type-1 or 2. Let
g,t
denote the prob-
ability per unit of time of nding a worker of type g, and J
g,t
the
present discounted value of a job held by a worker of type g, r the
interest rate, and dt an innitesimal interval of time. Knowing that
employees leave the rm at an exogenous rate of q, the value of a
vacancy of type g follows a Markov process:
VAC
g,t
= (1 rdt)[
g,t
dtJ
g,t+dt
+(1
g,t
dt)VAC
g,t+dt
]
with J
g,t
= y
g,t
w
g,t
+(1 rdt)[qdtVAC
g,t+dt
+(1 qdt)J
g,t+dt
] (7)
At the stationary equilibrium (VAC
g,t
=VAC
g,t+dt
and J
g,t
=J
g,t+dt
),
the rmhas no particular preference for one or the other kind of job
g (VAC
1t
=VAC
2t
). Eq. (8) formalizes the trade-off between vacancies
of types 1 and 2:

2
=

1
(r +q)
((y
2
w
2
)/(y
1
w
1
))(r +
1
+q)
1
(8)
Except for the asymptote, the graph of Eq. (8) is a continuous
curve, denoted A, on the (
1
,
2
) locus. It is upward-sloping under
specic conditions
8
(see Appendix A): a greater probability of ll-
ing a job must be compensated by a greater probability of lling
the other type of job. Its position depends on the ratio of individ-
ual prot made from both types of workers (y
2
w
2
)/(y
1
w
1
) =
G
2
/G
1
, making its position dependent on the social interactions
inside the rm (see Fig. 3).
The second equation of the model, which relates to the labor-
market externalities, represents the ow equilibrium in the labor
market. Following Saint-Paul (1996), we assume that the total num-
ber of positions in the economy, K, is xed. Eq. (9) indicates the
7
In this article wages are exogenous: workers in group 1 are paid at a constant
wage w
1
, andworkers of group2 are paidw
2
. The scope of this paper is not to explain
this difference inwages. Different theories explainwage inequalities associatedwith
social characteristics: discrimination theory, age-biased technological or organiza-
tional change, obsolescence of human capital for older workers, and the importance
of domestic work for women associated with a lower reservation wage. Alternative
sociological views highlight the role of social representation in the persistence of
wage inequalities: old people are rigid so they can not adapt easily to new stan-
dards of production, women nd low-paid work which is similar to the domestic
task etc. (Garner-Moyer, 2003). In our model it is possible to endogenize wages by
linking themto productivity, or by formalizing a wage bargaining process, as in Saint
Paul.
8
To obtain an upward-sloping curve, it is necessary for the instantaneous gain
G
2
to be upper-bounded following an innitesimal increase in
1
or
2
. Curve A
represents the equality of expected gains in the two segments of the labor mar-
ket. Curve A falls below the rst 45-degree line if y
2
w
2
> y
1
w
1
, and above if
the reverse holds. Some numerical simulations show that these conditions can be
relaxed (Jacques and Walkowiak, 2005).
J.-F. Jacques, E. Walkowiak / The Journal of Socio-Economics 38 (2009) 456463 461
Fig. 3. The equilibrium.
connection between the two segments of the labor market:
L
1
(
1
) +L
2
(
2
) = K
where (
i
) =

i
+q

i
+h(
i
)q
and h(
g
) = m

1,
U
g
V
g

1
(9)
Curve B illustrates Eq. (9). It is downward-sloping on the (
1
,

2
) locus (see Fig. 3). If the number of vacancies is constant at the
steady state, a rise in
1
(the probability of obtaining a job for a
worker of type-1) means that the number of unemployed workers
of type-1 is greater, as the total number of jobs in the economy is
xed (K). This implies that fewer type-2 workers are unemployed,
and so the probability of getting a type-2 job (
2
) is lower.
The stationary equilibrium is the solution of the system of Eqs.
(8) and (9). Graphically, curve A intersects curve B. The solution
(*
1
, *
2
) determines the proportion of each type of worker within
the rm (p and 1p), and their respective rate of unemployment
u
g
, which increases with respect to
g
:
u
g
=
qh(
g
)
qh(
g
) +
g
=
h(
g
)
h(
g
) +
g
/q
(10)
4. The properties of the equilibrium
In this model, even if the wage of a type-1 worker is lower than
that of a type-2 worker (w
1
< w
2
), type-1 workers canexperience a
higher rate of unemployment.
9
For example, if S >1 and p<1/2, hir-
ing a type-2 worker greatly increases the social capital of the type-2
workers, who already represent the majority group within the
rm. The improvement of their problem-solving abilities increases
their productivity. This implies that their productivity differen-
tial exceeds their wage differential relative to workers of type-1.
Therefore, when intra-group communication is more efcient than
inter-group communication, there is a segregation process within
the rm. Despite equal qualications, the lower-paid workers are
also less in demand because their collective problem-solving abil-
ity is lower. The opposite conclusionpertains whenS <1 andp>1/2.
Evenif w
1
< w
2
, thermcanhaveaninterest inhiringtype-2work-
ers. In fact, when type-1 workers are in the majority, hiring a type-1
worker causes a substantial deteriorationintheir overall social cap-
9
If y
1
w
1
< y
2
w
2
then the equilibrium is established for values such that

1
>
2
(see Eq. (8)) and u
1
>u
2
(see Eq. (10)).
ital and also of their problem-solving abilities. A rm should then
choose to post vacancies of type-2 and hire this category of worker
in order to gain from the positive consequences of a mix in worker
abilities. Finally, in underlining the collective and social nature of
abilities, this model can help to explain why, with equal quali-
cation levels, and lower wages, certain categories of workers may
experience a greater risk of unemployment. In addition, the rm is
not on its production efciency frontier because it does not con-
trol the peer group effect (S is exogenous) and there are frictions
in the labor market. In terms of X-efciency (Liebenstein, 1979),
the degree of X-inefciency comes from externalities inside the
rm (social interactions) and outside the rm (labor-market fric-
tions).
We now turn to comparative statics. Several shocks may affect
the equilibrium. We distinguishbetweenmovements inthe curve A
(which relates to the trade-off between the two types of vacancies
chosen by the rm) and movements in the curve B (which repre-
sents the equilibriumof job ows between the two segments of the
labor market).
Curve A is parameterized by the interest rate r, the quit rate q,
and the ratio of gains. We focus on the determinants of relative
gains (G
2
/G
1
). Any exogenous increase in this ratio implies a down-
ward movement in curve A (see Fig. 4). When the gains associated
with the hiring of a type-2 worker are relatively higher than those
from hiring a type-1 worker, the rm prefers to post vacancies of
type-2. It is then necessary that the probability of lling a type-1
vacancy increase, to compensate for this differential in gains. At the
equilibrium, the rmremains indifferent between the two types of
vacancies.
What are the determinants of G
2
/G
1
? Two parameters are key:
wages and the peer group effect (S). They both affect the unem-
ployment rate and the segregation process within the rm. First,
when labor supply and wages are exogenous, any change in relative
wages will affect G
2
/G
1
. For example, a fall intype-2wages (dw
2
<0)
raises the rms gain from a type-2 worker compared to a type-1
worker. Firms then post more type-2 vacancies and fewer type-1
vacancies. As labor supply remains unchanged, the probability of
providing a type-2 vacancy and the unemployment rate of this cat-
egorydecreases. Acontrario, the scarcityof type-1vacancies implies
an increase in the unemployment rate of these workers and in the
probability of providing this type of vacancy. Secondly, when S >1,
an increase in S raises segregation and the unemployment rate of
the minority employees inside the rm. In our model, the impact of
the peer group S leads to this kind of inertia and helps to explain the
difference in unemployment rates between the different groups of
Fig. 4. The effect of a shock on the curve A.
462 J.-F. Jacques, E. Walkowiak / The Journal of Socio-Economics 38 (2009) 456463
Fig. 5. Effect of the increase in the number of jobs in the economy.
workers. The peer group effect produces a hysteresis phenomenon
in segregation, which does not disappear.
Two shocks may impact the balance of job ows between the
two segments of the labor market. The rst is a change in the num-
ber of jobs in the economy (K). The effects are unambiguous: the
greater the number of jobs, the greater the number of vacancies.
If labor supply remains constant, the probability of lling these
vacancies falls, and curve B moves downward. Consequently, the
unemployment rates of both types of workers fall (see Fig. 5). The
second shock is an exogenous change in the social composition of
the labor supply. This implies a rotation of B. By leaving total labor
supply unchanged, a larger proportion of type-1 workers increases
the slope of B. It becomes more difcult to provide a type-2vacancy,
because there are fewer workers of this type. The unemployment
rate of type-2 workers falls and that of type-1 workers rises.
5. Conclusion
In the job-search model developed in this paper, workers are
matched to a job according to their problem-solving ability. The
latter is not only individual, as is generally assumed in human-
capital theory, but also mostly depends on social interactions
that generate collective learning. The efciency of these inter-
actions depends on the workers reference group. Consequently,
the socio-demographic composition of the rms workforce deter-
mines aggregate productivity. Taking into account the individual,
social and collective dimensions of abilities helps us to understand
why there is greater unemployment among low-wage workers.
Their productivity may be low because the socio-demographic
composition of the workforce does not favor them: they cannot
accumulate collective problem-solving abilities. As it is rational
from an employers point of view to hire people who communi-
cate more easily to solve problems, discrimination can persist in
the long run. When intra-group communication is more efcient
than inter-group communication, employers prefer perfect socio-
demographic homogeneity in their workforce. However, frictions
in the labor market prevent employers from hiring only one type
of worker. By way of contrast, when intra-group communication
is less efcient than inter-group communication, employers prefer
perfect socio-demographic heterogeneity in their workforce, and
consequently discrimination should disappear.
The key role of the socio-demographic composition of the work-
force underlines that it is important to distinguish the effects of
segregation from those linked to discrimination. Segregation and
discrimination both lead to unequal opportunities for workers as
a function of their social characteristics (gender, ethnicity, age,
etc.). The policies implemented to counter these inequalities are
not, however, the same. In our model, an afrmative-action policy
is not efcient if communication between communities of work-
ers is not improved. In this case, such a policy could generate
counter-productive effects and aggregate productivity might fall.
The key question raised is to understand how to favor communica-
tion between communities in order to encourage a transition from
segregation to integration.
The efciency of communication between workers is socially,
culturally and historically determined. However, economic factors
also play a role in improving inter-group interactions. Firstly, by
equippingall workers withnewcommunicationtechnologies, rms
could improve their inter-group communication. The communica-
tion network of a worker depends to a large extent on the use of
information technologies (Gant et al., 2002). Secondly, as shown
in the IBM case analyzed by Thomas (2004), the heterogeneity of
the consumer base requires an innovative spirit within the rm
that could be boosted by worker heterogeneity. Better access to
the right market could potentially provide incentives to rms to
increase the diversity of their teams, and consequently to improve
communications between the communities inside the rm.
Appendix A. Existence of the equilibrium
Let the curve A be dened by the equation:

2
=

1
(r +q)
R(r +q +
1
)
1
where R is equal to the ratio of the gain of type-2 workers G
2
and
type-1 workers G
1
:
R =
a(S(1 p) +p) w
2
a(Sp +(1 p)) w
1
.
This curve passes through the origin and has an asymptote:
If R > 1 lim

2
=
r +q
R 1
If R < 1 lim

1
=
R(r +q)
1 R
B is a downward-sloping curve with two asymptotes repre-
sented by the following implicit equations:
L
1
1 +

1
/q

1
/q +h(

1
)
= K and L
2
1 +

2
/q

2
/q +h(

2
)
= K
In order to obtain an equilibrium, the asymptote of the curve A
must be above the asymptote of curve B.
In the particular case where wages are identical and both popu-
lations are of the same size, p=1/2 is an equilibrium. Consequently
the slope of curve A is positive and equal to 1. We can conclude that
in the neighborhood of the equality of wages the curve is upward-
sloping by continuity and the equilibrium is locally stable.
References
Acemoglu, D., 2002. Technical Change, Inequality, and the Labor Market. Journal of
Economic Literature 40 (March), 772.
Alchian, A., Demsetz, H., 1972. Production, information costs and economic organi-
zation. American Economic Review 62 (December), 777795.
Altonji, J.G., Blank, R.M., 1999. Race and gender in the labor market. In: Ashenfelter,
O., Card, D. (Eds.), Handbook of Labor Economics, vol. 3, pp. 31433259 (chapter
48).
Arrow, K.J., 1973. The theory of discrimination. In: Ashenfelter, O., Rees, A. (Eds.),
Discrimination in Labor Markets. Princeton University Press, Princeton.
J.-F. Jacques, E. Walkowiak / The Journal of Socio-Economics 38 (2009) 456463 463
Bayard, K., Hellerstein, J., Neumark, D., Troske, K., 2003. New evidence on sex seg-
regation in wages from matched employee-employer data. Journal of Labor
Economics (October), 887922.
Becker, G.S., 1971. The Economics of Discrimination. University of Chicago Press.
Becker, G.S., Murphy, K., 2000. Social Economics. The Belknap Press of Harvard Uni-
versity Press Cambridge, Massachusetts and London.
Bourdieu, P., 1980. Capital social. Notes provisoires. Actes de recherche en sciences
sociales 31, 23.
Burt, R., 1992. Structural Holes. Harvard University Press, Cambridge.
Coleman, J., 1990. Foundations of Social Theory. HarvardUniversityPress, Cambridge.
Fryer, R.G., Jackson, M.O., 2003. Categorical cognition, a psychological model
of categories and identication in decision making. NBER working paper
9579.
Gant, J., Ichniowski, C., Shaw, K., 2002. Social capital and organizational change in
high involvement and traditional work organizations. Journal of Economics and
Management Strategy (no. 11), 289328.
Gant, J., Ichniowski, C., Shaw, K., 2003. Working smarter by working together: con-
nective capital in the workplace, mimeo.
Garner-Moyer, H., 2003. Discrimination et emploi : revue de la literature. Document
dtudes de la DARES, n

69, mai.
Glaeser, E.L., Laibson, D., Sacerdote, B., 2000. The economic approachtosocial capital.
NBER working paper, 7728.
Glaeser, E.L., Scheinkman, J., 2002. Non-market interactions. In: Dewatripont,
M., Hansen, L.P., Turnowsky, S. (Eds.), Advances in Economics, Eighth World
Congress. Cambridge University Press, p. 2002.
Goldin, C., 2002. A pollution theory of discrimination: male and female differences
in occupations and earnings. NBER working paper, 8985.
Greenan, N., Walkowiak, E., 2005. Informatique, organisation du travail et interac-
tions socials. Economie et Statistique (no. 387), 3563.
Hamilton, B., Nickerson, J., Owan, H., 2006. Diversity and productivity in production
teams. Olin School of Business, working paper.
Jacques, J.F., Walkowiak, E., 2005. Low wage and high unemployment: the role of
social interactions in hiring discriminations, Eurisco working paper, no. 2005-7.
Kochan, T., Bezrukova, K., Ely, R., Jackson, S., Joshi, A., Jehn, K., Leonard, J.,
Levine, D., Thomas, D., 2003. The effects of diversity on business performance:
report of the diversity research network. Journal of Human Resources 42 (1),
321.
Kremer, M., 1993. The O-Ring theory of economic development. Quarterly Journal of
Economics CVIII (3), 551575.
Kremer, M., Maskin, E., 1996. Wage inequality and segregation by skills. NBER work-
ing paper, no. 5718, August.
Leana, C., Van Buren, H.J., 1999. Organizational social capital and employment prac-
tices. Academy Management Review 24 (3), 538555.
Liebenstein, H., 1979. Abranch of economics is missing: micro-micro theory. Journal
of Economic Literature 17 (no. 2), 477502.
Lemieux, T., 2006. Increasing residual wage inequality: composition effect, noisy
data or rising demand for skill? A re-examination. American Economic Review
96 (June (3)), 461498.
Mintzberg, H., 1981. The Structuring of Organizations: A Synthesis of the Research.
Prentice Hall, Englewood Cliffs.
Mortensen, D.T., 2003. Wage Dispersion: Why are Similar Workers Paid Differently?
MIT Press.
OECD, 2003. OECD Employment Outlook. Paris.
Phelps, E., 1972. The statistical theory of racism and sexism. American Economic
Review 62 (no. 4), 659661.
Pissarides, C.A., 2000. Equilibrium Unemployment Theory, 2nd ed. MIT Press, Cam-
bridge, MA.
Rosen, S., 1982. Authority, control and distribution of earnings. Bell Journal of Eco-
nomics 13 (2), 311323.
Saint-Paul, G., 1996. Are the unemployed unemployable? European Economic
Review 40, 15011519.
Thomas, D.A., 2004. Diversity as a strategy. Harvard Business Review 82 (9),
98108.
Williams, K., OReilly, C., 1998. Demography and diversity in organizations: a
review of 40 years of research. Research in Organizational Behavior 20,
77140.

You might also like