Content uploaded by Stijn Baert
Author content
All content in this area was uploaded by Stijn Baert on Mar 11, 2017
Content may be subject to copyright.
Content uploaded by Nick Deschacht
Author content
All content in this area was uploaded by Nick Deschacht on Aug 21, 2015
Content may be subject to copyright.
DO EMPLOYER PREFERENCES CONTRIBUTE TO
STICKY FLOORS?
STIJN BAERT, ANN-SOPHIE DE PAUW, AND NICK DESCHACHT*
The authors investigate the importance of employer preferences in
explaining sticky floors, the pattern in which women are less likely,
as compared to men, to start to climb the job ladder. The authors
perform a randomized field experiment in the Belgian labor market
and test whether hiring discrimination based on gender is heteroge-
neous by whether jobs imply a promotion (compared to the appli-
cants’ current position). The findings show that women receive
33% fewer interview invitations when they apply for jobs that imply
a first promotion at the functional level. By contrast, the results
show that their hiring chances are not significantly affected by the
authority level of the job.
During the past decade, economists have attempted to identify sticky
floors in the labor market, thereby building on the seminal work by
Booth, Francesconi, and Frank (2003). Sticky floors can be described as the
pattern in which women are, compared to men, less likely to start to climb
the job ladder. Therefore, sticky floors complement the well-known concept
of glass ceilings, the pattern in which women are less like to reach the top
of the job ladder.
1
Evidence for the existence of sticky floors has been
found in countries such as Italy (Filippin and Ichino 2005), Spain (Gradı´n
and del Rı´o 2009), Thailand (Fang and Sakellariou 2011), and the United
1
For some seminal and recent contributions to the literature on glass ceilings, see Albrecht, Bjo
¨rklund,
and Vroman (2003); Arulampalam, Booth, and Bryan (2007); Booth (2009); Weinberger (2011);
Reuben, Rey-Biel, Sapienza, and Zingales (2012); Smith, Smith, and Verner (2013); Kassenboehmer and
Sinning (2014). For more general contributions on gender differentials in promotions, see Blau and
DeVaro (2007); Yap and Konrad (2009); Matteazzi, Pailhe´, and Solaz (2014).
*STIJN BAERT is an Assistant Professor affiliated with Ghent University, Antwerp University, Universite´
catholique de Louvain, and the Institute for the Study of Labor (IZA). ANN-SOPHIE DEPAUW is an
Assistant Professor at the IE
´SEG School of Management. NICK DESCHACHT is an Assistant Professor at KU
Leuven. We are grateful to Kate Neyts and Hanneke Van Den Steen for their research assistance. This
research was reviewed and approved by the Ethical Affairs Committee of the Faculty of Economics and
Business Administration of Ghent University. Additional results and copies of the computer programs
used to generate the results presented in the article are available from the author(s) at
Stijn.baert@UGent.be.
KEYWORDs: gender discrimination, hiring discrimination, labor market transitions, European labor mar-
kets, sticky floors
ILR Review, 69(3), May 2016, pp. 714–736
DOI: 10.1177/0019793915625213. ÓThe Author(s) 2016
Journal website: ilr.sagepub.com
Reprints and permissions: sagepub.com/journalsPermissions.nav
States (Baker 2003). In Belgium, the country where the present study was
done, Deschacht, Baerts, and Guerry (2011) concluded, based on their dis-
crete time-event historical analysis of the Panel Study of Belgian
Households data (1994 to 2001), that women near the top of the job ladder
face fewer obstacles to promotions than women on the lower- and
mid-career levels. In addition, Belgium is one of the 12 European countries
in which Christofides, Polycarpou, and Vrachimis (2013), investigating the
gender wage gap across 24 EU member states, found evidence for sticky
floors.
From a policy perspective, determining the nature of the phenomenon
of sticky floors is important to design adequate policy actions. The studies
mentioned, however, do not allow us to determine whether sticky floors
result from gender differences in human capital, preferences and behavior
on the employee side, or preferences (and unequal treatment) on the
employer side. In our study, we contribute to the literature by investigating
the importance of employer preferences (and thereby discrimination) in
explaining sticky floors, keeping employee characteristics constant.
2
Specifically, our aim is to answer the question: Is the unequal treatment
of equally productive male and female job candidates by employers hetero-
geneous by whether or not they are applying for jobs that imply a (first)
promotion?
3
The unfavorable treatment of females in the access to these
positions can be expected based on the theoretical models of discrimination
introduced by Becker (1957) and Arrow (1973).
4
Discrimination against
women applying for higher positions could result from the general distaste
that employers, coworkers, and customers may have for collaborating with
women in these higher positions, potentially related to the traditional over-
representation of men in these positions. Furthermore, employers could
expect a higher average productivity from males relative to females in lead-
ing positions. This may relate to the match between these positions and
more ‘‘masculine’’ characteristics such as dominance, autonomy, and
2
Throughout this article the terms employer preferences and employer discrimination are used to point to
the general concept of employers’ unequal appraisal and treatment of women and men in the labor
market. Therefore, these concepts do not refer to the narrower theoretical concept of Becker’s
(1957)‘‘employer discrimination.’’
3
This implies that we do not investigate unequal treatment in promotions within the firm. We come
back to this issue later in the article.
4
These theories are based on neoclassical assumptions and have been criticized concerning several
aspects; see, for example, Bergmann (1989) in the context of gender-based discrimination in the labor
market. In spite of the development of alternative approaches to theorizing discrimination, such as the
models of opportunistic discrimination, implicit discrimination, noncompeting groups, and lexico-
graphic search by employers (Darity and Mason 1998; Altonji and Blank 1999; Bertrand and
Mullainathan 2004), many recent contributions to the discrimination literature are still guided by the
theoretical framework of the models of taste-based and statistical discrimination (Charles and Guryan
2008; Borjas 2009). Moreover, whereas some newer models are particularly applicable to the case of eth-
nic discrimination, the models of Becker (1957) and Arrow (1973) can be applied to the case of gender
discrimination in a natural way.
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 715
assertiveness, on the one hand,
5
and the fact that females are documented
as having, on average, a less continuous and committed labor-market partic-
ipation, which is related to the traditional division of labor in the household
and maternity leave (Budig and England 2001; Blandford 2003; Baumle
2009; Romeu Gordo 2009), on the other. In addition to the theoretical sup-
port of the Becker (1957) and Arrow (1973) frameworks, potential gender
discrimination by employers when taking hiring decisions that imply a pro-
motion for the job candidate squares with the notion in the popular litera-
ture that men are promoted based on potential whereas women are
promoted based on prior performance (Carter and Silva 2011).
To the best of our knowledge, all previous studies on gender differences
in promotion opportunities have used regression-based methods on obser-
vational data (typically household-survey or labor-force data). The most
common approach has consisted of Oaxaca–Blinder-style decompositions of
wage differentials, with a part explained by observables related to human
capital and occupational characteristics and an unexplained part pointing
in the direction of discrimination (Blinder 1973; Oaxaca 1973). Residual
gaps at the top of the wage distribution were then interpreted as comple-
mentary with the glass ceiling concept, whereas residual gaps at the bottom
of the wage distribution pointed in the direction of sticky floors. In their
survey of experimental advances in the study of gender differentials in the
labor market, Azmat and Petrongolo (2014) formulated two important
drawbacks of using observational data to measure gender discrimination.
First, most observational data do not capture some of the determinants of a
workers’ productivity (e.g., motivation and intrinsic ability) that may be
observed—and valued—by employers. Stated otherwise, based on these
nonexperimental data, researchers cannot control for all the variables
employers take into account when making their hiring and remuneration
decisions. As a result, no conclusive evidence of discrimination can be pro-
vided based on these data. Second, if premarket investments in human capi-
tal or decisions to apply for a job implying a promotion are affected by the
expectation of future labor-market discrimination, the impact of discrimina-
tion is partly captured by observable human capital and occupational
characteristics.
To overcome these methodological problems, this study is novel in its
approach to gathering original experimental data on gender differences in
promotion opportunities and thereby, argued by Duflo, lets ‘‘the research
question determine the data to be obtained instead of the data determining
the questions that can be asked’’ (2006: 2). Specifically, we sent out 576
pairs of fictitious job applications for individuals with about five years of job
experience in their first and current job to vacancies for jobs requiring at
least some and at most five years of experience. The two applications in
5
For instance, Oakley (2000) argued that the less assertive forms of communication used by women
might be inappropriate for functioning in the upper echelons of many corporations.
716 ILR REVIEW
each pair are similar, except for the revealed gender of the candidate. By
monitoring the reaction from the employer side, we can identify unequal
treatment based on this single characteristic. This measure can be given a
causal interpretation because all the information the employer receives is
under the control of the researcher so that selection on individual unobser-
vables is eliminated. We sent fictitious applications to vacancies that implied
a promotion in terms of occupational level and/or job authority compared
to the current listed job of the candidate and to vacancies at the same func-
tional and authority levels as the current job. Thereby, we are able to test
whether unequal treatment of young men and women in the labor market
is heterogeneous by whether jobs imply a promotion compared to the appli-
cants’ current position.
6
If employer preferences contribute to sticky floors,
the unequal treatment of male and female job candidates should be rela-
tively more to the disadvantage of women who apply for a job implying a
promotion. This is our research hypothesis.
The Experiment
Causal Evidence of Gender Discrimination by Means
of a Correspondence Experiment
To answer our research question, we conducted a correspondence experi-
ment. In this experiment, pairs of fictitious job applications were sent to
real job vacancies. The applications in each pair were comparable except
for the gender of the applicant. By monitoring the subsequent callbacks
from the employer side, we can measure discrimination based on this single
characteristic and give it a causal interpretation.
Correspondence experiments are widely viewed as providing the cleanest
and most compelling evidence of unequal treatment in hiring (Riach and
Rich 2002). As we argue in the introduction, any bias introduced by individ-
ual unobservable characteristics is eliminated because all the information
received by the employer is controlled by the researcher. Strict equivalence
between the fictitious job candidates is realized, and unequal treatment by
employers can be disentangled from other explanations of differential hir-
ing outcomes, such as differences in human capital and preferences at the
employee side.
Concerning the identification of gender discrimination, the extensive
correspondence experiment conducted by Neumark, Bank, and Van Nort
(1996) is seminal. Neumark et al. showed that, in the 1994 Philadelphian
labor market for waiters and waitresses, female job candidates had, com-
pared to males, a substantially lower estimated probability of receiving an
6
Stated otherwise, we can argue that a potential gender differential in the chances of receiving a call-
back for a job that implies a promotion at a new firm is composed of a gender differential in the hiring
chances at a new firm and a gender differential in callbacks for jobs implying a promotion. As we apply
both for jobs implying a promotion and for jobs implying no promotion, our experimentally gathered
data allow us to isolate the latter gender differential.
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 717
interview in high-price restaurants and a substantially higher probability
of receiving one in low-price restaurants (where earnings were also lower).
In the same realm, Firth (1982), Riach and Rich (1987, 2006),
Weichselbaumer (2004), Petit (2007), Albert, Escot, and Ferna´ndez-Cornejo
(2011), Booth and Leigh (2010), and Carlsson (2011) studied hiring dis-
crimination based on gender in Australia, the United Kingdom, Austria,
France, Spain, and Sweden by means of correspondence experiments.
None of these studies, however, attempted to investigate heterogeneity in
labor-market discrimination by whether or not the jobs that the fictitious
applicants applied for implied a promotion.
We conducted our experiment between October 2013 and March 2014 in
the labor market of Flanders, the Dutch-speaking northern part of Belgium.
During this period, we randomly selected 288 vacancies for jobs targeting
Bachelors’ degrees (BA) in business administration and 288 vacancies for
jobs targeting Masters’ degrees (MA) in business economics in the database
of the Public Employment Agency of Flanders, the region’s major job-search
channel. We restricted ourselves to vacancies in the private sector requiring
at most five years of work experience.
7
Two job applications for individuals
with five years of work experience (in a first and current jobs), identical in
terms of productivity-relevant characteristics but different in their gender,
were sent to the selected vacancies. We opted for individuals with five years of
experience because sticky floors are related to the start of job-ladder climb-
ing. Moreover, job searches that target people with a tenure of five years can
be expected to be heterogeneous in promotion aspects compared to the
applicants’ current (starter) job because ‘‘at least five years of work experi-
ence’’ is typically asked for in announcements for (more) senior positions.
The data-gathering process was reviewed and approved by the Ethical
Committee of the Faculty of Economics and Business Administration of
Ghent University at its meeting of July 9, 2013.
Construction of Fictitious Applications
For each of the aforementioned academic degrees held by our fictitious
applicants, we created two template applications comprising a resume and a
motivation letter. We refer to these (template) applications as the Type A
and Type B applications. The two types were equal in all productivity-
relevant characteristics (e.g., educational qualifications, language skills, and
computer skills); however, to avoid detection of the experiment, they dif-
fered in layout and details. For instance, both candidates mentioned sports
and cultural activities, but the Type A applicant referred to general sports
7
The database of the Public Employment Agency of Flanders covers both the private and public sec-
tors. We were forced, however, to restrict ourselves to vacancies in the private sector because in Flanders
vacancies in the public sector almost always ask potential job candidates to provide the recruiter with for-
mal proofs of their educational degrees. Because we did not want to falsify these degrees, we were not
able to apply for public-sector vacancies.
718 ILR REVIEW
activities and (drama-club) acting whereas the Type B applicant referred to
fitness and tennis activities and playing the saxophone. To ensure that our
job applications were realistic and representative, example applications of
the Public Employment Agency of Flanders were used and calibrated for
our purposes.
The Type A and Type B applicants were single individuals born, studying,
and living in comparable suburbs of Ghent, the second largest city of
Flanders. These candidates graduated from tertiary education in June 2008
without any grade-retention experience. In line with the vacancy to which
the pair of applications was sent, a specialization was chosen.
8
All the appli-
cants had graduated from University College Ghent (BA in business admin-
istration) or Ghent University (MA in business economics). Since
September 2008, the fictitious applicants had been employed in a starter
job in line with their academic degree and specialization but with no job
authority (see the subsection on Variation in Occupational Level and Job
Authority for how we define job authority).
9
Furthermore, we added to the Type A and Type B applications the fol-
lowing characteristics: Belgian nationality, driver’s license, adequate
information-technology (IT) skills, Dutch mother tongue, adequate English
and French language skills, and sports and cultural activities. In their moti-
vation letters, the job candidates indicated they were highly interested, moti-
vated, and organized. Last, we appended a fictitious postal address (the
combination of a real street in a middle-class neighborhood and a nonexis-
tent street number) and a date of birth to all application templates.
Randomized Disclosure of Applicants’ Gender
To erase any dependence of gender differentials in the callbacks on the
application type, a typically male- and a typically female-sounding name was
alternately assigned to the Type A and Type B applications. An appendix
with the four combinations of resumes (Type A with female name, Type A
with male name, Type B with female name, and Type B with male name) for
one of the educational degrees (BA in business administration with accoun-
tancy and tax as the specialization) is available in the online Appendix.
10
We alternated the order in which we sent the applications to the selected
job postings, with a one-day delay in between. This resulted in a randomiza-
tion procedure in which we alternated among four pairs of applications: 1)
8
For the BA in business management, the specializations used were 1) accountancy and tax, 2) finance
and assurance, 3) management of small- or medium-size business, 4) logistics, 5) marketing, and 6) legal
practice. For the MA in business economics, the specializations used were 1) accountancy, 2) finance,
and 3) marketing.
9
The professions of accountant (International Standard Classification of Occupations [ISCO]-08 code
3313), file administrator assurances (code 4312), management assistant (code 3343), legal file adminis-
trator (code 3411), dispatcher (code 3331), and junior marketer (code 3322) were used.
10
See http://users.ugent.be/~sbaert/DoEmployerPreferencesContributetoStickyFloorsONLINE
APPENDIX.pdf.
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 719
first the Type A application with the female name and then (one day later)
the Type B application with the male name; 2) first the Type A application
with the male name and then the Type B application with the female name;
3) first the Type B application with the male name and then the Type A
application with the female name; 4) first the Type B application with the
female name and then the Type A application with the male name. To the
first vacancy, combination 1) was sent; to the second vacancy, combination
2) was sent; to the third vacancy, combination 3) was sent; to the fourth
vacancy, combination 4) was sent; to the fifth vacancy, again combination
1) was sent; and so on. This randomization procedure was carried through
for each of the two educational degrees separately.
We chose ‘‘Elke De Vos’’ and ‘‘Anke De Vriendt’’ as the names of the
female candidate holding a BA and a MA, respectively, and ‘‘Stijn Van
Lancker’’ and ‘‘Stijn Van De Voorde’’ as the names of the male candidates,
holding a BA and MA, respectively. We made sure that the chosen names
did not represent different social backgrounds, following Bloothooft and
Onland (2011), who linked the suffix of the name Dutch-speaking parents
choose for their children with these children’s later wages.
We matched to each assigned name a mobile phone number and an
e-mail address. The voicemail messages were the telecommunication provi-
ders’ standard (Dutch) messages (‘‘Hello, [Name of the fictitious candi-
date] cannot answer your call. Please leave a message after the tone’’). The
names of the fictitious candidates were recorded by young people with a
similar social-economic background (i.e., native students in the MA in [busi-
ness] economics program at Ghent University). The e-mail addresses were
registered with large and commonly used providers. All fictitious job appli-
cations were sent to the selected real vacancies by e-mail. To avoid detec-
tion, we applied to no more than one job posting from each employer.
Measurement of Callbacks
Callbacks were received by telephone voicemail and by e-mail. Because we
used postal addresses with nonexistent street numbers in the applications, we
could not measure reactions on the employer side sent by regular mail.
However, several Flemish human resource managers confirmed that, nowa-
days, employers rarely invite job candidates to selection interviews through
regular mail. To minimize the inconvenience to the employers, we immedi-
ately declined all positive callbacks. Following the corresponding testing liter-
ature, we discounted all reactions from the employer side received later than
30 days after sending out the applications. Because we did not receive any
positive reaction after 30 days, this turned out to be a redundant restriction.
In our data analysis, we distinguish between two definitions of positive call-
back. Positive callback in a strict sense is defined as getting an invitation for an
interview concerning the job for which the applicant applied; this definition
was adopted by, for example, Bertrand and Mullainathan (2004). Positive
720 ILR REVIEW
callback in a broad sense includes, in addition to the first definition, a request
to provide more information, a request to contact the employer, or the
receipt of an alternative job proposal; this definition was adopted by, for
example, Kaas and Manger (2012).
Variation in Occupational Level and Job Authority
Sticky floors may translate into lower probabilities for females to be pro-
moted in terms of four dimensions: occupational level, job authority, wages,
and nonstatutory benefits (Deschacht et al. 2011). Because the last two job
aspects are not mentioned in the job postings in the database of the Public
Employment Agency of Flanders, our experiment does not allow us to
observe these aspects at the vacancy level. Therefore, we focus here on pro-
motion in terms of occupational level and the degree of job authority. At
the end of the regression analysis, however, we include the average wage in
the occupation as a proxy for the wage in the posted job as a variable
explaining callback rates by gender.
Each of the 576 vacancies to which we sent the two fictitious job applica-
tions was matched with an indicator of the occupational level and with an
indicator of the job-authority level. This was done immediately after sending
out the first application to a selected vacancy (and ipso facto before any call-
back was received). First, based on the information in the vacancy, all vacan-
cies were matched with an occupation in the International Standard
Classification of Occupations (ISCO)-08 classification system; ISCO-08 is the
second-generation classification system, based on a resolution by experts on
labor statistics in December 2007 and endorsed by the Governing Body of
the International Labour Organization (ILO) in March 2008. To do this,
we searched for a best match between the name of the vacancy (i.e., the
name of the job) and the description of the tasks of the job in the vacancy,
on the one hand, and four elements mentioned in the ISCO-08 group defi-
nitions file (International Labour Organization 2013), on the other: defini-
tion, tasks included, included occupation names, and excluded occupation
names. A new job at a higher ISCO-08 one-digit level (called here its occu-
pational level) than the current job in the application was defined as a pro-
motion in terms of occupational level.
11
11
The reader will notice that, because of our choice of the candidates and current job and because of
our procedure of matching these candidates with vacancies, both the current job and the job for which
the candidates applied were closely related to their academic degree and their specialization. So,
observed (potential) occupational transitions are, as a rule, vertical movements (i.e., transitions in job
level) and not horizontal movements (i.e., transitions in job content). This is a desirable situation
because, therefore, promotions in occupational level following our definition may not reflect nonpromo-
tional horizontal transitions. The most common promotions in occupational level observed in our data
are those from accountant (ISCO-08 -level 3) and file administrator assurances (level 4) to financial pro-
fessional (level 2); from management assistant (level 3) and legal file administrator (level 3) to adminis-
tration professional (level 2); from dispatcher (level 3) to manufacturing, mining, construction, and
distribution manager (level 1); and from junior marketer (level 3) to sales, marketing, and public-
relations professional (level 2).
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 721
The jobs for which our fictitious job candidates applied were classified
into three categories of job authority, following Hachen (1990): jobs with
no supervision authority, that is, no supervision or coordination tasks con-
cerning the work of other workers (level 1); jobs with supervision authority
but no influence on the wage or promotion opportunities of other workers
(level 2); and jobs with both supervision authority and influence on the
wage or promotion opportunities of other workers (level 3). Because the
current job of the candidates was set at level 1, a new job with at least super-
vision authority (i.e., at level 2 or level 3) was defined as a promotion.
Limitations of the Experimental Design
Before addressing our research question by means of a statistical analysis of
the gathered data, we discuss some limitations of our experimental research
setting.
First, our design can be effective only in demonstrating gender discrimi-
nation in the first stage of the recruitment process. Because we simply ana-
lyze the content of first callbacks, we cannot translate our empirical results
into divergences in job offers. Bertrand and Mullainathan (2004), however,
argued that, even in the case in which the recruitment process has moder-
ate friction, one can expect that reduced chances of having an interview will
translate into reduced chances of receiving a job offer.
Second, our design is limited in its focus on promotion transitions to
other firms. Unequal opportunities in promotion transitions within firms
cannot be investigated by means of an experiment such as this one and may
be characterized by other dynamics.
Third, we test for unequal treatment only in the vacancies targeting individ-
uals with a BA or MA degree in business posted in the database of the Public
Employment Agency of Flanders. Possibly, unequal treatment based on gen-
der is more (or less) apparent among employers who rely on other channels
(such as social networks) for filling their vacancies or among employers who
have to fill vacancies targeting other profiles. We are specifically interested,
however, in the relationship between hiring discrimination against women
and the potential promotion implied by the jobs for which they apply.
Because this third limitation causes, by construction, a similar shift in the dis-
crimination measures for the selected jobs that imply a promotion and for
those that imply no promotion, our main research conclusions remain valid.
Fourth, our occupational-level and job-authority promotion dimensions
may correlate with each other and with other vacancy characteristics. As a
consequence, descriptive statistics on heterogeneity in discrimination using
these promotion dimensions of the tested jobs might reflect heterogeneity
of gender discrimination on other dimensions. Therefore, we merge the
gathered data with administrative data at the occupation, firm, and sector
levels and analyze the merged data by means of a regression analysis con-
trolling for potential confounders.
722 ILR REVIEW
Results
Descriptive Data Analysis
Tables 1 and 2 describe the experimentally gathered data, adopting the
strict and broad definition of positive callback, respectively. We provide the
reader with two frequently used statistical measures: the positive callback
ratio (PCR) and the net discrimination rate (NDR) (Neumark et al. 1996;
Riach and Rich 2002; Bertrand and Mullainathan 2004).
Overall, for 88 of the 576 vacancies at least one candidate received a posi-
tive callback in the strict sense (195 in the broad sense): 32 (98) vacancies
resulted in a positive callback for both candidates, 31 (51) in a positive call-
back for the male candidate only, and 25 (46) in a positive callback for the
female candidate only. The overall PCR is obtained by dividing the total
percentage of applications for which male candidates received a positive
callback by the corresponding percentage for the female candidates. At the
level of the total data set, the positive callback rate in the strict sense is
11%
12
(in the broad sense, 26%) for male candidates and 10% (25%) for
female candidates. The resulting PCR is 1.11 (1.03), indicating that the
male candidates received on average 11% (3%) more positive callbacks in
reaction to their applications. These ratios are not significantly different
from 1, and so we conclude that, overall, employers did not discriminate
based on gender.
To provide the reader with an initial answer to our research question, we
split our sample using the two dimensions of promotion. First, we find sig-
nificant evidence for unequal treatment of females when they applied for
jobs at a higher ISCO-08 level than the level of their current job. In this
case, the PCR is 1.50 (1.23), which means that males received about 50%
more invitations for a job interview (positive callback, strict sense) and 23%
more positive reactions (broad sense). Stated otherwise, when applying for
these jobs, women received about 33% fewer interview invitations (strict
sense) and 19% fewer positive reactions (broad sense). This finding is in
line with our first research hypothesis of more hiring discrimination against
women when they apply for jobs that imply a promotion. Second, we find
no significantly unequal treatment for both subsamples using the promo-
tion dimension related to job authority.
The overall NDR is calculated by subtracting the number of applications
for which the female candidate was preferred from the number of applica-
tions for which the male candidate was preferred and dividing the result by
the number of pairs in which at least one candidate received a positive call-
back. The result is a net measure of the number of discriminatory acts a
female applicant could expect to encounter per application for which at
least one candidate received a positive callback. Overall, the NDR is 0.07 for
positive callback in the strict sense (0.03, broad sense). Based on a standard
12
0.11 = (32 +31)/576.
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 723
Table 1. The Probability of a Positive Callback (Strict Sense): Descriptive Analysis
Data selection
Number
of jobs
Neither
candidate
had positive
callback
Both
candidates
had positive
callback
Only male
candidate
had positive
callback
Only female
candidate
had positive
callback PCR
a
t NDR
b
x
2
A. All jobs
576 488 32 31 25 1.105 0.802 0.068 0.643
B. Breakdown by occupational-level promotion dimension
Jobs at same level as current job 248 206 18 9 15 0.818 1.226 20.143 1.500
Jobs at higher level than current job 328 282 14 22 10 1.500** 2.133 0.261** 4.500
C. Breakdown by job-authority promotion dimension
Jobs with no job authority 328 271 21 20 16 1.108 0.666 0.070 0.444
Jobs with job authority 248 217 11 11 9 1.100 0.447 0.065 0.200
a
The positive callback ratio (PCR) is calculated by dividing the percentage of applications for which male candidates received a positive callback by the corresponding
percentage for the female candidates. The t-test for the PCR tests the null hypothesis that the probability of a positive callback is the same for candidates of both genders.
Because two applicants contacted the same firm, the probability of the male applicant receiving an invitation correlates with the probability of the female applicant
candidate receiving one. Therefore, standard errors are corrected for clustering of the observations at the vacancy level.
b
The net discrimination rate (NDR) is calculated by subtracting the number of applications for which the female candidate was preferred from the number of applications
for which the male candidate was preferred and dividing by the number of pairs in which at least one candidate received a positive callback. The x
2
test for the NDR tests
the null hypothesis that both candidates are treated unfavorably just as frequently.
***;**;*indicates significance at the 1%; 5%; 10% significance level, respectively.
Table 2. The Probability of a Positive Callback (Broad Sense): Descriptive Analysis
Data selection
Number
of jobs
Neither
candidate
had positive
callback
Both
candidates
had positive
callback
Only male
candidate had
positive callback
Only female
candidate had
positive callback PCR
a
t NDR
b
x
2
A. All jobs
576 381 98 51 46 1.035 0.507 0.026 0.258
B. Breakdown by occupational-level promotion dimension
Jobs at same level as current job 248 156 45 18 29 0.851 1.610 20.120 2.574
Jobs at higher level than current job 328 225 53 33 17 1.229** 2.277 0.155** 5.120
C. Breakdown by job-authority promotion dimension
Jobs with no job authority 328 209 60 28 31 0.967 0.390 20.025 0.153
Jobs with job authority 248 172 38 23 15 1.151 1.300 0.105 1.684
a
The positive callback ratio (PCR) is calculated by dividing the percentage of applications for which male candidates received a positive callback by the corresponding
percentage for the female candidates. The t-test for the PCR tests the null hypothesis that the probability of a positive callback is the same for candidates of both genders.
Because two applicants contacted the same firm, the probability of the male applicant receiving an invitation correlates with the probability of the female applicant
candidate receiving one. Therefore, standard errors are corrected for clustering of the observations at the vacancy level.
b
The net discrimination rate (NDR) is calculated by subtracting the number of applications for which the female candidate was preferred from the number of applications
for which the male candidate was preferred and dividing by the number of pairs in which at least one candidate received a positive callback. The x
2
test for the NDR tests
the null hypothesis that both candidates are treated unfavorably just as frequently.
***;**;*indicates significance at the 1%; 5%; 10% significance level, respectively.
x
2
test, we cannot reject the hypothesis that the candidates of both sexes
were treated unfavorably equally often. In contrast, if we break the NDR
down by the occupational-level and job-authority promotion dimensions, we
get results in line with those based on the PCR.
An explanation for the finding of no significant heterogeneity in the job-
authority dimension could be that, for supervising positions (i.e., middle
management rather than management), soft and social (leading) skills,
which are often related to women, outweigh the aforementioned more
‘‘masculine’’ characteristics for employers.
Regression Analysis
Because, by construction, both the male and female candidate exhibit the
same observable characteristics for each vacancy, regressing the positive call-
backs on the gender of the candidate (and on the vacancy and employee
characteristics) leads to the same empirical conclusion as the one based on
panel A of Tables 1 and 2. But, as previously mentioned, the occupational-
level and job-authority promotion dimensions may correlate with each other
and with other vacancy characteristics, and therefore, the results presented
in panels B and C of Table 1 may, in fact, reflect the heterogeneity of gen-
der discrimination on other dimensions. Moreover, because of the finite
size of our sample, the promotion indicators of the posted jobs may even
correlate with the application type (A or B) to which the female candidate
was assigned and with whether her application was sent before or after the
male candidate’s application. Therefore, we further explore the experimen-
tally gathered data by means of a regression analysis. Summary statistics of
the variables used in this analysis are outlined in Table 3.
Tables 4 and 5 present our regression results. We regress, by means of a
linear probability model, the probability of a positive callback on various
sets of key and control variables, both by themselves and in interaction with
Female candidate.
13
For comparability of the regression results, we normal-
ized all the explanatory variables that are interacted with Female candidate
by subtracting the mean among the subpopulation of female candidates
and, for continuous variables, by dividing the result by the standard devia-
tion of this subpopulation. The estimation results for all variables interacted
with Female candidate are presented in Table 4 (using positive callback in
the strict sense as an outcome) and in Table 5 (using positive callback in
the broad sense as an outcome).
14
For each of the presented models, we
computed multicollinearity diagnostics leading to variance inflation factors
less than five, except for some sector dummies.
13
We also performed estimations using vacancy fixed effects and probit estimations. The results for
these exercises are closely comparable to those presented in the main text, and they are available on
request.
14
For conciseness, we do not present the main effects for these variables here, but they are available
on request.
726 ILR REVIEW
Here, we first focus on the results of Table 4 and then compare these
results with the ones in Table 5. We regress positive callback on the dummy
Female candidate (Model 1); the dummy Female candidate and a dummy
indicating that the posted job implied a promotion in occupational level,
included both as such and in interaction with Female candidate (Model 2);
and the dummy Female candidate and a dummy indicating the posted job
implied a promotion in job authority, included both as such and in interac-
tion with Female candidate (Model 3). By construction, the estimations pre-
sented in Models 1 to 3 of Table 4 reflect those presented in Table 1,
panels A to C. For instance, based on Model 1, we learn that revealing a
female gender lowers a candidate’s chance for an invitation for a job inter-
view about 1 percentage point, which squares with the comparison of the
positive callback rates for males (11%) and females (10%) mentioned previ-
ously. Furthermore, Model 2 indicates that the gender gap in job-interview
invitations is about 6 percentage points higher (and relatively more to the
detriment of females) in jobs that imply a promotion in occupational level
than in jobs that imply no promotion in this respect.
Next, in Model 4, we combine the explanatory variables included in
Models 2 and 3. As a consequence, the finding of fewer positive callbacks
for women when they apply for jobs at a better ISCO level is further
enforced. In contrast, the coefficient for the interaction with the
Table 3. Regression Variables (Employee and Vacancy Characteristics):
Summary Statistics
Variable All jobs
Jobs implying a promotion
in occupational level
Jobs implying a promotion
in job authority
First application sent
within pair was the
female applicant’s
0.500 (0.500) 0.506 (0.501) 0.520 (0.501)
Application Type B was
the female
applicant’s
0.500 (0.500) 0.530 (0.500) 0.544 (0.499)
MA degree 0.500 (0.500) 0.616 (0.487) 0.540 (0.499)
Temporary contract 0.127 (0.333) 0.116 (0.321) 0.109 (0.312)
Part-time contract 0.075 (0.263) 0.061 (0.240) 0.052 (0.223)
Male recruiter 0.325 (0.469) 0.305 (0.461) 0.270 (0.445)
Log(Average FTE in
firm)
1.472 (1.048) 1.501 (1.085) 1.510 (1.003)
Percentage of female
workers in
occupation
0.550 (0.181) 0.497 (0.163) 0.490 (0.158)
Average wage in
occupation in EUR
4,098 (1,043) 4,787 (855.6) 4,813 (1,066)
Sector number (NACE,
two-digit-level)
65.54 (17.51) 66.86 (17.29) 66.32 (17.31)
Notes: Reported statistics are averages. Standard deviations are in parentheses. EUR, euro; FTE, full-time
equivalency; NACE, Nomenclature Generale des Activites Economiques dans I‘Union Europeenne.
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 727
Table 4. The Probability of a Positive Callback (Strict Sense): Regression Analysis
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Female candidate 20.010 (0.013) 20.010 (0.013) 20.010 (0.013) 20.010 (0.013) 20.012 (0.013) 20.012 (0.013)
Female candidate *Promotion in occupational
level (norm.)
20.061** (0.026) 20.092*** (0.035) 20.095** (0.040) 20.106** (0.049)
Female candidate *Promotion in job authority
(norm.)
0.004 (0.026) 0.055 (0.034) 0.067 (0.041) 0.063 (0.044)
Female candidate *First application sent within
pair (norm.)
0.065 (0.048) 0.068 (0.048)
Female candidate *Application Type B, MA degree
(norm.)
0.061 (0.063) 0.061 (0.063)
Female candidate *Application Type B, BA degree
(norm.)
0.046 (0.072) 0.044 (0.071)
Female candidate *MA degree (norm.) 20.030 (0.053) 20.032 (0.053)
Female candidate *Temporary contract (norm.) 0.071 (0.059) 0.071 (0.059)
Female candidate *Part-time contract (norm.) 0.019 (0.053) 0.019 (0.053)
Female candidate *Male recruiter (norm.) 20.003 (0.035) 20.002 (0.036)
Female candidate *Log(Average FTE in firm)
(norm.)
0.017 (0.017) 0.017 (0.017)
Female candidate *Percentage female workers in
occupation (norm.)
20.003 (0.017) 20.002 (0.017)
Female candidate *Average wage in occupation
(norm.)
0.009 (0.024)
Linear probability model Yes Yes Yes Yes Yes Yes
Dependent variable: Invitation to a job interview Yes Yes Yes Yes Yes Yes
Sector fixed effects interacted with Female
candidate
No No No No Yes Yes
Number of observations 1,152 1,152 1,152 1,152 1,074 1,074
Notes: Variables that are interacted with Female candidate are also included without interaction with this variable. Except for Female candidate, all variables are normalized
(norm.) by subtracting the mean among the population of female candidates. Continuous variables are further normalized by dividing by the standard deviation among this
subpopulation. Standard errors, corrected for clustering at the vacancy level, are in parentheses. In Models 5 and 6, 78 observations are dropped because of missing values for
the added control variables. FTE, full-time equivalency.
***;**;*indicates significance at the 1%; 5%; 10% level, respectively.
job-authority promotion dimension increases and even turns positive, albeit
insignificantly so. This is attributable to the positive correlation between the
variables indicating a promotion on the occupational-level and job-authority
dimensions (Pearson’s r= 0.56). Clearly, when we do not control for the for-
mer interaction, the latter interaction picks up part of the fact that hiring
chances are lower for women in jobs that imply a promotion in occupa-
tional level.
In Model 5, we extend the set of variables included, both as such and in
interaction with Female candidate. To control for the potential correlation
between these variables and the promotion indicators of the selected jobs,
we include variables indicating applications that were the first of the pair
sent to the vacancy and that used the Type B template (because the tem-
plates differ by academic degree, we adopt a Type B dummy for each
degree).
In addition, we include a set of variables capturing employee and vacancy
characteristics that may affect the callback rates and may be correlated, at
least because of the finite size of our data, with the promotion indicators of
the posted jobs (see Table 3). First, we include an indicator of the educa-
tion level of the applicant (a dummy indicating an MA degree) because
both theoretical and empirical studies have shown that discrimination
decreases with the applicant’s level of education (Taubman and Wales
1974; Bursell 2007; Wood et al. 2009; Baert, Cockx, Gheyle, and Vandamme
2015). Second, we adopt variables indicating temporary and part-time con-
tracts, which are uniformly mentioned in the vacancies in the database of
the Public Employment Agency of Flanders. If females have a relatively
higher access to temporary and part-time jobs and these kinds of jobs are
less present in the tested vacancies that imply a promotion in occupational
level—and Table 3 shows that this is the case—this could drive the results
presented in Table 1, panel B. Third, and related to the recent theory of
erotic capital as a personal asset in addition to economic, cultural, and
social capital and the empirical evidence on gender differentials in labor-
market discrimination (Hakim 2010; Baert 2014), we add a proxy for the
gender of the key decision maker in the first stage of the recruitment proce-
dure (i.e., the gender of the contact person mentioned in the vacancy
announcement). Fourth, we include, in Models 5 and 6, the natural loga-
rithm of the average number of workers in the firm in 2010 as a proxy of its
size.
15
We do this because some former contributions to the empirical litera-
ture on labor-market discrimination against ethnic minorities reported a
negative relationship between firm size and hiring discrimination, related to
the fact that larger firms are more likely to have a dedicated human
resources function and therefore, on average, more standardized proce-
dures for recruitment (Wood et al. 2009; Kaas and Manger 2012). Fifth, we
15
This proxy was obtained from Bureau Van Dijk (2014) after looking up the company number of the
firm in Trendstop (2014) and Unizo (2014).
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 729
include the fraction of female workers in the job because we might expect
that, because typically male (female) characteristics are perceived as particu-
larly productive in traditionally male (female) occupations, unequal treat-
ment should be more in favor of women in female-dominated occupations
(Weichselbaumer 2004; Booth and Leigh 2010).
16
Sixth, to erase any corre-
lation between the promotion indicators of the selected jobs and the
(other) characteristics of the sectors in which they were selected, we include
two-digit sector dummies.
17
In total, 57 sectors are represented in the data.
As is the case for all other mentioned variables, these dummies are included
both as such and in interaction with Female candidate.
Table 4 shows the estimation results for Model 5. We observe a substan-
tial but insignificant coefficient with the expected sign for the interaction
between Female candidate and the variable Temporary contract. More
important, however, is that the estimate for the interaction between Female
candidate and the dummy indicating occupational-level-promotion jobs
hardly changes when we introduce these variables. We also ran Model 5
without sector dummies, which led to very similar results.
18
Last, in Model 6 we extend the set of explanatory variables in Model 5
with the average wage in the job, following the ISCO-08 classification at the
three-digit level in 2010.
19
We have, however, no significant coefficient for
this variable. This may point in the direction of no heterogeneity existing in
gender discrimination by the wage of the offered job. But, equally, this find-
ing might be explained by the fact that the average wage in the job using
the ISCO-08 classification is not a good proxy for the wage in the particular
posted job. Moreover, even when employer discrimination as such is not
heterogeneous by the wage level in the occupation, we would expect that
our main finding of the unequal access for males and females to (better-
paid) positions at a higher functional level should translate into gender dif-
ferences in pay.
The results outlined in Table 5, using the broad sense definition of posi-
tive callback as an outcome variable, lead to the same conclusion concern-
ing the interactions between Female candidate and the promotion
indicators of the selected jobs. Thereby, we conclude that the regression
analysis presented confirms our previous descriptive findings.
16
This variable is proxied by the average proportion of females in each occupation, following the
ISCO-08 classification, in the European Social Survey (sixth round; 2012).
17
We use the Statistical Classification of Economic Activities in the European Community
(Nomenclature Generale des Activites Economiques dans I‘Union Europeenne [NACE]) at the two-digit
level. The sector code, based on the name and the address of the firm mentioned in the vacancy, was
found in the databases of Trends and UNIZO.
18
Results available on request.
19
Data from Directorate-General Statistics and Economic information of Belgium (2014). We use the
2010 statistics because those of later years are not yet available.
730 ILR REVIEW
Table 5. The Probability of a Positive Callback (Broad Sense): Regression Analysis
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Female candidate 20.009 (0.017) 20.009 (0.017) 20.009 (0.017) 20.009 (0.017) 20.010 (0.018) 20.010 (0.018)
Female candidate *Promotion in
occupational level (norm.)
20.093*** (0.035) 20.102** (0.043) 20.085*(0.051) 20.109*(0.062)
Female candidate *Promotion in job
authority (norm.)
20.041 (0.034) 0.015 (0.042) 0.034 (0.049) 0.025 (0.052)
Female candidate *First application sent
within pair (norm.)
0.160** (0.069) 0.163** (0.070)
Female candidate *Application Type B, MA
degree (norm.)
0.092 (0.097) 0.093 (0.097)
Female candidate *Application Type B, BA
degree (norm.)
0.053 (0.102) 0.051 (0.102)
Female candidate *MA degree (norm.) 20.042 (0.080) 20.046 (0.080)
Female candidate *Temporary contract
(norm.)
0.059 (0.089) 0.059 (0.090)
Female candidate *Part-time contract
(norm.)
20.005 (0.074) 20.006 (0.074)
Female candidate *Male recruiter (norm.) 20.067 (0.046) 20.066 (0.046)
Female candidate *Log(Average FTE in
firm) (norm.)
20.003 (0.022) 20.002 (0.022)
Female candidate *Percentage female
workers in occupation (norm.)
0.028 (0.020) 0.032 (0.021)
Female candidate *Average wage in
occupation (norm.)
0.021 (0.031)
Linear probability model Yes Yes Yes Yes Yes Yes
Dependent variable: Any positive reaction Yes Yes Yes Yes Yes Yes
Sector fixed effects interacted with Female
candidate
No No No No Yes Yes
Number of observations 1,152 1,152 1,152 1,152 1,074 1,074
Notes: Variables that are interacted with Female candidate are also included without interaction with this variable. Except for Female candidate, all variables are normalized (norm.)
by subtracting the mean among the population of female candidates. Continuous variables are further normalized by dividing by the standard deviation among this subpopulation.
Standard errors, corrected for clustering at the vacancy level, are in parentheses. In Models 5 and 6, 78 observations are dropped because of missing values for the added control
variables. FTE, full-time equivalency.
***;**;*indicates significance at the 1%; 5%; 10% level, respectively.
Conclusion
In this study, we have contributed to the literature by investigating the
importance of employer preferences in explaining sticky floors, the pattern
in which women are, compared to men, less likely to start to climb the job
ladder. Our study is novel in its approach to gathering original experimen-
tal data on gender differences in promotion opportunities. Specifically, we
conducted a field experiment in which pairs of fictitious job applications of
young employees, one male and one female, with a BA in business adminis-
tration or MA in business economics and with about five years of work expe-
rience in their first and current job were sent to companies with real job
postings in Belgium. Because we sent applications to companies both with
vacancies that implied a promotion in occupational level and/or job author-
ity and with vacancies at the same level as the current job, we were able to
test whether the unequal treatment of young men and women in the labor
market was heterogeneous by whether jobs imply a promotion from the
applicants’ current position. We found significant evidence for hiring dis-
crimination against females when they apply for jobs at a higher occupa-
tional level. For these jobs, female applicants received, compared to males,
about 33% fewer invitations for a job interview and 19% fewer positive reac-
tions (positive callback in the broad sense). In contrast, we found no signifi-
cant heterogeneity in hiring discrimination by the job-authority level of the
posted job. In addition, based on our regression analysis, we found no evi-
dence for heterogeneity in the equal treatment of males and females by the
average wage in the occupation mentioned in the vacancy.
From a policy perspective, we cannot deny that our results point in the
direction of discrimination against young women, albeit only when they
apply for jobs that imply a promotion in the applicant’s occupational level.
In Belgium, as in many other Organisation for Economic Co-operation and
Development (OECD) countries, the legal framework to punish labor-
market discrimination is available,
20
so that the main benefit seems to lie in
the more vigorous detection of the unequal chances of being hired for jobs
involving a promotion. Researchers could investigate whether this could
happen based on a systematic application of the method we have used in
this study. An alternative policy option, which could be introduced volunta-
rily by individual firms or imposed by the government, is anonymizing job
applications. Åslund and Skans (2012) showed, by means of a quasi-
experiment in Gothenburg, that removing the names from job applications
may indeed increase the first callbacks and job offers for female job candi-
dates. Krause, Rinne, and Zimmermann (2012) and Behaghel, Le
Barbanchon, and Cre´pon (2015), however, argued that policymakers
should be aware of the unintended effects of anonymized job applications
20
See Bassanini and Saint-Martin (2008) for a detailed description of the legal and institutional antidis-
crimination framework in 23 OECD countries.
732 ILR REVIEW
(e.g., with respect to affirmative action)
21
and that therefore the method of
implementation is crucial.
Because employees at a higher occupational level imply a higher invest-
ment and higher (temporary) exit costs, our results may be related to the
career penalty of motherhood (Budig and England 2001; Correll, Benard,
and Paik 2007; Kmec 2011). In this respect, broadening the concept of dis-
crimination to include dimensions regarding how jobs are structured and
rewarded that have a disparate impact on mothers could be fruitful. In addi-
tion, tax incentives could be provided to employers who reintegrate women
who return to their jobs after maternity leave.
Given our chosen experimental design, our results cannot not be
explained by theories of (‘‘old-boy’’) networking and sector segregation
(Groot and Maassen van den Brink 1996; Forret and Dougherty 2004;
Ludsteck 2014). Although we cannot rule out the possibility that these
mechanisms contribute to sticky floors, our results suggest that awareness
campaigns that target these mechanisms will not be sufficient in fighting
gender inequalities in promotion opportunities.
References
Albert, Rocı´o, Lorenzo Escot, and Jose´ A. Ferna´ndez-Cornejo. 2011. A field experiment to
study sex and age discrimination in the Madrid labour market. International Journal of
Human Resource Management 22: 351–75.
Albrecht, James, Anders Bjo
¨rklund, and Susan Vroman. 2003. Is there a glass ceiling in
Sweden? Journal of Labor Economics 21: 145–77.
Altonji, Joseph G., and Rebecca M. Blank. 1999. Race and gender in the labor market. In
Orley Ashenfelter and David Card (Eds.), Handbook of Labor Economics, pp. 3143–259.
Amsterdam: Elsevier.
Arrow, Kenneth J. 1973. The theory of discrimination. In Orley Ashenfelter and Albert Rees
(Eds.), Discrimination in Labor Markets, pp. 3–33. Princeton, NJ: Princeton University Press.
Arulampalam, Wiji, Alison L. Booth, and Mark L. Bryan. 2007. Is there a glass ceiling over
Europe? Exploring the gender pay gap across the wage distribution. ILR Review 60:
163–86.
Åslund, Olof, and Oskar N. Skans. 2012. Do anonymous job application procedures level the
playing field? ILR Review 65: 82–107.
Azmat, Ghazala, and Barbara Petrongolo. 2014. Gender and the labor market: What have we
learned from field and lab experiments? Labour Economics 30: 32–40.
Baert, Stijn. 2014. Career lesbians: Getting hired for not having kids? Industrial Relations Jour-
nal 45: 543–61.
Baert, Stijn, Bart Cockx, Niels Gheyle, and Cora Vandamme. 2015. Is there less discrimina-
tion in occupations where recruitment is difficult? ILR Review 68: 467–500.
Baker, Joe G. 2003. Glass ceilings or sticky floors? A model of high-income law graduates.
Journal of Labor Research 24: 695–711.
21
Although several developed countries (e.g., Australia, Belgium, France, Finland, Spain, New
Zealand, and Norway) have imposed gender quotas on the boards of big public institutions and/or pri-
vate companies, we are not aware of gender quotas to guarantee the access of women into middle-
management (or lower) positions (Fagan, Gonza´lez Mene´ ndez, and Go´mez Anson 2012; Bertrand,
Black, Jensen, and Lleras-Muney 2014).
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 733
Bassanini, Andrea, and Anne Saint-Martin. 2008. The price of prejudice: Labour market dis-
crimination on the grounds of gender and ethnicity. In OECD Employment Outlook: 2008,
pp. 139–202. Paris: Organisation for Economic Co-operation and Development.
Baumle, Amanda K. 2009. The cost of parenthood: Unraveling the effects of sexual orienta-
tion and gender on income. Social Science Quarterly 90: 983–1002.
Becker, Gary. 1957. The Economics of Discrimination. Chicago: University of Chicago Press.
Behaghel, Luc, Thomas Le Barbanchon, and Bruno Cre´pon. 2015. Unintended effects of
anonymous resume. American Economic Journal: Applied Economics 7(3): 1–27.
Bergmann, Barbara R. 1989. Does the market for women’s labor need fixing? Journal of Eco-
nomic Perspectives 3: 43–60.
Bertrand, Marianne, Sandra E. Black, Sissel Jensen, and Adriana Lleras-Muney. 2014. Break-
ing the glass ceiling? The effect of board quotas on female labor market outcomes in Nor-
way. NBER Working Paper No. 20256. Cambridge, MA: National Bureau of Economic
Research.
Bertrand, Marianne, and Sendhil Mullainathan. 2004. Are Emily and Greg more employable
than Lakisha and Jamal? A field experiment on labor market discrimination. American Eco-
nomic Review 94: 991–1013.
Blandford, John M. 2003. The nexus of sexual orientation and gender in the determination
of earnings. ILR Review 56: 622–42.
Blau, Francine, and Jed Devaro. 2007. New evidence on gender differences in promotion
rates: An empirical analysis of a sample of new hires. Industrial Relations 46: 511–550.
Blinder, Alan S. 1973. Wage discrimination: Reduced form and structural estimates. Journal of
Human Resources 8: 436–55.
Bloothooft, Gerrit, and David Onland. 2011. Socioeconomic determinants of first names.
Names 59: 25–41.
Booth, Alison L. 2009. Gender and competition. Labour Economics 16: 599–606.
Booth, Alison L., Marco Francesconi, and Jeff Frank. 2003. A sticky floors model of promo-
tion, pay, and gender. European Economic Review 47: 295–322.
Booth, Alison L., and Andrew Leigh. 2010. Do employers discriminate by gender? A field
experiment in female-dominated occupations. Economics Letters 107: 236–38.
Borjas, George J. 2009. Labor Economics. New York: McGraw-Hill.
Budig, Michelle J., and Paula England. 2001. The wage penalty for motherhood. American
Sociological Review 66: 204–25.
Bureau Van Dijk. 2014. Bel-first database. Accessed at https://belfirst.bvdinfo.com/version-
20151021/Home.serv?product=belfirstneo (August 21, 2014).
Bursell, Moa. 2007. What’s in a name? A field experiment test for the existence of ethnic dis-
crimination in the hiring process. Linnaeus Center for Integration Studies Working Paper
No. 2007–7. Stockholm: Stockholm University Linnaeus Center for Integration Studies.
Carlsson, Magnus. 2011. Does hiring discrimination cause gender segregation in the Swedish
labor market? Feminist Economics 17: 71–102.
Carter, Nancy M., and Christine Silva. 2011. Does Doing All the Right Things Really Get Women
Ahead? New York: Catalyst.
Charles, Kofi K., and Jonathan Guryan. 2008. Prejudice and wages: An empirical assessment
of Becker’s The Economics of Discrimination.Journal of Political Economy 116: 773–809.
Christofides, Louis N., Alexandros Polycarpou, and Konstantinos Vrachimis. 2013. Gender
wage gaps, ‘sticky floors’ and ‘glass ceilings’ in Europe. Labour Economics 21: 86–102.
Correll, Shelly J., Stephen Benard, and In Paik. 2007. Getting a job: Is there a motherhood
penalty? American Journal of Sociology 112: 1297–339.
Darity, William A., Jr., and Patrick L. Mason. 1998. Evidence on discrimination in employ-
ment: Codes of color, codes of gender. Journal of Economic Perspectives 12: 63–90.
Deschacht, Nick, Anneleen Baerts, and Marie-Anne Guerry. 2011. De m/v carrie
`rekloof. Carrie
`re-
verschillen tussen vrouwen en mannen in Belgie
¨. Ghent: Academia Press.
Directorate-General Statistics and Economic information of Belgium. 2014. Average wages
between 1999 and 2013. Accessed at http://statbel.fgov.be/nl/modules/publications/stati
stiques/arbeidsmarkt_levensomstandigheden/gemiddelde_bruto_maandlonen.jsp (August
21, 2014).
734 ILR REVIEW
Duflo, Esther. 2006. Field experiments in development economics. Unpublished manuscript,
MIT Economics.
Fagan, Colette, Maria Gonza´lez Mene´ndez, and Silvia Go´mez Anson. 2012. Women on Corpo-
rate Boards and in Top Management: European Trends and Policy. Chippenham, UK: Palgrave
Macmillan.
Fang, Zheng, and Chris Sakellariou. 2011. A case of sticky floors: Gender wage differentials
in Thailand. Asia Economic Journal 25: 35–54.
Filippin, Antonio, and Andrea Ichino. 2005. Gender wage gap in expectations and realiza-
tions. Labour Economics 12: 125–45.
Firth, Michael. 1982. Sex discrimination in job opportunities for women. Sex Roles 8: 891–901.
Forret, Monica L., and Thomas W. Dougherty. 2004. Networking behaviors and career out-
comes: Differences for men and women? Journal of Organizational Behavior 25: 419–37.
Gradı´n, Carlos, and Coral del Rı´o. 2009. Gender wage differentials in Spain: A distributional
approach by subpopulations. Hacienda Pu
´blica Espan
˜ola 189: 9–46.
Groot, Wim, and Henriette Maassen van den Brink. 1996. Glass ceilings or dead ends: Job
promotion of men and women compared. Economics Letters 53: 221–26.
Hachen, David S. 1990. Three models of job mobility in labor markets. Work and Occupations
17: 320–54.
Hakim, Catherine. 2010. Erotic capital. European Sociological Review 26: 499–518.
International Labour Organization (ILO). 2013. ISCO-08 group definitions—first draft.
Accessed at http://www.ilo.org/public/english/bureau/stat/isco/docs/gdstruct08.xls
(August 24, 2014).
Kaas, Leo, and Christian Manger. 2012. Ethnic discrimination in Germany’s labour market:
A field experiment. German Economic Review 13: 1–20.
Kassenboehmer, Sonja C., and Mathias G. Sinning. 2014. Distributional changes in the gen-
der wage gap. ILR Review 67: 335–61.
Kmec, Julie A. 2011. Are motherhood wage penalties and fatherhood bonuses warranted?
Social Science Research 40: 444–59.
Krause, Annabelle, Ulf Rinne, and Klaus F. Zimmermann. 2012. Anonymous job applications
in Europe. IZA Journal of European Labor Studies 1: article 5. doi:10.1186/2193-9012-1-5.
Ludsteck, Johannes. 2014. The impact of segregation and sorting on the gender wage gap:
Evidence from German linked longitudinal employer-employee data. ILR Review 67:
362–94.
Matteazzi, Eleonora, Ariane Pailhe´, and Anne Solaz. 2014. Part-time wage penalties for
women in prime age: A matter of selection or segregation? Evidence from four European
countries. ILR Review 67: 955–85.
Neumark, David, Roy Bank, and Kyle Van Nort. 1996. Sex discrimination in restaurant hir-
ing: An audit study. Quarterly Journal of Economics 111: 915–41.
Oakley, Judith G. 2000. Gender-based barriers to senior management positions: Understand-
ing the scarcity of female CEOs. Journal of Business Ethics 27: 321–34.
Oaxaca, Ronald. 1973. Male-female wage differentials in urban labor markets. International
Economic Review 14: 693–709.
Petit, Pascale. 2007. The effects of age and family constraints on gender hiring discrimina-
tion: A field experiment in the French financial sector. Labour Economics 14: 371–91.
Reuben, Ernesto, Pedro Rey-Biel, Paola Sapienza, and Luigi Zingales. 2012. The emergence
of male leadership in competitive environments. Journal of Economic Behavior & Organiza-
tion 83: 111–17.
Riach, Peter A., and Judith Rich. 1987. Testing for sexual discrimination in the labour mar-
ket. Australian Economic Papers 26: 165–78.
———. 2002. Field experiments of discrimination in the market place. Economic Journal 112:
480–518.
———. 2006. An experimental investigation of sexual discrimination in hiring in the English
labor market. B.E. Journal of Economic Analysis and Policy 5: 1–22.
Romeu Gordo, Laura. 2009. Why are women delaying motherhood in Germany? Feminist Eco-
nomics 15: 57–75.
DO EMPLOYER PREFERENCES CONTRIBUTE TO STICKY FLOORS? 735
Smith, Nina, Valdemar Smith, and Mette Verner. 2013. Why are so few females promoted
into CEO and vice president positions? Danish empirical evidence, 1997–2007. ILR Review
64: 949–80.
Taubman, Paul J., and Terence Wales. 1974. Higher Education and Earnings: College as an Invest-
ment and Screening Device. Cambridge, MA: National Bureau of Economic Research.
Trendstop. 2014. Trendstop database of Belgian companies. Accessed at http://trendstop
.knack.be/en/home.aspx (August 21, 2014).
Unizo. 2014. Tool for looking up company names. Accessed at http://www.unizo.be/tools/
ondernemingsnummer-opzoeken (August 21, 2014).
Weichselbaumer, Doris. 2004. Is it sex or personality? The impact of sex stereotypes on dis-
crimination in applicant selection. Eastern Economic Journal 30: 159–86.
Weinberger, Catherine. 2011. In search of the glass ceiling: Gender and earnings growth
among U.S. college graduates in the 1990s. ILR Review 64: 949–80.
Wood, Martin, Jon Hales, Susan Purdon, Tanja Sejersen, and Oliver Hayllar. 2009. A test for
racial discrimination in recruitment practice in British cities. DWP Research Report No.
607. London: UK Department of Work and Pensions.
Yap, Margaret, and Alison M. Konrad. 2009. Gender and racial differentials in promotions: Is
there a sticky floor, a mid-level bottleneck, or a glass ceiling? Relations Industrielles—Indus-
trial Relations 64: 593–619.
736 ILR REVIEW
