ArticlePDF Available

Blind dates: quasi-experimental evidence on discrimination

Authors:

Abstract and Figures

This paper provides evidence on discrimination in the hiring process. We use data generated from a “policy experiment” conducted at the Swedish public employment offices. Individuals registered at these offices can post their qualifications in a database available to employers over the Internet. Potential employers are free to search this database for job candidates and contacts between employers and candidates are recorded. We use two complementary identification strategies. First, since our data contain all information available to employers, we argue that selection on observables is viable. Second, we utilize the fact that individuals can choose not to reveal their name and gender to potential employers. Our main finding is that women have a 15 percent lower chance than men of getting contacted by employers and that this differential is fully explained by discrimination. Our results concerning ethnic discrimination are less conclusive, probably due to measurement errors.
Content may be subject to copyright.
Blind Dates: Quasi-Experimental Evidence on
Discrimination
*
by
Per-Anders Edin
a
and Jonas Lagerström
b
May 12, 2004
Abstract
This paper provides evidence on employer discrimination of women and ethnic
minorities in the hiring process. We use data generated from a “policy
experiment” conducted at the Swedish public employment offices. Individuals
registered at these offices can post their qualifications in a database available to
employers over the Internet. Potential employers are free to search this
database for job candidates and contacts between employers and candidates are
recorded. The important feature of this system is that individuals can choose to
“censor” some of the information available to potential employers. In
particular, individuals can choose not to reveal their name and gender. By
comparing the “contact rate” of censored and non-censored women and
minorities, we are able to investigate how employers use gender and “foreign
names” as a screening device in their hiring process.
Keywords: Discrimination.
JEL classification: J64, J71.
*
We gratefully acknowledge comments from Nils Gottfries, Peter Fredriksson, as well as
seminar participants at IFAU, Uppsala University and at the CEPR conference on Discrimination
and Unequal Outcomes held in Le Mans, France, 17-20 January 2002. We also thank AMS and
Claes-Göran Lock for providing us with the data.
a
Department of Economics, Uppsala University, and Institute for Labour Market Policy
Evaluation (IFAU). E-mail: Per-Anders.Edin@nek.uu.se.
b
Department of Economics, Uppsala University. E-mail: Jonas.Lagerstrom@nek.uu.se
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
1
Table of contents
1 Introduction ............................................................................................... 3
2 The Internet Search Database .................................................................... 5
3 The data ..................................................................................................... 6
4 Empirical results...................................................................................... 11
5 Concluding remarks................................................................................. 16
References.......................................................................................................... 17
Appendix 1: Comparison of the characteristics of the inflows.......................... 19
Appendix 2: Comparison of the selection into “blindness”............................... 20
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
2
1 Introduction
Like in many other Western economies, discrimination in the labor market is a
major issue in the Swedish policy debate. In spite of its well known equality of
outcomes, the Swedish labor market still produces large differentials in labor
market outcomes. The two groups that are most often mentioned in the
Swedish debate are immigrants and women. The key question, which is very
hard to answer, is how important labor market discrimination is to explain
these differences. This paper analyzes discrimination in the hiring process.
There is ample evidence that observed differentials are mainly driven by
differences in hiring and promotion, rather than by differences in wages within
jobs.
Immigrants in the Swedish labor market earn substantially less than native
Swedes and have actually been loosing ground over the last decade. In 1998,
the average non-OECD immigrant earned about 45 percent of what a native
Swede with similar observed characteristics earned per year (Edin and Åslund,
2001). Roughly a quarter of this difference was due to differences in hourly
wages. Another quarter was due to less working hours among those employed.
The remaining half of the earnings difference was due to lower employment
rates among immigrants.
Even though Swedish women are relatively high paid, compared to in most
other Western economies, they still earn only about 80 percent of men’s hourly
wage. A large share of the earnings gap is driven by occupational segregation.
Controlling for standard “human capital variables”, reduces the wage gap by
about half, e.g. Le Grand et al (1997) and Albrecht et al (2003). Most of the
remaining gap, though, is eliminated if detailed controls for occupations are
introduced (Meyerson and Petersen, 1997). Both these examples illustrate that
the sorting of workers to jobs, through hiring and promotion, is crucial for
generating the observed differences in outcomes across groups in the labor
market. Consequently, we need to get a better understanding for how this
sorting occurs to get a grip of the role of discrimination in the labor market.
The standard approach to analyzing discrimination, building on the seminal
work by Becker (1957), has been to estimate various outcome equations in the
spirit of Blinder-Oaxaca. Even though these analyses are informative, they
require very strong assumptions to infer anything about discrimination. For
instance, we have to assume that the unobservables are not systematically
different across groups.
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
3
One approach that tries to deal with this issue in the hiring process is the
“Audit method”, surveyed by Riach and Rich (2002). Here, observably similar
individuals from different groups, e.g. sex or ethnicity, apply for jobs at the
same firms. A recent example is Bertrand and Mullanaithan (2003) who found
that résumés carrying distinctively Black names are less likely to receive job
interviews. This approach seems to be a step forward, but also has it’s
limitations as discussed by Heckman (1998). He shows that the Audit studies
may actually be worse than regular observational studies under some
assumptions. For example, a man and a woman who share the same personal
characteristic may send a different signal in terms of anticipated productivity
which the researcher cannot control for. Also, Heckman argues that the
findings considering discrimination depends on differences in the variance of
uncontrolled characteristics between groups and/or the qualifications needed
for the applied job. In addition, of course, there are ethical issues: in these
experiments the firms cannot choose whether to participate and they get an
extra cost of recruiting applicants who have no intention of accepting a job
offer.
The most compelling evidence of discrimination in the recruitment process
has been produced in an analysis of what we refer to as a natural experiment.
Goldin and Rouse (2000) uses the introduction of blind auditions in U.S.
symphony orchestras to analyze discrimination of women in hiring. In a
differences-in-differences analysis, they find that the introduction of blind
auditions increased the probability that a woman will be hired by a substantial
amount. The probability that a woman would be advanced out of a preliminary
round was increased by 50 percent, and her likelihood of winning the final
round increased by 30 percent when blind auditions were introduced.
Our approach is heavily inspired by the Goldin and Rouse (2000) paper. We
use data generated from a “policy experiment” conducted at the Swedish public
employment offices. Individuals registered at these offices can post their
qualifications in a database available to employers over the Internet. Potential
employers are free to search this database for job candidates and contacts
between employers and candidates are recorded. The important feature of this
system is that individuals can choose to “censor” some of the information
available to potential employers. In particular, individuals can choose not to
reveal their name and gender. By comparing the “contact rate” of censored and
non-censored women and minorities, we are able to investigate how employers
use gender and “foreign names” as a screening device in their hiring process.
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
4
The rest of the paper is outlined as follows. In Section 2 we describe the
institutional features of the “experiment” we are using. We then turn to
describing the data collection procedure and our sample in Section 3. Section 4
contains our estimation strategy and the empirical estimates of discrimination.
In Section 5 we conclude by discussing the implications of our results for
outcomes in the labor market.
2 The Internet Search Database
Sweden has a long history of publicly provided employment exchanges.
Already in the 1930’s, there were public (municipal) employment offices
whose main objective was to improve the matching process in the labor market.
Nowadays, the employment offices are run by the National Labor Market
Board (AMS), who also administer the large supply of various active labor
market policies.
In the fall of 1997 AMS started up a new internet based search database to
further promote efficiency in the matching of job searchers and employers.
This new database, called the Search database (“Sökandebanken”), provides
the data for our study. The basic idea with this new tool is that all job
applicants (employed or not) can post their resumes on the search database free
of charge. Furthermore, there is no requirement to register at the employment
office before entering the database. Job searchers can present their job histories
and qualifications, as well as list their preferred occupations and other
aspirations. They are also required to write a more personal letter about
themselves. All this can be done either at one of the employment offices or
through internet. The software also provides examples of how to put up a CV
and similar practical issues. By the spring of 2001, about 50,000 individuals
were registered in the Search database. This corresponded to about 30 percent
of the number of unemployed according to the Labor Force Survey. The
monthly inflow of new individuals in the database was about 11,000
individuals.
The Search database is open for employers who are recruiting, provided that
they are registered employers in the public registers and in AMS’s internal
customer register. If an employer finds a potential candidate in the pool of job
searchers in the database, she is free to contact the candidate. In some cases the
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
5
contact can occur outside the system, e.g. by an e-mail to the job searchers
private address, and the contacts are not registered. The most usual way of
contacting, however, is by e-mail to the job searcher’s mailbox within the
Search database. These contacts are registered in the database.
The most important feature of the Search database, for our purposes, is that
the individual job searcher can choose not to disclose all personal information.
This option allows individuals to censor information on their name, sex and
age. In practice, since there is no separate entry for ethnicity, this means that
individuals can choose to censor information on age, sex and ethnicity. This
option was primarily introduced as a service to employed job searchers, who
did not want their employers to find out that they were looking for other jobs.
The presence of “blind” observations concerning some key variables is the
cornerstone of our identification strategy further discussed below. A second
important feature of the data is that we observe all the information that the
employers observe.
3 The data
The Search database has not been readily available for research purposes. In
order to get access to the data we had to obtain permission from each individual
job searcher. This was achieved, in cooperation with AMS, by adding an
introductory page to the Search database. This page contained a question about
whether the job searchers were willing to permit that data concerning
themselves was used for research purposes. All individuals that were or became
users of the search database got this question the first time they logged in to the
database from March 1, 2001. If they then agreed to “participate”, they got two
additional questions directly motivated by our research topic:
1. Are you a male or a female?
2. Do you think that employers in general perceive your name as Swedish
or foreign?
The answers to these questions were needed to get information on sex and
“ethnicity” for individuals who had exercised their option to censor these
entries in the search database.
The primary data used in this paper was collected in March 2001. It consists
of all individuals who accepted to participate among those who were in the
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
6
database and logged in to the database between March 1 and March 12.
Approximately 50 percent of those who logged in during this time period
accepted to participate, resulting in a sample of 8,666 individuals. Because we
did not want to include youth in secondary school in the sample, we excluded
all individuals aged below 20.
1
That gives us the sample used in this study
consisting of 8,043 individuals.
The sample characteristics are reported in Table 1. The first column refers
to the entire sample, while the second column refers to individuals who have
censored information on sex or name. In the full sample we note the average
duration in the database is over 33 weeks and that a third of the sample has
been contacted by an employer at least once during their “spell”. We also see
that half the sample is female and that 13 percent consider themselves having a
foreign name.
The sample of individuals that have concealed their name or sex (in column
2) consist of 1,472 individuals, corresponding to almost 17 percent of the full
sample. There are at least three differences between the sample with blind
observations and the full sample worth mentioning: i) they have shorter
duration in the database, ii) they have not received as many employer contacts,
and iii) they are to a larger extent low educated.
In most other respects, the two samples look pretty similar. In particular, it’s
worth nothing that the share of females and foreign names are fairly similar
across samples.
Table 1. Descriptive statistics, means
Variable Full sample Blind observations only
(name or sex)
Händel
Contacted
Duration (weeks)
0.341
34.551
0.293
25.694
-
58,7
Education:
1
Most of the applicants aged below 20 look for work during the summer break or temporary
work on school holidays etc. Therefore, it seems natural to exclude them in our empirical
investigation.
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
7
Primary 0.072 0.172 0.228
Secondary (gymnasium) 0.489 0.372 0.616
University 0.439 0.456 0.156
Good language skills:
Swedish
English
French, Spanish or German
Good computer skills
0.969
0.561
0.197
0.738
0.966
0.498
0.192
0.629
-
-
-
-
Managerial experience 0.343 0.344 -
Telecommuting experience
Research experience
5 years work experience
Drivers license
Region:
Stockholm
Uppsala
Södermanland
Östergötland
Jönköping
Kronoberg
Kalmar
Gotland
Blekinge
Skåne
Halland
Västra Götaland
Värmland
Örebro
Västmanland
Dalarna
Gävleborg
Västernorrland
Jämtland
Västerbotten
Norrbotten
Preferred occupations:
0.124
0.054
0.421
0.788
0.293
0.089
0.078
0.080
0.059
0.046
0.049
0.020
0.046
0.187
0.075
0.182
0.049
0.066
0.074
0.052
0.055
0.042
0.021
0.041
0.031
0.124
0.057
0.393
0.772
0.304
0.087
0.066
0.073
0.047
0.036
0.047
0.013
0.034
0.149
0.044
0.144
0.042
0.061
0.060
0.039
0.042
0.023
0.021
0.030
0.017
-
-
0.298
-
0.089
0.023
0.033
0.053
0.038
0.021
0.031
0.008
0.020
0.131
0.041
0.190
0.042
0.034
0.033
0.043
0.045
0.037
0.021
0.028
0.041
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
8
Elementary occupations (Amsyk 9)
Legislators, senior officials and
managers (Amsyk 1)
Professionals (Amsyk 2)
Technicians and associate
professionals (Amsyk 3)
Clerks (Amsyk 4)
Service workers and shop sales
workers (Amsyk 5)
Skilled agricultural and fishery
workers (Amsyk 6)
Craft and related trades workers
(Amsyk 7)
Plant and machine operators and
assemblers (Amsyk 8)
Foreign name
Female
Age
Age 20-25
Age 26-35
Age 36-50
Age 50-
Employed
Unemployed
University student
In other training
On child leave
Blind name
Blind gender
Blind age
Blind name * Foreign name
Blind gender * Female
Blind age * Age > 45 years
0.105
0.030
0.279
0.290
0.248
0.190
0.021
0.116
0.100
0.134
0.487
33.8
0.289
0.331
0.279
0.101
0.490
0.385
0.081
0.040
0.009
0.033
0.084
0.084
0.007
0.041
0.029
0.064
0.030
0.280
0.253
0.178
0.134
0.011
0.085
0.062
0.152
0.474
34.5
0.279
0.316
0.287
0.118
0.441
0.459
0.074
0.022
0.011
0.103
0.014
0.090
0.104
0.143
0.309
0.026
0.102
0.102
0.206
0.584
41.0
0.091
0.259
0.374
0.256
0.357
0.520
0.087
0.054
0.028
-
-
-
-
-
-
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
9
# Observations
8,043
922
26,532
An issue that arises naturally here concerns representativity. To what
population can we possible generalize our results? There are several steps in
the selection process on which we have very little information. First, both
employed and unemployed individuals choose whether to register in the
database. This selected sample may well be very different from the typically
used samples of unemployed. Second, individuals were free to choose whether
to release their data for research. We have no way of assessing this selection
process.
One way of assessing the specificity of our sample is to compare it with a
random sample of job searchers. In the third column of Table 1 we report the
mean characteristics of the stock of job searchers in 2001 using data from
LINDA (Edin and Fredriksson, 2000). There are some distinctive differences
between the two groups of job searchers. We find that our sample is younger,
more educated, and has more work experience. We also have a smaller share of
females and minorities in our sample.
One explanation of these differences is that the individuals in our sample
have much shorter job search duration, i.e. we compare high quality individuals
in the Search database to low quality individuals in LINDA. In Table A1, we
account for these effects by comparing inflows instead of stocks. The two first
columns show that the difference between the samples decreases if we compare
the inflow into the Search database to the inflow into LINDA. The similarities
are even more striking in the last two columns of Table A1, where we compare
the inflows of unemployed into the two bases. This is because an unemployed
individual who register at the Employment Office is requested by the
caseworker to join the Search database. Participation is not forced upon the
individual but simply recommended; there are no sanctions should the client
refuse. However, the vast majority of the people who register also choose to
join the base.
Concerning the representativity of our results, this indicates that our results
have some external validity to the unemployed population in Sweden.
However, there are other selection issues as well. For example, there may be
differences in the left-out variables between those who agreed to participate in
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
10
this study and those who did not. This should be kept in mind when drawing
inferences from our study to broader settings.
4 Empirical results
The empirical strategy of this paper is inspired by the work of Goldin and
Rouse (2000). We make use of the fact that some individuals have concealed
their sex and (foreign) name in a “differences-in-differences” framework. We
write our estimating equation as
P
i
= α + β’ F
i
+ γ’ B
i
+ δ’(F
i
B
i
) + θ’ X
i
where P is the probability of receiving at least one employer contact, F is a
vector of characteristics that we believe may be subject to discrimination
(female and foreign name), B is vector of variables showing what
characteristics are concealed (gender and name), and X is a vector of individual
characteristics including information on job preferences and a quadratic in
duration in the Search database.
The parameters of interest here are, δ, the vector of coefficients on the
interactions between F and B. There are two key interactions; between female
and concealed gender, and between foreign name and concealed name. Under
some additional assumptions, the coefficients of these two interactions measure
the change in the probability of receiving an employer contact that a female or
an individual with a foreign name experiences by concealing their sex and
name. The key assumption here is that there are no systematic differences in
the selection (on left-out variables) into “blindness” across groups. To get an
indication whether this assumption is valid, we have estimated linear
probability models of concealed identity. We present these results in Table A2
in the Appendix. The marginal effects of the observable characteristics are
similar across sexes; only four of the 55 are significantly different.
2
The fact
that the observable variables determine “blindness” in the same way across
2
Formally, including interaction terms for the sex variable and all the other explanatory variables
does not make our model significantly better (F-value of 1.28, p-value of 9 percent).
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
11
groups may support the assumption that the effect of potential left-out variables
is the same across groups as well.
The vector of coefficients on B, γ, captures the change in contact probability
that applicants face by not disclosing different parts of their identity (i.e. name,
gender or age). These effects probably consist of several things. For example,
they might reflect discrimination; given that discriminating employers
understand that a share of “blind” applicants consists of individuals from the
group that is discriminated against, these employers will be resistant to contact
an applicant who has not revealed his/her identity. In addition, noting that the
option of concealing the identity was introduced as a service to employed job
searchers who desired anonymity, the effects may partly capture employers’
preferences towards employed applicants.
We start our empirical analysis by showing some further descriptive
information. In Table 2 we report the share of individuals in four groups that
have been contacted at least once by an employer. It turns out that the share of
women that have been contacted is about 7 percentage points lower than for
men. Similarly, individuals with foreign names have a 3 percentage point lower
share than individuals with a Swedish name. The issue in the remainder of this
section is to what extent these differences in employer contacts reflect
discrimination of women and ethnic groups.
Table 2. Employer contacts by group
Group Contact # Observations
Males 0.378 4,127
Females 0.302 3,916
Swedish name 0.346 6,965
Foreign name 0.310 1,078
The main results of our analysis are presented as linear probability models of
employer contacts in Table 3.
3
The first column serves as a point of reference.
Here we restrict ourselves to the sub-sample of individuals with no concealed
information and run a standard regression. These results suggest that
3
Using Logit models we obtain the same qualitative results.
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
12
individuals with foreign names do not have significantly lower contact rates,
while women have a 4.7 percentage point lower incidence of contacts than
men.
In the second column we use the full sample and utilize the interactions
between characteristics and concealed information to identify potential
discrimination. (Here we also allow for an interaction between “Blind age” and
“Over 45 years” even though this effect is less clearly tied to discrimination.) A
first observation is that there seems to be no effects of concealing information
on the contact rate. None of the main effects (blind name, blind gender, and
blind age) is statistically significant and the point estimates are fairly small.
Turning to the parameters of interest, we see that only the interaction effect
for women is significant. It indicates that a woman’s chance of receiving an
employer contact increases by 6 percentage point if she conceals her sex. The
estimates for individuals with foreign names and those over 45 years of age are
similar in magnitude, but not statistically significant.
Several other things are worth noting. First, older searchers face a lower
probability than younger searchers to get contacted by an employer. Second,
education and labour market experience have the expected signs. A higher
level of completed education, or more labour market experience, has a clear
positive effect on the probability to get contacted. Third, employed applicants
face significantly lower probabilities of getting a contact.
4
In the final two columns of Table 3 we report separate estimates by level of
education. It turns out the evidence suggesting discrimination of women is
driven by the less educated sample. In the highly educated sample, the point
estimate on the interaction with foreign names is larger, though not significant.
Before drawing any conclusions we must consider the role of measurement
errors. It turns out that this may be a serious problem with the interaction with
foreign names, where only about 50 percent of the “blind foreign names” are
truly blind. The other half could be identified indirectly using for example rare
language skills or the personal letter in the database. This will of course
introduce potentially serious attenuation bias in our estimate of the effect of
having a foreign name. For the female applicants with “blind gender”, the share
4
Eriksson and Lagerström (2004) examines whether the fact the firms view employment status
as an important signal for productivity can explain the persistence of unemployment. (Ta bort alt.
skriv om)
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
13
that is truly blind is higher and the attenuation bias smaller since we have found
it harder to identify the gender using for instance working experiences or skills.
In addition, it is likely that our definition of minorities is seriously wanting.
Here we group labor immigrants from the Scandinavian countries with refugee
immigrants from Africa and the Middle East – groups we know have very
different outcomes in the labor market. For these reasons, we are inclined to
put much less weight on the estimates of the effect of foreign names than on
the effect of gender.
Table 3. Linear probability models of employer contact
Non-Blind
Sample
Full Sample High
Education
Sample
Low
Education
Sample
Duration (weeks)
Duration
2
/100
.011
(.0003)
-.004
(0.0002)
.010
(.0003)
-.004
(0.0002)
.011
(.0005)
-.004
(0.0003)
.010
(.0004)
-.003
(0.0003)
Foreign name -.010
(.015)
-.019
(.014)
-.030
(.021)
-.010
(.018)
Female
-.047
(.011)
-.051
(.011)
-.050
(.016)
-.053
(.015)
Over 50 years of age
-.113
(.022)
-.099
(.020)
-.090
(.033)
-.100
(.025)
36-50 years of age
-.079
(.016)
-.076
(.014)
-.071
(.023)
-.078
(.018)
26-35 years of age
-.032
(.013)
-.029
(.012)
-.029
(.019)
-.027
(.016)
Blind name
- .031
(.033)
.036
(.044)
.021
(.052)
Blind gender
- -.005
(.020)
-.010
(.037)
-.001
(.024)
Blind age
- -.013
(.024)
-.005
(.034)
-.019
(.034)
Blind name * Foreign name - .051 .111 -.060
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
14
(.068) (.087) (.108)
Blind gender * Female
- .057
(.029)
.022
(.050)
.085
(.035)
Blind age * Over 45 years
- .042
(.037)
.019
(.051)
.069
(.056)
Education:
Secondary (Gymnasium)
.014
(.019)
.022
(.017)
-
.024
(.018)
University .045
(.021)
.053
(.019)
- -
Good language skills:
Swedish
English
French, Spanish or
German
Good computer skills
.025
(.027)
.034
(.011)
.031
(.014)
.013
(.012)
.011
(.025)
.032
(.010)
.031
(.013)
.012
(.011)
-.023
(.043)
.039
(.017)
.043
(.016)
.003
(.020)
.027
(.031)
.027
(.013)
.009
(.020)
.018
(.014)
Managerial experience .037
(.012)
.052
(.011)
.060
(.016)
.038
(.016)
Telecommuting experience
Research experience
5 years work experience
No work experience
Labor market status:
Employed in preferred
occupation
University student
In other training
.026
(.017)
.015
(.024)
.034
(.013)
-.017
(.015)
.027
(.011)
-.032
(.020)
.024
(.025)
.025
(.015)
.005
(.022)
.024
(.012)
-.030
(.013)
.032
(.010)
-.025
(.018)
.024
(.023)
.031
(.020)
-.019
(.024)
.005
(.019)
-.029
(.020)
.036
(.017)
-.017
(.022)
.020
(.039)
.014
(.024)
.105
(.050)
.039
(.015)
-.037
(.018)
.031
(.013)
-.055
(.055)
.028
(.028)
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
15
On child leave
Drivers license
Constant
.069
(.057)
.011
(.013)
-.020
(.033)
.059
(.050)
.005
(.012)
-.003
(.031)
.013
(.077)
-.004
(.020)
.081
(.052)
.082
(.063)
.010
(.014)
-.015
(.037)
# observationer 6,657 8,043 3,530 4,513
R
2
0.2780 0.2819 0.2893 0.2765
Note: Standard errors in parentheses. Controls for regions of residence and preferred occupations
included. The high (low) education sample consists of individual with some (no) university
education.
5 Concluding remarks
In this paper we use data generated from a “policy experiment” conducted at
the Swedish public employment offices. Individuals registered at these offices
can post their qualifications in a database available to employers over the
Internet. The important feature of this system is that individuals can choose to
“censor” some of the information available to potential employers. In
particular, individuals can choose not to reveal their name and gender. By
comparing the “contact rate” of censored and non-censored women and
minorities we are able to investigate how employers use gender and “foreign”
names as a screening device in their hiring process.
Our empirical results show that women receive less job contacts than men
do even when controlling for qualifications. We also find that women that do
not reveal their gender receive as many job contacts as men with similar
characteristics. These results clearly demonstrate that employers use the gender
of the applicant as a screening device, and we interpret this as a clear sign of
discrimination.
Assessing the importance of discrimination for outcomes in the Swedish
labor market using these estimates is a much more difficult task. First, we have
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
16
no clear “structural” interpretation of our estimate. Second, we only observe
the first part of the chain of events that lead to a possible hiring. We have no
idea whether the mechanism we observe is reinforced or weakened in later
stages of the hiring process.
References
Albrecht, J, A Björklund and S Vroman (2003), “Is there a Glass Ceiling in
Sweden?”, Journal of Labor Economics, Vol. 21 (1), 145-177.
Becker G (1957), The Economics of Discrimination, University of Chicago
Press.
Bertrand, M and S Mullanaithan (2003), “Are Emily and Greg More Employ-
able than Lakisha and Jamal? A Field Experiment on Labor Market
Discrimination”, NBER Working Paper 9873.
Edin, P-A and O Åslund (2001), “Invandrare på 1990-talets arbetsmarknad”, in
Ofärd i välfärden, SOU 2001:54.
Edin, P-A and P Fredriksson (2000), “LINDA – Longitudinal Individual DAta
for Sweden”, Working Paper 2000:19, Department of Economics,
Uppsala University.
Goldin, C and C Rouse (2000), “Orchestrating Impartiality: The Impact of
“Blind” Auditions on Female Musicians”, American Economic Review,
Vol. 90, No. 4, 715-741.
le Grand, C (1997), “Kön, lön och yrke – yrkessegregering och lönediskrimine-
ring mellan män och kvinnor”, i I Persson and E Wadensjö (eds)
Kvinnors och mäns löner – varför så olika?, SOU 1997:136.
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
17
Heckman J (1998), “Detecting Discrimination”, Journal of Economic
Perspectives 12, 101 – 116.
Meyerson, E and T Petersen (1997), “Lika lön för lika arbete. En studie av
svenska förhållanden i internationell belysning”, in Kvinnors och mäns
löner – varför så olika?, SOU 1997:136.
Riach, P A and J Rich (2002), ”Field Experiments of Discrimination in the
Market Place”, Economic Journal 112, F480 – F518.
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
18
Appendix 1: Comparison of the
characteristics of the inflows
Table A1. Comparison of the characteristics of the inflow of unemployed in the
Applicant Database and the inflow of unemployed in Händel (in fractions)
Variable All
The Search
Database
All
Händel
Unemployed
The Search
Database
Unemployed
Händel
Highest level of completed
education:
Primary 0.17 0.34 0.29 0.34
Secondary 0.45 0.41 0.48 0.39
University 0.38 0.25 0.23 0.27
Work experience:
None 0.30 0.24 0.43 0.36
Some or long 0.70 0.66 0.57 0.64
Age:
Mean (years) 31.1 35.1 30.5 33.4
Age 20-25 0.39 0.32 0.43 0.38
Age 26-35 0.33 0.21 0.30 0.23
Age 36-50 0.22 0.32 0.21 0.26
Age 51- 0.06 0.15 0.06 0.14
Gender:
Female 0.49 0.47 0.41 0.43
Ethnicity:
Foreign name 0.16 0.28 0.19 0.34
Region:
Stockholm
Uppsala
Södermanland
Östergötland
Jönköping
Kronoberg
Kalmar
Gotland
Blekinge
Skåne
Halland
Västra Götaland
Värmland
Örebro
Västmanland
Dalarna
Gävleborg
Västernorrland
Jämtland
Västerbotten
Norrbotten
0.22
0.06
0.04
0.05
0.03
0.02
0.02
0.01
0.02
0.11
0.04
0.12
0.03
0.04
0.06
0.03
0.03
0.02
0.01
0.02
0.03
0.18
0.03
0.04
0.07
0.04
0.02
0.04
0.00
0.00
0.11
0.03
0.18
0.03
0.03
0.03
0.02
0.03
0.03
0.02
0.03
0.04
0.18
0.06
0.03
0.05
0.04
0.02
0.02
0.01
0.02
0.10
0.04
0.13
0.04
0.04
0.06
0.03
0.03
0.02
0.01
0.03
0.03
0.19
0.04
0.05
0.06
0.04
0.02
0.05
0.00
0.00
0.11
0.03
0.19
0.03
0.03
0.03
0.01
0.03
0.03
0.01
0.03
0.04
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
19
Preferred occupations:
Amsyk 1
Amsyk 2
Amsyk 3
Amsyk 4
Amsyk 5
Amsyk 6
Amsyk 7
Amsyk 8
Amsyk 9
0.02
0.21
0.19
0.17
0.15
0.02
0.08
0.07
0.10
0.03
0.15
0.08
0.12
0.26
0.02
0.10
0.12
0.13
0.01
0.16
0.18
0.18
0.20
0.02
0.11
0.10
0.14
0.04
0.17
0.07
0.11
0.25
0.02
0.11
0.09
0.14
Note: The data from the bases is for the inflow into unemployment in March 2001. The variable
“foreign name” in the Search database is compared to the variable “being born in a country other
than Sweden” in Händel.. The regions and the preferred occupations sum to more than one in the
Search Database, since it is possible to apply for several jobs.
Appendix 2: Comparison of the selection
into “blindness”
Table A2. Linear probability models of concealed sex, by sex.
Full sample Men Women
Duration (weeks)
Duration
2
/100
-.002
(.0002)
.000007
(.000001)
-.002
(.0003)
.000008
(.000002)
-.002
(.0003)
.000007
(.000002)
Foreign name -.014
(.009)
-.014
(.012)
-.017
(.013)
Female
-.002
(.007)
- -
Over 50 years of age
-.003
(.013)
-.006
(.018)
-.009
(.020)
36-50 years of age
-.006
(.010)
-.009
(.014)
.0001
(.014)
26-35 years of age
-.012
(.008)
-.023
(.012)
-.0004
(.012)
Education:
Secondary (Gymnasium)
-.118
(.013)
-.120
(.017)
-.113
(.019)
University -.106 -.116 -.094
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
20
(.014) (.019) (.021)
Good language skills:
Swedish
English
French, Spanish or German
Good computer skills
-.003
(.018)
-.019
(.007)
.001
(.008)
-.038
(.007)
.014
(.022)
-.018
(.010)
-.007
(.012)
-.030
(.011)
-.034
(.029)
-.020
(.010)
.008
(.011)
-.042
(.010)
Managerial experience .004
(.007)
-.013
(.010)
.023
(.011)
Telecommuting experience
Research experience
5 years work experience
No work experience
Labor market status:
Employed in preferred
occupation
University student
In other training
On child leave
Drivers license
Constant
.007
(.010)
.007
(.014)
-.002
(.008)
.064
(.010)
-.022
(.007)
-.047
(.013)
-.062
(.016)
-.003
(.032)
.004
(.008)
.302
(.023)
.015
(.012)
.027
(.018)
.008
(.011)
.091
(.013)
-.016
(.009)
-.080
(.019)
-.059
(.022)
.043
(.154)
-.003
(.012)
.299
(.029)
-.002
(.016)
-.023
(.022)
-.008
(.011)
.034
(.014)
-.028
(.010)
-.021
(.017)
-.068
(.022)
-.003
(.034)
.009
(.011)
.316
(.036)
# observationer 8,043 4,127 3,916
R
2
0.071 0.096 0.060
Note: Standard errors in parentheses. Controls for regions of residence and preferred
occupations included.
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination
21
... To avert age discrimination in the first phase of the screening procedure, policy makers as well as researchers recommended AAP (see Edin and Lagerström, 2006;Furtmueller et al., 2010;Åslund and Skans, 2012). Furtmueller et al. (2010, p. 10), for instance, concluded that "since employers are prohibited to select employees based on gender, birth date, nationality and marital status, resume forms should not ask for this personal information." ...
... Furtmueller et al. (2010, p. 10), for instance, concluded that "since employers are prohibited to select employees based on gender, birth date, nationality and marital status, resume forms should not ask for this personal information." Anonymous resume screening omits explicit demographic cues from resumes that are non-job-related, like date of birth, ethnic-sounding name, or gender (Edin and Lagerström, 2006;Åslund and Skans, 2012;Krause et al., 2012;Hiemstra et al., 2013). ...
Article
Full-text available
Anonymous resume screening, as assumed, does not dissuade age discriminatory effects. Building on job market signaling theory, this study investigated whether older applicants may benefit from concealing explicitly mentioned age signals on their resumes (date of birth) or whether more implicit/subtle age cues on resumes (older-sounding names/old-fashioned extracurricular activities) may lower older applicants’ hirability ratings. An experimental study among 610 HR professionals using a mixed factorial design showed hiring discrimination of older applicants based on implicit age cues in resumes. This effect was more pronounced for older raters. Concealing one’s date of birth led to overall lower ratings. Study findings add to the limited knowledge on the effects of implicit age cues on hiring discrimination in resume screening and the usefulness of anonymous resume screening in the context of age. Implications for research and practice are discussed.
... The authors concluded that although a college education was associated with being gainfully employed for people with disabilities in their 20s and 30s, it did not prevent major unemployment rates by the time they were in their 40s. However, studies like that of Edin and Lagerström (2006) showed that middle-aged individuals had the highest likelihood to be gainfully employed regardless of whether they had a disability or not. In the study, employment opportunities' odd was lowest among those aged 40 and above. ...
Article
Full-text available
The present study investigated the influence of disability types and socio-demographic status of people living with disabilities on employment opportunities. The study also examined how the career aspirations of people living with disabilities could impact the relationship between the variables in our first objectives. A descriptive survey was used to elicit data from 600 people living with disabilities in six Nigerian states. A logistic regression analysis revealed that having the communicative-vision type of disability, having no educational qualification, being a woman, not belonging to the privileged ethnic group, being middle-aged and having only numeracy skill without oral communication and literacy skills all affected negatively the chances of employment among people living with disabilities. Except for ethnic groups, career aspirations mediated the influence of disability types, gender, age, education and basic skills on employment opportunities. The implications of the findings were thereafter discussed.
... A possible reason for limited impacts of lower hiring costs is discrimination; i.e. employers are for some reason not indifferent between job seekers of varying characteristics at a given wage. Empirically, both trials using anonymous or internet based job applications (Edin & Lagerström, 2006;Eriksson & Lagerström, 2012;Åslund & Skans, 2012), so called correspondence studies Bursell 2014) and a stated preference study suggest that immigrants are discriminated against in the Swedish labour market. 50 Discrimination is a complex and much debated concept, and we will not get into the theoretical details here (ranging from classical taste-based models to broader structural perspectives). ...
... Instead, our results suggest that policies should focus at least as much on breaking down prejudice among natives. Additionally, governments should consider measures that level the playing field, such as anonymous job applications [17][18][19]. ...
Article
Full-text available
How pervasive is labor market discrimination against immigrants and what options do policymakers and migrants have to reduce it? To answer these questions, we conducted a field experiment on employer discrimination in Sweden. Going beyond existing work, we test for a large range of applicant characteristics using a factorial design. We examine whether migrants can affect their employment chances-by adopting citizenship, acquiring work experience, or signaling religious practice-or whether fixed traits such as country of birth or gender are more consequential. We find little systematic evidence that immigrants can do much to reduce discrimination. Rather, ethnic hierarchies are critical: callback rates decline precipitously with the degree of ethno-cultural distance, leaving Iraqis and Somalis, especially if they are male, with much reduced employment chances. These findings highlight that immigrants have few tools at their disposal to escape ethnic penalties and that efforts to reduce discrimination must address employer prejudice.
... At the interview stage, studies have found CV deidentification benefits both women and ethnically diverse candidates. However, at the job offer stage, this benefit persists largely for women only, and less so for those from minority ethnic backgrounds (Bøg & Kranendonk, 2011;Edin & Lagerström, 2006). ...
Technical Report
Full-text available
Recruit Smarter trialled four pilot interventions across a range of government departments and private sector organisations. The four interventions included two trials of targeted recruitment via modified language use in job advertisements, a CV de-identification program, and the provision of training to address unconscious bias. The pilot program has found evidence that these interventions are beneficial to improving equity of opportunity for diverse Victorian applicants.
... A possible reason for limited impacts of lower hiring costs is discrimination; i.e. employers are for some reason not indifferent between job seekers of varying characteristics at a given wage. Empirically, both trials using anonymous or internet based job applications (Edin & Lagerström, 2006;Eriksson & Lagerström, 2012;Åslund & Skans, 2012), so called correspondence studies Bursell 2014) and a stated preference study suggest that immigrants are discriminated against in the Swedish labour market. 50 Discrimination is a complex and much debated concept, and we will not get into the theoretical details here (ranging from classical taste-based models to broader structural perspectives). ...
Article
Full-text available
In 2002, the Danish government reduced the size of cash transfers to new refugees. We exploit the reform to study the effect of lower transfers on the educational outcomes of refugee children. First and foremost, the reduction in parental benefits has no negative effect on educational outcomes of the children, such as test scores, probability of completing the 9th grade or probability of enrollment in upper-secondary education. Likewise, the children of refugees affected by the reform do not work more in their youth to compensate for lower family income.
... Women choose to a higher extent than men to continue their studies at colleges and universities. Earlier research shows that, despite that, men establish themselves more quickly on the labor market after completing their studies, and that women are weeded out by employers to a greater extent when having equal merits [13]. A plausible interpretation of the results is that there seems to be gender related discriminating mechanisms that results in that for women with low education the possibility to gain a job is lower than for men with the same educational level. ...
Article
Full-text available
More knowledge is needed ofoccupational attainment of persons with disabilities, i.e. the relationship between their educational level and their profession, and factors of importance for this relationship.OBJECTIVE: To compare occupational attainment among persons with and without a disability. 3 396 informants with disabilities and 19 004 non-disabled informants participated (control group) in a survey study by Statistics Sweden.The informants with disabilities were divided into six groups. Occupational attainment did not differ between the disability groups, neither between persons with and without a disability. Follow-up analysis showed that men with disabilities with primary or secondary school had an occupation above their educational level to a significantlylarger extent than women with disabilities. This pattern was even clearer in comparison with the control group. Persons without disabilities, with secondary or higher education, were more successful in the labor market than persons with disabilities. Occupational attainment increased with age in both groups. Young women with disabilities who only have primary or secondary education run a higher risk of having a job that is below their educational level than men at the same educational level. This indicates discriminating mechanisms in the society related to gender and ability.
... Åslund and Nordström Skans (2012) study a trial with anonymous applications in Sweden, and find that woman and individuals of non-Western origin have a higher probability of advancing to the interview stage of the hiring process. Edin & Lagerström (2006) analyze a CV database where searchers are allowed to hide their names, and find that a woman has a 15 percent lower callback rate if she reveals her name. ...
Article
Full-text available
We study the recruitment behavior of Swedish employers using data from a stated choice experiment. In the experiment, the employers are first asked to describe an employee who recently and voluntarily left the firm, and then to choose between two hypothetical applicants to invite to a job interview or to hire as a replacement for their previous employee. The two applicants differ with respect to characteristics such as gender, age, education, experience, ethnicity, religious beliefs, family situation, weight, and health. Our results show that employers discriminate against applicants who are old, non-European, Muslim, Jewish, obese, have several children, or have a history of sickness absence. Moreover, increasing the firms’ cost of uncertainty in hiring – through more firm co-payment in the sickness benefit system – may reduce hiring, but does not affect the degree of discrimination. Also, there are only small differences in the degree of discrimination between different types of recruiters and firms. Overall, our results suggest that the discrimination, at least partially, should reflect statistical discrimination.
Article
The employment status of groups with different disabilities was analysed as were potentially important moderating factors (work ability, structural and individual factors). A secondary analysis was performed on 4359 respondents with disabilities from Statistics Sweden's Labour Market Investigation. The respondents were divided into six disability groups (communicative-hearing, communicative-speech-reading, communicative-vision, psychological disability, medical disability, physical disability). Logistic regression analyses showed that the probability of being employed was highest among respondents with hearing disabilities and respondents with psychological disabilities were least likely to be employed. Being a woman (very young or old) with only primary education and with partially or very impaired work ability, reduced employment opportunities. Higher education did not increase employment opportunities for respondents with impaired work ability. In summary, the type of disability is essential for employment opportunities, and differences between disability groups cannot be explained by differences in other variables. The moderating factors studied were found to be of equal importance in all groups.
Article
Discrimination data analysis has been investigated for the last fifty years in a large body of social, legal, and economic studies. Recently, discrimination discov-ery and prevention has become a blooming research topic in the knowledge discov-ery community. This chapter provides a multi-disciplinary annotated bibliography of the literature on discrimination data analysis, with the intended objective to pro-vide a common basis to researchers from a multi-disciplinary perspective. We cover legal, sociological, economic and computer science references.
Article
Full-text available
This paper presents LINDA - a register-based longitudinal data set for Sweden. LINDA consists of a large panel of individuals, and their household members, which is representative for the population during the period 1960 to 1998. As future years become available, this information will be added to the data set. LINDA also includes a specific sample of immigrants. This sample has the same design and covers the same time period as the population sample. We provide a description of the sources of data, the sampling frame as well as the sampling procedure. Moreover, to illustrate the usefulness and particular features of LINDA, we give the development of some of the key variables in the data set.
Article
This second edition of Gary S. Becker's The Economics of Discrimination has been expanded to include three further discussions of the problem and an entirely new introduction which considers the contributions made by others in recent years and some of the more important problems remaining. Mr. Becker's work confronts the economic effects of discrimination in the market place because of race, religion, sex, color, social class, personality, or other non-pecuniary considerations. He demonstrates that discrimination in the market place by any group reduces their own real incomes as well as those of the minority. The original edition of The Economics of Discrimination was warmly received by economists, sociologists, and psychologists alike for focusing the discerning eye of economic analysis upon a vital social problem—discrimination in the market place. "This is an unusual book; not only is it filled with ingenious theorizing but the implications of the theory are boldly confronted with facts. . . . The intimate relation of the theory and observation has resulted in a book of great vitality on a subject whose interest and importance are obvious."—M.W. Reder, American Economic Review "The author's solution to the problem of measuring the motive behind actual discrimination is something of a tour de force. . . . Sociologists in the field of race relations will wish to read this book."—Karl Schuessler, American Sociological Review
Article
We study race in the labor market by sending fictitious resumes to help-wanted ads in Boston and Chicago newspapers. To manipulate perceived race, resumes are randomly assigned African-American- or White-sounding names. White names receive 50 percent more callbacks for interviews. Callbacks are also more responsive to resume quality for White names than for African-American ones. The racial gap is uniform across occupation, industry, and employer size. We also find little evidence that employers are inferring social class from the names. Differential treatment by race still appears to still be prominent in the U. S. labor market.
Article
A change in the audition procedures of symphony orchestras--adoption of "blind" auditions with a "screen" to conceal the candidate's identity from the jury--provides a test for sex-biased hiring. Using data from actual auditions, in an individual fixed-effects framework, we find that the screen increases the probability a woman will be advanced and hired. Although some of our estimates have large standard errors and there is one persistent effect in the opposite direction, the weight of the evidence suggests that the blind audition procedure fostered impartiality in hiring and increased the proportion women in symphony orchestras.
Quasi-Experimental Evidence on Discrimination 17 rHeckmanDetecting Discrimination
  • Ifau Blind
IFAU – Blind Dates: Quasi-Experimental Evidence on Discrimination 17 rHeckman J (1998), “Detecting Discrimination”, Journal of Economic Perspectives 12, 101 – 116
Orchestrating Impartiality: The Impact of “Blind” Auditions on Female MusiciansKön, lön och yrke – yrkessegregering och lönediskrimine-ring mellan män och kvinnor”, i I Persson and E Wadensjö (eds) Kvinnors och mäns löner – varför så olika?
  • Goldin
Goldin, C and C Rouse (2000), “Orchestrating Impartiality: The Impact of “Blind” Auditions on Female Musicians”, American Economic Review, Vol. 90, No. 4, 715-741. le Grand, C (1997), “Kön, lön och yrke – yrkessegregering och lönediskrimine-ring mellan män och kvinnor”, i I Persson and E Wadensjö (eds) Kvinnors och mäns löner – varför så olika?, SOU 1997:136
Kön, lön och yrke – yrkessegregering och lönediskriminering mellan män och kvinnor " , i I Persson and E Wadensjö (eds) Kvinnors och mäns löner – varför så olika?
  • C Le Grand
le Grand, C (1997), " Kön, lön och yrke – yrkessegregering och lönediskriminering mellan män och kvinnor ", i I Persson and E Wadensjö (eds) Kvinnors och mäns löner – varför så olika?, SOU 1997:136.