Content uploaded by Peter A Riach
Author content
All content in this area was uploaded by Peter A Riach on Oct 16, 2014
Content may be subject to copyright.
Advances in Economic Analysis &
Policy
Volume 6, Issue 2 2006 Article 1
FIELD EXPERIMENTS
An Experimental Investigation of Sexual
Discrimination in Hiring in the English Labor
Market
Peter A. Riach
∗
Judith Rich
†
∗
Bloomsbury, London, Peterriach@aol.com
†
University of Portsmouth, Judy.Rich@port.ac.uk
Copyright
c
2006 The Berkeley Electronic Press. All rights reserved.
An Experimental Investigation of Sexual
Discrimination in Hiring in the English Labor
Market
∗
Peter A. Riach and Judith Rich
Abstract
Pairs of carefully-matched, written applications were made to advertisedjob vacanciesin Eng-
land to test for sexual discrimination in hiring. Two standard r
´
esum
´
es were constructed for each
occupation to control for all relevant supply-side variables, such as qualifications, experience and
age. Consequently any differential response recorded can be attributed to demand-side discrimi-
nation. Statistically significant discrimination against men was found in the ‘female occupation’ -
secretary, and against women in the ‘male occupation’ - engineer. Statistically significant, and un-
precedented, discrimination against men was found in two ‘mixed occupations’ - trainee chartered
accountant and computer analyst programmer.
KEYWORDS: discrimination, employment, field experiment, hiring, sex
∗
This project was funded by a grant from the Faculty of Business and Economics, Monash Uni-
versity. We wish to thank Anne Kennedy, Hayden Mathysen and Natalie-Chantal McCaughey for
their excellent research assistance. We wish to acknowledge the helpful comments of John List
and two anonymous referees on an earlier version of the paper. Dr. Peter A. Riach, Bloomsbury,
London WC1, United Kingdom. Address for correspondence: Dr. Judith Rich, Department of
Economics, University of Portsmouth, Portsmouth PO1 3DE, United Kingdom.
The first field experiment of discrimination in employment, involving written job
applications, was undertaken by the British sociologists Jowell and Prescott-
Clarke (1970). They developed the technique to investigate racial discrimination
in employment in England. The first application of the technique by economists
was by Riach and Rich, who investigated sexual discrimination (Riach and Rich
1987) and racial discrimination (Riach and Rich 1991), in employment in
Melbourne in the 1980s. More recently the technique has been applied by
Neumark, Bank and Van Nort (1996) to investigate sexual discrimination in
Philadelphia, and by Bertrand and Mullainathan (2004) to investigate racial
discrimination in Boston and Chicago. What follows is the first systematic field
experiment of sexual discrimination in the English labor market.
1. The Technique
Field experiments of labor market discrimination involve carefully-matched pairs
(male and female in this case) testing by telephone, postal or electronic mail
applications, and by attendance at interview. The methodological imperative of
the technique is that strict control is exercised over all relevant employment
characteristics, such as age, qualifications and experience, so that applicants differ
only in their sex. In those experiments where pairs of applicants attend in person
at interview there is a problem of ensuring, and demonstrating, such strict control.
Despite careful training of the participating pairs it is impossible to ensure that all
aspects of the applicants’ performance are identical during their interaction with
those performing the interview. In particular it is possible that, consciously or
unconsciously, minority applicants may be motivated to prove the existence of
discrimination, and thereby bias the results. A detailed discussion of this problem
is contained in Riach and Rich (2002). To avoid this uncertainty we chose to
make written applications to advertised vacancies. The precise nature of our
technique was as follows. Two standard résumés were sent in response to job
advertisements. In order to avoid detection they obviously could not be identical,
but in all essential job characteristics, such as age, qualifications and experience,
candidates were carefully matched, so that the only effective distinguishing
characteristic was sex. Sex was identified by name, and both male and female
were given traditional Anglo-Saxon/Celtic first names and surnames to avoid any
possibility that the test could be confounded by racial discrimination. Unlike in
the case of African Americans, sex can be unambiguously identified by the
candidate’s first name, provided, of course, that sexually-shared first names, such
as Lindsay, are avoided. It follows that only one pair of names is required per
occupation. For instance, in engineering our female name was Emma and our
male name was Philip. To control for the possibility that the style and contents,
such as hobbies and interests, of a particular letter might influence employer
Advances in Economic Analysis & Policy
1Riach and Rich: Testing for Sex Discrimination in the English Labor Market
response, letter-type was alternated regularly, and allocated equally between the
sexes; that is, if one hundred applications were made, then on fifty occasions the
female would be allocated letter A and the male letter B. On the other fifty
occasions this would be reversed and the male would get letter A. If both female
and male were invited to job interview this was treated as a case of no
discrimination, or ‘equal treatment’. If only one applicant was invited to
interview this was treated as a case of discrimination. Where neither applicant
was invited to interview this was treated as a non-observation, as it tells us
nothing about an employer’s use of sex as a screening device. Instead, it may
simply indicate that a short-list was determined before our applications arrived, or
that several other superior applications were received. In which case, the
employer’s penchant for discriminating on the basis of sex would have been put
to the test only if our applications had arrived earlier or contained qualifications
and experience of a different nature (McIntosh and Smith 1974, p. 24; Riach and
Rich 2002, pp. F487-8). Previous research has indicated that some employers
arbitrarily eliminate applications which do not include a contact phone number.
The advantages of written applications are that it is possible to exercise
precise control over the content of applications; to ensure that all relevant
characteristics other than sex are carefully matched; to ensure, by reversal of
letter-type, that no unintended bias is introduced by any stylistic difference which
is present to minimize the possibility of detection; and to demonstrate the
controlled and objective nature of this procedure to the reader. The disadvantage
is that it only tests for discrimination in labor-hiring at the initial stage of selection
for interview. It is possible that some employers may delay their discriminatory
activity until the interview, and in the final choice from amongst the interviewed
short-list. Thus written tests do not measure the full extent of discrimination in the
hiring decision, but, on the other hand, they do highlight one quite decisive form
of discrimination – that of denying the applicant the chance even to compete for a
job. In some cases field experiments of discrimination have involved a two-stage
investigation: applying for interview and then attending, if invited. In such studies
the bulk of discrimination has always been detected at the first (invitation to
interview) stage: for instance, in the International Labour Office (ILO)
investigations almost ninety per cent of the racial discrimination in employment
detected in Belgium, the Netherlands and Spain was at this first stage (Riach and
Rich 2002, pp. F493-F494).
The four occupations chosen for the study were computer analyst
programmer, engineer (electrical and mechanical), secretary and trainee chartered
accountant. Computer analyst programmer and chartered accountant are
occupations which have a majority of male workers, but, nevertheless have a
significant representation of women, as shown in Table1. We shall refer to them
as ‘mixed occupations’. The inclusion of these two occupations enables a useful
2 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
comparison with our Australian study (Riach and Rich 1987). Secretary is a
‘female-dominated’ occupation, whilst engineer is ‘male-dominated’: see Table1.
This range of occupations enables us to test the demand-side impact on the full
range of labor market outcomes. Neumark et. al. report that “…one restaurant
owner explained the lack of waitresses in his upscale restaurant as ‘a question of
us seeing an endless number of male applicants and few female applicants’”
(Neumark et. al. 1996, p. 917, footnote 5).
Table 1: Employee Numbers and Female Share of Occupations, UK, 2003
Occupation
Female
Employees
(number)
Male
Employees
(number)
Female
Share
( %)
Chartered and
Certified
Accountants
44,000
100,000
30.6
Computer Analysts
and Programmers
a
25,500 97,170 20.8
Engineering
Professionals
21,000 423,000 4.7
Secretarial and
Related
Occupations
911,000 25,000 97.3
Source: Office of National Statistics 2003. Labour Force Survey, Spring (March-May).
Office of National Statistics 1991. Economic Activity, Volume 1 of 2, Table 6.
Our bogus applicants were located in central London, therefore, in the
case of secretary, we confined our search to jobs within easy commuting distance
of central London, and took vacancies from the Saturday and Monday editions of
the Guardian newspaper and the Monday edition of the Evening Standard. In the
case of the three professional occupations we considered that it was more likely
that applicants would be assumed to be prepared to relocate, so we extended our
search geographically (see Table 2).
Analyst programmer vacancies were taken from Computing and
Computing Weekly and from an internet job-site. Engineering vacancies were
taken from an internet job-site. In the case of trainee chartered accountant
applications were made to all authorized training firms in England. Applications
were sent one day apart to minimize any risk of detection. Two standard résumés
were devised for each occupation and in these the marital status of the two
applicants was identical and age differed by one year. The résumés were prepared
Advances in Economic Analysis & Policy
3Riach and Rich: Testing for Sex Discrimination in the English Labor Market
with the assistance of senior academic colleagues in accounting, computing and
engineering. The résumés for secretary were prepared with the assistance of a
senior member of the human resource management department. British university
degrees are classified into four classes: first, upper second or two-one, lower
second or two-two, and third. Only a small percentage (8.9%) of elite students
obtain a first, so two-one is the acknowledged good quality degree obtained by
44.8 percent of students (Higher Education Statistics Agency 2000/1). In their
résumés all our graduates had two-one degrees, so they were at least equal to the
average calibre of applicants in accounting, computing and engineering. Two
samples of résumés are included in the Appendix. We were unable to apply for
public sector job vacancies, as nowadays they invariably demand the completion
of an application form, rather than the submission of a résumé. It follows that all
our observations are for the private sector.
We could not use the names of genuine educational institutions or
employers in the résumés for two reasons. First there is the risk of detection if an
employer were to make direct contact with a genuine company or university
falsely cited in a résumé. Secondly an educational institution might take legal
redress against a party falsely claiming to possess one of its awards. We decided
to deal with this difficulty by inventing bogus universities and employers. Just
such an approach had been adopted in the ILO’s investigation of racial
discrimination in the German labor market: bogus schools and universities were
invented. There are approximately eighty universities in Britain and all but a
handful have locational names; either of a city or a county. We therefore chose an
English city and county which did not have universities, but which quite plausibly
might, and have used them in the résumés of all occupations where a university
degree would be expected. There was a flood of new universities in the decade
prior to these tests. In 1992 twenty polytechnics became universities with names
like De Montfort, South Bank, Liverpool John Moores and London Guildhall.
Since then there has been a steady trickle of additions with Gloucestershire, Luton
and Thames Valley amongst those acquiring universities. If counties such as
Hertfordshire and Staffordshire have universities why might not Herefordshire
and Shropshire? If towns like Loughborough, Bournemouth and Brighton have
universities is it not conceivable that Ipswich, Salisbury and Winchester have
universities? We therefore believe that employers recruiting graduates in
computing, engineering and accounting/business studies, which traditionally were
strong in the former polytechnics, would be unlikely to have a definitive
knowledge of the current list of universities. We did obtain responses from very
large accounting firms and major recruitment agencies. The British system of
external-examining is intended to ensure that degrees are of comparable standard
across the university sector, so there should be no concern about the caliber of the
graduates from our bogus universities. In the case of former employers’ names we
4 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
have claimed always to have worked for small manufacturers or professional
firms and fabricated a name. The main precaution here was to check the Yellow
Pages to ensure we had not inadvertently used a real name.
Cultural and sporting interests were chosen to be interchangeable because
of the need to reverse résumés. They were also deliberately chosen to suggest that
neither applicant had stereo-typical sexual traits; the interchangeable interests
included classical and contemporary dance, marathon running, greyhound racing
and flute and language classes.
The two postal addresses which we used were approximately one mile
apart in comparable socio-economic districts of central London. We cited a
telephone number answered by a machine on one application and a fax number on
the other. We have always been very careful to retain documentation of our
research, so our procedure with the answering machine was to take details for
filing, but also to transfer the response to a master tape. The most fruitful source
of vacancies for engineers and analyst programmers were job sites on the internet.
Therefore we located vacancies by specified qualifications etc. and then submitted
résumés electronically to notified vacancies. Once again we were careful to print
off applications and replies, so hard copies could be filed.
When an invitation to job interview was received, a telephone call was
made early the following day declining the invitation, so as to minimize any
inconvenience to the employer. This deceptive experimental technique imposes
on the employer, briefly, a non-genuine transactor in a manner which is not
infrequent in the labor market, as participants carry out the process of search and
the acquiring of bargaining chips for negotiations with current and prospective
employers. The justification for this minor act of deception is that it is the only
effective way of discovering how employers actually do behave in practice, as
distinct from how they might claim to behave when questioned about their
employment practices. The distinction between actual and claimed behavior was
dramatically demonstrated over fifty years ago by La Piere (1934) in a classic
study. In an extensive trip through the USA with a Chinese couple, admittance
was gained to all except one of two hundred and fifty-one hotels and restaurants
approached, whereas, in response to questionnaires sent six months later to the
same establishments, over ninety per cent replied that they would not accept
Chinese guests. Employers’ replies have been kept strictly confidential to the
authors of this paper and the research assistants. For a more detailed discussion of
the ethical issues involved see Riach and Rich (2004).
The intention of this experimental method is to control strictly for all
productivity-determining variables, so that it is sex alone which distinguishes the
candidates and therefore accounts for any differential treatment. To allay any
possible concerns that personality differences between the sexes may lead to
perceived productivity differences, our résumé reversal procedure produces
Advances in Economic Analysis & Policy
5Riach and Rich: Testing for Sex Discrimination in the English Labor Market
women who are interested in running marathons and racing greyhounds, and men
interested in classical dance and learning languages; that is, our applicants are not
sexually typecast.
2. The Results
The results of the experiment are set out in Table 2 in a format which follows
McIntosh and Smith (1974, p.13) and which has since been adopted in field
experiments across Europe; e.g. Brown and Gay (1985); Bovenkerk (1992, pp.26,
31) (see Riach and Rich 2002, pp. F486-F491). Column 4 shows the number of
occasions when one or both applicants received a favorable response; by post,
telephone, fax or email. This total is divided as follows: column 5 shows
occasions when both received favorable responses (equal treatment); column 6
shows occasions when only the male received a favorable response
(discrimination against the female); and column 7 shows occasions when only the
female received a favorable response (discrimination against the male). Column 8
is net discrimination; that is 7 minus 6, so that it is negative when men
encountered more discrimination than women. Column 9 provides details of
responses by letter-type. The statistical significance of any finding of net
discrimination was determined by the application of the chi-square test. The data
were categorized as accepted /rejected for two applicants in a 2*2 contingency
table (Riach and Rich 2002, pp. F493 – F496).
The purpose of this paper is to use a carefully-controlled field experiment
to test for employment discrimination. The logical imperative of the experimental
technique is to design the experiment (in particular, the résumés) so as to control
for all factors, other than sex, which may influence selection for interview. The
résumés are identical in all aspects of human capital. Also, they are regularly
reversed, in case any unintended difference has been introduced by the need to
differentiate them, in order to avoid detection. Consequently we can conclude that
any systematic preference for one of the two candidates is attributable to the one
characteristic which is being controlled for. The data cannot be controlled for
more formally, such as, by regression analysis, after the experiment has been
conducted. If the experiment has been correctly-designed there can be no impact
from any, but the target, variable. On the other hand, if the style of one résumé
has influenced outcomes this should be detected very early in the running of the
experiment, and subsequently corrected. We tested the impact from letter type on
the final data, again using chi-square, and report the results in Table 2, note 2. Our
finding that no statistical significance attaches to letter type indicates that we had
successfully controlled for other factors, such as socio-economic background, in
the résumés. The scientific challenge in field experiments is careful ex ante
design; not ex post statistical manipulation.
6 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
Table 2: Results for the Sex Discrimination Tests
1
Occupation
2
Location
of test
3
Neither
invited
4
Usable
tests
5
Equal
treatment
6
Discrim-
ination
against
female
7
Discrim-
ination
against
male
8
Net
Discrim-
ination
a
9
Letter
type
A B
Chartered accountant
Total (number)
Per cent
Interview/Chat
Per cent
England 284 55
100
23
100
22
40
5
21.7
11
20
4
17.4
22
40
14
60.9
11
-20.0*
10
-43.5**
40 37
13 15
Computer analyst
programmer
Total (number)
Per cent
London
and South
East
96 34
100
14
41.2
4
11.8
16
47.1
12
-35.3**
23 25
Engineer
Total (number)
Per cent
London,
South, SE,
Home
Counties
134 39
100
12
30.8
18
46.2
9
23.1
9
23.1*
24 27
Secretary
Total (number)
Per cent
London 180 51
100
13
25.5
8
15.7
30
58.8
22
-43.1***
36 28
Note 1: Chi-squared tests were conducted on the response rates and the results are indicated in column 8 - * significant at the 0.05 level; **
significant at the 0.01 level; *** significant at the 0.001 level.
Note 2: Chi-squared tests were also conducted on the outcomes for letter type and in no case was a difference significant at the 0.05 level.
a. A negative value indicates discrimination against the male applicant.
Advances in Economic Analysis & Policy
7Riach and Rich: Testing for Sex Discrimination in the English Labor Market
In the 1980s, when applications and replies were by post, a favorable
response almost inevitably involved an invitation to a face-to face interview.
Nowadays with the widespread availability of telephone-answering machines,
faxes and computers, and with the increased involvement of recruitment agencies
in the hiring process, favorable responses were of a more variable nature. In the
case of secretary, where all applications were by post, and recruitment agencies
were largely avoided, responses came by post, fax and telephone. The majority of
responses were invitations to a formal interview, but a small minority involved a
request for a telephone discussion. In the case of trainee chartered accountant,
where all applications were by post, twenty-three responses involved an invitation
to interview or a request for a telephone discussion. The thirty-two other
responses involved returning an application form to the applicant or requesting
recontact at a latter date. In the case of computer analyst programmer, where most
applications were submitted electronically only two favorable responses involved
invitation to formal interview. The remainder was requests for a telephone
discussion or for more details transmitted electronically, which represents the
email equivalent of a ‘discussion’. This outcome reflected the heavy involvement
of recruitment agencies in this occupation. In the case of engineer, where all
applications were submitted electronically, all responses were for a discussion,
either in its telephone or electronic format, and again this reflected the heavy
involvement of recruitment agencies in this labor market.
In the two ‘mixed occupations’ there was statistically significant
discrimination against men. Men encountered discrimination four times more
frequently than women in analyst programming, and three and a half times more
frequently than women in chartered accounting, when it came to the serious
business of interviewing, rather than the mechanical process of dispatching
application forms. When training firms issued invitations to interview, men were
denied interviews on 60% of those occasions, whereas women were so denied at a
rate of only 17%. Men were denied interviews at a rate of 47% in analyst
programming, compared with 12% for women. In the ‘male-dominated’
occupation of engineering, women encountered discrimination at twice the male
rate; women were rejected for interview on 46% of the occasions when invitations
were issued, compared with a male rate of 23%. In the ‘female-dominated’
occupation of secretary men encountered discrimination at almost four times the
female rate; men were rejected for interview on 59% of the occasions when
invitations were issued, compared with a female rate of 16%.
It might be speculated that the results in analyst programming and
chartered accounting represent an informal instance of affirmative action, as
employers attempt to balance the sex ratio at interview in the face of a
preponderance of male applicants, with discrimination against women delayed
until the job offer stage, however affirmative action is not currently condoned by
8 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
the British sex discrimination legislation. Moreover, as explained in Section 2
above, two-stage field experiments of discrimination have always detected that
the vast bulk of discrimination occurs at the invitation to interview stage. This is
logical behavior, as if an employer does not want to appoint a particular
candidate, for whatever reason, it is tactically astute to deny them an interview,
because this leaves the candidate devoid of evidence on which to bring a
complaint. We return to this point in Section 5.
These results are all statistically significant: in the cases of secretary,
analyst programmer and interviews for chartered accountancy traineeships the
results are significant at the one per cent level, so the null hypothesis - that there
is no relationship between sex and employer hiring response - is rejected. This
technique captures discrimination of a particularly decisive form; the denial to the
individual of the opportunity even to present herself/himself in a competitive
fashion before an interview panel; the screening out of applicants at the very
outset of the hiring process, because of some personal characteristic, such as sex.
These results amount to a curate’s egg for the British Equal Opportunity
Commission (EOC). There is some encouragement for them in the preference
shown for women in the two mixed occupations, but the discrimination recorded
is at a disturbingly high rate (four to one), which must be of some concern in a
legal environment which precludes affirmative action. Also a continuation of this
rate of discrimination could see these occupations becoming ‘female-dominated’,
which is presumably not the EOC’s objective. The results in engineering and
secretary are clearly disturbing, as they indicate a lack of success in breaking
down sexual stereotypes.
Some interesting comparisons can be made with other experimental
studies. Whereas in this English experiment we recorded discrimination in favor
of women in chartered accounting, in our Australian study we recorded a net rate
of discrimination against women in management accounting of 7%, although it
was not statistically significant at the five per cent level (Riach and Rich 1987).
The Australian State of Victoria enacted sex discrimination legislation in 1977, so
the main differences, between the two studies, are the length of time for the
legislation to have had an impact, the seniority of the job, and employment in a
public company/commercial environment, rather than a professional office.
Without setting up simultaneous and identical experiments it is not possible to
distinguish between these possibilities, in explaining the difference between the
Australian and English results.
The results for computer analyst programmer are also contrary to research
abroad. We included this occupation in our Australian study and recorded net
discrimination against women of 12%, which was significant at the two per cent
level (Riach and Rich 1987): in marked contrast to the finding of net
discrimination against men of 35% in England. These experiments are separated
Advances in Economic Analysis & Policy
9Riach and Rich: Testing for Sex Discrimination in the English Labor Market
by fifteen years and twelve thousand miles, but this does indicate some dramatic
English success in female access to a relatively new area of professional
employment.
The secretarial results are consistent with Levinson (1975), who made
telephone inquiries to male and female-dominated jobs in Atlanta, USA. Fifty-
seven secretarial vacancies were amongst the one hundred and ten ‘female jobs’
to which he applied. He found ‘clear-cut’ discrimination against men in these jobs
of 44%, compared with our result of 30% for London secretarial vacancies.
Clearly, sex stereo-typing is enduring in this supportive activity, and is not
confined to England.
Graduate engineers have not been the subject of any other experiment, but
Nunes and Seligman (2000) tested automobile service jobs in San Francisco and
found discrimination against women which was significant at the one per cent
level. Levinson included thirty-seven auto mechanic and skilled serviceman
vacancies amongst the one hundred and forty-six ‘male jobs’ to which he applied.
He found ‘clear-cut’ discrimination against women of 28% in these ‘male jobs’.
These results in ‘male-dominated’ jobs are consistent with our finding of
statistically significant discrimination against women in applications to English
engineering vacancies.
Levinson noted a higher rate of discrimination against men in ‘female
jobs’, than against women in ‘male jobs’ (44% compared to 28%). He attributed
this to sex being a more salient characteristic of ‘female’ than ‘male’ jobs; for
example secretaries being expected to fill a decorative role. Also he suggests that
greater deviance attaches to male applicants for ‘female jobs’, than to female
applicants for ‘male jobs’; the implication is that such a man is an undesirable
underachiever, because he is seeking lower status ‘female employment’ (Levinson
1975, p. 540). Our English results show a similar pattern; a 43% net
discrimination rate against male applicants for secretarial vacancies, compared to
a 23% net discrimination rate against female applicants for engineering vacancies.
Certainly the most striking result of our English experiment is the high net
rate of discrimination against men for analyst programmer vacancies and in
interview offers for chartered accountancy traineeships. This suggests substantial
progress in opening up professional employment opportunities to women. On the
other hand, the secretarial results indicate that some sexually stereo-typed images
remain entrenched. This is reflected in the British media. We have long been
accustomed to seeing female police inspectors (Jane Tennyson) and female
forensic pathologists (Sam Ryan) on our television screens, but male secretaries
are nowhere to be seen.
10 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
3. Interpretation of the Findings
3.1 Personality or Discrimination?
Neumark et. al. and Weichselbaumer (2004) consider the possibility that
personality traits, rather than discrimination, might be the explanation for
differential treatment. We explained above that we have attempted to create
applicants who are not sexually typecast, by our choice of sporting and cultural
interests, and by our reversal of résumés. Therefore our written applications have
ensured a degree of control over this personality variable. Moreover, we are not
convinced that any differential treatment, which did arise from personality traits,
would represent a productivity-induced response, rather than discrimination.
Neumark et. al. concern themselves with - “…personality differences that differ
systematically by sex, and that are valued differently by restaurant owners…” and
go on to recommend that researchers - “…attempt to eliminate sex-related
personality differences that might influence employers’ decisions” (Neumark et.
al. 1996, p. 922, emphasis added). This begs the question of whether personality
does actually impact on the output of goods or services, or, instead, generates
discriminatory responses. The male personality might be valued differently by
restaurant owners because it appeals to customers’ prejudices about the style of
service they should experience in an expensive restaurant, in which case the
employer is an agent transmitting customer discrimination. Chiplin and Sloane
(1976) argued long ago that personality traits may be valued, although not
necessarily productivity-related. “Thus if past job holders were male, it may be
that they possessed attributes which were, in fact, unrelated to performance, but
noted by the employer on his job specification. A personnel manager given the
task of replacing a middle-aged white man may tend to favour an applicant of the
same age, race and sex, although the job could be performed equally well by a
young, black woman” (Chiplin and Sloane 1976, pp. 73-4).
Women may be preferred to men in the role of secretary, which, in
England, often involves making coffee, washing dishes etc., because they are
considered to be more obliging, and controllable than men. This personality
difference might influence employers’ decisions, but they so decide because they
expect it will make their life easier, in the belief that women are more acquiescent
secretaries; not because it will increase the output of goods or services. In this
case the discrimination is the employers’ own.
3.2 Supply or Demand?
Nelson and Polacheck (1995) argue that the experimental procedure ignores the
supply-side; in particular the relative proportion of male and female applicants.
Advances in Economic Analysis & Policy
11Riach and Rich: Testing for Sex Discrimination in the English Labor Market
Obviously these precise data cannot be obtained, because of the deceptive nature
of the procedure, but it is our practice to present the occupational sex mix, as in
Table 1. This enables some, admittedly imprecise, inferences to be drawn about
the sex mix of applications.
Nelson and Polacheck argue that there may be statistical discrimination,
because a low proportion of female applicants may imply a low acceptance
probability, in which case search cost minimization will weigh against female
candidates. It is not, however, explained why a low proportion of female
applicants implies a low acceptance probability. An analogy is drawn with the
receipt of a minority of ‘out-of-state’ applicants – “…the probability that an out-
of-state applicant will accept a job offer is low… If the firm finds it costly to offer
a job that is refused, prudent employers will tend to offer jobs only to in-state
residents, rather than an out-of-stater with otherwise identical characteristics”
(Nelson and Polacheck 1995, p. 396). But the very nature of the geography
suggests that, on average, ‘out-of-staters’ will have made more applications than
‘in-state’ applicants; that is, it is their ‘out-of-state’, rather than minority, status
which produces the expectation of a low acceptance rate and, therefore, statistical
discrimination. What would Nelson and Polacheck expect if ‘out-of-staters’ were
a majority of the applicants? They would still have a lower probability of
acceptance.
Nevertheless they continue “Like the case of out-of-staters, a firm can
easily conceive the probability of an offer being accepted to be related to the
proportion of applicants of a particular identifiable group. For example, relatively
few female applicants (all else constant) can easily signal low acceptance
probabilities and hence high hiring costs associated with keeping jobs vacant too
long” (Nelson and Polacheck 1995, p.396). Why? Minority status has certainly
not disadvantaged our female accountant and analyst programmer applicants;
Table 1 shows that women are a minority of the profession; Table 2 shows their
four-to-one success rate in obtaining positive responses from employers.
3.3 Statistical or Tastes?
Experimental studies such as this one are specifically designed to provide a direct
and unequivocal measure of discrimination in labor-hiring. Nevertheless we can
investigate the pattern of results for possible inferences about the sources of
discrimination. Why do some employers deny themselves the opportunity to
consider for employment a group of applicants distinguished only by their sex?
The two principal hypotheses available to explain such discrimination are
Becker’s (1971) ‘taste’theory and Phelp’s (1972) statistical’ theory.
The statistical theory stresses the incomplete information about the
productivity of individual workers, which induces employers to resort to
12 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
generalizations about the labor force characteristics of groups, as a screening
device to minimize the cost of information acquisition in the hiring process. In the
case of sex discrimination, in occupations with substantial fixed costs of
employment, this ‘statistical’ response is assumed to be particularly directed at the
maternal role of women; their job tenure under suspicion because of possible
pregnancy and their reliability suspect because of child-care responsibilities. We
found discrimination against women in engineering, but in their favour in
chartered accountancy and analyst programming. This pattern cannot be
reconciled with the use of crude screening procedures across the professional
labor market.
In these three professional occupations it is difficult to attribute the
discrimination which we found to the tastes of customers, as there is not a lot of
direct interaction between employees and customers in these professions. The
discrimination is more likely to have been employer driven. Could it be that, in
the professional occupations where women have established a presence,
employers have had the opportunity to recognise that they are more congenial,
industrious, reliable employees, and have consequently developed a ‘taste’ for
them as employees? (Chiplin and Sloane 1976, pp. 73-4). If so, it suggests a case
for temporary affirmative action, so as to ensure that employers in other areas
‘learn by experiencing’ that women do have desirable employment characteristics.
Now that several empirical studies have provided an empirical foundation
which indicates that discrimination is prevalent in the marketplace, it is important
to explore the origins of this discrimination. Findings from field studies appear to
be more consistent with the hypothesis that the majority white populations having
a general “distaste” for minorities in the sense of Becker (1971), or a general
“social custom” of discrimination in line with Akerlof (1980), although statistical
discrimination as described by Arrow (1973) and Phelps (1972), cannot be ruled
out, ex ante or ex post (Riach and Rich, 2002, pp. F499-F503). Some
experimental studies have recently attempted to explore the nature of
discrimination. For example, List (2004) provides a framework for parsing the
two forms of discrimination, by using a series of field experiments in an actual
marketplace. In addition, recent laboratory experiments have also provided
insights into discrimination. For example Fershtman and Gneezy (2001) found
that ethnic discrimination was due to (mistaken) stereotyping, rather than to a
taste for discrimination.
Important insights into labor market discrimination can also be gained
from ‘natural experiments’. Goldin and Rouse used naturally-occurring data,
rather than data generated by experiments, to test for sex discrimination in the
hiring of musicians by orchestras. They used the hiring outcomes when orchestras
did, and did not, use a screen to hide the musician auditioning, together with other
relevant information, to estimate a model explaining the probability of an
Advances in Economic Analysis & Policy
13Riach and Rich: Testing for Sex Discrimination in the English Labor Market
individual advancing in an audition (Goldin and Rouse 2000, p. 726). They found
persuasive evidence that a screen increased the probability that a female would be
hired by an orchestra.
4. Conclusion
Our results in the sex-stereotyped occupations: engineer and secretary are
consistent with the findings of experimental studies which have been conducted in
other countries. This reflects a lack of progress in undermining sexually-typecast
attitudes on the part of employers and customers. In particular, the supportive
(and decorative?) role of secretary is widely considered as one which needs to be
filled by a woman. On the other hand, our findings in the ‘mixed occupations’ of
chartered accountant and analyst programmer are without precedent (Riach and
Rich 2002, p. F505).
We do not believe the differential treatment recorded in our study reflects
any personality differences in the candidates which could influence productivity;
nor do we consider that it could be accounted for by supply-side variables. Instead
the nature of our controlled experiment enables us to conclude with confidence
that this differential treatment represents demand-side discrimination.
Being an “Investor in People”, which is a British accolade for good
employment practice, did not preclude discriminatory activity. For example, one
firm, which carried this insignia on its letterhead, wrote to the female applicant on
16 November “I regret to inform you that you have not been selected on this
occasion”. The same person from the same firm wrote to the male applicant on 10
December “I have pleasure in forwarding an application form…. if you would
like to apply for one of our vacancies please complete the application form and
return it”.
Sometimes the replies to applications demonstrated an alarming, but
inventive, lack of veracity as the following two examples demonstrate. In the first
example, on the 26 November the female applicant to a chartered accounting firm
was sent the following invitation - “Please could you telephone me in order to
discuss your CV in greater detail as we may possibly have a suitable vacancy”.
On the same day the same person wrote to the male applicant - “Unfortunately,
we have no vacancies for which we are able to consider you as we require staff to
have had a minimum of eighteen months’ experience”. Our candidates, of course,
had equivalent experience. In the second example, on 9 November the female
applicant was told “I am afraid that I don’t think I will be interested in your
application as I suspect that money will be a problem”. On 30 November the same
person wrote to the male applicant - “I would be grateful if you would either
telephone myself (sic) or my secretary to arrange an appointment for you to come
14 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
and visit us”. Neither candidate had indicated anything whatsoever about money
and were aged one year apart.
This dishonest concealment of discriminatory hiring activity has been
noted repeatedly in experimental studies; see Riach and Rich, 2004. Such
behaviour has important implications for the current British approach to equality
of opportunity policy, which is anti-discriminatory, complaint-based, with the
enforcement body cast in a passive role. The onus is on the rejected applicant to
activate the legislation and to demonstrate that sex was the reason for the
rejection. Experimental studies, such as ours, demonstrate the uninformative, and
sometimes dishonest, nature of rejection letters, which means that candidates may
not know, let alone be able to demonstrate, that they have encountered sexual
discrimination. This presumably explains the repeated finding, in two-stage
experiments, that the vast bulk of discrimination occurs at the “invitation to
interview” stage. The most efficacious, and safest, time to exercise any
discriminatory penchant is at the very outset, so that the rejected applicant is left
with zero evidence to seek redress.
This leads us to the following policy recommendations. The first policy
option would be to take action to enable this anti-discriminatory, complaint-
based, passive, enforcement process to operate effectively. Our recommendation
is that employers be required to provide unsuccessful applicants with a brief
résumé, listing qualifications and experience, of the successful applicant. This
would either provide prima facie evidence of discrimination, or allay the
suspicions of rejected applicants. Applicants would be required to assent to this
procedure when they submitted their application. Such a disclosure requirement
would make those responsible for the hiring decision much more cautious about
engaging in discriminatory activity, as they would realize that any applicant
treated in a discriminatory fashion would now possess prima facie evidence to
initiate a complaint to an employment tribunal.
The second policy option would be to turn the Equal Opportunities
Commission's role from a passive to an active one; that is to empower the Equal
Opportunities Commission to conduct random audits of hiring and personnel
practices. If employers were required to keep all records of job applications for a
period of twelve months, and obliged to justify decisions on short-listing for
interview and final choice of candidate, in the event of random audit, it would
reinforce the pressure for scrupulousness in the hiring decision, which derives
from the former proposal. An appropriate analogy can be drawn here with the
capital market. Public corporations have various duties with respect to reporting
to shareholders, potential shareholders and the business community at large. They
are also subject to independent financial audit, and they are usually required to
satisfy an independent commission about various aspects of their financial
activities. In effect, capitalist economies provide a range of regulations and
Advances in Economic Analysis & Policy
15Riach and Rich: Testing for Sex Discrimination in the English Labor Market
checks to protect the owners of financial capital against unscrupulous practices
and guard against the waste of this resource. Therefore it seems entirely
appropriate that similar protection be afforded the owners of human capital, and
that steps be taken to prevent it being wasted through employers using screening
devices, such as sex, for purposes unrelated to job performance. Barbara
Bergmann advocated such a policy in 1986 (Bergmann 1986, p. 158). The
American Equal Opportunity Commission now has power to initiate
investigations of discrimination, (US Equal Employment Opportunity
Commission 2005, p. 1)
The third policy option is to switch from a complaint-based procedure,
where the onus is on applicants as “claimants” to convince the employment
tribunal that they had been discriminated against, to a procedure where the onus
would be on employers to justify their employment decisions. Just such a
proposal was included in the policy of the British Labour Party (which has now
been in government since 1997) as far back as 1991 – “Instead of women having
to prove discrimination the employers will have to prove non-discrimination”
(Labour Party 1991). In November 2000 the European Parliament legislated to
shift the burden of proof from complainants to respondents in recognition of the
difficulty applicants have in obtaining evidence (ILO 2003, p. 61, footnote 45,
Council Directive 2000/78/EC).
There is room for debate on how to strengthen anti-discriminatory
legislation, but what is not in dispute is the need to strengthen it. Brown and Gay
concluded “…the heart of the problem is that employers know that cases rarely
get as far as legal action because the victim is very unlikely to be aware that he or
she has been discriminated against,” (Brown and Gay 1985, p.32). We would add
that even where an applicant suspects that he/she has been the victim of
discrimination, current labor market practices make it extremely difficult to
present a prima facie case to the courts.
Appendix
Two Résumés Used in Job Applications for Mechanical Engineering
RÉSUMÉ A
NAME P. B. (male name) /or E. M. (female name)
EMAIL: p.b.@hotmail.com /or e.m.@hotmail.com
BIRTH DATE 30-6-77
16 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
SINGLE
QUALIFICATIONS & TRAINING
GCSE passes in 8 subjects including Maths and English
Three A Levels Maths (C), Physics (B), Chemistry (C)
BEng (Mechanical) Upper Second, University of S
WORK HISTORY
Since 1999 I have been working for Mamma’s Foods which produces pre-packed
meals for supermarkets. I have gained experience in production planning and
control while at Mamma’s Foods where various dishes are prepared to suit stock
requirements for supermarket customers this includes “Just in Time techniques”.
In addition I have experience in operations management and manufacturing
technology as I had the responsibility for ensuring the production machinery runs
24 hours in three shifts and to adapt the production rates as necessary. I also have
experience in Quality Management, while at Mamma’s Foods I was a member of
a team responsible for the firm gaining ISO9003. Mamma’s Foods has been
expanding recently and this has ensured a challenging variety of work.
OUTSIDE INTERESTS
Marathon running, flute lessons and language classes.
RÉSUMÉ B
NAME: E. M. (female name) /or P. B. (male name)
EMAIL: e.m.@hotmail.com /or p.b.@hotmail.com
AGE: 25
MARITAL STATUS: Single
EDUCATION: 8 GCSE’s including English Language and Mathematics
3 A Levels: Maths B
Physics C
Biology C
Advances in Economic Analysis & Policy
17Riach and Rich: Testing for Sex Discrimination in the English Labor Market
2:1 Mechanical Engineering Degree from University of H
EMPLOYMENT:
For the last three and a half years I have been employed by Criterion Press, which
produces greeting cards, reception menus, wedding invitations, and similar
products. I have gained experience in computer aided design, where I have
designed modifications to our printing processes, doubling our output without any
loss in quality the requirements for ISO quality are paramount for Criterion Press
because our customers demand guaranteed quality. As a mechanical engineer I
have had the responsibility for engineering systems and instrumentation and
control in the factory, and have instigated novel controls throughout the printing
process. I have gained experience in inspection methods for incoming and
outgoing materials. This provides a challenging job for a mechanical engineer as
frequently there is a need to ensure large and small batch production and a quick
turnaround of material.
HOBBIES & INTERESTS:
Classical and contemporary dance and greyhound racing.
References
Akerlof, G. (1980). “The Theory Of Social Custom, Of Which Unemployment
May Be One Consequence”, Quarterly Journal of Economics 94, 749-775.
Arrow, K. (1973). “The Theory of Discrimination.” In Ashenfelter, O. and A.
Rees (eds) Discrimination in Labor Markets (Princeton: Princeton
University Press).
Becker, G. (1971). The Economics of Discrimination, 2
nd
edition (Chicago:
University of Chicago Press).
Bergmann, B. (1986). The Economic Emergence of Women (New York: Basic
Books).
18 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
Bertrand, M. and S. Mullainathan (2004). “Are Emily and Greg More Employable
Than Lakisha and Jamal? A Field Experiment on Labor Market
Discrimination”, American Economic Review 94, 991-1013.
Bovenkerk, F. (1992). Testing Discrimination in Natural Experiments: A manual
for international comparative research on discrimination on the grounds
of ‘race’ and ethnic origin (Geneva: International Labour Office).
Brown, C. and P. Gay (1985). Racial Discrimination 17 Years After the Act
(London: Policy Studies Institute).
Chiplin, B. and J. Sloane (1976). Sex Discrimination in the Labour Market
(London: Macmillan).
Fershtman, C. and U. Gneezy (2001). “Discrimination in a Segmented Society:
An Experimental Approach”, Quarterly Journal of Economics 116, 351-
377.
Goldin. C. and C. Rouse (2000). “Orchestrating Impartiality: The Impact of ‘Blind’
Auditions on Female Musicians”, American Economic Review, 90, 715-741.
Higher Education Statistics Agency (2000/1) http://www.hesa.ac.uk/holisdocs/
pubinfo/stud.htm
ILO (2003). Time for Equality at Work, Report I(B) (Geneva: International
Labour Office).
Jowell, R. and P. Prescott-Clarke (1970). “Racial Discrimination and White-
Collar Workers in Britain”, Race 11, 397-417.
Labour Party (1991). A New Ministry for Women (London: Labour Party).
La Piere, R. (1934). “Attitudes vs Actions”, Social Forces October, 230-237.
Levinson, R. (1975). “Sex Discrimination and Employment Practices: An
Experiment With Unconventional Job Inquiries”, Social Problems 22,
533-543.
List, J. (2004). “The Nature and Extent of Discrimination in the Market Place:
Evidence from the Field”, Quarterly Journal of Economics 119, 49-90.
Advances in Economic Analysis & Policy
19Riach and Rich: Testing for Sex Discrimination in the English Labor Market
McIntosh, N. and D. Smith (1974). The Extent of Racial Discrimination, Political
and Economic Planning Broadsheet no.547 (London: Political and
Economic Planning).
Nelson, P. and S. Polachek (1995). “Discerning Discrimination: Does
Interviewing Firms Make A Difference?”, Eastern Economic Journal 21,
393-398.
Neumark, D. with the assistance of R. Bank and K. Van Nort (1996). “Sex
Discrimination in Restaurant Hiring: An Audit Study”, Quarterly Journal
of Economics, 111, 915-941.
Nunes, A. and B. Seligman (2000). A Study of the Treatment of Female and Male
Applicants by San Francisco Bay Area Auto Service Shops, The Testing
Project of the Impact Fund (San Francisco: The Impact Fund).
Office of National Statistics (1991). Economic Activity, 1 of 2 (ONS, London).
Office of National Statistics (2003). Labour Force Survey, Spring (March-May)
(London: ONS).
Phelps, E. (1972). “The Statistical Theory of Racism and Sexism”, American
Economic Review 62, 659-661.
Riach, P. and J. Rich (2004). “Deceptive Field Experiments of Discrimination:
Are They Ethical?”, Kyklos 57, 457-470.
Riach, P. and J. Rich (2002). “Field Experiments of Discrimination in the Market
Place”, Economic Journal 112, F480-F518.
Riach, P. and J. Rich (1991). “Testing for Racial Discrimination in the Labour
Market”, Cambridge Journal of Economics 15, 239-256
Riach, P. and J. Rich (1987). “Testing for Sexual Discrimination in the Labour
Market”, Australian Economic Papers 26, 165 – 178.
US Equal Employment Opportunity Commission (2005). http://www.eeoc.gov/
Weichselbaumer, D. (2004). “Is it Sex or Personality? The Impact of Sex-
Stereotypes on Discrimination in Applicant Selection”, Eastern Economic
Journal 30, 159-86.
20 Vol. 6 [2006], No. 2, Article 1
http://www.bepress.com/bejeap/advances/vol6/iss2/art1
