Procedural Fairness and Disparate Impacts Testing for Discrimination Using Anonymous Entry Exams and Job

Performance Measures As Benchmarks

Dominique Meurs EconomiX, Université Paris-Ouest et INED

Patrick A. Puhani

Leibniz Universität Hannover; ERMES, Université Paris II; CreAM, University College London; SEW, University of St. Gallen; IZA, Bonn

October 2011 – Paper Sketch – not to be distributed

Abstract: To test for discrimination against women and immigrants in recruitment,

we exploit a natural experiment in the hiring process of higher management staff in the French public sector, which is based on anonymous and non-anonymous nationwide entrance exams. Subsequently, we also observe exam results and performance evaluations after the newly hired employees have passed a one-year trainee program.

The job applicants first have to pass an anonymous written exam and, if successful, an oral exam, which is not anonymous. The written test scores must by regulation not be known to the oral examiners, but they are known to us. The hiring contest exams turn out to be procedurally fair, that is, we do not find systematically worse oral exam results for women or immigrants once we hold the written test scores constant. Immigrants might even be positively discriminated.

However, at the end of the trainee program, that is the first year of employment, women outperform men, both conditionally and unconditionally on the previous entrance exam results. Hence, the entrance exams have a disparate impact on women and men. They put more emphasis on skills that men have a comparative advantage in but that are less relevant for the job. We conclude that although there is procedural fairness in the hiring process, there is functional discrimination against women through the design of the hiring contest. We find no such disadvantage for immigrants. JEL classification: J71, J45 Keywords:

Dominique Meurs Patrick A. Puhani EconomiX Leibniz Universität Hannover Université Paris-Ouest Institut für Arbeitsökonomik 200, avenue de la République Königsworther Platz 1 F-92001 Nanterre-Cedex D-30167 Hannover France Germany E-Mail: dominique.meurs@ined.fr E-Mail: puhani@aoek.uni-hannover.de Acknowledgement: This project was supported by the Agence Nationale de la Recherche (ANR) and Deutsche Forschungsgemeinschaft (DFG) under the joint project “Integration of First and Second Generation Immigrants in France and Germany”. We thank David Card, Lowell Taylor and seminar participants at the Center for Labor Economics, University of California at Berkeley, for helpful comments. All remaining errors are our own

1

“Tous les Citoyens … sont également admissibles à toutes dignités, places et employs

publics, selon leur capacité, et sans autre distinction que celle de leurs vertus et de leurs

talents” – Déclaration des droits de l’homme et du citoyen, 17891

1 Introduction Many studies analyzing gender and race discrimination in the labor market have relied on

descriptive comparisons, based, for example, on the Blinder (1973) – Oaxaca (1973) or Juhn,

Murphy and Pierce (1991) decompositions. Because it is usually unclear whether all

productivity-relevant characteristics correlated with gender and race have been controlled for

in these studies, there always remains a doubt whether the “unexplained” gap is due to

discrimination or due to unobserved productivity-relevant characteristics. Therefore, more

recently, experimental evidence, either in the form of natural experiments or audit studies,

have become popular to obtain a more direct measure of discrimination (Altonji and Blank,

1999, provide an overview of theoretical and empirical approaches on this topic, Newmark,

Bank und Van Nort, 1996, and Bertrand and Mullainathan, 2004, are examples of audit

studies, Goldin and Rouse, 2000, and Autor and Scarborough, 2008, are examples of natural

experiments that exploit changes in the recruitment process).

In this paper, we exploit a natural experiment in the hiring process of higher

management staff in the French public sector, which is based on anonymous and non-

anonymous nationwide entrance exams, to examine outcomes by gender and immigration

status. We also observe the performance of the hired employees after their first year, a trainee

program, after they have been hired. This allows us to test both for procedural fairness in the

oral exams and for functional discrimination by analyzing whether the hiring process is

testing candidates on skills required on the job.

1 “All citizens … are equally admissible to all honors, places and public employments, according to their

abilities, and without other distinctions than their virtues and their talents” – Declaration of the Rights of Man and Citizen, 1789” (translation by the authors).

2

Apart from the design of the hiring process and the subsequent performance measure,

the real-world situation we analyze is also interesting historically because it is part of a two-

century old tradition in France to hire workers based purely on merit, a principle established

during the French Revolution (see quote above). France has had such a system for virtually

all positions in the public sector as well as for entry into elite educational institutions (which

are mostly public) long before affirmative action, comparable worth or employment testing

became widespread in other parts of the western world. The public sector is the largest

employer in France, with a share of about 25 percent of total employment, compared to only

about 15 percent in the U.S. (Gregory and Borland, 1999). France also has a comparatively

low gender wage gap in the OECD, about 12 percent in France versus 19 percent in the U.S.,

respectively (OECD data base).

Several experiments related to the hiring opportunities of women or minorities exhibit

discrimination against these groups, also in industries that seem to be rather competitive

(Newmark, Bank and Van Nort, 1996, on restaurant hiring; Goldin and Rouse, 2000, on

orchestra musicians; Bertrand and Mullainathan, 2004, on service jobs advertised in

newspapers). Recent literature on employment testing, however, suggests that standardized

tests lead to fair recruitment with respect to the tested abilities, but that minorities perform

worse on these tests (Autor and Scarborough, 2008) raising the issue of efficiency versus

equity of the employment process.

In this paper, we use data on the population of applicants for entry into the higher

management levels of the French regional administration. About 90 percent of the applicants

have at least a college degree. The candidates first have to pass an anonymous written exam

and, if successful, an oral exam, which is not anonymous except that the written test scores

must by regulation not be known to the oral examiners: they only know that the written

exams were passed. We compare outcomes of anonymous written and non-anonymous oral

exams by gender and immigration status. Less than 40 percent of applicants pass the written

3

exam and less than half of the applicants passing the written test are finally hired after the

oral exam, so that both of these successive exams are highly selective. Unlike some audit

studies, our natural experiment therefore does not stop at the invitation to a job interview, but

covers the complete recruitment process.

Contrary to the oral examiners, we observe the written test scores and can evaluate the

oral examiners’ scores conditionally and unconditionally on the written scores. Although

there is a positive correlation between the written and oral test scores, this correlation is low

(with an R2 of only 3.5%) implying that oral examiners either examine different abilities of

the candidates or come to rather different conclusions than the written exam results. Despite

of the large variation in oral exam results, we do not find systematically worse oral exam

results for women or immigrants once we hold the written test scores, which are unobserved

to the oral examiners, constant. Compared to men, women perform significantly worse in the

oral exam, but this difference becomes statistically insignificant (and close to zero in the

point estimate) once the written test scores are controlled for. Hence, if the written exams can

be seen as a benchmark for judging the fairness of oral exams, the oral examiners about “get

it right” as concerns the average difference in female and male exam results and do not

discriminate against women. As concerns differences between immigrants and natives, there

is no significant difference in the unconditional performance between these two groups in the

oral exams. However, once we control for the written exam results, there seems to be some

positive discrimination in favor of immigrants, although these results are not robustly

statistically significant. It seems that this result may apply to both first- and second-

generation immigrants although the standard errors become too large to make precise

statements in these subsamples. In sum, we conclude that the oral examining process that we

analyze here (consisting of a jury of three people with diverse backgrounds) does not

discriminate against women or immigrants in the sense that it is procedurally fair. It may,

however, favor the employment of immigrants. The reason for the latter finding might be a

4

recent trend in France to stress the need for diversity in public sector employment. Note,

however, that among the pool of applicants, women and first generation immigrants still have

a lower chance of succeeding in the hiring process. This outcome is, however, caused by

worse results in the anonymous written exams for these groups and not by “discrimination”

in the non-anonymous oral exams (a result consistent with the findings of Autor and

Scarborough, 2008, for the recruitment process in a U.S. private sector retail firm). Hence,

the French hiring contest exams are procedurally fair.

This raises the question whether the entrance exams have a disparate impact on

women or immigrants in the sense that they test skills in which these groups may have a

comparative disadvantage which at the same time are not relevant for performing on the job

in question. We can test this, because successful candidates enter a one-year trainee program

and we observe their job-related test results (both anonymous and non-anonymous; written

and oral). It turns out that women outperform men both conditionally and unconditionally on

the written and oral entrance exam results. Hence, there is functional discrimination against

women in the hiring exams in that these exams place too much emphasis on skills less

relevant for the more job-related trainee program tests in which female trainees outperform

men.

The paper is structured as follows. Section 2 describes the French concours system,

which nationwide is the prevailing principle of recruitment into public sector positions and

elite public higher education institutions. We have obtained the complete data from the

examination process for entry into higher-level management careers in the French regional

administration (instituts régionaux d’administration, IRA) of the year 2008. In Section 3 we

describe the data sets that we have obtained (the administrative data from the application

process plus survey complementing the scarce information in the administrative data).

Second 4 presents the empirical results on oral exam outcomes unconditional and conditional

on the written exam scores and other characteristics that we observe due to the survey

5

information but which are not passed on to the oral examiners. Section 5 analyzed differences

between women and men and between immigrants and natives in the job-related exams

during and after the trainee period. Section 6 concludes.

2 Le Concours – Recruitment in the French Public Sector and the Natural Experiment

We have obtained data from the entry exams (concours) to the Regional Institutes of the

Administation (instituts régionaux d’administration, IRA). Acceptance to these institutes is

de facto equivalent to receiving a job in the higher management of the regional public

administration in France. To put the kind of jobs that candidates apply for into perspective,

there are about 80 persons recruited per year into the top management level trained by the

famous école nationale d’administration (ENA), which was attended by many French

presidents and other important figures in France. The IRA, which we consider here, are

institutions one hierarchical level below ENA, and recruit about 600-700 people per year. In

the data we have available, about 90 percent of the candidates have at least a Bachelor’s

degree from college, and 75 percent have at least a Master’s degree from university. Before

the successful candidates start their employment anywhere in France, they are trained at the

regional institutes in a two-year trainee program. During the program, the accepted

candidates already receive a small salary equivalent to a stipend (at around €1500 euros per

month). After the trainee program, virtually all hired candidates take up regular civil service

employment, only three out of 738 people obtained no such job due to their particularly bad

performance during the trainee program.

There is no other way to obtain the jobs the candidates apply for except through

passing the exams we analyze here and subsequently completing the trainee program. In

other words, it is impossible for a manager in the public sector to simply hire a person

without that person having completed the national entry exam (this is not only true for the

recruitment of the higher management that we consider here, but also for the hiring of a

6

secretary; public-sector secretaries, however, have to pass a different national entry exam,

tailored to their prospective jobs).

When taking their entrance exam, the candidates have to choose a place where they

want to attend the trainee program. There are five such places all over France, located in the

following cities (regions): Bastia (Corse), Lille (Nord), Lyon (Rhône-Alpes), Metz (Lorraine)

and Nantes (Pay de la Loire).2 Note that the place where the trainee program is completed

does not restrict the region of future employment. For example, a person graduating from

Bastia may take a job somewhere in Corsica (Corse), but also in Lille or in Strasbourg in the

North and East of France, respectively, or any other place in France. The distinction between

the five training centers is partly simply a matter of offering this type of training in different

parts of France and partly given by specializations on the type of administration trained for.

During the two-semester one-year training program, the first year is identical across training

centers, whereas in the second year, candidates may specialize in certain fields. The

candidates can change the training center which they apply for up to the last day of the

registration period and may see how many people have so far registered for the entrance

exam of each training center to gauge the level of competition. The entrance exams for those

five training centers take place jointly at the same time in various exam centers in the whole

of France (including the offshore départements and territories like Réunion or Guadeloupe).

The oral exams are undertaken by a jury consisting of groups of three people. In each

of the five regional training centers, there are about 20 examiners per regional training center,

who first correct the written exams anonymously. Out of these 20 people, about two-thirds or

more are chosen to build juries of three people for the oral examinations. These juries are

supposed to be equilibrated teams. For example, they must consist of at least one woman and

one man. They are also supposed to be balanced by professional status, work experience and

type of career. However, almost all jury members are civil servants at a hierarchical position

7

higher than the positions that successful candidates will take immediately after their training.

They are also supposed to be from a similar mix of backgrounds as the ministries for which

the candidates will work. Hence, half of the juries’ members are usually from the national

educational administration as also half of the candidates will work in this administration, but

the juries also contain people from the Departments of the Interior, Defense, Finance or

Transport. The jury members are trained for one day on how to conduct the interview. In

particular, they are sensitized on the issue of discrimination and they are told to create a table

to document their evaluation of the candidates’ responses during the oral exams. The names

and job positions of the jury members are published in advance on the web pages of the

regional training centers as well as in official government decrees. Examples of job positions

held by jury members are principal attaché of the administration, director of a préfecture (the

main administrative office of the central government in France’s 100 regional units called

départements), hospital director, or university professor.

There are separate entrance exams for internal candidates (who already work in the

administration but at a position at a lower level than the one they apply for) and external

candidates (those who do not work in the administration yet: they might be university

graduates or people working in the private sector).3

Only those who pass the written exams are admitted to the oral exam. Only those who

also pass the oral are offered a job at the regional administration (starting with the two-year

trainee program).

The written exams of the internal candidates consist of a dossier (counting four ninths

of the written mark) and a short questions exam (counting five ninths of the written mark). In

the four-hour dossier exam of 2008, which is specific to the internal candidates, the

examinees received a 40 page dossier of documents on policies on sustainable development

2 Because most of the candidates are supposed to work in the French regions, there is no such training center in

Paris.

8

(ten documents in total among them various decrees and government committee reports).

Based on these documents, the candidates have to pretend to be in charge at a prefecture in

the region Provence Alpes Côte d’Azur and develop a strategy for a program of policies

pertaining to sustainable environmental policies in the region. The second type of written

exam consists of a three hour exam containing ten short questions (three on Public Law, two

on European Politics, three on Public Finance and finally two on Economics.

The external candidates also take a three hour short questions exam with the same

structure as the internal candidates (here counting five twelfths of the written mark).

However, instead of a dossier, they have to write a four-hour essay on a given topic in the

field of Common Culture (counting four twelfths of the written mark). In our case, the topic

was “The Internet, A New Floor for Public Debate”. Unlike the internal candidates, the

external candidates have to take a third written exam, a “Summary Essay”, (note de synthèse)

counting three twelfths of the written mark.

The candidates who pass the written exam are not told their exact score, they are only

informed that they passed and are admitted to the oral exam. Candidates taking the oral exam

have a conversation with the jury (which we refer to as the “oral exam”).4 If a person fails the

interview with the jury, the candidate is rejected irrespective of his or her written score

(which is unknown to the jury at the time of the exam). If the oral exam is passed, a total

score is calculated to which the oral score counts five fourteenths5 for the internal and five

seventeenths for the external candidates. For each regional center, a cutoff for the total score

is then determined above which a candidate is admitted, based on the number of available

jobs.

3 There is also a third type of exam for people older than 40 years of age with at least 5 years of work

experience. Because only very few people take this exam (164 in the year 2008 for which have obtained the data) we ignore this group.

4 There is also a small (written ???) test in modern languages (counting one fifth of the oral score, the interview with the jury counting four fifths), but this is only mandatory for the external, but not for the internal candidates.

5 Or four thirteenths if the modern languages test is not taken.

9

The three-person jury interviews each candidate for about 25 minutes. During the 25-

minute interview, the first phase, which lasts about 8 to 10 minutes, consists of a discussion

of a text or a topic that the candidate has to prepare during the 30 minutes before the

interview starts (interview with JM Lenzi, page 7, Eberhard, Meurs and Simon, 2008: 89).

The rest of the interview consists of a discussion of the candidate’s motivation for applying

to this kind of job as well as previous work experience and training. Hence, during the

interview, the jury may learn something about the candidate’s background and education, but

they do not receive any documents like a CV, or any other formal information on degrees or

marks obtained. And they definitely must not know the scores of the written parts of the

concours. The jury does, however, know the candidate’s name which – apart from the

candidate’s appearance – may give further information on the candidate’s background.

In this paper, we exploit the natural experiment of non-anonymous oral and

anonymous written exams as a natural experiment to measure “discrimination” in the oral

examining process. This experiment is only valid if the groups analyzed here (women versus

men and immigrants versus natives) do not differ systemically in the skills required in the

oral and written exams. Building on this assumption, we ask whether unconditional

performance differences in the oral exam between women versus men and immigrants versus

natives deviate from the conditional performance differences, where written scores and other

productivity relevant characteristics act as control variables. If, for example, women perform

worse compared to men in the oral exam when controlling for written exam results (and other

productivity relevant characteristics), we will interpret this as evidence in favor of functional

discrimination against women.

After being hired, successful candidates enter a one-year (two-semester) trainee

program after which the candidates are ranked and move on to full-blown jobs. After each

semester of the trainee program, the new public sector employees take exams, of which the

10

first semester counts for one third and the second semester counts for two thirds of the total

score for the final ranking that determines where the new employee will work.

After the first semester, there are four tests: a written 5 hour exam based on a set of

documents on an administrative problem combining legal, organizational, financial and

communication issues [counting 5/20], a 20 minute oral exam based on an internship and

testing administrative management skills [counting 5/20], an internship mark based on the

supervisor’s judgement and the candidate’s internship report [counting 7/20], and finally the

defence of an administrative group report consisting of a 40 minute oral exam with the grade

awarded at the individual level [counting 3/20].

At the end of the second, final, semester, there are again 4 tests. First, a written 5 hour

exam consisting of solving one or several practical administrative cases [counting 5/20].

Second, a 30 minute oral exam on the field of specialization [counting 5/20]. Third, an

internship mark based on the supervisor’s judgement and the candidate’s internship report,

similar to first semester [counting 5/20]. Finally, a 20 minute oral foreign language exam

consisting of a discussion of a professional topic [counting 5/20].

In the following section, we provide descriptive statistics on the 2008 examination

process for which we have obtained both administrative and survey data.

We are currently in the process of collecting data from another examination process.

3 Data and Descriptive Statistics We have obtained data on the population of all candidates taking part in the contest

(concours) for entry into the French regional administration of 2008. The written part of

these exams took place in February 2008; for those who passed the written exams, the oral

exams took place in June 2008. The total number of participants in the external and internal

recruitment processes is 4,734, of which 3,129 are external candidates, 1,441 are internal

candidates and 164 are candidates from the so-called “third exam track”. The latter are

11

people who have at least five years’ work experience in the private sector and are older than

40 years of age; this group will be ignored here due to their small number.

Because of the tradition of anonymity of the exams, which goes back to the ideals of

the French revolution, the administrative data contain hardly any information on the

candidates: what is recorded is the exam results and the name and address of the person,

which also includes the gender. No more information is available.

Table 1 presents some summary statistics for pass rates and exam scores both for the

population of candidates and the survey participants. The survey questionnaire was

developed by Eberhard, Meurs and Simon (2008) and sent to the candidates by regular mail

before the written exam. Survey questionnaires were also given to the candidates not having

responded to the survey by mail immediately before the written exam took place. Altogether,

59 percent of participants of the written exam participated in the survey. The survey

questionnaire consisted of two-pages asking the candidates’ birth date, civil states,

immigration background, immigration background of their parents, parental occupation, and

educational attainment of the candidates. Because the administrative data contain – besides

the exam results – no socio-economic information except the gender of the candidate, we can

only compare the population and the survey sample by these variables. Table 1 suggests that

the sample means are roughly similar in the population and the survey sample, although

women are slightly over-represented in the latter (62 versus 57 percent). Note that only 40

percent of all applicants pass the written exam and only 20 percent pass both the written and

the oral exam (or are put on the waiting list). Hence, both exams have a high failure rate at 50

percent or more.

Table 2 compares the differences between women and men in the exam results

between the population of candidates and the survey sample. In Table 2, we only show

summary statistics for persons taking the oral exam (who must have passed the written exam;

this is why the standardized test scores of their written exam results are positive). When

12

considering the written exams in greater detail, it shows that women’s performance in the

written essay is about the same of men’s, but women do much worse on the written short

questions and but better in the summary essay (taken only by the external candidates).

Women do worse in the oral score than men. The distributions of the standardized scores of

the written exam results for the population of persons taking the written exams are plotted in

Figure 2.

Results for immigrants in Table 3 and Figure 2. TO BE WRITTEN

How do the oral scores correlated with the written scores for those who take the oral

exam (all of whom must have passed the oral)? Fig. 1 plots the oral test score (standardized

so that the standard deviation equals one) on the standardized essay test and short questions

test score. There is a positive, but only a small correlation between the oral score and the two

written scores, the correlation between the oral and short questions score being somewhat

higher than the correlation between the oral and exam questions score. This is not surprising,

given that a written short questions exam has more similarity to an oral exam than an essay

has to an oral exam. It is, however, quite remarkable, how high the variation in oral test

scores is even if we hold the written test scores constant.

4 Oral Exam Results Conditional on Written Results and Other

Productivity-Related Characteristics We will start by examining unconditional and conditional oral exam results by gender in

Section 4.1 and investigate the same outcomes by immigration status in Section 4.2.

4.1 Differences in Conditional Oral Exam Results Between Females and Males

Table 4 displays the coefficient of a dummy variable for “female” in a regression of the oral

exam result on the results in the written exams and potentially other control variables. The

estimating equation is:

13

resultorali = τ femalei + ρ1essayi + ρ2shortquestionsi + β ' xi + ε i

In the first set of equations, we set ρ1 = ρ2 = β = 0 , i.e. we simply compare the unconditional

oral exam results between women and men. In these estimates, we condition on what the oral

examiners observe at a minimum: the gender and the fact that the person has passed the

written exam (otherwise he or she would not be able to take the oral exam). By regulation,

the score of written exams must be kept unknown to the oral examiners.

TO BE WRITTEN, POINT ESTIMATES IN TABLE 4 SHOW THAT ALTHOUGH

WOMEN PERFORM WORSE THAN MEN IN THE ORAL EXAM, THIS GAP

BECOMES SMALLER ONCE WE CONTROL FOR THE ANONYMOUS WRITTEN

TEST SCORES AND/OR OTHER SOCIO-ECONOMIC CHARACTERISTICS.

POINT ESTIMATES IN TABLE 5 SHOW THAT IMMIGRANTS PERFORM

BETTER THAN NATIVES UNCONDITIONALLY AND EVEN BETTER

CONDITIONAL ON FOR THE ANONYMOUS WRITTEN TEST SCORES AND/OR

OTHER SOCIO-ECONOMIC CHARACTERISTICS.

WE WILL OBTAIN DATA FROM THE 2010 CONCOURS SHORTLY TO

DECREASE THE SIZE OF THE STANDARD ERROR.

5 Post-Recruitment Job-Performance Trainee Results TO BE WRITTEN, TABLES 6 AND 7 SHOW THAT HIRED WOMEN

OUTPERFORM MEN EVEN IF WE CONDITION ON WRITTEN TEST RESULTS

THIS IS NOT TRUE FOR IMMIGRANTS VERSUS NATIVES, SEE TABLES 8

AND 9.

6 Conclusions We have analyzed the performance of women versus men and immigrants versus

natives in oral exams for recruitment into higher management level of the French regional

14

administration, controlling for anonymous written exam outcomes (and other productivity-

relevant characteristics) that were unobservable for the oral examiners. Conditional on

anonymous written tests and on other human capital characteristics, the oral exam relevant

for the hiring decision seems to display no gender bias. Hence, contrary to some of the

previous literature, we cannot detect any statistically significant evidence for discrimination

against women or first- or second-generation immigrants so that in this sense, the exams are

procedurally fair. There is even some evidence for positive discrimination in favor of

immigrants.

However, the characteristics valued in the oral exam seem different from the ones

driving productivity “on the job“ (during the trainee program) and women are doing better in

the post-training tests than men conditional on previous test scores. Hence, by comparing

exam scores of the hiring contests with those after the trainee program, we find that the hiring

exams have a disparate impact by gender. Women do better in applied tests that are closer to

the job, but worse in the entrance exams that ask for more general knowledge. In this sense

there is functional discrimination against women. We find no evidence for functional

discrimination against immigrants.

References

Altonji, J. and R. Blank (1999): Race and Gender in the Labor Market, in: Ashenfelter, O. and D. Card: Handbook of Labor Economics Vol. 3C, Amsterdam, 3143-3259.

Anwar, Sh. and H. Fang (2006): An Alternative Test of Racial Prejudice in Motor Vehicle Searches: Theory and Evidence, American Economic Review 96: 127-151.

Autor, D.H. and D. Scarborough (2008): Does Job Testing Harm Minority Workers? Evidence from Retail Establishments, Quarterly Journal of Economics 123: 219-277.

Becker, G. S. (1957): The Economics of Discrimination. Chicago: University of Chicago Press.

15

Bertrand, M., Mullainathan, S. (2004): Are Emily and Greg More Employable than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination, American Economic Review 94: 991–1013.

Blinder, A. (1973): Wage Discrimination: Reduced Form and Structural Esti- mates, Journal of Human Ressources 8: 436-455

Combes, P.-P., Linnemer L. and M. Visser (2008): Publish or Peer-Rich? The Role of Skills and Networks in Hiring Economics Professors, Labour Economics 15: 456-474.

Duguet, E., Leandri, N., L’Horty, Y., and P. Petit (2008 ???): Are Young French Jobseekers of Ethnic Immigrant Origin Discrminated Against? A Controlled Experiement in the Ile de France Region, mimeograph, Université d’Evry-Val d’Essone.

Eberhard, M., Meurs, D. and P. Simon (2008): Construction d’une méthodologie d’observation de l’accès et du déroulement de carriére des generations issues de l’immigration dans la fonction publique, rapport final, Institut national d’études démographiques (INED), Paris.

Firth, M. (1981): Racial Discrimination in the British Labor Market, Industrial and Labor Relations Review 34: 265-272.

Goldin, C. and C. Rouse (2000): Orchestrating Impartiality: The Impact of “Blind” Auditions on Female Musicians, American Economic Review 90: 715-741.

Gregory, R.G. and J. Borland (1999): Recent Developments in Public Sector Labor Markets, in: Ashenfelter, O. and D. Card: Handbook of Labor Economics Vol. 3C, Amsterdam, 3573-3630.

Heckman, J.J. (1998): Detecting Discrimination, Journal of Economics Perspectives 12: 101–116.

Hirsch, B. T. 1993. “Trucking Deregulation and Labor Earnings: Is the Union Premium a Compensating Differential?” J. Labor Econ. 11(2): 279–301.

Hirsch, B. T. 1988. “Trucking Regulation, Unionization, and Labor Earnings: 1973–85.” J. Human Resources 23(3): 296–319.

Juhn, Ch., Murphy, K. and B. Pierce (1991): Accounting for the Slow- down in Black-White Wage Convergence, in: Kosters, Marvin (ed.), Workers and Their Wages, Washington D.C.

Kenney, G.M., Wissoker, D.A. (1994): An Analysis of the Correlated of Discrimination Facing Young Huspanic Job-Seekers, American Economic Review 84: 674–683.

McCrary, J. (2007): The Effect of Court-Ordered Hiring Quotas on the Composition and Quality of Policy, American Economic Review 97: 318-353.

Newmark, D., Bank, R.J. and K.D. Van Nort (1996): Sex Discrimination in Restaurant Hiring: An Audit Study, Quarterly Journal of Economics 111: 915–941.

Oaxaca, R. (1973): Male-Female Wage Differentials in Urban Labor Markets, International Economic Review 14: 693-709.

16

Riach, P.A., Rich, J. (1991): Testing for Racial Discrimination in the Labour Market, Cambridge Journal of Economics 15: 239–256.

Peoples, J., and R. Robinson. 1996. “Market Structure and Racial and Gender Discrimination: Evidence from the Telecommunication Industry.” American J. Econ. and Sociology 55(3): 309–325.

Peoples, J., and L. Saunders (1993): Trucking Deregulation and the Black/White Wage Gap, Industrial and Labor Relations Review 47(1): 23–35.

Peoples, J., and W. K. Talley. 2001. “Black-White Earnings Differentials: Privatization versus Deregulation.” American Econ. Review, Papers and Proceedings 91(2): 164–168.

Rose, N. L. (1987): Labor Rent Sharing and Regulations: Evidence from the Trucking Industry; Journal of Political Economy 95(6): 1146–1178.

Wozniak, A. K. (2007): Product Markets and Paychecks: Deregulation’s Effects on the Compensation Structure in Banking, Industrial and Labor Relations Review 60(2): 246–267.

17

Table 1 Sample means of administrative information for the population of participants and

survey participants Population n Survey n

binary outomceshired 0.16 4492 0.17 2548hired incl. waiting list 0.20 4492 0.21 2548passed written, adm. to oral 0.40 4511 0.39 2561scoreswritten essay 0.00 4509 0.01 2560written short questions 0.00 4509 0.02 2560summary essay 0.00 3139 0.02 1719oral 0.00 1539 0.03 906other characteristicsfemale 0.57 4526 0.62 2561in survey 0.59 4526 1.00 2561first generation imm. - 0.05 2561second generation imm. - 0.14 2561

18

Table 2 Summary Statistics for Male and Female Candidates Admitted to the Oral Exam (i.e. Passed Written)

Men WomenEssay Score 0.66 / 0.63 0.66 / 0.62Short Q Score 0.90 / 0.92 0.74 / 0.80Summary Essay Score 0.57 / 0.61 0.72 / 0.75Oral Score 0.07 / 0.08 -0.06 / -0.02Internal 0.40 / 0.41 0.35 / 0.31Age 31.7 29.0Bachelor's 0.23 0.15Master 1st year 0.27 0.31Master 2nd year 0.36 0.44Law 0.33 0.50Other Humanities 0.27 0.14Student 0.34 0.47prep. course 0.40 0.47Father pub sect 0.33 0.31Mother pub sect 0.30 0.37observations 681 / 318 857 / 477

19

Table 3 Summary Statistics for Native and Immigrant Candidates Admitted to the Oral Exam (i.e. Passed Written)

Natives ImmigrantsEssay Score 0.67 0.38Short Q Score 0.85 0.86Summary Essay Score 0.69 0.79Oral Score 0.01 0.08Internal 0.32 0.50Age 29.5 33.0Bachelor's 0.18 0.17Master 1st year 0.30 0.29Master 2nd year 0.42 0.36Law 0.43 0.43Other Humanities 0.21 0.14Student 0.45 0.27prep. course 0.45 0.37Father pub sect 0.34 0.20Mother pub sect 0.37 0.17observations 666 129

20

Table 4 Regression Coefficients for Women’s versus Men’s Oral Exam results Specification Coeff. s.e. obsunconditional all -0.131 (0.05) 1538unconditional survey -0.104 (0.07) 795cond on internal/center -0.085 (0.07) 795cond on internal/center X -0.064 (0.07) 795cond on internal/center WT -0.064 (0.07) 795cond on internal/center X WT -0.035 (0.07) 795

Note: X: conditions on oberved education and other socio-economic charactistics; WT: conditions on anonymous written hiring exam scores. Sources: administrative and survey data. Table 5 Regression Coefficients for Immigrants’s versus Native’s Oral Exam results Specification Coeff. s.e. obsunconditional all - - -unconditional survey 0.070 (0.09) 795cond on internal/center 0.083 (0.09) 795cond on internal/center X 0.110 (0.09) 795cond on internal/center WT 0.081 (0.09) 795cond on internal/center X WT 0.134 (0.09) 795

Note: X: conditions on oberved education and other socio-economic charactistics; WT: conditions on anonymous written hiring exam scores. Sources: administrative and survey data.

21

Table 6 Post-Recruitment Performance Women Versus Men 1st Semester Specification Oral Pre-Tr. Written Oral Internship Defense Total 1st Semunconditional survey -0.17 0.35 0.15 0.21 0.10 0.33cond on internal/center -0.13 0.31 0.12 0.12 0.06 0.25cond on internal/center X -0.09 0.28 0.07 -0.02 0.09 0.15cond on internal/center WT -0.14 0.33 0.13 0.11 0.05 0.26cond on internal/center WT O 0.34 0.17 0.12 0.06 0.28cond on internal/center X WT -0.12 0.30 0.08 -0.02 0.08 0.16cond on internal/center X WT O 0.31 0.11 -0.01 0.08 0.18 Note: X: conditions on oberved education and other socio-economic charactistics; WT: conditions on anonymous written hiring exam scores; O: conditions on oral exam scores; sample size: 365; the standard error of the displays coefficients varies between 0.09 and 0.12. Sources: administrative and survey data. Table 7 Post-Recruitment Performance Women Versus Men 2nd Semester Specification Written Oral Internship Language Total 2nd Sem Totalunconditional survey 0.23 0.01 0.22 0.15 0.24 0.32cond on internal/center 0.25 0.02 0.10 0.08 0.19 0.25cond on internal/center X 0.11 -0.02 0.00 0.07 0.06 0.11cond on internal/center WT 0.25 0.02 0.09 0.09 0.19 0.25cond on internal/center WT O 0.25 0.04 0.11 0.10 0.21 0.28cond on internal/center X WT 0.12 -0.02 0.00 0.07 0.06 0.11cond on internal/center X WT O 0.13 -0.01 0.01 0.08 0.07 0.13 Note: X: conditions on oberved education and other socio-economic charactistics; WT: conditions on anonymous written hiring exam scores; O: conditions on oral exam scores; sample size: 365; the standard error of the displays coefficients varies between 0.09 and 0.12. Sources: administrative and survey data. Table 8 Post-Recruitment Performance Immigrants Versus Natives 1st Semester Specification Oral Pre-Tr. Written Oral Internship Defense Total 1st Semunconditional survey 0.15 -0.32 -0.09 -0.06 -0.10 -0.23cond on internal/center 0.11 -0.25 0.02 0.01 -0.02 -0.09cond on internal/center X 0.12 -0.20 0.04 0.04 -0.01 -0.05cond on internal/center WT 0.09 -0.20 0.03 0.01 0.01 -0.07cond on internal/center WT O -0.21 0.01 0.00 0.00 -0.09cond on internal/center X WT 0.10 -0.16 0.05 0.03 0.02 -0.03cond on internal/center X WT O -0.16 0.02 0.03 0.01 -0.04 Note: X: conditions on oberved education and other socio-economic charactistics; WT: conditions on anonymous written hiring exam scores; O: conditions on oral exam scores; sample size: 365; the standard error of the displays coefficients varies between 0.12 and 0.14. Sources: administrative and survey data. Table 9 Post-Recruitment Performance Women Versus Men 2nd Semester Specification Written Oral Internship Language Total 2nd Sem Totalunconditional survey -0.29 -0.04 -0.13 0.35 -0.18 -0.23cond on internal/center -0.21 0.00 0.02 0.38 -0.05 -0.08cond on internal/center X -0.13 -0.01 -0.01 0.42 -0.03 -0.04cond on internal/center WT -0.21 0.00 0.03 0.40 -0.05 -0.07cond on internal/center WT O -0.22 -0.02 0.02 0.39 -0.06 -0.08cond on internal/center X WT -0.13 -0.01 -0.01 0.43 -0.03 -0.03cond on internal/center X WT O -0.14 -0.02 -0.02 0.43 -0.04 -0.05 Note: X: conditions on oberved education and other socio-economic charactistics; WT: conditions on anonymous written hiring exam scores; O: conditions on oral exam scores; sample size: 365; the standard error of the displays coefficients varies between 0.12 and 0.14. Sources: administrative and survey data.

22

23

Fig. 1. Plot of Oral Versus Written Standardized Test Score for Participants of the Oral Exam

24

A. Essay Score

B. Short Question Score

25

C. Summary Essay Score

Fig. 2. Differences in between females and males in unconditional standardized entrance exam scores

26

A. Essay Score

B. Short Question Score

27

C. Summary Essay Score

Fig. 3. Differences in between immigrations and natives in unconditional standardized entrance exam scores

28

A. Unconditional Oral Score Distributions

B. Conditional Oral Score Distributions

Fig. 4. Differences in between females and males in unconditional and conditional standardized oral exam scores

29

A. Unconditional Oral Score Distributions

B. Conditional Oral Score Distributions

Fig. 5. Differences in between natives and immigrants in unconditional and conditional standardized oral exam scores

30

Appendix Table A1 Example for Full Regression Results (Women versus Men)

cond on internal/center cond on internal/center WT cond on internal/center X WTFemale -0.08 (0.07) -0.06 (0.07) -0.04 (0.07)Essay Score 0.20 (0.05) 0.18 (0.05)Short Questions Score 0.23 (0.05) 0.19 (0.05)Summary Essay Score 0.16 (0.05) 0.17 (0.05)Internal 0.21 (0.07) 0.34 (0.08) 0.67 (0.13)Training Center (Lyon)Bastia 0.40 (0.11) 0.42 (0.11) 0.49 (0.11)Lille 0.28 (0.10) 0.16 (0.10) 0.20 (0.10)Metz 0.08 (0.11) 0.10 (0.11) 0.12 (0.11)Nantes 0.38 (0.10) 0.28 (0.11) 0.30 (0.10)Age 0.04 (0.05)aAesq 0.00 (0.00)Age High School Graduation 0.10 (0.09)Age HSG2 0.00 (0.00)Degree (Bachelor's)Master 1st year 0.07 (0.10)Master 2nd year 0.12 (0.11)Ph.D. 0.33 (0.21)Grandes Ecoles 0.09 (0.15)Field of Major (Law)Admin Econ Social 0.02 (0.14)Economics -0.11 (0.16)Public Admin -0.04 (0.18)Other Humanities 0.33 (0.09)Language Literature Arts -0.12 (0.16)Science -0.26 (0.20)Others 0.07 (0.11)Student 0.29 (0.10)Preparatory Course 0.01 (0.07)Father Public Sector -0.05 (0.07)Mother Public Sector -0.13 (0.07)Constant -0.229 (0.09) -0.641 (0.11) -2.70 (1.27)Adj R2 0.033 0.072 0.110n 795 795 795