Are Emily and Greg more employable than Lakisha and Jamal? a field experiment on labor market discrimination

Digitized

by

the

Internet

Archive

in

2011

with

funding

from

Boston

Library

Consortium

Member

Libraries

B31

M415

M'Obr

P-

7

"

Massachusetts

Institute

of

Technology

Department

of

Economics

Working

Paper

Series

Are

Emily

and

Greg

More

Employable

than

Lakisha

and

Jamal?

A

Field

Experiment

on

Labor

Market

Discrimination

Marianne

Bertrand

Sendhil

Mullainathan

Working

Paper

03-22

May

27,

2003

RoomE52-251

50

Memorial

Drive

Cambridge,

MA

02142

This

paper

can

be

downloaded

without

charge from

the

Social

Science

Research Network

Paper

Collection

at

MASSACHUSETTS

INSTITUTE

OF

TECHNOLOGY

SEP

1 9

2003

Are

Emily

and Greg

More

Employable

than Lakisha

and Jamal?

A

Field

Experiment

on Labor

Market

Discrimination

Marianne

Bertrand

and

Sendhil

Mullainathan

*

May

27,

2003

Abstract

We

performafieldexperimenttomeasureracialdiscriminationinthelabor market.

We

respondwith

fictitious resumestohelp-wanted adsin Boston and Chicagonewspapers.

To

manipulateperception of race, eachresume israndomlyassigned eithera very African American sounding

name

oravery

White

sounding name.

The

results

show

significant discrimination against African-American names:

White

names

receive 50 percentmore callbacksfor interviews.

We

also find that raceaffects thebenefits ofa

better resume. For

White

names, ahigher qualityresume elicits30 percent

more

callbackswhereasfor

AfricanAmericans,itelicitsafarsmallerincrease. Applicantslivinginbetterneighborhoodsreceive

more

callbacks but, interestingly, thiseffectdoes notdifferby race.

The amount

of discriminationisuniform

across occupations and industries. Federal contractors and employers

who

list "Equal Opportunity Employer"in theiraddiscriminateas

much

asotheremployers.

We

find littleevidence that ourresults

aredrivenbyemployersinferringsomethingotherthanrace,suchas socialclass, fromthenames. These

resultssuggestthat racialdiscriminationisstill aprominent featureofthe labor market.

•University ofChicagoGraduateSchool of Business,

NBER

and

CEPR;

MIT

and

NBER.

Addresses: 1101E.58thStreet,

R0

229D,Chicago, IL 60637; 50MemorialDrive,E52-380a, Cambridge,

MA

02142. E-mail: marianne.bertrand@gsb.uchicago.edu;

mullain@mit.edu. DavidAbrams,VictoriaBede,Simone Berkowitz,HongChung,AlmudenaFernandez,Mary Anne

Guedigu-ian,ChristineJaw, RichaMaheswari, Beverley Martis, Alison Tisza, Grant Whitehorn, andChristine Yeeprovided excellent

1

Introduction

Every

measure

of

economic

success reveals significant racialinequality in the

US

labormarket.

Compared

toWhites, African

Americans

are twice as likely to be

unemployed and

earn nearly25 percent less

when

they are

employed

(Councilof

Economic

Advisers, 1998).

This

inequality hassparked a debate

on whether

employers discriminate

by

race.

When

facedwith observablysimilarAfrican

American

and White

applicants,

do

theyfavor the

White

one?

Some

argueyes, citingeither

employer

prejudice or

employer

perceptionthat

race signals lower productivity. Others argue that discriminationis a relicofthe past, eliminated

by

some

combination of

employer

enlightenment, affirmative action

programs

and

the profit-maximization motive.

In fact,

many

in this later

camp

even feel that stringent enforcement of affirmative action

programs

has

produced

an environment

of reverse discrimination.

They

would

argue thatfacedwith identicalcandidates,

employers

might

favorthe African

American

one.1

_Data

limitations

make

itdifficultto empiricallytestthese

views. Sinceresearchers possess far less data

on

workers

than

employers do,

White and

African

American

workers that

appear

similarto researchers

may

look verydifferent toemployers.

So

any

racialdifference in

labor

market outcomes

couldjust as easily

be

attributed to thesedifferences

unobserved by

researchersas

todiscrimination.

We

conduct

a field

experiment

to circumvent this difficulty.

We

send resumes in response to

help-wanted

ads in

Chicago

and Boston newspapers and measure

the

number

ofcallbackseach

resume

receives

forinterviews.

We

experimentally manipulate perceptionofraceviathe

name

on

theresume.

We

randomly

assign very

White sounding

names

(such as

Emily Walsh

or

Greg

Baker) to half the resumes

and

very

African

American sounding

names

(such as

Lakisha Washington

or

Jamal

Jones) tothe otherhalf.

Because

we

are also interested in

how

credentials affect discrimination,

we

experimentally vary the quality ofthe

resumes used inresponse to a givenad. Higherquality applicantshave

on

average alittle

more

labor

market

experience

and

fewerholesin their

employment

history;they arealso

more

likelyto

have

an email address,

have

completed

some

certificationdegree,possess foreign

language

skillsorhave been

awarded

some

honors.2

In practice,

we

typicallysendfour

resumes

in responsetoeachad:

two

higherquality

and two

lowerquality

ones.

We

randomly

assign to

one

ofthe higher

and

oneofthe lowerqualityresumes

an

African

American

sounding

name.

In total,

we

respond

toover 1300

employment

ads in the sales, administrative support,

clerical

and customer

servicesjob categories

and

send nearly

5000

resumes.

The

ads

we

respond to covera

large

spectrum

ofjob quality,

from

cashier

work

at retailestablishments

and

clerical

work

in a

mailroom

to

office

and

sales

management

positions.

We

find large racial differences in callback rates.3 Applicants with

White

names

need to send about

J

This

camp

often explains thepoorperformanceofAfricanAmericansintermsofsupplyfactors. IfAfricanAmericanslack

manybasicskillsenteringthe labormarket,then theywill perform worse, even with parityor favoritismin hiring.

Increating the higher quality resumes,wedeliberatelymadesmallchangesincredentialsso as tominimizethechanceof over-qualification.

3_{For ease of exposition,we}

10

resumes

to get one callback

whereas

applicants with African

American

names

need to send

around

15

resumes

to getonecallback. This 50 percent

gap

in callback rates isstatisticallyverysignificant.

Based on

ourestimates, a

White

name

yieldsas

many

more

callbacks as

an

additional eight years of experience. Since

applicants'

names

are

randomly

assigned, this

gap

can only

be

attributed tothe

name

manipulation.

Race

alsoaffects the

reward

to having a betterresume.

Whites

with higher quality

resumes

receive 30

percent

more

callbacks

than

Whites

with lowerqualityresumes, astatisticallysignificant difference.

On

the

otherhand, having a higherquality

resume

hasa

much

smallereffect forAfricanAmericans. Inotherwords,

the gap

between

White and

African-

Americans widens

with

resume

quality.

While

one

may

have expected that

improved

credentials

may

alleviate employers' fear that African

American

applicants are deficient in

some

unobservableskills, this is notthe case in

our

data.4 Discrimination thereforeappears to bite twice,

making

itharder not onlyforAfrican

Americans

to find ajob

but

alsoto

improve

their employability.

The

experiment

also reveals several other aspects of discrimination. First, since

we

randomly

assign

applicants' postal addresses to the resumes,

we

can

study the effect of

neighborhood

of residence

on

the

probability of callback.

We

find that living ina wealthier (or

more

educated or

more White) neighborhood

increases callback rates. But,interestingly, African

Americans

are not helped

more than Whites by

living

ina "better" neighborhood. Second, the

amount

ofdiscrimination

we measure by

industrydoes not

appear

correlated to Census-based

measures

ofthe racial

gap by

industry.

The same

is true for the

amount

of

discrimination

we

measure

in differentoccupations. Infact,

we

findthatdiscriminationlevelsarestatistically

indistinguishable across all the occupation

and

industrycategoriescoveredin theexperiment.

We

also find

that federal contractors,

who

are thought to

be

more

severely constrained

by

affirmative action laws,

do

not discriminate less;neither

do

largeremployersoremployers

who

explicitly statethat

they

are

an

"Equal

Opportunity Employer"

intheir ads. In Chicago,

we

findthat employers locatedin

more

African

American

neighborhoods

are slightlyless likelyto discriminate.

The

rest of the

paper

is organized as follows. Section 2

compares

this experiment to prior

work

on

discrimination,

and most

notably tothelabor

market

audit studies.

We

describethe experimental design

in Section 3

and

present the results in Section 4.1. In Section 5,

we

discuss possible interpretations of

our results, focusing especially

on

two

issues. First,

we

examine whether

the race-specific

names

we

have

chosen

might

also

proxy

for social class

above

and beyond

the raceofthe applicant.

Using

birthcertificates

data on mother's education for the different

names

usedin our sample,

we

find littlerelationship

between

social

background

and

the

name

specific callback rates.5 Second,

we

discuss

how

our results

map

back

effectsareabout theracialsoundingnessofnames.

We

brieflydiscussthepotentialconfounds between

name

andracebelow andmoreextensivelyinSection5.1.

4_These

resultscontrastwith the view, mostly basedon non-experimentalevidence, that AfricanAmericansreceivehigher returns toskills. Forexample,estimating earnings regressionsonseveral decadesofCensusdata, Heckmanetal. (2001) show

that AfricanAmericansexperience higher returns to a high school degreethanWhites.

5

_We

alsoarguethatasocial classinterpretationwouldfindithardtoexplainall ofourresults,such as

why

livinga better

to the different

models

of discrimination

proposed

in the economics literature. In doingso,

we

focus

on

two

important results: thelowerreturns tocredentials forAfrican

Americans

and

therelative

homogeneity

of discrimination across occupations

and

industries.

We

conclude that existing

models

do

a

poor

job of

explainingthe fullsetoffindings. Section6 concludes.

2

Prior

Research

on

Discrimination

With

conventional laborforce

and

householdsurveys,it isdifficultto

measure

racialdiscriminationoranalyze

itsmechanics.6

With

surveydata, researchersusually

measure

discrimination

by comparing

the labor

market

performance

of

Whites and

African

Americans

who

reportasimilarsetofskills.

But

such

comparisons

can

be

quite misleading.

Standard

laborforcesurveys

do

not containall thecharacteristicsthatemployers observe

when

hiring,

promoting

or settingwages.

So one

can never

be

sure that the African

American and White

workers being

compared

aretruly similar

from

the employer'sperspective.

As

a consequence,

any measured

differencesin

outcomes

could

be

attributed tothese

unobserved

(tothe econometrician)factors

and

not to

discrimination.

This difficulty with conventional data has led

some

authors to use pseudo-experiments.' Goldin

and

Rouse

(2000), for example,

examine

the effect ofblindauditioning

on

the hiringprocessof orchestras.

By

looking atthe treatment offemale candidates before

and

aftertheintroductionof blind auditions, theytry

to

measure

the

amount

ofsex discrimination.

When

such pseudo-experiments

can be

found, the resulting

study

can be

veryinformative,

but

findingsuch experiments has

been

extremely difficult.

A

different set of studies,

known

as audit studies,

attempt

to place

comparable

minority

and White

subjects into actual social

and economic

settings

and measure

how

each

group

fares in these settings.8

Labor market

audit studiessend

comparable

minority (African

American

orHispanic)

and

White

auditors

in forinterviews

and measure whether

one is

more

likely toget thejob than the other.9

While

the results

vary

somewhat

across studies, minority auditors tend to perform

worse

on

average: they are less likelyto

get called

back

for a second interview and,conditional

on

getting calledback, less likelytoget hired.

These

audit studiesprovide

some

ofthe cleanestnon-laboratory evidenceoflabor

market

discrimination.

But

they also

have

weaknesses,

most

of

which

have

been

highlighted in

Heckman

and

Siegelman (1992)

and

Heckman

(1998). First,these studies requirethat

both

members

ofthe auditorpair are identical in all

dimensions that

might

affectproductivityinemployers'eyes, exceptfor race.

To

accomplish this,researchers

6_{See Altonji}

and Blank 1999)fora detailedreview of the existingliteratureonracialdiscriminationinthelabormarket.

7_Darity

andMason(1998) describeaninterestingnon-experimentalstudy. Prior totheCivilRightsActof1964,employment

ads would explicitlystateracial biases, providing adirect measureofdiscrimination. Ofcourse, as Arrow (1998) mentions,

discrimination wasatthattime "afacttoo evidentfordetection."

8_{Fix and Turner}

(1998) provide a survey of

many

suchauditstudies. 9

Earlier hiringaudit studies include

Newman

(1978) and Mclntyre et al (1980). Three more recentstudies are Crosset al(1990), Turneretal (1991), Jamesand DelCastillo (1991). Altonjiand Blank (1999),Heckman and Siegelman(1992) and

typically

match

auditors onseveral characteristics (height, weight, age, dialect, dressingstyle, hairdo)

and

train

them

forseveral daystocoordinate interviewing styles. Yet, criticsnote that this is unlikelyto erase

the

numerous

differencesthat exist

between

theauditors in

a

pair.

Another weakness

oftheaudit studiesisthattheyarenotdouble-blind. Auditors

know

the

purpose

ofthe

study.

As

Turner

et al (1990) note:

"The

first

day

oftraining alsoincluded

an

introductionto

employment

discrimination, equal

employment

opportunity,

and

a reviewofprojectdesign

and

methodology." This

may

generate consciousorsubconsciousmotives

among

auditors togenerate dataconsistent or inconsistentwith

racialdiscrimination.

As

psychologists

know

verywell,these

demand

effectscan

be

strong. Itisverydifficult

to insurethat auditors willnot

want

to

do

"a

good

job." Sincethey

know

thegoal oftheexperiment, they

can alter their behavior in front ofemployers to express (indirectly) their

own

views.

Even

a small belief

by

auditors that employerstreat minorities differently can resultin apparent discrimination. This effect is

further magnified

by

the fact that auditors are not in fact seekingjobs

and

are therefore

more

free to let

theirbeliefs affect the interview process.

Finally, audit studies are extremely expensive,

making

it difficult to generate large

enough

samples to

understand

the

nuances

and

possiblemitigatingfactorsof discrimination. Also, these

budgetary

constraints

worsen

the

problem

of

mismatched

auditorpairs. Costconsiderations forcetheuseofafewpairs of auditors,

meaning

that

any

one

mismatched

paircan easily drivethe results. Infact,these studies generallytendto

find significant differences in

outcomes

across pairs.

Our

study

circumventstheseproblems. First,because

we

only rely

on resumes

and

notpeople,

we

can

be

suretogenerate comparabilityacrossrace. Infact,since raceis

randomly

assignedtoeach resume, the

same

resume

will

sometimes

be associated with

an

African

American

name

and sometimes

with a

White

name.

This guarantees that

any

differences

we

findare

due

solelyto theracemanipulation. Second,theuseofpaper

resumes

insulates us

from

demand

effects.

While

the research assistants

know

the

purpose

ofthestudy,our

protocolallowslittle

room

forconsciousorsubconsciousdeviations

from

thesetprocedures. Moreover,

we

can

objectively

measure whether

the randomization occurred asexpected.

This

kindof objective

measurement

is impossiblein the caseofthe previous auditstudies. Finally, because of relatively low marginalcost,

we

can send out alarge

number

ofresumes. Besides giving us

more

precise estimates, this larger

sample

size

also allowsusto

examine

the

mechanics

ofdiscrimination

from

many

more

angles.10

1

_A

_{similar "correspondence"techniquehasbeen used}

ina fewU.K.studies. See Joweiland Precott-Clarke (1970),Brown

and Gay (1985)andHubbockand Carter (1980). Theseearlierstudies have verylimitedsamplesizeandfocus exclusivelyon

documenting gapincallbacks betweenthe minorityand non-minoritygroups.

Some

of these studiesfailtofullymatchskills

3

Experimental Design

3.1

Creating

a

Bank

of

Resumes

The

firststep oftheexperimental designisto generate templatesfortheresumesto besent.

The

challenge

isto

produce

a setofrealistic

and

representative

resumes

without using

resumes

that belongto actualjob

seekers.

To

achieve this goal,

we

start with resumes of actualjob searchers but alter

them

sufficiently to

create distinct resumes.

The

alterations maintain the structure

and

realismofthe initial

resumes

without

compromising

theirowners.

We

begin withresumes posted

on two

jobsearchwebsitesasthebasisforour artificialresumes.11

While

the

resumes

posted

on

these websites

may

not be completely representative ofthe average jobseeker,they

provide apracticalapproximation.12

We

restrictourselves topeople seeking

employment

inour experimental

cities (Boston

and

Chicago).

We

alsorestrictourselves to fouroccupationalcategories: sales, administrative

support,clericalservices

and customer

services. Finally,

we

furtherrestrictourselves to

resumes

posted

more

than

six

months

prior tothestart ofthe experiment.

We

purge the selected

resumes

ofthe person's

name

and

contactinformation.

During

this process,

we

classify the

resumes

within each occupational category into

two

groups: high

and

low quality. Injudging

resume

quality,

we

usecriteriasuch as labor

market

experience, careerprofile,

existence ofgaps in

employment and

skills listed.

Such

a classification is admittedly subjective but it is

made

independently ofany race assignment

on

the

resumes

(which occurs laterintheexperimentaldesign).

To

further reinforcethe quality

gap between

the

two

setsofresumes,

we

add

toeach high quality

resume

a

subset ofthefollowingfeatures:

summer

or while-at-school

employment

experience,volunteeringexperience,

extra

computer

skills, certification degrees, foreign languageskills,

honors

or

some

militaryexperience. This

resume

quality manipulation needs to

be

somewhat

subtle to avoid

making

a higher qualityjob applicant

over-qualified for a given job.

We

try toavoid this

problem by

making

sure that thefeatures listed

above

are not all

added

at oncetoa given resume. Thisleaves us with a high quality

and

alow quality pool of

resumes.13.

To

minimizesimilarityto actualjobseekers,

we

use

resumes

from Boston

jobseekers to

form

templates

forthe resumes to

be

sentout in

Chicago

and

used the

Chicago

resumes

to

form

templatesforthe

resumes

to

be

sent out in Boston.

To

implement

this migration,

we

alter the

names

ofthe schools

and

previous

employers

on

theresumes.

More

specifically,foreach

Boston

resume,

we

usethe

Chicago resumes

toreplace

a

Boston

school

by

a

Chicago

school.14

We

also usethe

Chicago

resumes

toreplace a

Boston employer

by a

Thesitesarewww.careerbuilder.

com

andwww.americasjobbank.com.

12_{lnpractice,wefound}

largevariation inskilllevelsamongpeople postingtheir resumes onthesesites.

In Section4.2and Table3,weprovide a detailedsummaryofresumecharacteristicsby qualitytype. 14

_We

Chicago employer

inthe

same

industry.

We

useasimilarprocedure to migrate

Chicago

resumestoBoston.15

This produces distinct but realistic looking resumes, similarin their education

and

career profiles to this

sub-populationofjob searchers.16

3.2

Identities

of

Fictitious

Applicants

The

next step is to generate identities forthe fictitiousjob applicants: names, telephone

numbers,

postal

addresses

and

(possibly) e-mail addresses.

The

choice of

names

is crucialto our experiment.17

To

decide

on which names

areuniquely African

American and which

areuniquely

White,

we

use

name

frequency

data

calculated from birth certificates ofall babies

born

in Massachusetts

between

1974

and

1979.

We

tabulate

these

data by

race to determine

which

names

are distinctively

White

and which

are distinctively African

American.

Distinctive

names

are those that have the highest ratio of frequency in one racial

group

to

frequencyinthe otherracial group.

As

a checkof distinctiveness,

we

conducted

a surveyinvariouspublicareasinChicago.

Each

respondent

was

given a

resume

with a

name

and

asked to assess features ofthe person,

one

of

which

being race. In

general, the

names

led respondents to readily attribute the expected race for the person but there

were

a

few exceptions

and

these

names

were

disregarded.18

The

final list offirst

names used

for thisstudyarereportedin

Appendix

Table 1.

The

table alsoreports

the frequency ofthese

names

in the Massachusetts birth certificates data.19

The

African

American

first

names

usedinthe experimentare

remarkably

common

inthe population. This suggests that

by

usingthese

names

as

an

indicator ofrace,

we

are actually coveringalarge

segment

ofthe African

American

population.20

Applicantsineach race/sex/ city/resumequalitycellare allocatedthe

same

phone number.

This

guaran-teesthat

we

canpreciselytrack

employer

callbacksineachofthesecells.

The

phone

lines

we

useare virtual

ones

with

only avoicemail

box

attached toit.

A

similaroutgoing

message

isrecorded

on

each ofthevoice

mail

boxes

buteach

message

is recorded

by

someone

oftheappropriaterace

and

gender. Since

we

allocate

the

same

phone

number

forapplicants with different

names,

we

cannot use a person

name

inthe outgoing message.

15_Notethat

forapplicants with schooling orworkexperience outside of theBostonorChicagoareas, weleavethe school or

employernameunchanged.

16

_We

alsogenerate asetofdifferentfonts,layoutsandcoverlettersto furtherdifferentiatethe resumes. Theseareapplied at

thetimetheresumesare sentout.

17

_We

_chosename

over other potential manipulationsofrace, suchasaffiliationwith a minority group,becausewefeltsuch

affiliations

may

especiallyconveymorethanrace.

18_{Forexample, MauriceandJeromearedistinctivelyAfricanAmerican namesina frequency sense}yetarenot perceived as suchby

many

people.

19

_We

also triedtousemore WhitesoundinglastnamesforWhiteapplicantsandmoreAfricanAmerican soundinglastnames

forAfrican American applicants. The lastnames usedforWhite applicantsare: Baker, Kelly, McCarthy, Murphy, Murray,

O'Brien, Ryan, Sullivan and Walsh.

The

last names used forAfrican American applicants are: Jackson, Jones, Robinson,

Washington andWilliams.

20

_One

_{mighthowever wonder whether}

this isan atypical segmentof the African Americanpopulation. InSection 5.1,

we

While

we

do

not expect positivefeedback

from

an

employer

totake place via postal mail,resumes still

needpostal addresses.

We

thereforeconstructfictitiousaddressesbased

on

realstreetsin

Boston

and Chicago

using the

White

Pages.

We

select

up

to3 addressesineach 5-digitzip

code

in

Boston and

Chicago.

Within

cities,

we

randomly

assign addressesacrossallresumes.

We

alsocreate 8emailaddresses, 4for

Chicago

and

4 for Boston.21

These

email addresses are neutral with respect to

both

race

and

sex.

Not

all applicants

aregiven an email address.

As we

explained above, the email addresses areused almost exclusivelyfor the

higherqualityresumes. This procedureleavesuswith a

bank

of

names,

phone numbers,

addresses

and

e-mail

addresses

which

we

can

assigntothe

template

resumes

when

respondingtothe

employment

ads.

3.3

Responding

to

Ads

The

experiment

was

carried

on between

July 2001

and January 2002

in

Boston

and between

July 2001

and

May

2002in Chicago.22

Over

thatperiod,

we

collected all

employment

ads in the

Sunday

editions of

The

Boston

Globe

and The

Chicago Tribune inthesales, administrative support,

and

clerical

and customer

servicessections.

We

eliminate

any ad where

applicants are askedtocall orappearin person. In fact,

most

ads

we

surveyed inthesejob categoriesasked forapplicants to faxin or

(more

rarely) mail in theirresume.

We

log the

name

(when

available)

and

contact information for each employer, along with

any

information

on

the positionadvertised

and

specificrequirements (such aseducation,experience, or

computer

skills).

We

alsorecord

whether

ornot the

ad

explicitlystatesthatthe employeris

an

equal opportunity employer.

For eachad,

we

use the

bank

ofresumes to

sample

fourresumes (twohigh-quality

and two

low-quality)

that fit the job description

and

requirements as closely as possible.23 In

some

cases,

we

slightly alterthe

resumes

to

improve

thequalityofthe match, suchas

by

adding

the

knowledge

ofaspecificsoftwareprogram.

One

ofthehigh

and

oneofthe lowqualityresumesselected arethen

drawn

at

random

to receiveAfrican

American

names, the other high

and

low

resumes

receive

White

names.

24

We

use

male and

female

names

forsalesjobs,

whereas

we

use nearlyexclusively female

names

foradministrative

and

clericaljobsto increase

callback rates.25

Based on

sex, race, city

and resume

quality,

we

assign a

resume

the appropriate

phone

number.

We

also select at

random

a postal address. Finally, e-mail addresses are

added

to

most

ofthe

high qualityresumes.26

The

final resumesareformatted, with fonts, layout

and

coverletterstylechosenat

Thee-mailaddresses are registeredon Yahoo.com,Angelfire.com or Hotmail.com.

22_{Thisperiodspans}

bothtight andslacklabormarkets. Inour data,thisisapparentascall-backrates (andnumberofnew

ads) dropped precipitously after September 11th, 2001. Interestingly, however, the amount of discrimination wemeasure in

thesetwoperiodsisthesame.

23

Insomeinstances,ourresumebankdoes not havefourresumesthatareappropriatefora givenad. Insuchinstances,we

send onlytworesumes.

Thoughthesamenamesare repeatedly used inourexperiment,weguarantee that nogivenad receivesmultipleresumes

with thesamename.

25_{Male nameswere used}

fora few administrative jobsinthefirst monthofthe experiment.

26

In thefirst month ofthe experiment, a few high quality resumes weresent withoutemail addressand afewlow quality

random.

The

resumes

arethen faxed (orin a fewcasesmailed) tothe employer.27 All inall,

we

respond to

more than

1300

employment

ads overthe entire

sample

period

and send

close to

5000

resumes.

3.4

Measuring Responses

We

measure whether

a given

resume

elicits a callback or e-mail

back

for

an

interview. For each

phone

or

email response,

we

use the content ofthe

message

left

by

the

employer

(name

ofthe applicant,

company

name,

telephone

number

for contact) to match,the response to the corresponding

resume-ad

pair.28

Any

attempt by

employers to contact applicants via postal mailcannot

be

measured

in our

experiment

sincethe

addresses are fictitious. Several

human

resource

managers

confirmed to us that employers rarely, ifever,

contact applicantsvia postal mailtoset

up

interviews.

3.5

Weaknesses

of

the

Experiment

We

have

already highlighted thestrengths ofthis

experiment

relative toprevious audit studies.

We

now

discuss its weaknesses. First, our

outcome measure

is crude, even relative to the previous audit studies.

Ultimately,one cares

about whether

an applicantsgets the job

and about

the

wage

offered conditional

on

getting thejob.

Our

procedure, however, simply

measures

callbacksfor interviews.

To

the extent that the

search process has even

moderate

frictions,

one would

expect that reduced interview rates

would

translate

intoreduced joboffers. However,

we

arenotable to translateourresultsintogapsinhiring ratesorearnings.

Another weakness

is that the resumes

do

not directly report race but instead suggest race through

personal names. This leadsto varioussources ofconcern. First, while the

names

arechosen to

make

race

salient,

some

employers

may

simply notnoticethe

names

ornot recognize theirracial content.

As

aresult,

ourfindings

may

under-estimate the extent ofdiscrimination. Relatedly,because

we

arenot assigning race

but only race-specific

name,

our results are not representative ofthe average African

American (who

may

not

have

such a racially distinct name).29

Finally,

and

this is

an

issue pervasive in

both

our

study

and

the pair-matchingaudit studies,

newspaper

adsrepresentonly

one

channelforjob search.

As

iswell

known

from

theexistingliterature,social networks

are a

common

means

through

which

people find jobs

and

one

that clearly cannot

be

studied here. This

omission

would

affect our results if African

Americans

use social networks

more

or ifless discriminating

employersrely

more on

networks.30

27

Aspartofthefaxingprocess,westripal!identifiersfromtheoutgoingfaxtoguaranteethatemployerscould notseethat

allfour faxes originatefromthesamelocale.

28_{Veryfewemployersused email to contactanapplicant back.}

29_{As AppendixTable}

1indicates,the AfricanAmericannames

we

usearehoweverquite

common among

AfricanAmericans,

makingthislessofaconcern.

We

return tothis issue inSection5.1.

4

Results

4.1

Is

There

Discrimination?

Table 1 tabulates average callback rates

by

racial soundingness ofnames. Included in brackets

under

each

rate is the

number

of

resumes

sent in that cell.

Row

1 presents ourresults for the full

data

set.

Resumes

with

White

names

have a 10.08 percent chance ofreceiving a callback. Equivalent

resumes

with African

American

names

have a 6.70percent chance ofbeingcalled back. This representsa difference in callback

rates of 3.35 percentage points, or 50 percent, that can solely

be

attributed to the

name

manipulation.

Column

4

shows

thatthisdifferenceisstatistically significant.31

Put

inotherwords,these resultsimply that a

White

applicantshould expect

on

average

one

callbackfor every 10ads sheorheapplies to; on the other

hand,

an

African

American

applicant

would

needto applyto 15 different adsto achievethe

same

result.32

How

large are theseeffects?

While

the costofsending additionalresumes

might

not

be

largeperse, this

50

percent

gap

could

be

quite substantial

when

compared

totherate of arrival of

new

job openings. In our

own

study,thebiggestconstraining factor

was

thelimited

number

of

new

job openings each week.

Another

way

to

benchmark

the

measured

return to a

White

name

is to

compare

it to thereturns to other

resume

characteristics. For example, in Table5,

we

will

show

that, atthe average

number

ofyears ofexperiencein

our sample,

an

extra yearofexperienceincreasesthelikelihood ofacallback

by

.4percentagepoint.

Based

on

thispoint estimate,the return toa

White

name

isequivalent to about 8 additionalyears of experience.

Rows

2

and

3break

down

thefull

sample

of sent

resumes

intothe

Boston

and Chicago

markets.

About

20 percent

more

resumes

were

sent in

Chicago than

in Boston.

The

average callback rate (across races)

is lowerin

Chicago

than in Boston. This

might

reflect differences in labor

market

conditions across cities

over theexperimental periodor

maybe

differencesin theability ofthe

MIT

and Chicago teams

ofresearch

assistantsinselecting

resumes

that

were

good matches

for a given help

wanted

ad.

The

percentagedifference

incallback ratesis

however

strikinglysimilaracross

both

cities.

White

applicantsare48 percent

more

likely

than African

American

applicants to receive a callback in

Chicago and

52 percent

more

likely in Boston.

These

racial differencesarestatisticallysignificant in both cities.

Finally,

rows

4 to 7 break

down

the full

sample

into female

and male

applicants.

Row

4 displays the

averageresultsforallfemale

names

while

rows

5

and

6break the female

sample

into administrative (row5)

and

salesjobs (row 6);

row

7displaysthe averageresultsforall

male

names.

As

notedearlier,female

names

were

usedin

both

sales

and

administrativejobopenings

whereas

male

names were

used closeto exclusively

forsalesopenings.33

Looking

acrossoccupations,

we

findasignificantracial

gap

in callbacksforboth males

31

Thesestatistical testsassume independenceofcallbacks.

We

havehowever verifiedthat theresultsstaysignificant when

weassumethatthe callbacks are correlatedeitherattheemployerorfirst

name

level.

32_{ThisobviouslyassumesthatAfricanAmericanapplicantscannotassessa}

prioriwhichfirmsaremorelikelyto discriminate againstthem.

33_{Only about6percent ofallmale resumes weresent}

inresponse to an administrative job opening.

(49 percent)

and

females (50 percent).

Comparing

males to femalesin sales occupations,

we

find a larger

racial

gap

among

males (49percent versus25percent). Interestingly,females in salesjobsappear to receive

more

callbacksthan males; however, this (reverse) gender

gap

isstatistically insignificant

and

economically

smaller

than

any

ofthe racial gaps discussed above.

Rather

than

studying the distribution of callbacks at the applicant level, one can also tabulate the

distribution of callbacks at the

employment ad

level. InTable 2,

we

compute

thefraction ofemployers that

treat

White

and

African

American

applicantsequally,thefraction ofemployersthat favor

White

applicants

and

thefraction ofemployersthat favorAfrican

American

applicants.

Because

we

send

up

to four

resumes

in response to each

sampled

ad, the 3 categories

above

can each take 3 different forms.

Equal

treatment

occurs

when

either

no

applicant gets calledback,

one White

and

one African

American

get called

back

or

two Whites

and two

African

Americans

get calledback.

Whites

are favored

when

eitheronly one

White

gets

called back,

two Whites

and no

African

American

get called

back

or

two

Whites and one

African

American

get calledback. African

Americans

arefavoredin the othercases.

As

Table2 indicates, equaltreatment occursfor

about

87.5percent ofthehelp-

wanted

ads.

As

expected,

the

major

sourceofequaltreatment

comes from

the high fractionofadsfor

which no

callbacks arerecorded

(82.5 percentoftheads).

Whites

are favored

by

nearly 9percent oftheemployers,with a majorityofthese

employers contacting exactly one

White

applicant. African Americans,

on

the other

hand,

arefavored

by

only

about

3.7 percentofemployers.

We

formally test

whether

thereis

symmetry

in thefavoring of

Whites

over African

Americans

and

African

Americans

over

Whites by

employers.

We

find that the difference

between

thefractionofemployersfavoring

Whites and

the fraction ofemployersfavoringAfrican

Americans

(5.11 percent) is statisticallyverysignificant (p

=

.0000).

4.2

Do

African

Americans

Receive

Different

Returns

to

Resume

Quality?

Our

results so far suggest a substantial

amount

of discrimination in job recruitment. Next,

we

would

liketo learn

more

about

how

employers discriminate.

More

specifically,

we

ask

how

employers respond to

improvements

in African

American

applicants' credentials.

To

answer this question,

we

examine

how

the

racial

gap

in callback ratesvaries

by

resume

quality.

As we

mentioned

in section 3, for

most

ofthe

employment

ads

we

respond

to,

we

send four different

resumes:

two

higher quality

and

two

lower quality ones. Table 3 gives a better sense of

which

factors

enter into this subjective classification. It displays

means and

standard deviations of the

most

relevant

resume

characteristics for thefull

sample (column

1) aswell asbroken

down

by

race

(columns

2

and

3)

and

resume

quality (columns4

and

5). Sinceapplicants'

names

are randomized, thereis

no

differencein

resume

characteristics

by

race.

Columns

4

and

5

document

theobjective differences

between resumes

subjectively

classifiedashigh

and

low. Higherquality applicants

have

on average an extra yearoflabor

market

experience,

fewer

employment

holes (where

an employment

hole is defined as a period of at least 6

months

without a

reportedjob), are

more

likelyto have

worked

whileat school

and

to report

some

military experience. Also,

higher quality applicants are

more

likelytohave

an

email address,tohavereceived

some

honors

and

tolist

some

computer

skills

and

otherspecialskills(such asacertificationdegreeor foreignlanguageskills)

on

their

resume.

Note

thatthehigher

and

lower quality

resumes do

not differon average with regard toapplicants'

educationlevel.34

About 70

percentofthesent

resumes

report acollegedegree.35

The

last 5 rows of Table 3

show

summary

characteristics of the applicants' zip

code

address.

Using

1990

Census

data,

we compute

thefractionhigh-schooldropouts, fraction collegeeducatedormore,fraction

Whites, fractionAfrican

Americans

and

log(median percapita

income)

for eachzipcode usedin the

exper-iment. Sinceaddressesare

randomized

within cities, these

neighborhood

quality

measures

areuncorrelated

with

race or

resume

quality.

The

differencesincallback rates

between

high

and

lowresumesarepresentedin Table4.

The

firstthing to

noteisthatthe

resume

qualitymanipulation works: higherqualityresumesreceivehigher callbackrates.

As

row

1 indicates,

we

recordacallback rate of

more than

11 percentfor

White

applicantswith a higherquality

resume,

compared

to 8.8 percent for

White

applicants with lower quality resumes. This is a statistically

significantdifference of2.51percentagepoints,or30percent.

Most

strikingly,African

Americans

experience

much

less of

an

increase in callback rate forsimilar

improvements

in their credentials. African

Americans

with

higher qualityresumesreceiveacallback 6.99percent ofthe time,

compared

to 6.41percentforAfrican

Americans

with lowerquality resumes. This is only a .58 percentagepoint, or 9 percent, increase

and

this

increaseis notstatisticallysignificant.

Instead of relying

on

the subjective quality classification,

Panel

B

directly uses

resume

characteristics

to classify the resumes.

More

specifically,

we

use a

random

subsample

of one-third of the

resumes

to

estimate a probit regression ofthe callback

dummy

on

the

resume

characteristics listed in Table 3.

We

further control for a sex

dummy,

a city

dummy,

6 occupation

dummies

and

a vector ofjob requirements

as listed in the

employment

ads.36

We

then use the estimated coefficients

on

the

resume

characteristics to

rank

the remaining two-thirds ofthe

resumes by

predicted callback.

We

classify as "high"

resumes

those

that have

above

median

predicted callback; similarly,

we

classify as "low"

resumes

those that have

below

median

predicted callback.

As

one

can see

from

Panel B, qualitatively similar results

emerge from

this

analysis.

While

African

Americans

do appear

to significantly benefitfrom higherquality

resumes

under

this

alternativeclassification,they benefit

much

less

than

Whites.

The

ratio of callback rates forhigh versuslow

qualityresumes is 1.66forAfricanAmericans,

compared

to 2.81 for Whites.

34_This

reflects the factthat all sent resumes, whetherhigh or lowquality, arechosen tobegood matches fora given job opening.

35_This

varies from about 50 percent fortheclerical andadministrativesupport positions to more than 80 percentfor the executive,managerial andsalesrepresentativespositions.

36_{Seesection 4.4 formore}

detailsontheseoccupationcategoriesandjob requirements.

In

Table

5,

we

directly report the results of race-specificprobit regressions ofthe callback

dummy

on

resume

characteristics.

We

however

start in

column

1 withresults for thefull

sample

ofsent resumes.

As

one

can see,

many

ofthe

resume

characteristics have the expected effect

on

the likelihood ofa callback.

The

additionofan email address, honors

and

special skills all

have

a positive

and

significant effect

on

the

likelihood of

a

callback.37 Also,

more

experienced applicants are

more

likely to get called back: at the

average

number

ofyears ofexperienceinour

sample

(8years), each extra yearofexperienceincreasesthe

likelihoodofacallback

by about

.4percentagepoint.

The

most

counterintuitiveeffects

come

from computer

skills,

which appear

tonegativelypredict callback,

and

employment

holes,

which appear

to positively predict callback.

The same

qualitativepatternsholdin

column

2

where

we

focus

on

White

applicants.

More

importantly,

the estimated returns to

an

email address, additional

work

experience, honors

and

special skills

appear

economically stronger for that racial group. For example, at the average

number

ofyears ofexperience in

our sample, each extra yearofexperienceincreasesthelikelihoodofacall

back

by

about.7percentagepoint.

Also,

working

whileatschool, whilenot astatisticallysignificant determinantof callbackin thefullsample,

yields positivereturns for

White

applicants.

As

might have been

expected

from

the

two

previous columns,

we

findthat the estimatedreturns

on

these

resume

characteristicsarealleconomically

and

statistically

weaker

forAfrican

American

applicants

(column

3). Infact, alltheestimatedeffectsforAfrican

Americans

arestatisticallyinsignificant,exceptforthe return

tospecialskills.

Resume

characteristicsthus

appear

lesspredictive of callback ratesforAfrican

Americans

than

theyarefor Whites.

To

see this

more

clearly,

we

predict callback rates usingeither the regressionin

column

2 orthe regressionin

column

3.

The

standard deviation ofthe predicted callback

from

column

2is

.064,

whereas

it is only 0.034

from

column

3. In

summary,

employers simply

seem

to

pay

less attention or

discount

more

thecharacteristicslisted

on

the

resumes

with African

American

sounding names.

Taken

at

face value, these results suggest that African

Americans

may

face relatively lower individual incentivesto

investinhigherskills.38

4.3

Applicants'

Address

An

incidental feature ofour experimental design is the

random

assignment ofaddresstothe resumes. This

allows usto

examine whether

and

how

an

applicant's residentialaddress, all elseequal, affectsthelikelihood

ofacallback. In addition,

and most

importantlyforour purpose,

we

canalsoask

whether

African

American

applicants arehelped

more

by

residingin

more

affluentneighborhoods.

37_Note

that the e-mail addressdummy,becauseitisclose toperfectly correlatedwith the subjectiveresumequalityvariable,

may

inpartcapturesomeotherunmeasuredresumecharacteristicsthat

may

haveledus to categorize a givenresumeas higher

quality.

38_{This of course assumesthatthechanges}

inwageoffersassociatedwith higherskills arethesameacrossraces,or at least

not systematicallylarger forAfrican Americans.

We

performthisanalysisinTable6.

We

start

(columns

1,3

and

5)

by

discussingtheeffectof

neighborhood

of residence across all applicants.

Each

of these

columns

reports the results ofa probit regression ofthe

callback

dummy

on

a specific zip

code

characteristic

and

a city

dummy.

Standard

errors are corrected

for clustering of the observations at the

employment ad

level.

We

find a positive

and

significant effect of

neighborhood

quality

on

thelikelihoodofacallback. Applicantslivingin

Whiter (column

1),

more educated

(column

3) orhigher

income (column

5)

neighborhoods

have a higherprobability of receivingacallback. For

example, a 10 percentage pointincreasein the fraction ofcollege-educated in zipcode ofresidence increases

thelikelihoodofacallback

by

.6percentagepoint.

In

columns

2, 4

and

6,

we

further interact each ofthe zip

code

characteristic

above

with a

dummy

variable for

whether

the applicant is African

American

or not.

Each

ofthe probit regressions in these

columns

also includes

an

African

American

dummy,

a city

dummy

and an

interaction ofthe city

dummy

with

the African

American

dummy.

There

is

no

evidence that African

Americans

benefit

any

more than

Whites from

livingin a

more

White,

more

educated zipcode.

The

estimated interactions

between

fraction

White

and

fraction college educated with the African

American

dummy

are economically very small

and

statistically insignificant.

We

do

find

an

economically

more

meaningful effect of zip

code

median income

level

on

the racial

gap

incallback; that effect, however, isstatistically insignificant.

In

summary,

while the quality of the

neighborhood

ofresidence is a significant factor in employers'

decision to contact a job applicant, African

Americans do

not

appear

to benefit

more than Whites from

livinginbetterneighborhoods.

These

findingsare interesting. Indeed, ifghettos

and

bad neighborhoods

are

particularly stigmatizingfor African Americans,

one might have

expected African

Americans

to

be

helped

more

by

having a "good" address.

Our

results

do

not lendsupporttothis hypothesis.

4.4

Job

Characteristics

Inthissection,

we

turntojobcharacteristics

and

ask

whether

discriminationvaries based

on

thespecificjob

requirements oroccupation listed in the

employment

ads.

Column

1 ofTable7 givesadescription of thespecificrequirementsstatedinthe

sample

ofads

we

respond

to.

About

80 percentofthe adsstate

some

form ofrequirement.

About

43 percent ofthe ads require

some

minimum

experience, of

which

roughly 50 percent simply askfor

"some

experience," 25 percent less

than

two

years

and

25 percent at least 3 years of experience.

About

44 percent ofads

mention

some

computer

knowledge

requirement,

which

can range from Excel or

Word

to

more

esoteric software programs.

Good

communication

skillsareexplicitlyrequiredin

about

12percentoftheads. Organizationskillsare

mentioned

7 percentofthe time. Finally, only

about

11 percent ofthe ads list an expliciteducation requirement.

Of

these, 8.8 percent require a high school degree, 49 percent

some

college (such as an associate degree)

and

therest at leasta 4-yearcollegedegree.39

How

do

these differentjob requirements affect the racial

gap

in callback?

To

answer

thisquestion,

we

show

in

column

2theresultsof variousprobit regressions.

Each

entryin this

column

isthemarginal effect

ofthe requirement

on

the racial

gap

in callback. Specifically, each entry

comes from

a separate probit

regression

where

we

regressacallback

dummy

on an

African

American

dummy,

therequirement

dummy

and

theinteraction oftherequirement

dummy

with the African

American

dummy. The

reportedcoefficientisthat

on

theinteraction term.

The

point estimates

show

no

consistent

economic

pattern

and

are allstatistically

insignificant.

Measures

ofjob quality, such as experience,

computer

skills or education requirements

do

not predict the extent ofdiscrimination.

Equally

surprising,

communication

or other inter-personal skill

requirements have

no

effect

on

thelevelofdiscrimination. In short,discrimination appearstovarylittle

by

job requirements.40

Panel

A

ofTable 8 investigates

whether

discrimination varies significantly across occupations.

As

we

mention

earlier, the specificsubsectionsofthe

Sunday

newspapers

help-wantedsectionsthat

we

sample

for

thisstudy broadly relate tosales

and

administrativepositions. Thisisreflected in theoccupational

distri-butionreportedin

column

1.

To

form

this classification,

we

usethejobdescription listed inthe

employment

ad

and

map

it into

one

ofsix following categories: executives

and

managerial occupations; administrative

supervisors; sales representatives; sales workers; secretaries

and

legal assistants; clerical occupations.41 In

the 19905 percent Census, these occupations accountfor

about

46.5 percent of totalsalariedprivate sector

employment

in

Chicago

and

Boston. African

Americans

accountfor

about

4percentof

employment

inthese

occupations in

Boston

and

10.5percent in Chicago. For comparison, African

Americans

account for about

4.8 percent of total salaried private sector

employment

in

Boston and

11.8 percent in Chicago.

We

report inthesecondtolast

column

of

Table

8average earnings

measures

foreachoftheseoccupation

categories.

These measures

are

computed from

the

1990

5 percent

Census

for

Boston

and

Chicago.

More

specifically,

we

first estimate

a micro

wage

regression oflog(annual earnings)

on

8 education

dummies,

a

quadratic in age, a sex

dummy

and

a city

dummy.

We

then

compute

the

mean

residual earnings in each

occupationcategory.

As

expected, the

sampled

occupationcategories areverydifferent.

The

first

two

(about

22%

ofthe data) correspond to better jobs

on

average that involve,

among

other things, the supervision

ofother workers; these

two

categories are associated with the highest earnings. Sales representatives

and

secretaries (about

50%

ofthe data)involvelesssupervisionbutarestillfairlywellpayingjobs.

The

last

two

occupations (salesworkers

and

clerical workers) correspond tothe lowestpaying jobs in our sample.

Otherrequirements sometimes mentioned include typing skills for secretaries (withspecific words perminute minimum

thresholds)and,morerarely,foreignlanguageskills.

40_{Otherways ofestimating these}

effectsproduce a similar non-result.

Among

otherthings,weconsidered including a city

dummy

orestimatingtheeffectsseparatelybycity;wealsoestimatedonesingleprobit regressionincludingallrequirementsat

once.

41

Thesecategories respectivelycorrespondtothe followingCensus 1990occupation codes: 000-042; 303-307; 253-262; 263-302;

234and313; 308-402(excluding313).

The

second

and

third

columns

ofPanel

A

ofTable 8 respectivelyreport the callback ratesfor

White and

African

American

applicants in eachofthese occupation categories.

Column

4 reports theratioofcallback

rateswhile

column

5 reportsthe difference.

The

averagecallback rate across races

seems

to decreasewith

the level ofjob quality.

Whites

receive

more

callbacks

than

African

Americans

in alljobcategories.

While

the racial

gap

in callback rates varies

somewhat

across occupations,

we

cannot reject the null hypothesis

thatthe

gap

isthe

same

across all categories.

We

however

brieflydiscuss the point estimates

by

occupation

and

how

they relate to

Census

measures

ofaverage earnings.

The

smallest

gap

appears to

be

for executivejob positions,

where Whites

only face

a

33%

higher chance

than

African

American

of being called back.

The

second lowest

gap

is for clerical

jobs, with

Whites

being called

back

38%

more

often. Interestingly, these

two

job categories areat opposite

ends of the job quality spectrum, as captured

by

our earnings measure. This suggests

no

monotonous

relationship

between

jobquality

and

the extentofdiscrimination.

The

highest discriminationratio

happens

for administrativesupervisors, thesecond highest job quality category in ourdata. For suchjobs,

Whites

are

64%

more

likelytoget acallback.

In thelast

column

ofTable 8,

we

reporttherace differences in earnings

by

occupation.

These

areagain

computed

from

the 19905percent

Census

for

Boston

and

Chicago.

The

racialgapsinearningaredefined as

the

White-

African

American

differencesin

mean

residual log(annualearnings)

by

occupation,

where

residual

earningsareestimatedasexplained above. Interestingly,theredoesnot

seem

to

be

a

monotonous

association

between

theseestimatedracialgapsin earnings

and

theracial gapsin callbacks. For example,executive

and

managerial positions have the lowest racial

gap

in callbacks but thesecond highest racial

gap

inearnings;

similarly,administrativesupervisorshave the highest racial

gap

incallbacksbut the secondlowestracial

gap

inearnings.

On

the otherhand, sales representatives are associated with thesecond-highest racial

gap

in

callbacks

and

the highest racial

gap

in earnings.

These

results are interesting

and

potentially suggest that

readily available

measures

ofracialdifferences,suchasthe racialgapin earningsused above,

may

notgivea

reliable depiction ofdiscrimination patterns in the

economy.

However, because the occupationaldifferences

incallbackarenot significant,these resultsshould not

be

overstated.

To

summarize,

we

findthatthelevel ofdiscriminationis remarkably uniformacross avariety of

occupa-tions

and

job requirements. This occursdespitethefact that

some

ofthejobs

we

study

arehigher earnings,

such asmanagerial positions orhigh

end

salesjobs.42

4.5

Employer

Characteristics

The

finalset ofpossibledeterminantsofdiscrimination

we

consider areemployer characteristics. Collecting

such

employer

information is not obvious as

most

of this information is not readily available from the

42_{Obviously,wecannot}

ruleout thatdifferentpatternsmightexistwhenonemoveseven higherinthe job qualitydistribution.

employment

ads

we

respondto. In fact,the only pieceofemployer information

we

can

directlycollect

from

the

employment ad

is

whether

ornotthe

employer

explicitlystatesbeing

an

"Equal

Opportunity

Employer."

In severalcases,the

name

ofthe

employer

isnot even

mentioned

inthe

ad

and

theonlypiece ofinformation

we

can

rely

on

isthefax

number

which

applications

must be

submittedto.

We

proceededasfollows. For

employment

adsthat

do

notlistaspecificemployer,

we

usethefax

number

to identifythe

company name

via

web

reverse-lookupservices.

Based

on

the

company

names,

we

usethree

differentdata sources (Onesource Business Browser,

Thomas

Register

and

Dun

and

Bradstreet Million Dollar

Directory, 2001) to track

company

information such as total

employment,

industry

and

ownershipstatus.

Using

this

same

setofdatasources,

we

alsotrytoidentifythespecificzipcodeofthe

company

(or

company

branch) that

resumes

areto

be

sentto. Finally,

we

use theFederal

Procurement

and

Data

Center websiteto

findalistof

companies

thathavefederalcontracts.43

The

racialdifferenceincallback ratesforthe

subsample

where employer

characteristicscould

be determined

isverysimilarin

magnitude

tothatinthe fullsample.

Column

1 of

Table

9reports

summary

employer

characteristics for theavailable data.

Sample

sizes for

eachvariable arereportedinparentheses.

Twenty-nine

percentofallemployersexplicitlystatethat theyare

"Equal

Opportunity

Employers". Eleven percent are federalcontractors and, therefore,

might

facegreater

scrutiny

under

affirmativeaction laws.

The

average

company

sizeis

around

2000

employees

butthereisalot

of variation across firms. Finally, 73 percent ofthe firms are privately held, 15percent are publiclytraded

and

12 percent are non-profits.

The

second

column

ofTable 9presentsthemarginal effectofeach ofthese characteristics

on

discrimina-tion.

As

before, each entry correspondsto a separate probit regression

where

we

regress acallback

dummy

on an

African

American

dummy,

the

employer

characteristicinthat

row and

theinteractionofthe

employer

characteristicwith the African

American

dummy. The

reported coefficient is that

on

the interaction term.

First, neitherthe "Equal

Opportunity Employers"

nor the federal contractors

appear

to discriminateless.

Infact, each ofthese employercategories is associated with

more

discrimination,even

though

theseeffects

are noisily estimated. Second,

we

find

no

effect of

employer

size

on

the degree of discrimination.44 Point

estimates indicatethat publicly traded firms

seem

to discriminate

more,

while not-for-profit organizations

seem

to discriminateless; however, theseeffects areagain noisily estimated.45

Panel

B

of

Table

8

documents

how

the racial

gap

in callbacks varies across

broad

industry categories.

Our

experiment

coversavarietyof industries.

Around

8percentofthe jobsare inmanufacturing, 3 percent

in transportation

and communication,

22 percentare inwholesale

and

retailtrade, 8 percent arein finance

insurance

and

real estate, 27 percent are in business

and

personal services

and

15 percent are in health,

43_This

website (www.fpdc.gov)isaccurateuptoandincludingMarch21,2000.

44_Similar

resultsholdwhen wemeasure employersizeusing atotal salesmeasureratherthananemploymentmeasure.

45

Ofcourse, ourmeasurementofdiscriminationbyfirmtype

may

notbe agoodindicatorofthe fractionofAfricanAmericans

actually employed in these firms. For example, "Equal Opportunity Employers"

may

receive a higher fraction of African

Americanresumes. Theiractual hiring

may

therefore lookdifferent from thatofnon "Equal Opportunity Employers" when

oneconsiders thefullsetofresumes theyreceive.