Du yea, Suzanne; Millán-Quijano, Jaime; Mo ison, Judi h; O iedo Gil, Yani a
Ma cela
Wo king Pape
Measu ing acial bias in employmen se ices in Colombia
IDB Wo king Pape Se ies, No. IDB-WP-01594
P o ided in Coope a ion wi h:
In e -Ame ican De elopmen Bank (IDB), Washing on, DC
Sugges ed Ci a ion: Du yea, Suzanne; Millán-Quijano, Jaime; Mo ison, Judi h; O iedo Gil, Yani a
Ma cela (2024) : Measu ing acial bias in employmen se ices in Colombia, IDB Wo king Pape
Se ies, No. IDB-WP-01594, In e -Ame ican De elopmen Bank (IDB), Washing on, DC,
h ps://doi.o g/10.18235/0012870
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/299410
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/3.0/igo/
Measu ing Racial Bias in Employmen
Se ices in Colombia
Suzanne Du yea
Jaime Millan-Quijano
Judi h Mo ison
Yani a O iedo
WORKING PAPER No IDB-WP-01594
In e -
A
me ican De elopmen Bank
Gende and Di e si y Di ision
Ma ch 2024
Measu ing Racial Bias in Employmen
Se ices in Colombia
Suzanne Du yea (IDB)
Jaime Millan-Quijano (NCID, Uni e sidad de Na a a and CEMR)
Judi h Mo ison (IDB)
Yani a O iedo (Econome ia S.A.)
In e -
A
me ican De elopmen Bank
Gende and Di e si y Di ision
Ma ch 2024
Ca aloging-in-Publica ion da a p o ided by he
In e -Ame ican De elopmen Bank
Felipe He e a Lib a y
Measu ing acial bias in employmen se ices in Colombia / Suzanne Du yea,
Jaime Millan-Quijano, Judi h Mo ison, Yani a O iedo Gil.
p. cm. — (IDB Wo king Pape Se ies ; 1594)
Includes bibliog aphical e e ences.
1. Mino i ies-Economic aspec s-Colombia. 2. Gende mains eaming-
Colombia. 3. Equali y-Colombia. 4. Disc imina ion in employmen -Colombia.
5. Race disc imina ion-Colombia. I. Du yea, Suzanne. II. Millan-Quijano, Jaime.
III. Mo ison, Judi h A. IV. O iedo, Yani a. V. In e -Ame ican De elopmen
Bank. Gende and Di e si y Di ision. VI. Se ies.
IDB-WP-1594
JEL codes: J15, J21, J71.
Keywo ds: Implici s e eo ypes, labo ma ke disc imina ion, de eloping
coun ies.
h p://www.iadb.o g
Copy igh © 2024 In e -Ame ican De elopmen Bank ("IDB"). This wo k is subjec o a C ea i e
Commons license CC BY 3.0 IGO (h ps://c ea i ecommons.o g/licenses/by/3.0/igo/legalcode). The
e ms and condi ions indica ed in he URL link mus be me and he espec i e ecogni ion mus be
g an ed o he IDB.
Fu he o sec ion 8 o he abo e license, any media ion ela ing o dispu es a ising unde such license
shall be conduc ed in acco dance wi h he WIPO Media ion Rules. Any dispu e ela ed o he use o
he wo ks o he IDB ha canno be se led amicably shall be submi ed o a bi a ion pu suan o he
Uni ed Na ions Commission on In e na ional T ade Law (UNCITRAL) ules. The use o he IDB's name
o any pu pose o he han o a ibu ion, and he use o IDB's logo shall be subjec o a sepa a e
w i en license ag eemen be ween he IDB and he use and is no au ho ized as pa o his license.
No e ha he URL link includes e ms and condi ions ha a e an in eg al pa o his license.
The opinions exp essed in his wo k a e hose o he au ho s and do no necessa ily e lec he iews o
he In e -Ame ican De elopmen Bank, i s Boa d o Di ec o s, o he coun ies hey ep esen .
Measu ing Racial Bias in Employmen Se ices in
Colombia.*†
Suzanne Du yea
IDB
Jaime Millán-Quijano
NCID, Uni e sidad de Na a a and CEMR
Judi h Mo ison
IDB
Yani a O iedo
Econome ía S.A.
Ma ch 22, 2024
Abs ac
In his pape , we documen de ac o, implici , and explici acial biases wi hin he public
employmen se ice in Colombia. By combining adminis a i e da a abou job seeke s and
job openings wi h di ec su eys o job counselo s, including a Race Implici Associa ion
Tes , we compu e di e en ypes o acial bias. We ind ha while job counselo s do no
sel - epo biased a i udes agains A o-descendan indi iduals, he majo i y exhibi high
le els o implici bias, which also co ela es s ongly wi h obse ed lowe e e al a es o
A o-descendan s o job openings. In addi ion, we andomly p o ide in o ma ion o job
counselo s abou hei implici bias and es i his in o ma ion changes hei e e al be-
ha io . While we demons a e ha he implici bias o counselo s is a majo con ibu o o
acial gaps in labo ou comes, we do no ind ha p o iding eedback on his unconscious
bias changes hei e e al beha io .
JEL classi ica ion: J15, J21, J71
Keywo ds: Implici s e eo ypes, labo ma ke disc imina ion, de eloping coun ies.
*Con ac au ho s: Suzanne Du yea. [email p o ec ed]. In e -Ame ican De elopmen Bank 1300 New Yo k
A enue, N.W. Washing on, DC 20577. Jaime Millán-Quijano. jmillanq@una .es. ICS, Uni e sidad de Na a a.
Edi icio de Biblio ecas - En ada Es e, 2ª Plan a. Pamplona, Spain 31009. AEA RCT Regis y AEARCTR-0008376
†We a e g a e ul o ASOCAJAS, pa icula ly o Rena a Samaca, who helped us wi h he implemen a ion o
ou s udy, and he Unidad del Se icio de Empleo o p o iding he adminis a i e da a. We also hank he da a
collec ion eam in Econome ía S.A., and Lau a Goyeneche, Lau a Gómez Cely, Ma ía Camila Á ias, Juan José
Rincón, and So ia Vaca o hei assis ance and da a cleaning. The au ho s acknowledge inancial suppo om
he Spanish Minis y o Economy and Compe i i eness G an PID2020-120589RA-I00, and IDB g an RG-E1724.
The opinions exp essed in his publica ion a e hose o he au ho s and do no necessa ily e lec he iews o he
In e -Ame ican De elopmen Bank, i s Boa d o Di ec o s, o he coun ies hey ep esen .
1
1 In oduc ion
Ex ensi e li e a u e exis s on labo ma ke gaps by ace in high-income coun ies, including
he p o ound impac o disc imina ion in he hi ing p ocess (Cha les and Gu yan, 2008, 2011;
Kline e al., 2022). A global e iew o audi s udies inds ha disc imina ion is ubiqui ous
agains non-majo i y ace-e hnic g oups. Ac oss 97 ield expe imen s, whi es ecei e abou
50% mo e callbacks han non-Whi es (Quillian e al., 2019).
In he case o La in Ame ica (LAC), app oxima ely 24% o he o al popula ion iden i ies as
A o-descendan . While employmen a es do no di e subs an ially ac oss ace, job quali y
is lowe o A o-descendan s. A gap by ace in o mal employmen o o e 10 pe cen age
poin s is obse ed o B azil, Colombia, and U uguay, wi h gaps pe sis ing o e he las 15
yea s (A aujo e al., 2022). Howe e , in con as o high-income coun ies, ew s udies exis
ha a emp o ca e ully measu e labo ma ke disc imina ion by ace o e hnici y, mainly due
o he lack o da a (Ñopo e al., 2010; Ñopo, 2012).
This pape p esen s di e en es ima es o acial bias in he publicly inanced employmen sys-
em in Colombia. To achie e his, we collec ed ex ensi e da a om a ious sou ces o measu e
de ac o, implici , and explici acial bias o job counselo s wi hin he sys em.1We ga he ed
adminis a i e da a o o e 300,000 job applican s and 70,000 job pos s om ac oss he coun y
o e one yea . This da a allowed us o obse e who was e e ed o which job, enabling us o
compa e e e al a es by ace. We combined he adminis a i e da a wi h de ailed in e iews
wi h 349 job counselo s in 81 job cen e s. Du ing he in e iews, we conduc ed a Race Implici
Associa ion Tes (IAT), ollowing G eenwald e al. (2009). Addi ionally, we asked pa icipan s
abou hei explici p e e ences owa d A o-descendan wo ke s. Finally, ollowing Alesina
e al. (2018), we andomized eedback on he IAT esul s ac oss job cen e s o es whe he
in o ma ion abou implici , some imes unconscious, bias agains / owa ds A o-descendan s
could in luence job e e al beha io wi hin he employmen se ices.
We ind ha job counselo s a e 15% less likely o e e A o-descendan s o an opening
han non-A o-descendan s. Wi h he use o he IAT, we also show ha wo- hi ds o job
counselo s e ealed s ong p e e ences o whi es o e blacks. Ne e heless, when asked
abou pe cep ions abou whi e and black indi iduals, job counselo s do no di ec ly e eal
p e e ences owa ds whi e indi iduals. By combining he adminis a i e da a wi h he su ey
o job counselo s, we ind ha he di e ence in he o wa ding a es be ween A o- and non-
A o-descendan s co ela es wi h he implici bias o job counselo s. In ac , A o-descendan s
a e less likely o be e e ed o an opening only i he median job counselo o a gi en job cen e
has a high le el o bias agains A o-descendan s. Finally, by using he andom eedback o
he IAT sco e, we ind ha in o ma ion abou hei own implici bias does no change job
1De ac o bias e e s o p e e en ial ea men as measu ed by ealized beha io s. Implici bias, as measu ed
by he IAT is unde s ood o be o an unconscious o unknown na u e. Explici bias is measu ed by p e e ences
indi iduals a e willing o epo .
2
counselo s’ beha io .
We con ibu e o se e al s ands o he li e a u e. Fi s ly, we show he dis ibu ion and co -
ela ion o de ac o, implici , and explici acial bias among job counselo s o he Public Em-
ploymen Se ices in Colombia, a middle-income coun y. We a e no he i s aiming o
unde s and he ole o ace in Colombia’s labo ma ke s. P e ious wo ks, ypically ocus
on la ge ci ies, include s udies in Cali, he ci y wi h he la ges numbe o A o-descendan s
(Diosa Ramí ez, 2015; Ma ulanda and Rod íguez, 2014; Paz Mo eno and Delgado Co ez, 2017;
He edia e al., 2010), and Ca agena, he ci y wi h he highes p opo ion o A o-descendan s
(Rome o-P ie o, 2007). The e a e also some s udies o he wo bigges Colombian ci ies, Bo-
go á (Ga a i o e al., 2013), and Medellín (Ál a ez Ossa e al., 2014; Ga cía Sánchez, 2010).
Ou esea ch is he i s o analyze acial di e ences a he na ional le el.
Fu he mo e, we joined ecen e o s o comp ehend acial disc imina ion in de eloping coun-
ies, pa icula ly in LAC. Fo ins ance, a ecen s udy in B azil ound ha employe p e e -
ences o whi e wo ke s explain app oxima ely 6% o 7% o he acial wage gap (Ge a d e al.,
2021). In ano he s udy, using da a om B azilian i ms in he o mal sec o , (Mille and
Schmu e, 2023) inds s ong pa e ns o co- acial hi ing. New i ms ha a e disp opo ion-
a ely comp ised o whi e employees will ini ially end o hi e whi e wo ke s, al hough wi h
pe sis en g ow h he hi ing becomes mo e di e se. Addi ionally, in Mexico, A ceo-Gomez
and Campos-Vazquez (2014) shows ha indigenous-looking-women ecei ed ewe in e iew
eques s han mes izo- o caucasian-looking women.2
Secondly, ou s udy adds o ecen wo ks measu ing implici bias agains di e en mino i y
g oups and i s co ela ion wi h disc imina o y beha io . Glo e e al. (2017) examines he
dynamics be ween manage s wi h high le els o unconscious bias agains immig an s and
immig an wo ke s in F ance. Ca lana (2019) measu es he implici gende bias o eache s
in I aly and shows ha he gende gap in ma h pe o mance inc eases when s uden s a e as-
signed o ma h eache s wi h s ong gende s e eo ypes. Addi ionally, s udies ha e also used
he in o ma ion on unconscious bias as a low-cos in e en ion o p o oke beha io change.
Alesina e al. (2018) ound ha I alian eache s in o med abou hei implici bias agains
immig an s changed hei ela i e g ades wi h espec o immig an and non-immig an s u-
den s. Teache s who had epo ed explici bias agains immig an s did no change how hey
g aded he s uden s. This sugges s ha being in o med abou unconscious bias can help
change beha io ela ed o di e en ial ea men , o example, by job counselo s. O he s ud-
ies include Alan e al. (2021) on eache s’ e hnic p ejudice in Tu key, and Co no e al. (2022)
on acial s e eo ypes in s uden do ms in Sou h A ica.
Finally, ou esea ch con ibu es o he li e a u e on unde s anding he ole o in e media-
ion se ices in educing labo ma ke dis o ions and inc easing wel a e o mino i y g oups.
2O he s udies include Sala di (2016); Co nwell e al. (2017); Hi a a and Soa es (2020) in B azil, Canelas and
Gisselquis (2018) in Gua emala, and Ga a i o e al. (2013) in Colombia, and c oss-coun y analyses such as Chong
e al. (2008); Woo-Mo a (2022)
3
Recen li e a u e has ound posi i e e ec s o labo ma ke in e media ion se ices on he
p obabili y o employmen o job seeke s (C épon and Van Den Be g, 2016), wi h job sea ch
assis ance p og ams mo e success ul o indi iduals wi h less access o con ibu o y bene-
i s (Ca d e al., 2018). Fu he mo e, he ole o public employmen se ices in comba ing
disc imina ion has been unde sco ed by he In e na ional Labo O ganiza ion (ILO). Indeed,
ILO Con en ion C111, conside ed a co e labo s anda d and a i ied by all coun ies in La in
Ame ica and he Ca ibbean, add esses disc imina ion in employmen and es ablishes a se ies
o ac ions o Go e nmen s, o p omo e equali y and elimina e disc imina ion in oca ional
aining and placemen se ices (In e na ional Labo O ganiza ion (ILO), 1958). Public em-
ploymen se ices a e c ucial o ollow hese ecommenda ions. Howe e , he ole o acial
bias in labo ma ke in e media ion se ices has ecei ed li le a en ion. A no able excep ion
is a ecen s udy in Swi ze land ha ound ha a es o con ac by ec ui e s in he Swiss pub-
lic employmen se ice a e 4-19% lowe o indi iduals in mino i y e hnic g oups (Hanga ne
e al., 2021). Ano he s udy ha examined he ole o e hnici y in labo ma ke in e media ion
se ices in Pe u, in e ed he e hnici y o he job applican and did no ind biased ea men
on he pa o job counselo s (Mo eno e al., 2012). Despi e inding null e ec s o IAT in o -
ma ion on counselo s’ beha io , ou expe imen al amewo k p o ides aluable in o ma ion
on he limi s o e ealing unknown biases in changing a i udes and ac ions.
This pape is o ganized as ollows. A e his in oduc ion he nex sec ion desc ibes he public
employmen sys em in Colombia, ollowed by o he da a in Sec ion 3. Sec ion 4 p esen s ou
di e en es ima es o acial bias and hei co ela ions. Sec ion 5 summa izes he esul s o
ou expe imen al in e en ion and Sec ion 6 concludes.
2 Ins i u ional backg ound – A o-descendan s in he Colombian
Labo Ma ke and Public Employmen Se ices (PES).
Acco ding o he la es epo by he Colombian Na ional S a is ics Depa men (DANE), di -
e ences be ween whi es and A o-descendan s in he labo ma ke a e mo e ela ed o he
quali y o jobs han o pa icipa ion i sel (DANE, 2023).3Using da a om a na ional house-
hold su ey he epo ound ha he occupa ion a es o whi es and A o-descendan s a e
simila , 89% and 87%, espec i ely. Howe e , 70.8% o A o-descendan wo ke s a e in he
in o mali y, while 55% o whi e wo ke s ha e in o mal jobs. Also, A o-descendan s a e less
p e alen among p o essional and di ec o ial jobs. These di e ences help o explain why he
a e age household mon hly income o an A o-descendan is only 65% o he a e age income
o he household o a whi e indi idual (USD 220 and USD 334, o A o-descendan s and
whi es, espec i ely).
3In he epo whi es a e desc ibed as non-E hical, which e e s o non-A o-descendan and non-Indigenous
indi iduals.
4
Unde his amewo k, he Colombia Public Employmen Se ice (PES) was es ablished in
2013 o educe he high unemploymen a es in he coun y and enhance he quali y o em-
ploymen .4The PES p o ides labo ma ke in e media ion se ices h ough employmen
agencies o job cen e s. The se ice is o e ed by local go e nmen s and non-p o i p i a e
ins i u ions called Cajas de Compensación Familia (CCF), which a e unded by con ibu ions
om i ms.
Job counselo s in employmen agencies a e esponsible o egis e ing and guiding job seeke s
and employe s. They also unde ake p e-selec ion and e e al o candida es o job pos ings.
In e ac ions among applican s, i ms, and job counselo s a e mos ly online, and all CVs and
job pos ings a e uploaded o a common online pla o m. Howe e , job seeke s and i ms can
also a end hei local o ices o access di ec ad ice om a counselo (Núñez Méndez and
Oso io, 2015).
Once job seeke s and po en ial employe s egis e on he pla o m, job counselo s e alua e
each p o ile and pos ing o iden i y s eng hs and weaknesses o CVs o di e en job lis ings.
They hen ma ch job seeke s wi h job lis ings o minimize in o ma ion gaps. Ne e heless,
applican s and i ms can also sea ch and apply o jobs o in i e job seeke s o in e iews.
Addi ionally, job counselo s can link job seeke s o wo kshops (e.g., mo i a ion, so skills,
suppo o sel -employmen ), job ai s, and o he e en s ha could enhance hei employ-
abili y. Employe s also ecei e ad ice o acili a e and op imize hei ec ui men p ocess,
p o iding in o ma ion on bes p ac ices.
3 Da a
To documen di e en es ima es o bias agains A o-descendan s in he con ex o labo ma -
ke in e media ion wi hin PES we use h ee di e en da a sou ces.5
3.1 PES adminis a i e eco ds
Ou i s sou ce o da a comes om each CV and job acancy pos ed a PES by seeke s/ i ms
a ilia ed o a CCF, which a e p i a e p o ide s o social se ices o i ms and wo ke s, unded
by public esou ces collec ed h ough he pay oll o wo ke s (Minis e io de T abajo and ILO,
2014). We use all CVs om adul s 20 o 65 yea s old, and lis ings uploaded om Ma ch
1s , 2021 o Feb ua y 28 h, 2022. Fo each CV we obse e demog aphic in o ma ion such as
gende , age, educa ion, and sel -iden i ied ace. F om each acancy, we obse e wage o e ,
con ac ype, sec o , and each CV ha was e e ed and hi ed. In o al, we obse e 348,919
4F om 2001 o 2018 he a e age unemploymen a e in Colombia was 11%, 3 pp highe han he a e age o LAC
(Ramos and Ál a ez Ga cía, 2020).
5Figu e E.1 in Appendix E summa izes he s udy imeline including all da a collec ions and in e en ions.
5
Figu e 2: Di e en ial es ima ed p obabili y o e e and hi e an A o-descendan CV s a
non-A o-descendan CV by he median job cen e ’s IAT sco e, and o he job counselo s’
cha ac e is ics.
(A) By job cen e ’s Race IAT Median
(B) P obabili y o e e ing a CV (A o - non-A o) by job cen e ’s
counselo s’ IAT and quali y
(C) P obabili y o hi ing a e e ed CV (A o - non-A o) by job cen e ’s
counselo s’ IAT and quali y
No es: Panel A shows he es ima ed di e en ial p obabili ies ollowing es ima es o he equa ion Re iJ =α0+α1A oi+
α2g(IATJ) + α3A oi×g(IATJ) + α5A oi×Xi+Mi+CCFJ+RJ +µiJ . Whe e Re iJ =1 i he CV i, managed by job cen e J,
in mon h was e e ed o a job pos . A oi=1 i he indi idual iiden i ies himsel as A o-descendan , and g(IATJ)is a unc ion
o he IAT sco es o he job counselo s a job cen e J, o example, he median IAT o he counselo s o a gi en job cen e . Xiis
a ec o o cha ac e is ics o he job-seeke , Mia e municipali y ixed e ec s, CCFia e CCF ixed e ec s, RJ a e egion-mon h
ixed e ec s, and µiJ is a andom unobse able e o e m. The a ea e lec s a 95% con idence in e al. The e ical dashed
lines indica e he c i ical h esholds sugges ed by G eenwald e al. (2009). Panels B and C epo he di e en ial p obabili ies
ollowing es ima es o he equa ion YiJ =α0+α1A oi+α2DJ+α3A oi×DJ+α5A oi×Xi+Mi+CCFJ+RJ +µiJ , whe e
Yis e e ing (Panel B) and hi ing condi ional on being e e ed (Panel C), Dhas h ee de ini ions, and akes alue o 1 i (i) job
cen e ’s IAT is abo e 0.65, (ii) job cen e has a leas one counselo wi h 5 yea s o mo e o expe ience, and (iii) job cen e has a
leas one counselo wi h pos g adua e educa ion. Dashed lines ep esen a 95% con idence in e al.
12
o non-A o-descendan CVs by job counselo ’s IAT and quali y (a he job cen e le el). The
igu e shows ha job cen e s wi h high bias a e he ones whe e A o-descendan s ha e a
lowe p obabili y o be e e ed. When we look a job counselo quali y, job cen e s wi h less
educa ed and less expe ienced counselo s a e he ones ha show a nega i e bias owa d A o-
descendan CVs. Addi ionally, Panel C shows he di e en ial p obabili y o hi ing o CVs
ha we e e e ed by job counselo s. In he igu e, we can see ha he le el o implici bias
o counselo ’s quali y does no c ea e a ia ion in he hi ing bias agains A o-descendan s.
The e o e, ou e idence does no suppo he idea ha he disc imina ion o job counselo s
agains A o-descendan s comes om an e o o in e nalize he bias o he i ms ha a e
hi ing. Wha is mo e, once a CV has been e e ed, job counselo s’ cha ac e is ics seem o
play no ole in a ec ing i ms disc imina ion agains A o-descendan wo ke s.
5 E ec o in o ma ion abou implici bias on ace bias
Finally, we use he andom alloca ion o in o ma ion abou IAT esul s o es ima e he impac
o such in o ma ion on A o-descendan e e al and hi ing a es. We combine da a om
UAESPE and ou igne e s udy. Un o una ely, all ou e idence sugges s ha knowing abou
hei implici associa ion bias (IAB), does no change he e e ing beha io o counselo s.
Be o e discussing ou esul s, i is wo h no ing Table C.1 in Appendix C summa izes he
balanced es wi h espec o job counselo s and CV cha ac e is ics. The i s piece o e idence
comes om Figu e 3, which shows ha o bo h A o- and non-A o-descendan job seeke s
he e e al a es did no change a e he IAT eedback o he ea men g oup inished.
We also es ima e o mally he e ec by using da a om UAESPE we can use he iming o
ou in e en ion o es ima e he e ec o in o ma ion on e e al a es using a di e ence-in-
di e ence s a egy aking ad an age o he andom alloca ion o he IAT eedback. The de ails
o hese es ima ions a e in Appendix C. Summa izing, in line wi h Figu e 3, we do no ind
any changes in e e al beha io induced by he bias in o ma ion.
6 Conclusion
Labo ma ke disc imina ion by ace is a opic o g ea in e es wi h a la ge knowledge gap
in low- and middle-income economies, due mos ly o a lack o da a. In his pape we docu-
men ed s ong de ac o and implici biases agains A o-descendan job seeke s in he Colom-
bian public employmen sys em. While he IAT eedback did no ha e a s ong impac , di -
e en om he case o Alesina e al. (2018) in e en ion in I aly, ou s udy highligh s a key
pa o he labo ma ke whe e acial bias plays a ole in con ibu ing o gaps in employmen
ou comes o A o-descendan s in Colombia. Be e unde s anding his bias, a he le el o
in e media ion counselo s o a he le el o employe s, can imp o e he design o u u e in-
13
Figu e 3: Re e al a e by mon h, ace, and expe imen al a m
(A) A o-descendan applican s
(B) non-A o-descendan applican s
No es: Dashed lines ep esen a 95% con idence in e al. S anda d e o s clus e ed a he job cen e le el. The pe iod o s udy
implemen a ion be ween g ay e ical lines ( om i s con ac o CCFs o IAT eedback o he ea men job counselo s).
14
e en ions ha aim o p e en o educe his bias. Mo eo e , he e is a po en ial o a b ie
in e en ion o ha e sus ainable impac s as shown by Mille (2017) who inds ha sho - un
p og ams can ha e long- un e ec s on he acial composi ion o i ms, such as in e en ions
ha induce pe sis en changes in ec ui men policies.
15
Re e ences
Alan, S., E. Duysak, E. Kubilay,and I. Mumcu (2021): “Social Exclusion and E hnic Seg-
ega ion in Schools: The Role o Teache ’s E hnic P ejudice,” The Re iew o Economics and
S a is ics, 1–45.
Alesina, A., M. Ca lana, E. LaFe a a,and P. Pino i (2018): “Re ealing s e eo ypes:
E idence om immig an s in schools,” Tech. ep., Na ional Bu eau o Economic Resea ch.
Ál a ez Ossa, L. e al. (2014): “Muje es, pob es y neg as, iple disc iminación: una mi ada a
las acciones a i ma i as pa a el acceso al me cado labo al en condiciones de abajo decen e
en Medellín (2001-2011),” .
A aujo, M. C., S. Du yea,and L. E che e y (2022): “Gende and Di e si y Sec o F ame-
wo k Documen ,” Tech. ep., In e -Ame ican De elopmen Bank (IDB), h ps://www.iadb.
o g/documen .c m?id=EZSHARE-1011213690-92.
A ceo-Gomez, E. O. and R. M. Campos-Vazquez (2014): “Race and ma iage in he labo
ma ke : A disc imina ion co espondence s udy in a de eloping coun y,” Ame ican Eco-
nomic Re iew, 104, 376–80.
Blundell, R. W. and J. L. Powell (2004): “Endogenei y in semipa ame ic bina y esponse
models,” The Re iew o Economic S udies, 71, 655–679.
Canelas, C. and R. M. Gisselquis (2018): “Human capi al, labou ma ke ou comes, and
ho izon al inequali y in Gua emala,” Ox o d De elopmen S udies, 46, 378–397.
Ca d, D., J. Klu e,and A. Webe (2018): “Wha wo ks? A me a analysis o ecen ac i e
labo ma ke p og am e alua ions,” Jou nal o he Eu opean Economic Associa ion, 16, 894–931.
Ca lana, M. (2019): “Implici s e eo ypes: E idence om eache s’ gende bias,” The Qua -
e ly Jou nal o Economics, 134, 1163–1224.
Cha les, K. K. and J. Gu yan (2008): “P ejudice and wages: an empi ical assessmen o
Becke ’s The Economics o Disc imina ion,” Jou nal o poli ical economy, 116, 773–809.
——— (2011): “S udying Disc imina ion: Fundamen al Challenges and Recen P og ess,”
Annual Re iew o Economics, 3, 479–511.
Chong, A., H. Ñopo, L. Ronconi,and M. U quiola (2008): “The mys e y o disc imina ion
in la in ame ica [wi h commen s],” Economía, 8, 79–115.
Co no, L., E. LaFe a a,and J. Bu ns (2022): “In e ac ion, s e eo ypes, and pe o mance:
E idence om Sou h A ica,” Ame ican Economic Re iew, 112, 3848–75.
16
Co nwell, C., J. Ri e a,and I. M. Schmu e (2017): “Wage disc imina ion when iden i y
is subjec i e e idence om changes in employe - epo ed ace,” Jou nal o Human Resou ces,
52, 719–755.
Cou s, A. (2020): “Racial bias a ound he wo ld,” .
C épon, B. and G. J. Van Den Be g (2016): “Ac i e labo ma ke policies,” Annual Re iew o
Economics, 8, 521–546.
DANE (2023): “Me cado Labo al de los G upos É nico-Raciales En Colombia an es y
después de los con inamien os po el Co id-19,” Tech. Rep. 2, Depa amen o Ad-
minis a i o Nacional de Es adís ica, h ps://www.dane.go .co/ iles/in es igaciones/
no as-es adis icas-casen/ab -2023-Me cado-Labo al-E nico-Raciales.pd .
Diosa Ramí ez, J. (2015): “¿ Causan di e enciación sala ial la ca ac e ís ica é nica, el géne o y
la ubicación espacial en el me cado labo al de Cali?” Tech. ep.
Ga a i o, C. A. R., J. C. Cá denas, J. D. O. M., and S. Villamiza (2013): La disc iminación
acial en el abajo: Un es udio expe imen al en Bogo á, Dejus icia.
Ga cía Sánchez, A. (2010): “Espacialidades del des ie o y la e-exis encia: A odescendi-
en es des e ados en Medellín, Colombia,” .
Ge a d, F., L. Lagos, E. Se e nini,and D. Ca d (2021): “Asso a i e ma ching o exclu-
siona y hi ing? he impac o employmen and pay policies on acial wage di e ences in
B azil,” Ame ican Economic Re iew, 111, 3418–57.
Glo e , D., A. Pallais,and W. Pa ien e (2017): “Disc imina ion as a Sel -Ful illing
P ophecy: E idence F om F ench G oce y S o es,” The Qua e ly Jou nal o Economics, 1219–
1260.
G eenwald, A. G., T. A. Poehlman, E. L. Uhlmann,and M. R. Banaji (2009): “Unde s and-
ing and using he Implici Associa ion Tes : III. Me a-analysis o p edic i e alidi y.” Jou nal
o Pe sonali y and Social Psychology, 97, 17–41.
Hanga ne , D., D. Kopp,and M. Siegen hale (2021): “Moni o ing hi ing disc imina ion
h ough online ec ui men pla o ms,” Na u e, 589, 572–576.
He edia, J. A. C., C. H. O. Que edo, F. U. Gi aldo, C. A. V. López, N. J. S. Al a ado,
H. Bonilla, D. M. Co aza , D. I. Oso io,and N. Páez (2010): “In o me inal del p oyec o:
Desigualdad de opo unidades educa i as y segmen ación labo al en la población de 15 a
29 años de b asil y colombia, según au oclasi icación acial,” .
Hi a a, G. and R. R. Soa es (2020): “Compe i ion and he acial wage gap: E idence om
B azil,” Jou nal o De elopmen Economics, 146, 102519.
In e na ional Labo O ganiza ion (ILO) (1958): “C111 Disc imina ion (Employ-
men and Occupa ion) Con en ion, Gene a,” h ps://www.ilo.o g/dyn/no mlex/en/ ?p=
NORMLEXPUB:12100:0::NO::P12100_Ilo_Code:C111.
17
Kline, P., E. K. Rose,and C. R. Wal e s (2022): “Sys emic disc imina ion among la ge US
employe s,” The Qua e ly Jou nal o Economics, 137, 1963–2036.
Maccini, S. and D. Yang (2009): “Unde he wea he : Heal h, schooling, and economic
consequences o ea ly-li e ain all,” Ame ican Economic Re iew, 99, 1006–1026.
Ma ele o, L. J. (2012): “Educa ional inequali y by ace in B azil, 1982–2007: s uc u al
changes and shi s in acial classi ica ion,” Demog aphy, 49, 337–358.
Ma ulanda, L. P. and J. J. M. Rod íguez (2014): “La calidad del empleo en la población
a odescendien e colombiana: una ap oximación desde la ubicación geog á ica de las co-
munas,” Re is a de Economía del Rosa io, 17, 315–347.
Mille , C. (2017): “The Pe sis en E ec o Tempo a y A i ma i e Ac ion,” Ame ican Economic
Jou nal: Applied Economics, 152–190.
Mille , C. and I. Schmu e (2023): “The Dynamic E ec s o Co-Racial Hi ing,” Tech. ep.
Minis e io de T abajo and ILO (2014): “Lib o Blanco del Sis ema de Subsidio Familia ,”
Tech. ep., In e na ional Labo O ganiza ion.
Mo eno, M., H. Ñopo, J. Saa ed a,and M. To e o (2012): “De ec ing gende and acial
disc imina ion in hi ing h ough moni o ing in e media ion se ices: he case o selec ed
occupa ions in Me opoli an Lima, Pe u,” Wo ld De elopmen , 40, 315–328.
Ñopo, H. (2012): New cen u y, old dispa i ies: Gende and e hnic ea nings gaps in La in Ame ica
and he Ca ibbean, Wo ld Bank Publica ions.
Ñopo, H., J. A al,and N. Winde (2010): “New cen u y, old dispa i ies: Gende and e hnic
wage gaps in La in Ame ica,” IZA discussion pape .
Nosek, B. A., A. G. G eenwald,and M. R. Banaji (2007): “The Implici Associa ion Tes a
age 7: A me hodological and concep ual e iew.” .
Núñez Méndez, J. and A. F. Oso io (2015): “E aluación Ins i ucional y de Ges ión del Se i-
cio Público de Empleo,” Tech. ep., Fedesa ollo.
Paz Mo eno, D. P. and S. J. Delgado Co ez (2017): “Inclusión labo al y di e sidad de la
población a odescendien e: los desa íos del sec o emp esa ial en San iago de Cali,” .
Quillian, L., A. Hea h, D. Page , A. H. Mid bøen, F. Fleischmann,and O. Hexel (2019):
“Do some coun ies disc imina e mo e han o he s? E idence om 97 ield o acial dis-
c imina ion in hi ing,” Sociological Science, 6, 467–496.
Ramos, C. I. and R. D. Ál a ez Ga cía (2020): “La asa na u al de desempleo en Colombia
2001-2018: e olución y es imaciones,” En amado, 16, 76–93.
18
Rome o-P ie o, J. E. (2007): “¿ Disc iminación labo al o capi al humano?: de e minan es del
ing eso labo al de los a oca agene os,” Documen os de T abajo Sob e Economía Regional y
U bana; No. 98.
Sala di, P. (2016): “The E olu ion o Gende and Racial Occupa ional Seg ega ion Ac oss
Fo mal and Non-Fo mal Labo Ma ke s in B azil, 1987 o 2006,” Re iew o Income and Weal h,
62, S68–S89.
Woo-Mo a, G. (2022): “Un eiling he Cosmic Race: Skin one and e hno acial inequali ies in
La in Ame ica,” Tech. ep., Wo king pape , Pa is School o Economics.
19
A IAT esou ces
• Con en : Implici Race Associa ion Tes (IAT) – In e na ional S anda d Ve sion
• Sou ce: h ps://implici .ha a d.edu/implici /S udy? id=-1
• Wo ds:
Ca ego y i ems
A. In English:
Good Joy, Lo e, Peace, Wonde ul, Pleasu e, Glo ious, Laugh e , Happy
Bad Agony, Te ible, Ho ible, Nas y, E il, Te ible, Failu e, Pain
B. In Spanish
Good Aleg ía, Amo , Paz, Ma a illoso, Place , Glo ioso, Risa, Feliz
Bad Agonía, Te ible, Ho ible, Desag adable, Mal ado, Malísimo, Fallo, Dolo
• Pic u es:
–Black aces:
–Whi e aces:
20
B IAT esul s eedback
The eedback o job counselo s happened in wo ins ances. Fi s we sen an e-mail ( ollowing
Alesina e al., 2018), including he exac nume ical alue o he es , he ca ego y he a e placed
acco ding o G eenwald e al. (2009). We complemen he e-mail wi h a phone call in o de o
cla i y doub s. In his sec ion we show he e-mail empla e and he sc ip ollowed by ou call
cen e . Gi en ha bo h, e-mail and calls, we e pe sonalized, in his sec ion you will ind XXX
whe e he sco e was placed. In addi ion, in he email, we place an "X" in he espec i e box in
he scale igu e, and gi e one o he ollowing PERSONALIZED MESSAGE, depending on
he job counselo ’s esul :
• High p e e ence owa ds whi e people: You ha e a high unconscious endency o as-
socia e posi i e hings wi h whi e people and nega i e hings wi h A o-descendan
people. This could lead o a s ong p e e ence o whi e people o e A o-descendan
people in di e en a eas o li e ( o example, wo k, pe sonal, among o he s).
• In e media e p e e ence owa ds whi e people: You ha e an in e media e and uncon-
scious endency o associa e posi i e hings wi h whi e people and nega i e hings wi h
A o-descendan people. This could lead o a p e e ence o whi e people o e A o-
descendan people in di e en a eas o li e ( o example, wo k, pe sonal, among o he s).
• Low p e e ence owa ds whi e people: You ha e a low unconscious endency o associa e
posi i e hings wi h whi e people and nega i e hings wi h A o-descendan people.
This p e e ence could lead o a sligh p e e ence o whi e people o e A o-descendan
people in di e en a eas o li e ( o example, wo k, pe sonal, among o he s).
• Neu al: You a e neu al o do no ha e unconscious endencies o associa e posi i e
o nega i e hings wi h whi e o A o-descendan people. This could lead o no ha -
ing p e e ences be ween A o-descendan o whi e people in di e en a eas o li e ( o
example in wo k, pe sonal, among o he s).
• Low p e e ence owa ds A o-descen people: You ha e a low unconscious endency o
associa e posi i e hings wi h A o-descendan people and nega i e hings wi h whi e
people. This could lead o a sligh p e e ence o A o-descendan people o e whi e
people in di e en a eas o li e ( o example, wo k, pe sonal, among o he s).
• In e media e p e e ence owa ds A o-descen people: You ha e an in e media e and un-
conscious endency o associa e posi i e hings wi h A o-descendan people and neg-
a i e hings wi h whi e people. This could lead o a p e e ence o A o-descendan
people o e whi e people in di e en a eas o li e ( o example, wo k, pe sonal, among
o he s).
• High p e e ence owa ds A o-descen people: You ha e a high unconscious endency
o associa e posi i e hings wi h A o-descendan people and nega i e hings wi h whi e
21
Table D.3: Fi s s age es ima es o di e en de ini ions o he ins umen
Ou come: P obabili y o sel -iden i y A o-descendan
Re e ence popula ion: All adul s Economic ac i e Employed
adul s adul s
Age cell size (yea s): 5 10 5 10 5 10
(1) (2) (3) (4) (5) (6)
Panel A. All applican s
A o-descendan p opo ion by cell 0.090∗∗ 0.091∗∗ 0.091∗∗ 0.091∗∗ 0.092∗∗ 0.092∗∗
(0.022) (0.027) (0.022) (0.027) (0.022) (0.027)
F-Tes 16.610 11.124 17.003 11.319 16.857 11.209
N 156053 156053 156053 156053 156053 156053
Panel B. Only o wa ded applican s
A o-descendan p opo ion by cell 0.104∗∗ 0.106∗∗ 0.104∗∗ 0.104∗∗ 0.108∗∗ 0.108∗∗
(0.036) (0.040) (0.035) (0.039) (0.036) (0.041)
F-Tes 8.585 6.947 8.649 6.915 8.853 7.067
N 59905 59905 59905 59905 59905 59905
No es: S anda d e o s clus e ed a municipali y-gende -age cell le el in pa en heses. +p<0.1, ∗p<0.05, ∗ ∗ p<0.01. Only
includes CVs ha en e ed he sys em be ween Ma ch 1s and Augus 5 h, 2021. All es ima ions con ol o age, expe ience, and
educa ional le el, and include municipali y and CCF-mon h o egis y ixed e ec s. The F-Tes ep esen s he F s a is ic om
he null hypo hesis ha ϕ1=.
28
Table D.4: E ec o A o-descendan CV on he p obabili y o being o wa ded by Sec o and
job- ype. P e-s udy
By sec o
All Se ices Comme ce Manu ac u e O he s
(1) (2) (3) (4) (5)
Reduced o m es ima es -0.220∗∗ -0.237∗∗ -0.017∗∗ -0.027∗∗ 0.033∗∗
(0.035) (0.034) (0.005) (0.005) (0.010)
IV es ima es -2.449∗∗ -2.637∗∗ -0.194∗∗ -0.297∗∗ 0.368∗
(0.756) (0.780) (0.071) (0.109) (0.152)
Mean non-a o 0.280 0.197 0.036 0.040 0.087
FS-F Fi s S age F-Tes 16.610 16.610 16.610 16.610 16.610
R2-0.459 -0.697 -0.001 -0.024 -0.013
Obse a ions 156053 156053 156053 156053 156053
By job ype
All By wage By con ac ype
≤1 MMLW >1 MMLW Sho e m Long e m
(1) (6) (7) (8) (9)
Reduced o m es ima es -0.220∗∗ -0.032+-0.215∗∗ -0.090∗∗ -0.158∗∗
(0.035) (0.018) (0.035) (0.019) (0.036)
IV es ima es -2.449∗∗ -0.360 -2.386∗∗ -0.996∗∗ -1.759∗∗
(0.756) (0.220) (0.745) (0.343) (0.613)
Mean non-a o 0.280 0.176 0.172 0.184 0.171
FS-F Fi s S age F-Tes 16.610 16.610 16.610 16.610 16.610
R2-0.459 0.032 -0.634 -0.084 -0.321
Obse a ions 156053 156053 156053 156053 156053
No es: S anda d e o s clus e ed a municipali y-gende -age cell le el in pa en heses. +p<0.1, ∗p<0.05, ∗ ∗ p<0.01. Only
includes CVs ha en e ed he sys em be ween Ma ch 1s and Augus 5 h, 2021. All es ima ions con ol o age, expe ience, and
educa ional le el, and include municipali y and CCF-mon h o egis y ixed e ec s. A o-descendan iden i ica ion ins umen ed
wi h he p opo ion o a o-descendan popula ion a he municipali y-gende -age cell (5 yea s) a he 2018 census – As in Table
D.3 Panel A column 1. The F-Tes ep esen s he F s a is ic om he null hypo hesis ha ϕ1=.
29
Table D.5: IV-P obi es ima ed e ec o A o-descendan CV on he p obabili y o being
o wa ded by Sec o and job- ype.
Es ima ed di e ence
P(A o-descendan ) - P(Non-A o)
Coe icien s.e. Non-A o mean Obse a ions
(1) (2) (3) (4)
All -0.054 (0.008)∗∗ 0.283 154441
By Wage
Less han 1 MMLW -0.023 (0.007)∗∗ 0.179 153337
Mo e han 1 MMLW -0.000 (0.008) 0.174 153039
By Con ac ype
Sho e m -0.014 (0.008)+0.187 153297
Long e m -0.009 (0.007) 0.174 152888
By Sec o
Se ices -0.006 (0.008) 0.200 153751
Comme ce -0.022 (0.003)∗∗ 0.040 141501
Manu ac u e 0.016 (0.006)∗∗ 0.043 144929
O he s 0.023 (0.006)∗∗ 0.090 150569
No es: Boo s ap s anda d e o s a e 1000 epe i ions in pa en heses. +p<0.1, ∗p<0.05, ∗ ∗ p<0.01. Only includes CVs
ha en e ed he sys em be ween Ma ch 1s and Augus 5 h, 2021. All es ima ions con ol o age, expe ience, and educa ional
le el, and include municipali y and CCF-mon h o egis y ixed e ec s. A o-descendan iden i ica ion ins umen ed wi h
he p opo ion o a o-descendan popula ion a he municipali y-gende -age cell (5 yea s) a he 2018 census. To accoun o
endogenous epo ing we use a con ol unc ion app oach by in oducing he e o e m om he i s s age es ima ions in o he
P obi es ima ion.
30
E Addi ional ables and igu es
Figu e E.1: S udy imeline
2021 · · · · · ·• Feb ua y
(1s ) UAESPE da a i s en y.
· · · · · ·•
Augus
(5 h) In i a ion o CCFs.
(27 h) Baseline su ey s a s.
· · · · · ·•
Sep embe
(13 h) Baseline su ey ends.
(20 h) Random alloca ion o job cen e s.
(23 h o 29 h) IAT sco e eedback o ea ed job counselo s.
· · · · · ·• Oc obe
(12 h) Vigne e s udy s a s.
· · · · · ·• No embe
(9 h) Vigne e s udy ends.
2022 · · · · · ·•
Feb ua y
(1s ) UAESPE da a las en y.
(21s o 25 h) IAT sco e eedback o con ol job counselo s.
Figu e E.2: IAT sco e dis ibu ion by job counselo s’ gende and ace
(A) By gende (B) By ace
No es: Race IAT sco es. A posi i e alue indica es a s onge associa ion be ween "whi e"-"good" and "black"-"bad". The e ical
dashed-lines indica e he c i ical h esholds sugges ed by G eenwald e al. (2009).
31
Figu e E.3: P opo ion o A o-descendan job seeke s by Depa men
No es: P opo ion o A o-descendan s among all seeke s and hi ed seeke s by Depa men o esidence o he job seeke . We use
da a om UAESPE o all job seeke s 20 o 65 yea s old egis e ed be ween 01/Ma ch/2021 and 28/Feb ua y/2022. We excluded
he CCF COMFAGUAJIRA as i s beha io dis ances om all o he CCFs wi h espec he A o-descendan CVs egis a ion and
e e als.
Table E.6: E ec o ace on he p obabili y o CV- o wa ding o hi ed by job-o e ype.
P e-s udy
Any job By wage By con ac ype
≤1 MMLW >1 MMLW Sho e m Long e m
(1) (2) (3) (4) (5)
A. P obabili y o being o wa ded
A o-descendan -0.041+-0.029 -0.018 -0.016 -0.026
(0.021) (0.018) (0.015) (0.015) (0.016)
Mean non-a o 0.280 0.176 0.172 0.184 0.171
R20.172 0.163 0.127 0.155 0.112
Obse a ions 156,053 156,053 156,053 156,053 156,053
B. P obabili y o being hi ed | being e e ed
A o-descendan -0.055∗∗ -0.074∗∗ -0.008 -0.065∗∗ -0.011
(0.020) (0.026) (0.005) (0.022) (0.008)
Mean non-a o 0.239 0.260 0.040 0.214 0.047
R20.181 0.248 0.045 0.230 0.067
Obse a ions 59,905 35,915 133,456 39,215 131,624
No es: S anda d e o s clus e ed a job cen e le el in pa en heses. +p<0.1, ∗p<0.05, p<0.01. Only includes CVs ha en e ed
he sys em Augus 5 h, 2021. All es ima ions con ol o age, zone, expe ience and educa ional le el, and include egion-mon h
o egis y and ci y o esidence ixed e ec s.
32
Figu e E.4: Explici bias. E alua ion o whi es and A o-descendan s wi h espec o
in elligence and laziness
(A) In elligence (B) Laziness
No es: In o ma ion om di ec in e iews o job counselo s.
Figu e E.5: P opo ion o job counselo s ha would no like o ha e speci ic popula ions as
neighbo s
No es: P opo ion o indi iduals who eply yes o he ques ion – among he ollowing g oups, who would you no like o ha e
as neighbo s?.
33
Figu e E.6: Di e en ial es ima ed p obabili y o o wa d an A o-descendan CV s a
non-A o-descendan CV by he minimum, maximum, and mean IAT sco e in a gi en job
cen e .
(A) Minimum
(B) Mean
(C) Maximum
No es: Base on he es ima ed p obabili ies ollowing es ima es o he equa ion Re iJ =α0+α1A oi+α2g(IATJ) + α3A oi×
g(IATJ) + α5A oi×Xi+Mi+CCFJ+RJ +µiJ , whe e gis he minimum, mean, and maximum espec i ely. A ea e lec s a
95% con idence in e al. Only includes CVs ha en e ed he sys em Augus 5 h, 2021. All es ima ions con ol o age, zone,
expe ience and educa ional le el, and include egion-mon h o egis y and ci y o esidence ixed e ec s. The e ical dashed-
lines indica e he c i ical h esholds sugges ed by G eenwald e al. (2009).
34
Figu e E.7: Di e en ial es ima ed p obabili y o o wa d a CV wi h espec o ace, gende
and age, by job cen e IAT sco e. P e-s udy
No es: Repo he di e en ial p obabili ies ollowing es ima es o he equa ion YiJ =α0+α1CATi+α2DJ+α3CATi×DJ+
α5CATi×Xi+Mi+CCFJ+RJ +µiJ , whe e Yis e e ing a CV, and CAT is A oi o Race, Womani o gende , and [Age <32]
o Age. Dashed lines ep esen 95% con idence in e al using he Del a me hod.
Figu e E.8: Re e ing and hi ing p obabili ies by job cen e cha ac e is ics
(A) P obabili y o e e ing a CV by job cen e
cha ac e is ics
(B) P obabili y o being hi ed by job cen e
cha ac e is ics and e e ing s a us
No es: Dashed lines ep esen 95% con idence in e al. S anda d e o s clus e ed a job cen e le el in pa en heses. P edic ed
condi ional p obabili ies con olling by ace, gende , age, zone, expe ience, educa ion, and municipali ies, CCF and mon h ixed
e ec s.
35
