ArticlePDF Available

Abstract and Figures

In this paper we investigate gender differences in workers’ career developments within and outside the firm in order to explain the existence of gender wage gaps. Using Danish employer-employee matched data, we find that good female workers are more likely to move to better firms than men, while they are less likely to be promoted. Furthermore, these differences in career advancements widen after the first child is born. Our findings suggest that career impediments in some firms cause the most productive female workers to seek better jobs in firms where there are less gender biases.
Content may be subject to copyright.
HAL Id: hal-01687343
https://hal.archives-ouvertes.fr/hal-01687343
Submitted on 18 Jan 2018
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Gender Dierences in Sorting
Luca Paolo Merlino, Dario Pozzoli, Pierpaolo Parrotta
To cite this version:
Luca Paolo Merlino, Dario Pozzoli, Pierpaolo Parrotta. Gender Dierences in Sorting. Industrial
Relations, Wiley, A Paraître. <hal-01687343>
Gender&Differences&in&Sorting*&
!
Luca&Paolo&Merlino^& & Dario&Pozzolié& & Pierpaolo&Parrotta¬&
!
!
&
Abstract(
In&this&paper,& we&investigate&gender&differences&in&workers'&career&development&within&and&
outside& the& firm& to& explain& the& existence& of& gender& wage& gaps.& Using&Danish& employer-
employee& matched& data,& we& find& that& good& female& workers& are&more& likely& to& move& to&
better&firms&than&men&but&are&less&likely&to&be&promoted.&Furthermore,&these&differences&in&
career& advancement& widen& after& the& first& child& is&born.& Our&findings& suggest& that& career&
impediments&in&certain&firms&cause&the&most&productive&female&workers&to&seek&better&jobs&
in&firms&where&there&is&less&gender&bias.&
&
&
&
&
&
JEL(Classification:&J16,&J24,&J62.&
&
Keywords:&Sorting,&Assortative&Matching,&Gender&Gap.&
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
*&We&thank&Estelle&Cantillon,&Francis& Kramarz,&Fane&Groes,&Rafael&Lalive,&Patrick& Legros,&Nina&Smith&
and&the&participants&of&the&COSME&workshop&(2013),&the&Search&and&Matching&Conference&(2014)&and&
the&workshop&on&Gender&and&Ethnic&Differentials&(2014)&for&their&helpful&comments&and&suggestions.&
We& gratefully& acknowledge& funding& from& the& Danish& Council& for& Independent& Research|Social&
Sciences,& Grant& no.& x12-124857/FSE;& LMDG,& the& Department& of& Economics& and& Business,& Aarhus&
University;& the& ERC& through& grant& 208535;& the& Wiener& Wissenschafts-,& Forschungs-& und&
Technologiefonds&(WWTF)&under&project&fund&MA& 09-017;& the& Carlsberg& Foundation;& and& the& Swiss&
National& Centre& of& Competence&in& Research& LIVES& and& FNRS.& The& data& used& in& this& paper& build& on&
anonymized&micro&data&sets&owned&by&Statistics&Denmark&(SD).&In&the&interest&of&scientific&validation&
of& analyses& published& using& DS& micro& data,& the& Department& of& Economics& and& Business,& Aarhus&
University,&will&assist&researchers&in&obtaining&access&to&the&data&set.&The&usual&disclaimer&applies.&
&
^&University&of&Antwerp&and&Université&Paris&1-&Panthéon&Sorbonne.&Email:&LucaPaolo.Merlino@uantwerpen.be.&
!
é!Corresponding!author.!Copenhagen!Business!School,!Department!of!Economics,!Porcelaenshaven!16!A,!DK-
2000!Frederiksberg.!IZA,!the!Institute!for!the!Study!of!Labor.!Tuborg!Research!Centre!for!Globalisation!and!
Firms,!Aarhus!University.!InSTED,!Department!of!Economics!at!University!of!Exeter!Business!School.!Email:!
dp.eco@cbs.dk.!
!
¬ ICN&Business& School.& IZA,& the& Institute& for&the&Study&of&Labor.& Tuborg& Research& Centre& for& Globalisation&and&
Firms,& Aarhus& University.& BETA,& Bureau& d'Économie& Théorique& et& Appliquée.& NoCeT,& Norwegian& Center& for&
Taxation.&ROA,&Research&Centre&for&Education&and&the&Labour&Market.&E-mail:&Pierpaolo.Parrotta@icn-groupe.fr.
!
1"
"
1" " Introduction"
Recent"studies"exploit"the"availability"of"matched"employee-employer"data"to"report"that"wage"
gaps" between" male" and" female" workers" arise" as" a" result" of" segregation" in" lower-paying"
occupations," in" less-productive" establishments" and" in" lower-paying" occupations" within"
establishments" (Bayard," Hellerstein," Neumark," and" Troske," 2003;" Hellerstein" and" Neumark,"
2008)."Card,"Cardoso,"and"Kline"(2016)"find"that,"while"women"negotiate"lower"wages"than"men"
with"the"same"employer,"they"are"also"less"likely"to" work"in"firms"that"pay"higher"premiums"to"
either"gender.
These"studies"show"that"wage"gaps"in"the"labor"market"are"related"to"gender"differences"in"
the" extent" to" which" better" workers" are" employed" in" better" positions," i.e.," in" the" degree" of"
positive"assortative"matching"(Becker,"1973).
Our" paper" differs" from" the" existing" literature" because" it" explains" the" gender" gap" from" a"
new"perspective"by"examining"workers’"career"development"within"and"outside"the"firm.
When"we" examine" gender" differences"in" sorting" in" Denmark,"surprising" patterns" emerge."
Female" workers" of" good" types" (proxied" by" the" residual" predicted" from" a" Mincerian" log-wage"
regression)" are" more" likely" to" move" to" a" better" firm" than" similarly" ranked" male" workers,"
conditional"on"workers"changing"employers."This"is"quite"puzzling" given" the" significant" gender"
wage"gap"in"Denmark"(around" 15" percent" in" the"Danish"private"sector—see"Gallen,"2015)"and"
that"the"representation"of"women"among"legislators,"senior"officials"and"managers"is"relatively"
low.i
We"then"turn" our" attention"to"transitions"within"the"firm"to" determine" whether"there"are"
gender"biases"in"promotions." There," a" quite" different"pattern"emerges:"male"workers"of" good"
types" are" more" likely" to" be" promoted" than" similarly" ranked" female" workers" conditional" on"
workers"not"changing"employers.
The"objective" of"this"paper"is"to"explain"this"apparently"conflicting"evidence"by" accounting"
for"the"fact"that"workers"move"up"the"job"ladder"in"two"ways:"either"they"are"promoted"by"the"
current" employer," or" they" find" a" better" employer." Studying" career" development" within" and"
outside"the" firm"as"a"unified"sorting"problem"allows" us"to"relate"the"patterns"that"we" observe."
Our"findings"suggest"that"career"impediments"in"certain"firms"cause"the"most"productive"female"
workers"to"seek"better"jobs"in"firms"with"less"gender"bias."In"our"interpretation,"segregation"and"
gender"gaps"emerge"because"of"the"costs"associated"with"changing"employers"combined"with"
the"career" impediments" that" female"workers"face." As" a"result,"consistent" with" the" findings"of"
Card,"Cardoso,"and"Kline"(2016),"only"the"best"female"workers"can"pursue"career"advancement"
via"job"transitions,"and"they"climb"the"occupational"ladder"more"slowly"than"men.
" To"detect"sorting"patterns,"we" apply" and" extend" the" methodology"proposed"by"Bartolucci"
and"Devicienti"(2013)"to"study"both"internal"and"external"carreer"advancements"across"genders."
In" particular," (i)" we" exploit" within-firm" variation" in" wages" to" rank" workers" within" firms" and"
conditional" on" observables." Intuitively," while" a" worker’s" type" is" not" observable" to" the"
econometrician,"it"is"observable"to"the"firm,"which"then"pays"her/him"accordingly."Furthermore,"
(ii)"we"use"profits"to" rank" firms," as"maximizing"profits"is"an"objective"for"all"firms."In" a" stylized"
theoretical" framework," we" demonstrate" that" this" estimation" strategy" allows" us" to" analyze"
gender"differences"in"sorting"both"within"and"across"firms"when"workers"care"about"wages.
2"
"
We" then" use" these" rankings" to" predict" separately" by" gender" (a)" the" probability" that" a"
worker"moves" to" a" better"firm," i.e.," a" firm"that" makes" higher" profits,"conditional" on" changing"
firms"(being"a"mover),"and"(b)"the"probability"of"being"promoted"to"a"higher"occupational"level"
conditional"on"staying"employed"in"a"given"firm"(being"a"stayer)."In"our"analysis,"we"use"Danish"
employer-employee"matched" data"for"two"important"reasons."First,"a"representative"and"large"
sample" of" both" workers" and" firms" allows" us" to" trace" workers’" career" developments." Second,"
Denmark" has" a" very" flexible" labor" market," similar" to" that" in" the" U.S." Hence," our" analysis" is"
relevant"beyond"the"case"of"Denmark.
In"line"with"the"predictions"of"our"theoretical"framework,"when"we"plot"the"probability"of"a"
job" transition" against" a" worker’s" type," we" find" a" U-shaped" relationship" that" is" steeper" for"
transitions"to"a"better" firm:" while" bad" workers" are" likely"to"be"replaced,"the"best" workers" are"
likely"to"move"to"better"firms"(Figure"1).
The"findings"depicted"in"Figures"2"and"3"that"positive"sorting"is"stronger"in"job"transitions"
but"weaker"in"promotions"for"female"workers"are"confirmed"by"our"regression"models"in"which"
we" add" several" controls." Specifically," a" one-standard-deviation" increase" in" the" log" of" lagged"
wages"increases"the" probability" of" moving"to"a" better" firm" by"
2
" percent"for"female" workers"
and" by" approximately" half" as" much" for" male" workers," while" it" increases" the" probability" of"
promotion" by"
19
" percent" for" female" workers" and" by"
31
" percent" for" male" workers." These"
gender" differences" are" sizable," significant" and" stable," as" they" arise" in" a" number" of" different"
specifications" and" tests." Most" important," the" same" patterns" emerge" when" we" use" other"
methods"to"rank"workers"and"firms.
Note,"however," that" if" all"firms" had" the"same" attitude" toward" female"workers," we" would"
observe"the"same"sorting"patterns"in"transitions"both"within"and"across"firms."Thus,"our"findings"
strongly"indicate"that"female"workers"face"more"career"impediments"in"certain"firms"and"that"
they"attempt" to"overcome"these"barriers" by"searching"for"better"jobs"in"fairer"companies."This"
interpretation"is"further"corroborated"by"the"fact"that"negligible"gender"differences"arise"when"
we" examine" promotions" to" higher" occupational" levels" in" the" firms" to" which" good" female"
workers"tend"to"move."Such"firms"are"also"highly"profitable,"which"suggests"that"the"best"firms"
are"those"with"non-discriminatory"policies.ii
The" gender" differences" in" sorting" that" we"uncover" for" job" transitions" are" possibly"
explained" by" female" and" male" workers" pursuing" different" routes" to" achieve" career"
advancement" due" to" career" impediments" in" certain" firms." This" overall" interpretation" of"our"
main" findings" is" further" supported" by" the" fact" that" gender" differences" in" job" transitions"
disappear"when"transitions"are"not"voluntary"(e.g.,"when"they"are"driven"by"a"firm’s"closure).
Another" interesting" result" is" that" gender" differences" in" promotions" appear" to" emerge"
especially" after" workers" become" parents," a" fact" that" is" not" related" to" firms’" observable"
characteristics"such"as"sector"or"size.
Our" analysis" is" related" to" Card," Cardoso," and" Kline" (2016)." Using" Portuguese" matched"
worker-firm"data,"the"authors"decompose"firm-specific"pay"differentials"into" a" combination" of"
sorting"and"bargaining"effects"by"implementing"the"methodology"proposed"by"Abowd,"Kramarz,"
and"Margolis"(1999),"which"we" discuss" further" in"Section"2."Specifically,"Card"et"al."(2016)"find"
that,"compared"to"men,"women"are"less"likely"to"be"employed"at"higher-wage"firms"and"to"gain"
a"significant" share"of"the"surplus"associated"with"their"jobs."Although"our"study"focuses"on" the"
Danish" labor" market," which" differs" from" the" Portuguese" one" along" several" dimensions," we"
3"
"
interpret"the"interplay"of"career"paths"within"and"outside"firms"in"our"analysis"as"corroborative"
evidence" in" favor" of" the" gender" differences" in" bargaining" power" documented" by" Card" et" al."
(2016).
Groes,"Kircher,"and"Manovskii"(2015)"show"that"both"low"and"high"wage"earners"within"an"
occupation"are"more"likely"to"leave"their"occupation"and"that"the"high"earners"tend"to"move"to"
new" occupations" with" higher" average" wages." Interestingly," we" find" similar" patterns" when" we"
examine" workers" changing" firms." In" addition," we" focus" on" how" the" strength" of" this"
phenomenon"differs"across"genders.
Our"insights" on"promotions"are"also"consistent"with"Booth,"Francesconi,"and"Frank" (2003)."
They" find" that" men" and" women" are" equally" likely" to" be" promoted" but" women" receive" lower"
wage" increases." This" is" the" sticky0 floor" phenomenon." Yet," Booth" et" al." (2003)" do" not" look" at"
transitions" across" employers." Hence," our" study" represents" a" significant" advancement" with"
respect"to"theirs.
Our"finding"that" career" impediments" for" women" emerge"after"motherhood"is"in" line" with"
the"evidence"presented"by"Smith,"Smith,"and"Verner"(2013)"for"CEOs" and" with" that" of" Kleven,"
Landais," and" Søgaard" (2015)," who" show" that" motherhood" is" a" career" impediment" in" certain"
firms"but"not"in"others."In"particular,"Kleven"et"al."(2015)"find"that"most"of"the"gender"wage"gap"
can"be"explained" by" a" parenthood" penalty"that"affects"women"but" not" men." In" our"study,"we"
focus" on" the" link" between" gender" biases" in" promotions" in" certain" firms" and" job" transitions."
Hence," we" restrict" our" attention" to" sorting" patterns" of" full-time" workers," and" we" therefore"
model"only"one"of"the"dynamic"effects"of"having"children"on"career"developments.
Several"authors"find"that" women" have" a" less" elastic" across-firms"labor" supply" than" men"
(Webber," 2016;" Hirsch," 2016)." This" can" result" in" wage" gaps" when" firms" set" wages"
monopsonistically" (Manning," 2003)." Differently" from" those" studies," we" exploit" within-firm"
variation"to"identify"gender" differences" among" (i)" separated" workers" (movers)"who"move"to"
better" firms" and" (ii)" non-separated" workers" (stayers)" who" are" promoted" to" a" better"
occupation." Thus," we" are" able" to" relate" gender" biases" in" promotions" to" the" allocation" of"
workers"to"firms.
The"structure" of" the" article"is" as" follows." Section"2" describes" our" estimation"strategy," the"
data" and" the" institutional" background." Section" 3" includes" the" main" results." Section" 4" reports"
robustness"checks"and"additional"results."Section"5"offers"concluding"remarks.
2" " The"Estimation"Strategy"
The" model" of" Merlino" (2012)" shows" that" negative" biases" against" certain" group" of" workers"
translates"into"differences"in"sorting"that"generates"unemployment"and"wage"gaps.
Detecting"sorting"patterns" is" however" not"a"straightforward" task," especially" when" agents’"
types"are"not"observable"(Filippin"and"Ours,"2015)."Different"approaches"have"been"proposed"to"
overcome" this" problem." Abowd," Kramarz," and" Margolis" (1999)," henceforth" AKM," use" the"
correlation" between" workers’" and" firms’" fixed" effects" derived" from" wage" equations."
Identification" of" the" individual" fixed" effects" relies" upon" the" assumption" that" workers’"
movements"between"firms"are"conditionally"random."In"other"words,"after"conditioning"on"the"
observable"worker"and"firm"effects,"when"a"worker"moves"to"a"new"firm,"she"draws"at"random"
from" the" existing" firms" in" the" economy." This" assumption" is" challenged" theoretically" by"
4"
"
on-the-job"search"(Eeckhout"and"Kircher,"2011)."Intuitively,"when" workers" voluntarily" move" to"
another"firm,"their"compensation"should"improve.
Bartolucci" and" Devicienti" (2013)" address" this" issue" by" exploiting" within-firm" variation" in"
wages"to" rank"worker"types"(within" firms)"and"profits"to"rank"firms."All"firms"seek"to"maximize"
profits,"and"thus,"it"is"natural"to"consider"profits"as"a"measure"of"firm"“quality”."Workers"instead"
might" care" about" many" job" characteristics" beyond" wages." Hence," wages" might" be" a" noisy"
measure"of"workers’"quality.
While" Card," Heining," and" Kline" (2013)" propose" tests" to" check" the" conditional" exogenous"
mobility" assumption" of" the" original" AKM" method," Bartolucci," Devicienti," and" Monzón" (2015)"
design" a" method" that" does" not" rely" on" wages:" they" measure" the" variance" of" firm" ranking"
(proxied" by" the" arrival" firm’s" profits" per" worker)" that" can" be" explained" by" the" movers’" types"
(proxied"by"the" sending"firm’s"profits"per" worker)."The"smaller"the"variance" of"firm"types"for"a"
given"worker"type"is"relative"to"the"unconditional"variance"of"firm"types,"the"more"intensively"
workers"sort"into"firms."Yet,"this"method"cannot"be"used"to"assess"sorting"in"promotions.
In" this" paper," we" then" use" the" methodology" of" Bartolucci" and" Devicienti" (2013)" to" rank"
workers"and"firms"because"it"allows"us"to"analyze"gender"differences"in"sorting"both"within"and"
across" firms" under" the" assumption" that" workers" care" mostly" about" wages." In" Section" 4," we"
show"that"our"main"results"do"not"depend"on"this"assumption,"as"implementing"several"other"
approaches"to"measure"workers’"ranks"provides"similar"findings.
Our"baseline"empirical"strategy"is"rooted"in"a"simple"theoretical"framework"with"learning,"
job"search"and"a" production" technology" in" which" skills"and"capital"are"complements,"i.e.," that"
induces"positive"assortative" matching." While"the"model"is"formally"presented" in"the"Appendix,"
we"now"present"the"main"mechanisms"it"embeds.
In"the"model," there"is"a"unit"mass" of"workers"and"firms,"with" type"indexed"by"
e
"and"
,"
respectively."The"productivity" of"the"match"is"increasing" in"both"the"type"of" the"worker"and"of"
the" firm," and" we" consider" a" super-modular" production" function" that" induces" positive"
assortative"matching"(Becker,"1973).
First," workers" are" randomly" matched" to" firms." This" assumption" captures" the" idea" that"
workers"have"little"information"regarding"employer"types" at" the" beginning" of" their" careers."As"
they"acquire"experience,"workers"might"learn"the"characteristics"of"all"firms"in"the"market,"and"
they"also"might"become"more"productive.
We" introduce" learning" by" doing" in" the" following" way:" during" period"
0
," workers" acquire"
relevant"experience"that"improves"their" skills" as" long" as"
fe ³
."In"other"words,"there" are" skill"
requirements:"only"agents"that" are" sufficiently" qualified"understand"the"technology"enough"to"
improve" their" productivity." As" a" result," in" the" second" period," some" workers" are" more"
productive,"i.e.,"they"increase"their"type."Hence,"a"worker"
e
" in"a"match"in"period"0"with"a"firm"
" such"that"
fe ³
" becomes"of"type"
e
t
"in"period"1.
However,"some"firms"do"not"allow"female"workers"who"learned"to"express"their"acquired"
potential:"for" example," the"suggestions" they" make" to"improve" the" productivity"of" the" current"
match" are" not" implemented," or" they" are" not" assigned" to" better" tasks." Hence," good" male"
workers"are"more"likely"to"be"promoted"in"the"firm"where"they"are"currently"employed"than"are"
female"workers"of"similar"type.
Subsequently," each" pair" can" decide" whether" to" stay" together" or" to" search" for" a" better"
5"
"
partner."Search"is"costly,"but"it"allows"that"workers"are"matched"to"their"best"partner"available."
Hence,"only"agents"that"are"sufficiently"mismatched"will"change"employers.
In"equilibrium,"female"workers"who"(a)"improve"their"type"through"learning"but"(b)"are"not"
promoted" by" the" current" employer" are" more" likely" to" move" to" better" firms" with" fairer"
promotion"policies."More"formally,"we"have"the"following:
Empirical" Predictions."Consider0
m
e
0and0
f
e
0such0 that0
m
e
0is0 male,0
f
e
0is0 female0 and0
eee fm ==
.0Then,
(i)0the0higher0the0type0of0a0worker0is,0the0higher0the0probability0of0moving0to0a0better0firm;
(ii)0the0probability0of0moving0to0a0better0firm0from0a0given0firm0for0
f
e
0 with0respect0 to0
D-
f
e
0
is0higher0than0for0
m
e
0 with0respect0to0
D-
m
e
,0for0
0>D
;0and
(iii)0the0probability0 of0 being0promoted0in0a0given0 firm0 for0
m
e
0 with0respect0to0
D-
m
e
0is0higher0
than0for0
f
e
0 with0respect0to0
D-
f
e
,0for0
0>D
.
While" wages" are" non-monotonic" in" firm" type," in" this" model," wages" and" (both" total" and"
average)"profits"are"increasing"in"own"type." Hence,"we0can0use0 within-firm0variations0in0wages0
to0rank0 workers0 and0profits0 to0 rank0 firms."In" our" analysis,"we" therefore" use" these"rankings" to"
test"the"empirical"predictions"made"above.
The"theoretical"framework"explicitly"shows"that"if"female"workers"were"to" have" the" same"
career"prospects"in"all"firms,"it"would"not"be"necessary"to"separately"study"gender"differences"in"
sorting" within" and" across" firms," as" these" should" be" comparable." This" is" true" even" if" female"
workers’" types" are" drawn" from" a" worse" distribution," as" in" that" case" the" outside" option" is"
relatively"more"attractive"for"them"than"for"men.
If"instead"firms"are"heterogeneous"in"their"gender"bias"in"promotions,"female"workers"who"
learn"but"are"not"promoted"in"non-female-friendly"firms"are"more"likely"to"move"to"better"firms"
that"are"female"friendly."Conversely,"good"male"workers"are"more"likely"to"be"promoted"in"the"
firm" where" they" are" currently" employed." Hence," the" association" between" the" probability" of"
moving"to"a" better"firm"and"wages"should" be"higher"for"female"workers,"while" the"association"
between"the"probability"of"being"promoted"and"wages"should"be"stronger"for"male"workers.
Let" us" stress" that" our" estimation" strategy" is" premised" on" the" presence" of" mismatches"
between"workers"and"firms"as"with"perfect" sorting" no" transitions" would" occurr." Yet," even" in"a"
flexible"labor" market,"such"as"the"Danish"one,"mismatches"and"frictions"are"likely"to"arise"for"a"
variety"of"reasons,"such"as"commuting"distances"or"non-monetary"factors"(Manning,"2003).
2.1" " Empirical"Approach"
Based"on"this"theoretical"framework,"we"estimate"the"following"linear"probability"model,"which"
is" conditional" on" workers’" movements" (i.e.," for" the" sample" of" movers)," separately" for" each"
gender:"
"
etft
t
'
f
ftetet
'
et uzzzxfwageffupmove
eggbaa
+++++++ --- 2111110 ''')(=),(_
"(1)"
where"
),(_ '
et ffupmove
" is"a"dummy"variable"that"is"equal"to"1"if"an"employee"of"type"
e
,"who"
6"
"
has"worked"in"a"sending"firm"of"type"
,"moves"to"a"“better”"receiving"firm"of"type"
ff '>
"at"
time"
t
." All" agents" are" indexed" by" their" type." The" term"
)(
1fwageet -
"is" the" log" of" the" wage"
earned"in"sending"firm"
" by"employee"
e
."As"there"are"many"worker"characteristics"that"may"
influence" wages" and" mobility," such" as" demographic" characteristics," and" it" is" unclear" to" what"
extent" the" monotonicity" assumption" on" payoffs" is" fulfilled" when" comparing" coworkers" in"
different" occupations," we" augment" equation" (1)" with" the" vector"
e
x
." The" latter" consists" of"
relevant" worker" characteristics," such" as" age," tenure," work" experience,iii"ethnicity," marital"
status,"parental"status,"education,"occupation"and"a"family"network"dummy"(i.e.,"a"dummy"that"
records"whether"a"worker"has"had"at"least"one"parent"employed"as"a"manager)."The"vectors"
f
z
"
and"
'
'
f
z
" include" the" share" of" white-collar" women" and" the" size" of" the" sending" and" receiving"
firm,iv" respectively," while" the" vector"
" represents" time" fixed" effects." Finally,"
f
u
" captures"
the"fixed"effects"of"firm"
,"and"
et
e
" is"a"mean-zero"error"term.v
The" extent" and" sign" of" sorting" in" job" transitions" are" tested" by" investigating" whether" the"
coefficient"
1
a
" is"different"from"zero."Specifically,"if"
0>
1
a
,"better"workers"(i.e.,"those"workers"
who"receive" higher" wages" in"a" given" firm" after"controlling" for" observables)" are"more"likely" to"
move" to" firms" that" earn" higher" profits," i.e.," there" is" positive" sorting." Furthermore," a" more"
positive" coefficient" indicates" a" relatively" stronger" tendency" toward" positive" assortative"
matching.
"""The"focus"of"this"paper"is"to"test"whether"the"degree"and"sign"of" sorting" in" job" transitions"
vary"according" to"gender"by"estimating" equation"(1)"separately"by"gender"and"testing"whether"
1
a
" significantly"varies"across"the"female"and"male"sub-samples."We"test"for"sorting"differences"
between" men" and" women" by" comparing" estimations" from" each" sub-sample" since" within-firm"
rankings"are"highly"gender-specific."
With" regard" to" the" sample" of" stayers" and" their" probability" of" being" promoted," a" similar"
model"is"implemented"separately"by"gender:" "
"
etftftetetet vuzzxfwagefprom ++++++ --
gbaa
'')(=)( 1110
"(2)"
where"
)( fpromet
" is"a"dummy"variable"that"is"equal"to"1"if"employee"
e
,"who"has"worked"in"a"
specific" occupation" in" firm"
," is" promoted" to" a" higher" occupational" level." Because" of" data"
constraints," we" consider" three" main" occupational" groups:" managers," middle" managers" and"
other"white"collar"positions,"and"blue-collar"workers."The"term"
f
u
" captures"within-firm"fixed"
effects." As" in" the" previous" model," the" vectors"
1-et
x
"and"
ft
z
"include" worker" and" firm"
characteristics"while"the"vectors"
" are"time"dummies"and"
et
v
" is"an"error"term.vi
2.2" " Data"
The"data"set,"provided"by"Statistics"Denmark,"is"a"merged"employer-employee"panel"sample"of"
Danish"firms"observed"over"the"1996-2005"period.
The" firm-level" data" include" information" on" sales," employment," value" added," materials,"
profits,"fixed"assets"and"a"two-digit"NACE" identifier"(further"details"are"provided"in"section"A-1"
7"
"
of" the" Online" Appendix)." We" consider" firms" in" the" private" sector" that" have" more" than" 20"
employees.vii" All"firms"with"imputed"accounting"variables"are"omitted"from"the"analysis.
The"individual-level"data"cover"the"working-age"population"from"1980"onward."These"data"
include"information"on"the"wage,"age," gender," marital" status," number" of"children,"experience,"
tenure,"highest"completed" education,"occupation"and"family"background"characteristics." Apart"
from"deaths" and" permanent" migration," there"is"no" attrition" in" the"data"set." The" labor" market"
status"of" each" person" as"of"the" last" week" in"November"is" recorded" as" the"relevant"datum" for"
each"person"for"that"year."Therefore,"if"a"worker"changes"his"or"her"main"job,"then"we"observe"
only" the" year" in" which" this" change" occurred." However," we" observe" whether" a" worker"
experiences"unemployment"and"its"duration"(in"weeks)"in"a"given"calendar"year."For"individuals"
with"multiple"jobs,"only"the"main"occupation"is"considered.
In" the" analysis" that" follows," we" include" only" individuals" with" a" positive" annual" salaryviii""
and" individuals" younger" than" 60." Furthermore," apprentices" and" part-time" employees" are"
excluded"from"the"main"analyses.
The" empirical" estimations" are" based" on" two" samples." The" first" sample" considers" only"
“movers”," i.e.," those" workers" who," within" the" 1996-2005" period," changed" at" least" once" from"
one" firm" (the" sending" firm," according" to" our" terminology)" to" another" firm" (the" current" or"
receiving" firm)" in" the" data" set" within" the" 1996-2005" period.ix ""An" important" challenge"
regarding"this"data"set"is"that,"because"of"changes"in"firms’"ownership,"there"appear"to"be"some"
“false”"transitions" in" the" data." To"minimize"miscoded" movers," transitions" involving"more"than"
50
" percent" of" the" size" of" the" same" sending" firm" are" excluded" from" the" final" sample."
Furthermore,"we"exclude" from" the" sample"of"movers"those"workers"who"changed" jobs" after" a"
firm"closure."In"total," our" sample" includes" 479,161" yearly"observations"of"357,487"job"movers"
(i.e.," approximately" 15" percent" of" the" entire" population" of" job" movers)" and" approximately"
17,000"firms.x" The"second"sample"excludes"the"movers"and"consists"of"4,658,374"observations,"
617,513"“stayers”"(i.e.,"approximately"25"percent"of"the"entire"population"of"stayers)"and"nearly"
18,000"firms."The"sample"of"movers"is"a"lower"share"of"the"underlying"population"compared"to"
the"sample"of"stayers"because"of"the"removal"of"“false”"and"involuntary"transitions"as"described"
above."As" we" discuss" later"in"the" results" section,"we"also" estimate" all" of"our" models" on" more"
general"samples"obtained"by"relaxing"all"of"our"sample"selection"criteria.
2.3" " Descriptive"Analysis"
Table" 1" separately" lists" the" descriptive" statistics" for" the" two" samples" according" to" gender,"
measured"at"both"the"worker"and"firm"level.
The"average"male"job"mover"is"39"years"of"age"and"has"16"years"of"experience,"whereas"the"
average" female" job" mover" is" 38" years" of" age" and" has" 14" years" of" experience." The" average"
tenure"for" both" women"and" men" is"approximately" three" years."The" majority" of"workers" have"
secondary" or" post-secondary" diplomas," and" 6" percent" of" male" job" changers" have" at" least" a"
university"degree,"whereas" 30" percent" have" completed"only"primary"education." In" addition," 7"
percent"of"female"job"changers"have"at"least"a"university"degree,"and"37"percent"have"a"primary"
education."Most"men"and"women"are"classified"as"blue-collar"workers"(72"percent),"followed"by"
middle" managers" (24-26" percent)." Significantly" more" male" movers" have" managerial" jobs"
compared"with" their" female" counterparts"(4"percent" versus" 2" percent,"respectively)." For" both"
8"
"
genders,"approximately" 5" percent" are"foreigners,"nearly" 15" percent" have"at"least" one" child" of"
0-3"years"of"age,"and"around"4-5"percent"have"at"least"one"parent"working"as"a"manager"at"the"
time"of"the"job"transition"or"before,"i.e.,"a"“family"network”."In"comparison,"the"average"stayer"
is" approximately" two" years" older" and" has" two" more" years" of" tenure," with" a" slightly" lower"
educational"level." The"average"stayer"is"also"more"likely"to"be"married"and"less"likely"to"have"a"
child"between"0"and"3"years"of"age,"regardless"of"the"gender"of"the"individual."The"percentage"
of"foreigners"is"reasonably"comparable"across" the" two" samples." During" the" period"covered"by"
our" sample," the" wage" of" an" average" male" and" female" job" mover" was" approximately" 250" and"
200" thousand" Danish" krones," respectively," or" approximately" 40" and" 30" thousand" USD" per"
annum,"respectively."The"salary"of"an"average"stayer"was"approximately"10"percent"above"that"
figure."Turning"to"the"firm-level"characteristics,"we"find"that"the"average"firm"size"is"fairly"similar"
across"the"two"samples," although" the" share"of"white-collar"women"and"profits"per"worker"are"
higher"in"the"sample"of"movers,"regardless"of"the"gender"of"the"employee.
The"bottom"part" of" Table" 1" includes"the"mean" of" the" main" outcome-dependent"variables"
used"in"our"empirical"analysis."For"the"sample"of"job"movers,"we"calculate"an"indicator"function"
that"takes"value"one"(zero)"if"a"worker"moves"to"a"receiving"firm"that"is"of"higher"(lower)"quality"
than"the"sending"firm."Firm"type"is"defined"in"terms"of"profits."Given"that"the"measure"of"profits"
is" firm" specific" and" might" be" affected" by" measurement" error," we" calculate" a" set" of" indicator"
variables" that" are" based" on" alternative" improvements" in" profits" (i.e.," the" profit" differential"
between" sending" and" receiving" firms" is" at" least" either" 5" or" 10" percent)." The" means" of" these"
outcome" variables," also" reported" in" Table" 1," allow" us" to" conclude" that" women" have" higher"
probabilities"of"moving"to"a"receiving"firm"of"higher"quality,"regardless"of"the"definition"of"firm"
quality"that"we"use."In"addition,"for"the"sample"of"stayers,"we"also"examine"the"probability"of"
promotion"to"a"higher"occupational"level"and"to"a"managerial"position."We"find"that"women"are"
generally"less"likely"to"be"promoted"than"men.
As" we" report" in" Figures" A1-A3" of" the" Online" Appendix," whereas" a" somewhat" decreasing"
gender"wage"gap"characterizes"workers"who"changed"employers,"stayers"present"a"stable"wage"
differential" between" men" and" women" over" time." More" important," the" wage" developments"
have" almost" identical" slopes" across" genders," suggesting" that" the" career" profiles" of" men" and"
women"are"very"likely"to"feature"very"similar"initial"conditions"and"match"quality."Women"earn"
substantially" less" than" their" male" peers," but" the" differences" are" particularly" marked" for" the"
sample"of"stayers."This"descriptive"evidence"seems"to"suggest"that"more"productive"women"are"
very"likely"to"change"workplaces"to"achieve"higher"wages"and"promotion"rates."This"evidence"is"
not" driven" by" educational" and" occupational" levels" or" by" the" gender" composition" of" the"
workforce.
2.4" " Institutional"Background"
As"institutional"constraints" may"hamper"the"degree"of"assortativeness" in"the"labor"market,"we"
outline" the" main" features" of" the" Danish" labor" market," which" are" represented" by" the"
combination" of" high" flexibility" and" social" security" (“flexicurity”)," the" role" of" family-friendly"
policies"and"decentralized"wage"setting.
The"cornerstones"of"the"Danish"model"are"a"high"level"of"labor"mobility"and"generous"social"
security" schemes." The" absence" of" severance" pay" reduces" hiring" and" firing" costs" and" labor"
9"
"
market"frictions" and" facilitates" firms’"efforts" to" adjust" the"quality" and" size" of"their"workforce."
Moreover," although" workers" are" not" protected" by" stringent" employment" rules," they" bear"
relatively" low" costs" of" changing" employers" and" have" easy" access" to" unemployment" or" social"
assistance" benefits." In" fact," Danish" replacement" ratios" are" among" the" most" generous" in" the"
world." Therefore," a" notable" part" of" the" observed" labor" mobility" is" also" associated" with" wage"
mobility"(Eriksson"and"Westergaard-Nielsen,"2009).
A"further"key"feature"of"the"Danish"labor"market"is"the"wide"coverage"of"publicly"provided"
childcare."Combined"with"the"length"and"flexibility"of"parental"leave"schemes,"this"has"favored"
female" labor" market" participation" and" full-time" employment" without" dramatic" consequences"
for" the" fertility" rate" (OECD," 2005)." While" many" of" the" jobs" held" by" women" were" initially"
part-time" occupations" in" the" public" sector," at" present," a" notable" proportion" of" women" is"
employed"in"the"private"sector"and"works"full-time."Nonetheless,"the"descriptive"statistics"show"
that"women"in"the"private"sector"earn"a"5-percent"higher"wage"and"are"slightly"more"educated"
compared" to" their" counterparts" in" the" public" sector." These" statistics" may" suggest" that" the"
sample" used" in" this" study" is" relatively" selected," consisting" of" slightly" more" motivated" and"
career-oriented"women."Yet," we" will" see" that"even"this"sample" of" female" workers" encounters"
significant"career"impediments"in"some"firms.
For" the" purposes" of" our" analysis," a" brief" description" of" wage" bargaining" in" the" Danish"
private"sector"is"also"important."Denmark"experienced"a"shift"in"wage"bargaining"from"a"highly"
centralized"system"to"a"substantially"decentralized"system."Since"the"early"1980s,"an"increasing"
share"of"wage"bargaining"has"been"devolved"to"the"firm"(individual"employee)"level."Currently,"
the" within-firm" wage" variability" in" Denmark" represents" more" than" 80" percent" of" the" total"
variability"observed"among"all"workers"(Shaw"and"Lazear,"2008).
Given" the" key" characteristics" of" the" Danish" labor" market," the" evidence" of" gender" gaps"
arising" from" our" empirical" analysis" may" be" generalized" to" other" labor" markets" with" similar"
degrees"of"flexibility.
3" " Results"
Given" the" large" volume" of" results," we" discuss" them" in" two" separate" sub-sections." The" first"
sub-section"describes" the" main"results" of" sorting" in"job" transitions" and"promotions," while" the"
second"sub-section"discusses"some"additional"analysis"and"alternative"specifications.
3.1" " Main"Results"
We" first" analyze" the" main" patterns" present" in" the" data" using" some" intuitive" figures" that"
concisely"describe"the"sorting"patterns"of"workers"of"different"types.
According" to" our" theoretical" framework," conditional" on" observables," the" probability" of"
leaving"the"current"firm"should"be"high"for"workers"with"low"and"high"wages,"while"it"should"be"
lower"for"workers"who"are"ranked"neither"too"high"nor"too"low"in"the"wage"distribution"of"the"
firm."Furthermore," this"probability"should"be" higher"for"women,"for"whom"outside"options" are"
more"attractive.
In" Figure" 1," we" plot" the" probability" of" leaving" the" current" firm" for" men" and" women" as" a"
10"
"
function"of"their"type,"proxied"by" the"residual"predicted"from" a"Mincerian"log-wage"regression"
in" which" we" control" for" both" observable" individual" characteristics" and" firm" fixed" effects." We"
find"that"high"wages"increase"the"probability"of"leaving"the"firm"to"a"greater"extent"than"do"low"
wages,"and"this"is"particularly"true"for"female"workers.
To"further"understand"these"mobility"patterns,"we"now"investigate"which"workers"are"more"
likely"to"move"to"a"better"firm"conditional"on"leaving" a"firm"and"which"workers"are"more"likely"
to" be" promoted" conditional" on" not" leaving," again" depending" on" the" residuals" of" a" Mincerian"
log-wage"equation.
Figure"2"shows"that"workers"who"change"firms"are"more"likely"to"move"to"a"better"firm"the"
higher"their" rank"in"the"sending"firm"is."Furthermore,"good"female"workers"are" generally"more"
likely"to"leave"the"current"firm,"expect"possibly"for"very"high"types."Similar"results"are"obtained"
for"promotions"(Figure"3),"although"in"this"case,"male"workers"of"good"types"are"more"likely" to"
be"promoted"than"similarly"ranked"female"workers.
This" suggests" that" women" change" firms" because" they" are" less" likely" to" pursue" career"
advancement" in" the" current" firm." To" better" understand" these" patterns," we" now" present" the"
results"of"the"estimation" of" a"linear"probability"model"(1)."This"approach"has" the" advantage"of"
allowing"for" an" analysis" of"the" different" factors"driving"gender" differences" in"sorting"and" how"
they"change"for"different"sub-samples"of"the"population.
The" main" results" pertaining" to" transitions" to" better" firms" are" reported" in" the" first" two"
columns"of"Table" 2." For"both"men"and"women,"there"is" a"significantly"positive"elasticity"of"the"
probability"of"moving"to"a"better" firm"with"respect"to"the"logarithm"of"the"wage"earned"in"the"
previous"firm." These" results" are"consistent" with" those" of"Bagger,"Sørensen," and" Vejlin"(2013),"
who"document"strong"positive"sorting" in" Denmark." This"evidence"justifies"the"assumption"of"a"
production"function"that"induces"positive"assortative"matching.
Female" workers" display" a" substantially" stronger" tendency" toward" positive" sorting" in" job"
transitions" compared" to" men." Specifically," a" one-standard-deviation" increase" in" the" log" of"
lagged"wages" raises" female" workers’"probability"of" moving" to"a"better" firm" by" 2"percent." The"
same"increase"has"half"as"much"of"an"effect"on"the"male"workers’"probability"of"moving."
Hypothesis" testing," reported" at" the" bottom" of" Table" 2," confirms" that" the" coefficient"
estimated" on" women’s" lagged" wages" is" statistically" higher" than" that" associated" with" men’s"
wages." These" empirical" associations" suggest" that" among" movers" better-ranked" women" are"
more"likely"to"move"to"companies"with"higher"profits"compared"to"men.xi
Interestingly,"a" rise" in" the"share"of" white-collar" women" in"the"sending" firms" has" opposite"
effects" on" sorting" in" the" two" sub-samples:" it" decreases" the" probability" of" moving" for" female"
workers,"whereas"it"increases"the"corresponding"probability"for"men.
This" analysis" of" job" transitions" gives" little" support" for" the" classical" version" of" the" glass"
ceiling"hypothesis,"which"would"imply"that"sorting"is"weaker"for"women"than"for"men.
Let"us" now" turn" our" attention" to" promotions," which"are" a" proxy"for" career" advancement"
within" firms" (columns" 3-4" of" Table" 2)." We" find" a" general" positive" relationship" between" the"
lagged" wage" and" the" probability" of" being" promoted" for" both" genders.xii""However," this"
association"is"stronger"for"men,"as"confirmed"by"the"hypothesis"tests"reported"at"the"bottom"of"
Table" 2." A" one-standard-deviation" increase" in" the" log" of" lagged" wages" raises" female" workers’"
promotion" probability" by" 19" percent." The" same" increase" triggers" a" rise" in" male" workers’"
promotion"probability"of"31"percent."A"greater"share"of"white-collar"women"is"associated"with"
11"
"
a"higher" conditional" probability"for" both" women"and" men." Thus"the" within" firm" increase"in"
the" share" of" female" workers" per" se" is" not" an" indication" of" unbiased" promotion" policies."
Gender" differences" in" favor" of" men" persist" when" we" focus" on" substantial" career" advances,"
i.e.,"promotions"to"positions"at"the"managerial"level"(see"the"last"two"columns"of"Table"2).xiii""
These"results" pertaining" to"promotions" might" qualify"the" findings" indicating"that" a" rise"in"
the" share" of" white-collar" women" in" sending" firms" has" a" negative" correlation" with" women’s"
probability" of" moving" to" a" better" firm." As" women" are" more" likely" to" be" promoted" in"
female-friendly"firms,"i.e.,"firms"with"many"white-collar"female"workers,"they"are" less" likely" to"
seek" a" job" elsewhere" if" the" sending" firm" is" female" friendly." Furthermore," men" have" fewer"
incentives" to" seek" a" job" outside" of" their" current" firm," especially" in" non-female-friendly" firms,"
which"implies"stronger"positive"sorting"for"women"in"job"transitions.
This" explanation" presumes" the" existence" of" firms" that" have" no," or" less," gender" bias" in"
promotions." We" will" now" investigate" whether" such" firms" exist." To" do" so," we" define"
female-friendly"firms"as"companies"characterized"by"a"share"of"women"in"white-collar"positions"
higher" than" the" industrial" median." Furthermore," we" define" “female-sought”" firms" as" the"
female-friendly"firms" that" are" destinations"in"the" job" transition" of"at" least" one" female"worker"
coming"from"a"worse"firm.
The" first" panel" of" Table" 3" shows" that" the" sorting" parameter" in" job" transitions" to"
female-friendly"firms"is"larger."There"are"instead"smaller"and"less"significant"gender"differences"
for"transitions" to"firms"that"are"not"female"friendly."Hence,"transitions"to"female-friendly" firms"
drive"the"stronger"positive"sorting"for"female"workers"in"job"transitions.
Positive" sorting" in" promotions" is" stronger" for" women" when" we" examine" promotions" in"
female-friendly"and"female-sought"firms"(second"panel"of"Table"3)."However,"the"difference" is"
consistent"with"the"baseline"results,"but"stronger,"in"firms"that"are"not"sought"after"by"females."
Hence,"these"findings"strongly"indicate"that"good"female"workers"seek" career" advancement" in"
female-friendly" firms" because" promotion" opportunities" in" these" firms" do" not" depend" on"
gender.
Nonetheless," gender" differences" in" favor" of" men" for" promotions" to" managerial" positions"
seems" to" emerge" in" all" firms" (third" panel" of" Table" 3)." Hence," only" few" female" workers" reach"
managerial"positions."These"results"nicely"complement"those"of"Gayle,"Golan,"and"Miller"(2012),"
who"find" that"female"CEOs"are"more"likely"to" exit"their"occupations"but"are"also"more"likely"to"
become"CEOs"when"they"have"not"exited.
According"to" our" empirical"strategy,"female-friendly" and" female-sought"firms" perform," by"
construction,"better"than"sending"firms."In"Table"A4"of"the"Online"Appendix,"we"explicitly"assess"
the"correlation"between"firm"performance" and" female" friendliness" by" estimating"a"set"of"firm"
performance" equations" with" several" control" variables" and" an" indicator" variable" for" female"
friendliness," which" is" alternatively" measured" with" either" a" “female-friendly" firm”" or"
“female-sought" firm”" dummy." We" find" a" positive" and" significant" correlation" between" being"
either"a"“female-friendly"firm”"or"a"“female-sought"firm”"and"firms’"profits"per"employee,"sales"
per"employee"and"value-added"per"employee.
Overall," the" empirical" results" presented" thus" far" are" not" consistent" with" the" view" that"
women"have"better"non-market" opportunities" (Lazear" and" Rosen," 1990)."In"that"case,"women"
would" be" less" likely" to" be" promoted" than" men" in" all" firms" but" more" likely" to" receive" higher"
wages"if"promoted"and"more"likely"to"quit"to"pursue"non-market"opportunities."Instead,"female"
12"
"
workers"are"less"likely"to"be"promoted"and"are"more"likely"to"move"to"a"better"firm"when"career"
advancement" is" not" too" difficult" to" achieve." In" fact," the" evidence" of" gender" differences" in"
promotion" strongly" suggests" that" women" who" cannot" climb" the" occupational" ladder" within" a"
firm"because"of"discriminatory"promotion"policies"attempt"to"overcome"these"gender"barriers"
by"searching"for"better"jobs"offered" by" fairer" firms.However," great" career" advancement" tends"
to"be"easier"to"achieve"for"men"than"for"women"in"all"firms.
4" " Robustness"and"Additional"Results"
In"this"section,"we"provide" further" evidence" of" the" robustness"of"gender"differences"in"sorting"
and"of"the"mechanisms"generating"them.
4.1" " Definitions"of"Workers’"Types"
With"renegotiation"and"endogenous"search"intensity,"wages"provide"a"noisy"ranking"of"workers"
(Bagger"and" Lentz,"2014)."To"evaluate" whether"this"issue"affects" our"analysis,"we"rank"workers"
using" employee" fixed" effects," which" are" estimated" from" a" gender-specific" wage" equation."
Specifically," the" individual" fixed" effects" are" obtained" by" estimating" the" following" wage"
regression:" "
"
eftftettefeeft ZXwln
ebbya
++++ ,= 21),(
"(3)"
where"
eft
w
" is"the"gross"annual"wage"earned"by" individual"
e
"in"firm"
"in"year"
t
."
et
X
"is"a"
vector" of" individual-specific" controls" that" change" over" time." Following" Card" et" al." (2013)," we"
include" in"
et
X
" a" set" of" interactions" between" year" dummies" and" educational" attainment" and"
interaction" terms" between" quadratic" and" cubic" terms" in" age" and" educational" attainment." In"
addition," we" also" control" for" other" factors" that" might" affect" wages" such" as" experience" and"
tenure." The" vector"
ft
Z
" contains" firm-specific" controls," such" as" value" added" and" capital" per"
employee."The"parameters"
e
a
"and"
),( tef
y
"are"the"individual-"and" firm-specific" fixed" effects,"
respectively." We" estimate" this" additive" “two-way”" worker-firm" effects" model" using" the"
methodology"developed"in"AKM.xiv""The"findings"reported"in"Table"4"confirm"the"results"of"our"
main"analysis.
We"then"measure"the"strength"of"sorting" in"job"transitions"using" the"method"proposed"by"
Bartolucci"et"al."(2015)."While"this"approach"cannot"be"used"to"evaluate"sorting"in"promotions,"
it"does"not"rely"on"wages."It"measures"the" variance" of" firm" rankings" (proxied" by" the" arrival" or"
current"firm’s"profits" per" worker)" that"can"be" explained" by" the"movers’"types" (proxied" by" the"
sending"firm’s"profits"per"worker)." The" smaller" the" variance"of"(firm)"partner"types"for"a"given"
worker"type"relative" to"the"unconditional"variance"of"firm" types,"the"more"intensively"workers"
sort" into" firms." Specifically," the" strength" of" sorting" is" defined" by" the" correlation" ratio"
,]]/|[[= 2
f
efEvar
sh
" where"
]]|[[ efEvar
" represents" the" partners’" variance" for" a" given"
worker" of" type"
e
," whereas"
2
f
s
" is" the" variance" of" firm" types"
." We" then" estimate" the"
correlation"ratio"
h
" as"the"mean" of" the" correlation"between"the" type" of" the"sending"firm"
"
13"
"
and" the" arrival" firm" f" for" all" worker" types" e"(represented" by" the" firm’s" profit" per" worker),"
defined" as"
).()/|,(=)|,( fvareffcoveff ''
r
"As" in" Bartolucci" et" al." (2015)," we" estimate"
"
using"only"transitions"mediated"by"an"interim"unemployment"spell"to"make"
),( ef
" and"
),( ef '
"
independent"conditional" on" worker" type"e."The" strength" of" sorting"
h
" provides"a" measure" of"
the"association"between"firm"type"f"and"worker"type"e."Finally,"Bartolucci"et"al."(2015)"estimate"
the"sign"of"sorting"by"considering"the"empirical"association"between"the"sending"firm’s"ranking"
and"the"mover’s"wage"earned"in" current" employment" or," alternatively," between" the"receiving"
firm’s"ranking"and"the"mover’s"wage"earned"in"the"previous"employment.
When"we"implement"these"methodologies"on"job"transitions,"we"find"that"the"results"are"
consistent"with"our"main"findings,"as"the"estimated"parameter"
h
" is"larger"for"women"than"for"
men"and"the"sign"of"sorting"is"positive"for"both"genders"(Table"5).xv""
4.2" " Definitions"of"Firms’"Types"
Profits"are"the"objective" of" all" firms." Furthermore,"a"precise"estimate"of"mean"profits"for"each"
firm"can"be"recovered"as"long"as"there"are"a"large"number"of"workers"per"firm."However,"our"
results"for"job"transitions"could"be" sensitive" to" the" particular" definition" of"firm"quality"we"use"
(i.e.,"a"firm"is"better"than"another"when"it"has"at"least"5"percent"higher"profits)."It"is"important"
to"consider" other"ranking"measures"for" firm"types"than"profits"also"because"these" may"be"also"
determined"by"monopoly"power"or"taxation.
We"address" this" issue" in"different" ways." First," we"strengthen" the" conditions" on"profits" by"
defining"a" transition" to" a"better" firm" as" a"transition"to" a" firm"with"profits" that" are" at"least" 10"
percent" higher" than" those" of" the" sending" firm." This" stronger" requirement" mitigates" eventual"
measurement"errors"and"corroborates"the"findings"of"the"main"specification"(first"two"columns"
of"the"upper"panel"of"Table"6).
Second,"we"estimate"equation"(1)"using"alternative"methods"to"rank"firms."As"reported"in"
Table" A5" of" the" Online" Appendix," we" obtain" qualitatively" similar" results" using" average" profits"
over"the"sample"period;"past"profits"that"were"made"before"the"job"transition"occurred;"profits"
per"worker;"value"added"and"total"factor"productivity"(TFP)"estimated"separately"by"industry"as"
in"Parrotta"and"Pozzoli"(2012).
4.3" " Definitions"of"Job"Transitions"
Jobs"differ"in"many"dimensions"beyond"the"type"of"the"firm"that"offers"them,"and"workers"may"
take"into"consideration" these" other" characteristics"when"deciding"whether" to" change" jobs." To"
control" for" this," we" restrict" the" definition" of" job" transitions" to" a" better" firm." Specifically," we"
impose"the"condition"that"movers"earn"higher"wages"or"are"employed"at"a"higher"occupational"
level" after" a" transition" to" a" better" firm." Using" these" restrictive" definitions" of" career"
improvements"in"job"transitions," we" find" strong" gender" differences"in"sorting,"in"line"with"the"
baseline"results"(the"first" two" columns" of" the"bottom"panel"of"Table" 6)." This" corroborates" the"
appropriateness"of"our"theoretical"framework.
As"we" mentioned" in" Section" 2.2,"because" of" changes" in" firms’"ownership,"there" may" be"
some" “false”" transitions" in" the" data." We" now" assess" the" robustness" of" our" main" results" by"
14"
"
setting"alternative"criteria" in" order" to"minimize"miscoded"movers."First"we" exclude" from" the"
sample" transitions" involving" more" than" either" 20"or" 10"percent" of" the" size" of" the" same"
sending"firm."The"first"panel"of"Table"A14"of"the"Online"Appendix"shows"that"the"coefficients"
estimated"on"the"worker’s"type"hardly"changes"across"the"two"sub-samples"of"movers"and"are"
very"similar" to"the"ones"reported"in"Table"2." Second,"given"that"when" two"or"more"firms"are"
merged"there"can"be"a"change"of"the"firm"identifier"but" not" of" the" establishment’s" number,"
we"identify" job" transitions" by" requiring"a"change" of" both"the"firm" and" of" the"establishment"
identifiers." In" an" additional" robustness" check" we" further" narrow" the" definition" of" job"
transition"by"adding"as"a"requirement"a"change"of"the"workers’"municipality"of"residence."We"
report"the" results"from"these"restricted"samples" of"movers"in"the"second" panel"of"Table"A14"
of" the" Online" Appendix." Despite" the" coefficients" are" slightly" different" compared" to" those"
reported"in" the"main"analysis,"gender" differences"in"favor"of" women"persist"and"confirm" the"
main"findings.
4.4" " Firm"Exit"
Our" results" could" arise" because" women" are" more" skilled" at" finding" a" job" outside" the" current"
firm."This"could"be"because"of"a"lower"search"cost"or"because"of"a"higher"investment"in"general"
versus"specific"human"capital"with"respect"to"men.
We"test"these"hypotheses"by"focusing" solely" on" transitions" from" a" firm’s"closure"(last"two"
columns"of" the" bottom"panel" of" Table"6)" since" these"mobility" patterns" do"not" stem" from"the"
voluntary" choices" and" career" concerns" of" employees." Indeed," in" such" situations," all" workers,"
including"men,"are"forced"to"seek"jobs"outside"of"their"current"firms.
Interestingly,"we" find"gender"differences"in"sorting"in"favor"of"men"in"these"job"transitions"
(albeit"not"always"significant"differences)."These"results"lend"additional"support"to"the"fact"that"
positive" sorting" in" job" transitions" is" stronger" for" women" because" of" voluntary" transitions"
triggered"by"gender"biases"in"promotions"in"the"sending"firm.
4.5" " Results"by"Cohort"
The"sorting" patterns"that"we"document"could"be"due" to"a"worse"initial"allocation"of"women"to"
firms"with" respect" to"men." This" would"imply" that" positive" sorting"in" job" transitions"should" be"
stronger" for" women" and" especially" so" for" younger" female" workers." Furthermore," wage"
increases" should" be" steeper" for" women." Alternatively," there" could" be" intrinsic" biological"
differences"(Ichino"and"Moretti,"2009):" since" women’s" rate" of" absenteeism" is"generally"higher"
than"that"of"men,"the"former"are"less"productive"(or"their"productivity"is"less"observable)"at"the"
beginning" of" their" careers." In" that" case," gender" gaps" in" sorting" should" be" smaller" for" older"
workers," if" not" in" favor" of" women," in" all" transitions." Finally," it" could" be" that" female" workers"
learn"more"slowly"than"men."Then,"females"would"be"less"likely"to"be"promoted"in"all"firms"and"
to"move"to"better"firms.
To"check" the" relevance"of" these" hypotheses,"we" analyze" gender"differences" in" sorting"by"
selecting" workers" aged" 25-30," 30-40," 40-50," or" 50-60" in" 1995" and" following" them" separately"
along"the"sample"period.
In"the"first"panel"of"Table"7,"we"present"results"for"job"transitions"by"cohort."We"find"strong"
15"
"
positive" sorting" parameters" for" women," large" gender" differences" for" younger" cohorts" (25-20;"
30-40)" and" weak" evidence" of" sorting" for" workers" 40-50" years" old." Sorting" is" negative" or"
negligible" for" women" and" men" in" the" oldest" age" cohorts." These" results," together" with" the"
gender-specific" wage" developments," allow" us" to" rule" out" the" hypothesis" that" the" gender"
differences" in" sorting" patterns" are" merely" driven" by" a" higher" extent" of" initial" mismatches" for"
women"with"respect"to"men."However,"the"case"of"movers"aged"50-60"is"very"peculiar"because"
a"large"share"of"these"workers"was"likely"approaching"early"retirement,"which"at"that"time"was"
strongly"supported"by"generous"public"programs.
The" second" panel" of" Table" 7" investigates" the" promotion" probability" of" stayers" by" cohort."
We"find" that" coefficients"on" the" previous" wage" are" similar"between" men" and" women" for" the"
youngest"cohort"but"a"gap"emerges"for"older"cohorts:"the"coefficient"for"males"is"twice"(one"and"
one-half"times)"larger"than"that"for"females"for"the"cohort"50-60"(40-50)."The"last"panel"of"Table"
7"examines"promotions"to"the"managerial"level:"gender"differences"are"significant"in"all"cohorts,"
although"the" probability" of"better" workers" being"promoted" is" much"higher" for" men"when" we"
consider" older" cohorts." Hence," contrary" to" the" biological" differences" hypothesis," gender"
differences"in"promotions"in"favor"of"men"are"stronger"for"middle-aged"workers.
The" discrepancies" between" men" and" women" in" job" transitions" are" also" confirmed" in" the"
sub-samples"that"refer"to"different"age"groups"(see"Table"A6"of"the"Online"Appendix).
Overall," the" analysis" by" age" groups" and" cohorts" yields" limited" support" for" the" hypothesis"
that" biological" differences" explain" gender" gaps." Indeed," while" gender" differences" are" more"
important"when"career"advancements"mostly"take"place,"i.e.,"for"workers"aged"between"35"and"
50" years," in" line" with" our" baseline" results," such" differences" are" in" favor" of" women" in" job"
transitions"and"in" favor" of"men"in"promotions."These"patterns"appear" instead"to"be"consistent"
with"the"idea"that"women"tend"to"climb"the"career"ladder"at"a"slower"pace"than"men;"hence,"
women" exhibit" an" increasing" gap" with" respect" to" men." This" lowers" a" woman’s" probability" of"
reaching"top-level"positions.
4.6" " Parenthood"
Several" studies" suggest" that" career" advancement" is" more" difficult" for" women" due" to"
motherhood" (Datta" Gupta," Smith," and" Verner," 2008;" Smith" et" al.," 2013;" Kleven" et" al.," 2015;"
Gallen,"2015)."We"now"test"this"hypothesis.
While"parenthood"per0se"does"not"appear"to"be"relevant"to"job"transitions"(see"Tables"A17"
and" A18" of" the" Online" Appendix)," the" impact" of" parenthood" on" career" advancement" varies"
substantially"across"firms.
Indeed,"Table" 8"reports"the"results" by"firm"type"(female-," not"female-friendly,"female-"and"
not" female-sought" companies)" for" the" sub-samples" before" and" after" the" first" child" is" born."
Female-friendly"firms"show"no"gender"differences"in"promotions"to"better"occupations" before"
the"first"child"is"born,"while"a"small"bias"in"favor"of"females"emerges"after"the"first"child"is"born."
However," women" who" work" in" other" firms" encounter" a" significant" penalty" in" promotion,"
especially" after" bearing" a" child." Regarding" promotions" to" managerial" positions," gender"
differences"appear"in"all"firms"independent"of"parenthood."However,"the"parenthood"penalty"is"
harsher"in"not"female-sought"and"not-female-friendly"firms"after"the"first"child"is"born.
Overall,"this"set"of"results"provides"evidence"of"an"interplay"between"motherhood"and"the"
16"
"
glass"ceiling"phenomenon"in"certain"firms.
4.7" " Further"Checks"
Our" results" could" be" due" to" a" different" distribution" of" skills," endowments" and" impediments"
across" genders." Alternatively," women" may" find" good" jobs" in" specific" occupations" or" sectors"
where" their" skills" are" more" valued." Yet," gender" gaps" in" job" transitions" also" emerge" when" we"
sub-sample"by"occupation" and" by"education"(Table"A7"of" the" Online"Appendix):"men"generally"
show"weaker"positive"sorting"patterns"in"job"transitions,"and"the"difference"between"genders"is"
larger"for"blue-collar"workers"and"for"workers"with"primary"education,"whereas"it"decreases"for"
more" educated" workers" and" for" those" with" better" occupations." The" latter" result" is" fairly"
consistent"with"Card"et"al."(2016),"who"conclude"that"sorting"differences"across"genders"are"less"
important"for"highly"educated"workers"and"managers.
Manning" (2003)," Ch." 7," documents" that" women" in" the" UK" are" more" constrained" in" their"
opportunities"to"change"jobs."We"test"whether"the"costs"associated"with"job"mobility"affect"our"
results"by"focusing"on"transitions" without" a" change" of" residence"and"for"single"women,"as"we"
expect" such" costs" to" be" lower" in" these" samples." We" find" that" sorting" in" job" transitions" is"
stronger"for"women"although"slightly"less"so"than"in"baseline"regression"(Table"A8"of"the"Online"
Appendix)."This"suggests"that"our"main"results"do" not" entirely" depend" on" the" costs" associated"
with"changing"employers"but"rather"on"career"concerns."Conversely,"the"reductions"in"the"labor"
supply" that" are" represented" by" shifts" from" full-time" to" part-time" employment" are" not"
associated" with" positive" sorting," as" changes" in" the" number" of" hours" worked" are" likely" to" be"
triggered" by" family" considerations." Further," the" finding" that" the" sorting" coefficient" in" job"
transitions"is"significantly"higher"for"women"with"a"family"network"might"reflect"the"importance"
for"women"of"having"good"connections.
It"does"not"appear"to" be" relevant" whether" movers" find" a"job"in"the"same"industry"or" in" a"
different" industry," which" emphasizes" that" the" results" are" not" driven" by" women" self-selecting"
into" particular" industries" (Table" A9" of" the" Online" Appendix)." Consistent" with" this" view," our"
results"on"job"transitions"do"not"depend"on"firm"size"(Table"A10"of"the"Online"Appendix).
Regarding" promotions," estimations" conducted" separately" by" education" show" that" gender"
differences" in" promotion" are" lower" for" workers" with" mandatory" and" tertiary" education"
compared"to"workers"with"secondary"education"(Table"A11"of"the"Online"Appendix)."Results"by"
industry" indicate" that" the" same" pattern" generally" emerges" in" all" sectors" (Table" A12" of" the"
Online"Appendix).
While"our"model"predicts"that"better"workers"are"either"more"likely"to"quit"jobs"in"order"
to"move"to"better"firms"or"to"be"promoted"within"the"current"firm,"it"is"silent"on"the"curvature"
of"these"relationships."However,"the"descriptive"analysis"in"Figure"2"and"3"seems"to"suggest"a"
convex"relationship"between" workers’" type"and"either"the"probability"of" transition" to"better"
firms" or" the" probability" of" promotion," respectively." We" now" assess" the" presence" of"
non-linearities"by"adding"to"our"main"specification"the"square"of"workers’" wages" (Table" A13"
of" the" Online" Appendix)." We" find" suggestive" evidence" in" support" of" a" convex" relationship"
between"workers’"type,"sorting"and"promotion"for"female"workers."For"men"the"results"with"a"
non"linear" specification" are" not" always"precisely" estimated." Hypothesis" testing" confirms"the"
gender"differences"highlighted"in"the"main"analysis.
17"
"
Finally,"our"main" results" reported"in"Table"2"may"be" driven" by"the"fact"that"we"focus" on" a"
selected"sample" of" workers" because"we" rule" out" part-timers"and"employees" of" small"firms"or"
firms" with" imputed" accounting" variables." The" first" two" panels" of" Table" A16" of" the" Online"
Appendix" allow" us" to" dismiss" these" concerns." We" obtain" a" similar" impression" of" the" gender"
differences"in"sorting"and"promotion"even"when"we"estimate"our"main"models"on"less-selected"
samples,"either"including"part-time"employeesxvi" or"employees"of"small"firms.
Another"concern"is"that"our"analysis" is" based" exclusively" on" workers" in" the"private"sector."
Indeed," female" workers" are" more" likely" to" take" a" more" “family-friendly”" occupation" in" the"
public" sector" after" the" first" child" is" born" than" their" male" counterparts." While" the" analysis" of"
sorting" cannot" be" extended" to" the" public" sector" due" to" the" absence" of" a" measure" of" firms’"
performance" for" public" firms," we" can" extend" the" analysis" of" promotion" to" the" public" sector."
When"we"do"so,"there"is"a"considerable"gender"gap"in"promotion"in"favor"of"men"even"for"this"
broader"sample"of"workers"(third"panel"of"Table"A16"of"the"Online"Appendix).
5" " Conclusions"
"
In" this" paper," we" measure" gender" differences" in" sorting" by" using" Danish" employer-employee"
matched" data" to" study" gaps" in" labor" market" outcomes." Our" methodology" is" centered" on" the"
relationship" between" workers’" ability," measured" by" their" position" in" the" wage" hierarchy" in" a"
given"firm,"and"the"probability"of"moving"to"a"better"firm"or"the"probability"of"being"promoted."
We"find"that" the" degree"of"positive"sorting"is"higher"for" women" than"for"men"in"voluntary"job"
transitions," while" it" is" higher" for" men" than" for" women" in" promotions," especially" in" firms" that"
have"fewer"female"workers"in"white"collar"positions"than"their"respective"industry"mean.
Our"detailed"account"of"gender"differences"provides"support"for"the"hypothesis"that"female"
workers" encounter" glass" ceilings" in" some" firms." This" obstacle" is" likely" to" lead" good" female"
workers"to"seek"firms"that"will"reward"their" talents" in" a" fair" manner." As" a" result," good"female"
workers"are"more"mobile"than"male"workers"in"the"direction" of" better" firms," while" it" is" easier"
for"good"male"workers"to"be"promoted."Nonetheless,"gender"differences"in"promotion"persist"
and"are"similar"in"all"firms"when"we"focus"on"large"career"advances.
The"gender"differences"in"sorting"that"we"document"are"broadly"consistent"with"an"overall"
gender"gap"in"labor"market"outcomes"and"an"under-representation"of"women"in"top"positions,"
as" observed" in" the" case" of" Denmark." Since" mobility" is" a" way" to" circumvent" gender-biased"
promotion"policies"in"certain"firms,"we"expect"gender"gaps"to"be"even"more"severe"in"countries"
with"less"flexible"labor"markets.
Furthermore," our" findings" suggest" that" women" who" become" mothers" have" difficulties"
advancing"in"their"careers"in"certain"firms."These"hurdles"may"be"associated"with"the"significant"
generosity" of" parental" leave" policies," as" suggested" by" Datta" Gupta" et" al." (2008)." Thus," it" is"
important"to"conduct"further"research"to"determine"why"these"effects"emerge"and"why"they"do"
so"only"in"some"firms.
18"
"
Figure"1:"Probability"of"changing"firms"against"the"deviation"of"estimated"residual"wages"earned"
in"the"sending"firm.
" "
Notes:"Residual"wages"are"predicted"from"a"Mincerian"regression"in"which"we"include"the"following"variables:"age"and"age"squared,"tenure"and"
tenure" squared," marital" status," education" level," family" network," experience" and" experience"squared," occupational" dummies," share" of" white"
collar"women"employed"in"the"sending"firm,"firm"fixed"effects"and"dummies"for"having"children,"foreigners,"sending"firm’s"size"and"years."
Figure"2:"Conditional"probability"of" moving" to" a"better"firm"against"the"deviation"of" estimated"
residual"wages"earned"in"the"sending"firm.
""
" "
Notes:"Residual"wages"are"predicted"from"a"Mincerian"regression"in"which"we"include"the"following"variables:"age"and"age"squared,"tenure"and"
tenure" squared," marital" status," education" level," family" network," experience" and" experience"squared," occupational" dummies," share" of" whi te"
collar"women"employed"in"the"sending"firm,"firm"fixed"effects"and"dummies"for"having"children,"foreigners,"sending"firm’s"size"and"years."
"
Figure" 3:" Probability" of" promotion" against" the" deviation" against" the" deviation" of" estimated"
residual"wages"earned"in"the"current"firm.
" "
Notes:"Residual"wages"are"predicted"from"a"Mincerian"regression"in"which"we"include"the"following"variables:"age"and"age"squared,"tenure"and"
tenure" squared," marital" status," education" level," family" network," experience" and" experience"squared," occupational" dummies," share" of" white"
collar"women"employed"in"the"sending"firm,"firm"fixed"effects"and"dummies"for"having"children,"foreigners,"sending"firm’s"size"and"years."
19"
"
Table"1:"Descriptive"statistics"
"
""Notes:"All"the"variables"are"averages"from"1996"to"2005."
"Variables" "
"Sample"of"movers" "
"Sample"of"stayers"""
"
"Women" "
"Men" " "
"Women" " "
"Men"""
"
"Mean" "
"S.d." "
"Mean" "
"S.d." "
"Mean" "
"S.d." "
"Mean" "
"S.d." "
Log"of"wage"(in"sending"firm)" "
"12.206" "
"0.506" "
"12.430" "
"0.522" "
"12.293" "
"0.429" "
"12.542" "
"0.426" " "
Age" "
"37.748" "
"9.055" "
"38.594" "
"9.345" "
"39.819" "
"9.281" "
"40.983" "
"9.554" " "
Tenure" "
"3.561" "
"3.795" "
"3.487" "
"3.809" "
"5.567" "
"4.877" "
"5.931" "
"5.126" " "
Labor"market"experience" "
"14.118" "
"8.358" "
"16.602" "
"9.152" "
"15.638" "
"8.278" "
"18.884" "
"9.161" "
Manager" "
"0.024" "
"0.155" "
"0.041" "
"0.199" "
"0.018" "
"0.131" "
"0.046" "
"0.208" " "
Middle"manager" "
"0.260" "
"0.438" "
"0.239" "
"0.427" "
"0.299" "
"0.458" "
"0.258" "
"0.437" " "
Blue"collar" "
"0.716" "
"0.451" "
"0.719" "
"0.449" "
"0.683" "
"0.465" "
"0.697" "
"0.460" " "
With"at"least"a"child"(0-3)" "
"0.149" "
"0.356" "
"0.148" "
"0.355" "
"0.125" "
"0.331" "
"0.125" "
"0.331" " "
Primary"(1,"if"with"primary"
education)" "
"0.366" "
"0.482" "
"0.300" "
"0.458" "
"0.380" "
"0.485" "
"0.292" "
"0.455" " "
Secondary"(1,"if"with"secondary"
and"post-secondary"education)" "
"0.561" "
"0.496" "
"0.644" "
"0.479" "
"0.552" "
"0.497" "
"0.650" "
"0.477" " "
Tertiary"(1,"if"with"tertiary"
education)" "
"0.073" "
"0.260" "
"0.056" "
"0.230" "
"0.068" "
"0.252" "
"0.058" "
"0.233" " "
Foreigner" "
"0.051" "
"0.220" "
"0.049" "
"0.216" "
"0.048" "
"0.213" "
"0.046" "
"0.209" " "
Family"network"(1,"if"one"parent"is"
manager)" "
"0.050" "
"0.217" "
"0.041" "
"0.198" "
"0.049" "
"0.217" "
"0.041" "
"0.199" " "
Married"or"cohabiting" "
"0.740" "
"0.439" "
"0.732" "
"0.443" "
"0.783" "
"0.412" "
"0.767" "
"0.423" " "
Share"of"white-collar"women"in"
the"sending"firm" "
"0.091" "
"1.772" "
"0.065" "
"1.637" " "
"
"
" "
" "
Share"of"white-collar"women"in"
the"current"firm" "
"0.186" "
"2.864" "
"0.109" "
"1.617" "
"0.031" "
"0.050" "
"0.017" "
"0.036" " "
Sending"firm"size"less"than"50"
employees" "
"0.118" "
"0.323" "
"0.166" "
"0.372" "
"""
" "
" "
" "
Sending"firm"size"between"51"and"
100"employees" "
"0.102" "
"0.303" "
"0.131" "
"0.338" " "
" "
"""
"
" "
Sending"firm"size"more"than"100"
employees" "
"0.780" "
"0.415" "
"0.703" "
"0.457" " "
"""
"""
" " "
""""
Current"firm"size"less"than"50"
employees" "
"0.140" "
"0.347" "
"0.188" "
"0.391" "
"0.147" "
"0.354" "
"0.194" "
"0.395" " "
current"firm"size"between"51"and"
100"employees" "
"0.112" "
"0.315" "
"0.139" "
"0.346" "
"0.117" "
"0.321" "
"0.144" "
"0.351" " "
Current"firm"size"more"than"100"
employees" "
"0.748" "
"0.434" "
"0.673" "
"0.469" "
"0.736" "
"0.441" "
"0.662" "
"0.473" " "
Sending"firm"accounting"profit"
before"taxes"per"worker" "
"86.292" "
"288.338" "
"87.682" "
"278.075" "
" "
" "
" "
" "
Current"firm"accounting"profit"
before"taxes"per"worker" "
"100.746" "
"482.845" "
"96.862" "
"440.232" "
"71.103" "
"3288.197" "
"87.630" "
"2311.018" " "
Prob(profits"of"current"firm"
>profits"of"previous"firm"by"5%)" "
"0.401" "
"0.493" "
"0.378" "
"0.490" "
"""
"""
"""
"""
Prob(profits"of"current"firm"profits"
>"of"previous"firm"by"10%)" "
"0.356" "
"0.412" "
"0.349" "
"0.402" "
"""
"""
"""
"""
Promotion"(better"occupation)" "
" "
" "
" "
" "
"0.030" "
"0.109" "
"0.032" "
"0.144" "
Promotion"(manager)" "
" "
" "
" "
" "
"0.033" "
0.028" "
"0.035" "
"0.055" "
Obs" "
"126,676" "
"294,073" "
"1,329,800" "
"2,773,928" " "
Number"of"individuals" "
"97,502" "
"218,542" "
"368,810" "
"663,237" "
Number"of"firms" "
"16,764" "
"18,034"
20"
"
Table"2:"Sorting"in"job"transitions"and"promotions:"main"results
" "
"Sorting"in"job"transitions" "
"Promotions" "
"Promotions"to"manager"""
"
"Women" "
"Men" "
"Women" "
"Men" "
"Women" "
"Men" "
Log"of"wage"(in"sending"firm)" "
"0.016***" "
"0.007***" "
"0.013***" "
"0.023***" "
"0.002***" "
"0.007***" "
"
"(0.000)" "
"(0.001)" "
"(0.002)" "
"(0.003)" "
"(0.000)" "
"(0.001)" "
Percentage"of"white-collar" "
" -0.078***" "
"0.097***" "
"""
" "
"""
" " "
women"in"sending"firm"
"(0.000)" "
"(0.000)" "
"""
"""
"""
" " "
Percentage"of"white-collar" "
"0.062***" "
"0.091***" "
" "
"""
"""
" " "
women"in"receiving"firm"
"(0.000)" "
"(0.000)"
"""
" "
"""
" "
Share"of"white-collar"women" "
" "
" "
"0.780***" "
"0.676***" "
"0.017***" "
"0.041***" "
in"the"firm"
" "
" "
"(0.039)" "
"(0.021)" "
"(0.003)" "
"(0.004)" "
N" "
"126,676" "
"294,073" "
"1,329,800" "
"2,773,928" "
"1,329,800" "
"2,773,928" "
2
R
"
"0.124" "
"0.130" "
"0.019" "
"0.021" "
"0.004" "
"0.011" "
Hypothesis"test"[
2
c
;"p-value]:"
"
"
""
menwomen
11 =
aa
"
"149.50;"0.000" "
"110.78;"0.000" "
"323.27;"0.000" "
Notes:"*Statistically" significant" at" the" .10" level,"**at" the" .05" level," and" ***at" the" .01" level." For" job" transitions," the" dependent" variable" is" a"
dummy"that"takes"value"1"if"the"worker"moves"to"a"firm"with"profits"that"are"at"least"5%"higher"than"those"of"the"previous"firm."For"promotions,"
the" dependent" variable" is" a" dummy" that" takes" value" 1" if" the" worker" is," within" the" same" firm," promo ted" to" a" higher" occupational" level." For"
promotions"to"managerial"positions,"the"dependent"variable"is"a"dummy"that"takes" value"one"if"the"worker"is,"within" the"same"firm," promoted"
to"a" managerial"occupational" level."All" specifications"include" age"and" age"squared," tenure"and" tenure"squared," marital"status,"having"children,"
education" level," family" network," a" dummy" for" foreigners," experience" and" experience" squared," firm" fixed" effects," firm" size" dummies" (both"
receiving" and" sending" firm" in" the" regressions" regarding" job" transitions)," year" and" o ccupational" dummies." The" standard" errors" reported" in"
parentheses"are"clustered"at"the"sending"firm"level"and"at"the"individual"level."
"
Table"3:"Sorting"in"job"transitions"and"promotions"by"female"friendliness"of"firms""
" "
"Sorting"in"job"transitions"""
"
""Transition"to"female-friendly"firms" "
""Transition"not"to"female-friendly"firms" "
"
"Women" "
"Men" "
"Women" "
"Men" "
Log"of"wage"in"sending"firm" "
"0.023***" "
"0.005***" "
"0.006***" "
"0.007***"
"
"(0.003)" "
"(0.000)" "
"(0.002)" "
"(0.002)" "
N" "
"77,383" "
"157,585" "
"49,293" "
"136,488"
2
R
" "
"0.098" "
"0.109" "
"0.165" "
"0.150"
Hypothesis"test"[
2
c
;"p-value]:"
""
"
""
"
menwomen
11 =
aa
"
"27.51;"0.000" "
"0.02;"0.897" "
" "
"Promotions"""
"
""Female-friendly"firms" "
""Non-female-friendly"firms" "
""Female-sought"firms" "
"
"Women" "
"Men" "
"Women" "
"Men" "
"Women" "
"Men" "
Log"of"wage" "
"0.015***" "
"0.012***"
"0.011**" "
"0.025***" "
"0.014***" "
"0.009***" "
"
"(0.003)" "
"(0.003)" "
"(0.001)" "
"(0.003)" "
"(0.003)" "
"(0.002)"
N" "
"459,405" "
"574,378" "
"870,395" "
"2,199,550" "
"391,628" "
"449,077" "
2
R
" "
"0.019" "
"0.023" "
"0.009" "
"0.022" "
"0.027" "
"0.033" "
Hypothesis"test"[
2
c
;"p-value]:"
""
"
"
" " "
menwomen
11 =
aa
"
"31.57;"0.000" "
"134.60;"0.000" "
"62.82;"0.000" "
"
"Promotions"to"managerial"occupation"""
"
""Female-friendly"firms" "
""Not"female-friendly"firms" "
""Female-sought"firms" "
"
"Women" "
"Men" "
"Women" "
"Men" "
"Women" "
"Men" "
Log"of"wage" "
"0.003***" "
"0.007***"
"0.002***" "
"0.007***" "
"0.003***" "
"0.006***"
"
"(0.000)" "
"(0.000)"
"(0.000)"
"(0.000)"
"(0.000)" "
"(0.000)" "
N" "
"459,405" "
"574,378" "
"870,395" "
"2,199,550" "
"391,628" "
"449,077"
2
R
" "
"0.005" "
"0.009"
"0.003" "
"0.012" "
"0.005" "
"0.009"
Hypothesis"test"[
2
c
;"p-value]:"
""
"
""
"
""
"
menwomen
11 =
aa
"
"69.26;"0.000" "
"376.79;"0.000" "
"93.37;"0.000" "
Notes:"*Statistically"significant"at"the".10"level," **at"the".05" level,"and"***at" the".01"level." For"promotions,"the"dependent"variable"is"a"dummy"
that"takes"value"1"if"the"worker"is"promoted"to"a" better"occupation"in"the"same"firm"or"if"the"worker"is"promoted"to"a"managerial"occupational"
level"in" the"same"firm." All"specifications" include"the"same" controls"as" the"regressions"in" Table"2." The"standard"errors" reported"in" parentheses"
are"clustered"at"the"sending"firm"level"and"at"the"individual"level.
21"
"
Table"4:"Sorting"in"job"transitions"and"promotions"using"fixed"effects"from"AKM
""
" "
"Sorting"in"job"transitions" "
"Promotions" "
"Promotions"to"manager" "
"
"Women" "
"Men" "
"Women" "
"Men" "
"Women" "
"Men" "
Individual"fixed"effects" "
"0.032***" "
"0.011***" "
"0.038***" "
"0.065***" "
"0.005***" "
"0.020***" "
"
"(0.002)" "
"(0.001)" "
"(0.003)" "
"(0.001)" "
"(0.000)" "
"(0.001)" "
N" "
"123,154" "
"287,100" "
"1,310,132" "
"2,735,987" "
"1,310,132" "
"2,735,987" "
2
R
" "
"0.126" "
"0.130" "
"0.023" "
"0.032" "
"0.005" "
"0.017" "
Hypothesis"test"[
2
c
;"
p-value]"
menwomen
11 =
aa
"
" "
"""""""""1201.36;"0.000" """
"
"""146.55;"0.000"
"
160.28;"0.000" "
"
Notes:"*Statistically" significant" at" the" .10" level," **at" the" .05" level," and" ***at" the" .01" level."All" specifications" include" the" same"dependent"
variables"and" controls"as" the"regressions" in"Table" 2."The" standard"errors" reported"in"parentheses"are"clustered"at"the"sending"firm"and"at"the"
individual"level."
Table"5:"Sorting"in"job"transitions"using"the"methodology"from"BDM
" "
"Strength"of"sorting" "
"Sign"of"sorting"(1)" "
"Sign"of"sorting"(2)" "
"
"Women" "
"Men" "
"Women" "
"Men" "
"Women" "
"Men" "
Sending"firm’s"profits"per" "
"0.345***" "
"0.253***" "
" "
" "
" "
" "
worker"
"(0.003)" "
"(0.002)" "
" "
" "
" "
" "
h
"as"in"BDM" "
"0.587" "
"0.503" "
" "
" "
" "
" "
Log"of"wage"in" "
" "
" "
"0.173***" "
"0.094***" "
" "
" "
receiving"firm"
" "
" "
"(0.011)" "
"(0.007)" "
" "
" "
Log"of"wage"in" "
" "
" "
" "
" "
"0.039***" "
"0.010***" "
sending"firm"
" "
" "
" "
" "
"(0.006)" "
"(0.003)" "
N" "
"78,842" "
"199,233" "
"78,842" "
"199,233" "
"78,842" "
"199,233" "
2
R
" "
"0.464" "
"0.359" "
"0.146" "
"0.063" "
"0.693" "
"0.625" "
Hypothesis"test"[
2
c
;"
p-value]"
menwomen
11 =
aa
"
"
777.86;"0.000" "
"""
"""""
"
Notes:"*Statistically"significant" at" the" .10" level,"**at" the" .05" level," and" ***at"the".01" level." In" the" first"and"last" two" columns," the" dependent"
variable" is" the" receiving" firm’s" profits" per" worker." In" the" third" and" fourth" columns," the" dependent" variable" is" the" sending" firm’s" profits" per"
worker."
Table"6:"Sorting"in"job"transitions"by"type"of"transitions"
Notes:"*Statistically"significant" at"the".10" level,"**at"the" .05"level,"and" ***at"the".01" level."All"specifications" include"the"same"controls" as" the"
regressions"in"Table"2."The"standard"errors"reported"in"parentheses"are"clustered"at"the"sending"firm"level"and"at"the"individual"level."
" "
""Profits">"10%"
""Wage"improvement" "
"
"Women" "
"Men" "
"Women" "
"Men" "
Log"of"wage"in"sending"firm" "
"0.014***" "
"0.005***" "
"0.020***" "
"0.009***"
"
"(0.000)" "
"(0.001)" "
"(0.003)" "
"(0.002)"
N" "
"126,676" "
"294,073" "
"50,943" "
"97,662" "
2
R
" "
"0.122" "
"0.130" "
"0.204" "
"0.169"
Hypothesis"test"[
2
c
;"p-value]:"
menwomen
11 =
aa
"
"
67.72;"0.000"
"""
47.08;"0.000"
"
""Better"occupational"level" "
""Transitions"from"firm"exit" "
"
"Women" "
"Men" "
"Women" "
"Men" "
Log"of"wage"in"sending"firm" "
"0.017***" "
"0.003***" "
"0.009***" "
"0.011***" "
"
"(0.001)" "
"(0.001)" "
"(0.000)" "
"(0.000)"
N" "
"17,520" "
"39,878" "
"26,083" "
"57,820"
2
R
" "
"0.112" "
"0.118" "
"0.084" "
"0.117" "
Hypothesis"test"[
2
c
;"p-value]:"
"""
"
"""
"
menwomen
11 =
aa
"
"250.99;"0.000" "
"11.05;"0.000" "
22"
"
Table"7:"Sorting"in"job"transitions"and"promotions:"results"by"cohort
" "
"Sorting"in"job"transitions"""
"
""Cohort"25-30" "
"Cohort"30-40" "
""Cohort"40-50" "
""Cohort"50-60" "
"
"0Women" "
Log"of"wage"in"sending"firm" "
"0.026***" "
"0.029***" "
"0.005***" "
" -0.020***" "
"
"(0.001)"
"(0.004)" "
"(0.001)" "
"(0.000)" "
N" "
"2,876" "
"3,679" "
"1,945" "
"630" "
2
R
" "
"0.143" "
"0.142" "
"0.134" "
"0.211" "
"
"0Men" "
Log"of"wage"in"sending"firm" "
"0.002***" "
"0.009***" "
" -0.001" "
" -0.001***" "
"
"(0.001)" "
"(0.002)" "
"(0.001)")" "
"(0.000)" "
N" "
"11,124" "
"16,982" "
"9,133" "
"4,027" "
2
R
" "
"0.153" "
"0.142" "
"0.129" "
"0.123" "
Hypothesis"test"[
2
c
;"p-value]:"
""""
"
""""
"
menwomen
11 =
aa
"
"1046.97;"0.000" "
"92.37;"0.000" "
"8.49;"0.000" "
"683.65;"0.000" "
"
"Promotions"""
"
""Cohort"25-30" "
""Cohort"30-40" "
""Cohort"40-50" "
""Cohort"50-60" "
"
"0Women" "
Log"of"wage" "
"0.018***" "
"0.029***" "
"0.039***" "
"0.044***" "
"
"(0.001)" "
"(0.002)" "
"(0.002)" "
"(0.002)" "
N" "
"121,018" "
"199,285" "
"175,239" "
"52,682" "
2
R
" "
"0.025" "
"0.030" "
"0.033" "
"0.032" "
"
"0Men" "
Log"of"wage" "
"0.018***" "
"0.041***" "
"0.059***" "
"0.091***" "
"
"(0.002)" "
"(0.001)" "
"(0.002)" "
"(0.003)" "
N" "
"271,866" "
"477,224" "
"451,131" "
"169,002" "
2
R
" "
"0.024" "
"0.037" "
"0.047" "
"0.062" "
Hypothesis"test"[
2
c
;"p-value]:"
"""
"
"""
"
menwomen
11 =
aa
"
"0.01;"0.90" "
"25.50;"0.000" "
"34.18;"0.000" "
"102.02;"0.000" "
"
"Promotions"to"managerial"occupations"""
"
""Cohort"20-30" "
""Cohort"30-40" "
""Cohort"40-50" "
""Cohort"50-60" "
"
"0Women" "
Log"of"wage" "
"0.002***" "
"0.005***" "
"0.007***" "
"0.007***" "
"
"(0.000)" "
"(0.000)" "
"(0.000)" "
"(0.001)" "
N" "
"121,018" "
"199,285" "
"175,239" "
"52,682" "
2
R
" "
"0.004" "
"0.015" "
"0.011" "
"0.007" "
"
"0Men" "
Log"of"wage" "
"0.003***" "
"0.013***" "
"0.019***" "
"0.038***" "
"
"(0.000)" "
"(0.000)" "
"(0.000)" "
"(0.003)" "
N" "
"271,866" "
"477,224" "
"451,131" "
"169,002" "
2
R
" "
"0.005" "
"0.025" "
"0.024" "
"0.035" "
Hypothesis"test"[
2
c
;"p-value]:"
"""
"
"""
"
menwomen
11 =
aa
"
"15.09;"0.000" "
"169.49;"0.000" "
"123.23;"0.000" "
"72.21;"0.000" "
Notes:"*Statistically" significant" at" the" .10" level," **at" the" .05" level," and" ***at" the" .01" level."All" specifications" include" the" same"dependent"
variables"and"controls" as"the"regressions" in"Table"2." The"standard"errors" reported"in"parentheses" are"clustered"at"the"sending"firm"level"and"at"
the"individual"level."
"
"
"
"
"
"
"
23"
"
Table"8:"Sorting"in"promotions"by"female"friendliness"of"firms"before"and"after"children
" "
""Promotion"to"better"occupation" " "