Union membership and the wage gap between the public and private sectors: Evidence from China

Ma, Xinxin
A icle
Union membe ship and he wage gap be ween he public
and p i a e sec o s: E idence om China
Jou nal o Labou Ma ke Resea ch
P o ided in Coope a ion wi h:
Ins i u e o Employmen Resea ch (IAB)
Sugges ed Ci a ion: Ma, Xinxin (2024) : Union membe ship and he wage gap be ween he public
and p i a e sec o s: E idence om China, Jou nal o Labou Ma ke Resea ch, ISSN 2510-5027,
Sp inge , Heidelbe g, Vol. 58, Iss. 1, pp. 1-25,
h ps://doi.o g/10.1186/s12651-024-00361-2
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/308533
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/4.0/
Ma J Labou Ma ke Res (2024) 58:3
h ps://doi.o g/10.1186/s12651-024-00361-2
ORIGINAL ARTICLE Open Access
© The Au ho (s) 2024. Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License, which
pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long as you gi e app op ia e c edi o he
o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons licence, and indica e i changes we e made. The images o
o he hi d pa y ma e ial in his a icle a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line
o he ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons licence and you in ended use is no pe mi ed by s a u o y
egula ion o exceeds he pe mi ed use, you will need o ob ain pe mission di ec ly om he copy igh holde . To iew a copy o his
licence, isi h p:// c ea i eco mmons. o g/ licen ses/ by/4. 0/.
Jou nal o Labou Ma ke Resea ch
Union membe ship and hewage gap
be ween hepublic andp i a e sec o s:
e idence omChina
Xinxin Ma1*
Abs ac
As ade unions a e ac i e in co po a ions wo ldwide, hei e ec s on he labo ma ke ha e a ac ed global a en-
ion. Howe e , he e is sca ce empi ical e idence ega ding how ade unions’ e ec s on he wage gap di e
be ween he public and p i a e sec o s, especially in he Chinese con ex . Using na ional longi udinal su ey da a
om he China Family Panel S udies o he yea s 2010, 2012, 2014, 2016, and 2018 and he Blinde -Oaxaca decompo-
si ion me hod, his s udy es ima es unions’ e ec on he wage gap be ween he public and p i a e sec o s in China.
The esul s om he O dina y Leas Squa es (OLS) me hod indica e he exis ence o a signi ican posi i e union wage
p emium na ionwide. Addi ionally, he p emium in he public sec o is g ea e han ha in he p i a e sec o . How-
e e , his e ec becomes insigni ican a e accoun ing o indi idual he e ogenei y using he ixed e ec s model. The
decomposi ion esul s based on he OLS me hod indica e ha he union co e age di e ence ( he endowmen e ec )
widens he wage gap be ween he public and p i a e sec o s; con e sely, he union wage p emium di e ence ( he
p ice e ec ) na ows he wage gap. These esul s indica e ha a policy expanding union co e age in he p i a e sec-
o may e ec i ely na ow he wage gap be ween bo h sec o s.
Keywo ds Wage gap, Public and p i a e sec o , Union co e age, Union wage p emium, China
JEL Classi ica ion J51, J53, J31, P21, P31, P41
1 In oduc ion
As ade unions (he eina e , unions) a e ac i e in co -
po a ions wo ldwide, hei e ec s on he labo ma ke
ha e a ac ed global a en ion in de eloped (Lewis 1963;
F eeman 1980; Pa sley 1980; Hi sch and John 1986; Ca d
1996; Chang and Huang 2016; Fa be e al. 2021; Kulka ni
and Hi sch 2021; Masso e al. 2022), de eloping (Casale
and Posel 2010; Gunde son 2016; Ke and Wi enbe g
2021), and eme ging ma ke economies (Yuan 2015;
Magda e al. 2016; Li and Song 2017; Yang e al. 2018; Sun
and Liu 2019). A union can p o ec i s membe s h ough
collec i e ba gaining and by ob aining en om a co -
po a ion. Consequen ly, his would lead o an inc ease
in he union wage p emium (e.g., Lewis 1963; F eeman
1980). Highe wages o union membe s can lead o a
wage gap o income inequali y (Ca d 1996). Nume ous
s udies ha e examined he “union wage p emium” in
de eloped coun ies (Lewis 1963; F eeman 1980; Ca d
1996; Blanch lowe and B yson 2010; Rosen eld and Den-
ice 2019; Choi and Ramos 2021; Ke and Wi enbe g
2021) and de eloping o eme ging ma ke economies,
including China (Yao and Zhong 2013; Li and Xu 2014;
Yuan 2015; Gunde son e al. 2016; Magda e al. 2016; Li
and Song 2017; Yang e al. 2018; Sun and Liu 2019). How-
e e , he e is sca ce empi ical e idence ega ding how
unions’ e ec s on he wage gap di e be ween he pub-
lic and p i a e sec o s (Blanch lowe and B yson 2010;
Rosen eld and Denice 2019). This s udy is he i s o
*Co espondence:
Xinxin Ma
[email p o ec ed]
1 Facul y o Economics, Hosei Uni e si y, 4342 Machi a-Shi Aiha amachi,
Tokyo 194-0298, Japan
3 Page 2 o 25
X.Ma
in es iga e union membe ship’s e ec s on he wage gap
be ween he public and p i a e sec o s in China, he la g-
es coun y unde going an economic sys em ansi ion.
The wage gap be ween he public and p i a e sec o s
is a ho opic in bo h de eloped and eme ging ma ke
economies (Ca dozo and Cunha 2019; Gindling e  al.
2020; Ma ia and Ch is o ides 2020; Sławinska 2020; Ma
and Li 2022). Nume ous s udies ha e a gued ha because
he public sec o [e.g., s a e-owned en e p ises (SOEs)]
is usually suppo ed by he go e nmen , i can easily
become a monopoly sec o ha ob ains a highe monop-
oly o poli ical en (Shlei e and Vishny 1994, 1997;
Lin e al. 2020; Iwasaki e al. 2022; Jin e al. 2022). Con-
sequen ly, i can se a highe wage le el han ha in he
p i a e sec o , which leads o a wage gap be ween bo h
sec o s.
Many coun ies ha e aimed o na ow he wage gap
be ween he public and p i a e sec o s, ye his issue pe -
sis s wo ldwide (Ca dozo and Cunha 2019; Gindling e al.
2020; Ma ia and Ch is o ides 2020; Sławinska 2020; Ma
and Li 2022). Using wage decomposi ion me hods, Chen
e al. (2005), Zhang and Xue (2008), Ye e al. (2011), Dem-
u ge e al. (2012), Zhang (2012), Ma (2018), and Ma and
Li (2022) ound ha di e ences in wo ke s’ endowmen
(e.g., educa ion) and wage-se ing sys ems (e.g., e u n
o educa ion o wo k expe ience) gene a e a wage gap
be ween he public and p i a e sec o s. Howe e , p io
s udies ha e no in es iga ed unions’ e ec s on ha gap.
Blanch lowe and B yson (2010) and Rosen eld and
Denice (2019) epo ed ha bo h union membe ship
densi y, collec i e ag eemen co e age and union wage
p emiums di e ences exis be ween he public and p i-
a e sec o s in he US and he UK. Howe e , hey ha e
no in es iga ed how he wo componen s—(i) he di -
e ence in union co e age ( he endowmen e ec ) and
(ii) he di e ence in union wage p emium ( he p ice
e ec )—a ec he wage gap simul aneously. Fu he ,
policy implica ions di e be ween he wo componen s:
when he di e ence in union co e age is he main com-
ponen , expanding union co e age in he p i a e sec o
is expec ed o na ow he wage gap; when he di e ence
in he union wage p emium is he main componen , he
policy o imp o ing he collec i e ba gaining unc ion in
he p i a e sec o may con ibu e o na owing he wage
gap. The e o e, bo h componen s’ magni udes mus be
examined simul aneously and compa ed. Thus, a concu -
en in es iga ion o each componen ’s magni ude has
bo h academic and p ac ical signi icance. The pu pose o
his s udy is o p o ide some empi ical e idence o cla i y
bo h componen s, which could o e aluable insigh s o
policymaking.
I chose China as he esea ch se ing o wo easons.
Fi s , he p opo ion o wo ke s in he public sec o is
inconsequen ial in de eloped coun ies. In con as ,
China ins i u ionally ansi ioned om a planned
economy o a ma ke -o ien ed economy in 1978. Fu -
he mo e, as i s go e nmen implemen ed a g adual-
is e o m s a egy aiming o main ain he exis ence
o la ge-size SOEs (Lin e al. 1994, 2020; Iwasaki e al.
2022; Jin e al. 2022), China has he g ea es p opo ion
o wo ke s in he public sec o wo ldwide. Hence, a su -
icien ly la ge sample is a ailable o ully examine his
issue. Second, because cu en economic heo ies on
union e ec s a e based on capi alis ma ke s in de el-
oped coun ies, hey may no explain union e ec s in
China, which is he wo ld’s la ges de eloping coun-
y and is unde going an economic sys em ansi ion.
Hence, empi ical esul s om China can p o ide new
insigh s ha en ich economic heo ies conce ning
unions.
This s udy con ibu es signi ican ly o he li e a u e in
wo ways. Fi s , al hough some s udies ha e e ealed ha
he union membe ship densi y and union wage p emium
di e be ween he public and p i a e sec o s (Blanch-
lowe and B yson 2010; Rosen eld and Denice 2019),
hey ha e no in es iga ed how di e ences in union
membe ship densi y and union wage p emiums a ec
he wage gap be ween he public and p i a e sec o s.
This s udy is he i s o explo e he e ec s o wo com-
ponen s o he wage gap simul aneously, based on he
Blinde -Oaxaca (B-O) decomposi ion me hod (Blinde
1973; Oaxaca 1973). Second, ega ding he es ima ions
o he union wage p emium, mos s udies ha e used sin-
gle- ime-poin c oss-sec ional su ey da a and ha e no
add essed he indi idual he e ogenei y p oblem, excep
o some s udies om de eloped coun ies (e.g., F ee-
man 1980). Based on na ional longi udinal su ey da a
om i e wa es (2010, 2012, 2014, 2016, and 2018) o he
China Family Panel S udies (CFPS), his s udy a emp s
o add ess his issue, using a ixed-e ec s (FE) model ha
has been used in empi ical s udies o de eloped coun-
ies (e.g., F eeman 1980). I also use a selec ion- bias-
adjus ed model wi h union lagged a iables (LV), and
p opensi y sco e ma ching (PSM) o add ess o he endo-
genei y issues. Al hough his s udy does no ackle he
endogenei y issue comple ely, i p o ides aluable empi -
ical da a, compa ed wi h p e ious s udies, especially in
he Chinese con ex .
The emainde o his pape is o ganized as ollows.
Sec ion 2 in oduces he ins i u ional backg ound and
de elops he wo es able hypo heses. Sec ion 3 in o-
duces he me hodology including he models, da a, and
a iables used. Sec ion 4 desc ibes he esul s o he
desc ip i e s a is ics. Sec ion5 epo s and explains he
esul s o he econome ic analysis. Finally, Sec .6 con-
cludes he s udy.
Page 3 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
2 Ins i u ional backg ound, heo y,
andhypo hesis de elopmen
2.1 Ins i u ional backg ound
The unc ion o unions in China changed wi h he SOE
e o m. Du ing he 1949–1977 planned-economy pe iod,
he All-China Fede a ion o T ade Unions(ACFTU), a
na ional union o ganiza ion, was o med and managed
by he go e nmen . Subsequen ly, he Chinese go e n-
men p omulga ed he i s T ade Union Law in June
1950. The Resolu ion on he Wo k o he All-China Fede -
a ion o T ade Unions, published on Decembe 22, 1951,
s a ed he ollowing:
I is necessa y o make all ade union wo ke s u he
clea ly unde s and he pa y’s impo ance o unions,
and each le el o he ACFTU mus wo k unde he uni-
ied leade ship o he commi ees o he Communis Pa y
o China (CPC) a he same le el…Unde he people’s
democ a ic sys em, he mos basic and impo an ask
o unions is o o ganize and educa e wo ke s o inc ease
labo p oduc i i y, ensu e he comple ion o he na ional
p oduc ion plan, and s i e o o e ul ill; on he basis o
imp o ing he p oduc ion, i is necessa y o equen ly
ca e o he daily needs o wo ke s, p o ide se ices o
ul ill hese needs, and lead he wo ke s in he s uggle o
socialism.
Du ing he planned-economy pe iod, he ACFTU
es ablished b anches a he egional o indus ial le el,
while egional unions managed indus y-le el unions
(Guo and Dai 2022). Thus, all unions we e managed by
he ACFTU. Howe e , unlike unions in Wes e n coun-
ies, he ACFTU did no ha e he igh o collec i ely
ba gain wi h employe s o bo h employees’ wage le els
and employmen because wage-se ing and employmen
we e managed by he cen al go e nmen du ing his
pe iod (Ma 2018).
Subsequen ly, he Chinese go e nmen implemen ed
ma ke -o ien ed e o ms, beginning in 1978, and, since
he la e 1990s, has en o ced he SOE e o m (which p o-
mo es he p i a iza ion o middle- and small-sized SOEs,
as well as en o ces o es ablish mode n co po a e go -
e nance in SOEs). The go e nmen has gi en SOEs mo e
au onomy o wage-se ing and employmen while simul-
aneously p omo ing he de elopmen o en e p ises
in he p i a e sec o . Since hen, p i a ely owned en e -
p ises (POEs) and o eign in es men en e p ises (FIEs)
ha e de eloped conside ably (Lin e al. 1994, 2020).1
Wi h he p og ess o ma ke -o ien ed e o ms, labo
dispu es abou wage le els and employmen ha e
inc eased in China.2 To add ess such new p oblems in
he labo ma ke , he Chinese go e nmen p omulga ed
he Second T ade Union Law in Ap il 1992, he Labo
Law o he People’s Republic o China in 1995, and he
Labo Con ac Law o he People’s Republic o Chinain
2008. These egula ions s a ed ha “Unions shall ake
measu es o p omo e he implemen a ion o he Labo
Con ac Law and he de elopmen o ha monious and
s able employmen ela ionships.” The 24 h mee ing o
he Commi ee o he Nin h Na ional People’s Cong ess
passed a e ision o he T ade Union Law on Oc obe
27, 2001, and he 32nd mee ing o he Commi ee o he
Thi een h Na ional People’s Cong ess passed a e ision
o he T ade Union Law on Decembe 24, 2021. A icle
2 o he 2021 T ade Union Law added he p o ision ha
“all o ganiza ions o he ACFTU ep esen he in e es s o
employees and sa egua d he legi ima e igh s and in e -
es s o employees in acco dance wi h he law.” Fu he -
mo e, A icle 6 added he p o ision ha “p o ec ing he
legi ima e igh s and in e es s o employees is one o he
basic esponsibili ies o he ACFTU.” The en e p ises eli-
gible o hese egula ions include hose in bo h he pub-
lic and p i a e sec o s (You 2017; Guo and Dai 2022).
Addi ionally, in con as wi h de eloped coun ies (e.g.,
manage s canno become union membe s), acco ding o
China’s T ade Union Law, bo h manage s and non-man-
age employees can choose o join he i ms o o ganiza-
ions’ unions olun a ily.
Du ing he cu en ma ke -o ien ed e o m pe iod,
acco ding o he egula ions men ioned abo e, al hough
i m-and indus ial-le el ade unions a e s ill managed
by he ACFTU, hey ha e ob ained collec i e ag eemen
igh s and can in luence wage-se ing and employmen
a ce ain le els h ough collec i e ag eemen s wi h
employe s. The nego ia ion usually akes place a he
i m le el. A gi en i m’s union mus epo he collec i e
ag eemen esul s o a highe -le el ACFTU b anch (e.g.,
egional b anch o he ACFTU).
Fu he mo e, he e has been a di e ence in union
membe ship densi y be ween he public and p i a e sec-
o s du ing his pe iod. As unions ha e been es ablished
in he public sec o , including SOEs and public o ganiza-
ions (e.g., public school, public hospi al, public esea ch
ins i u e), and ha e been suppo ed by he go e n-
men 3 since he planned-economy pe iod, mos egula
1 Fo example, he numbe o employees in non-SOEs (including POEs,
FOEs, and sel -employed indi iduals) inc eased om 86.67 million in 1990
o 522.13 million in 2019, while ha o SOEs dec eased om 103.46 million
in 1990 o 82.61 million in 2019 (Na ional Bu eau o S a is ics 2020).
2 Based on he China Labo S a is ics Yea book (Na ional Bu eau o S a is-
ics and Minis y o Labou and Social Secu i y 1992–2022), he o al num-
be o labo e s in ol ed in labo dispu es inc eased om 17,140 in 1992 o
1214,328 in 2008 and 1404,754 in 2021.
3 Fo example, he union und is composed o i e pa s: (1) union membe
ees om wo ke s (app oxima ely 10 CNY pe mon h); (2) union ees om
co po a ions o o ganiza ions (2% o he o al employees’ wage bills); (3)
p o i s om SOEs and public o ganiza ions ha a e sen o he go e nmen ;
(4) go e nmen subsidies; and (5) miscellaneous i ems.
3 Page 4 o 25
X.Ma
wo ke s in his sec o a e likely o join unions. Al hough
he ACFTU has en o ced he es ablishmen o union
b anches in POEs and FIEs since 2000, he union mem-
be ship densi y in he p i a e sec o is s ill lowe han
ha in he public sec o (You 2017; Guo and Dai 2022).
2.2 Hypo hesis de elopmen
In his sec ion, I o mula e wo hypo heses based on eco-
nomic heo ies and he ins i u ional backg ound men-
ioned o he empi ical s udy abo e.
When conside ing he union wage p emium and i s
dispa i y be ween he public and p i a e sec o s, wo
ad e se e ec s can be seen, which will be explained
below.
Fi s , based on monopoly ba gaining heo y (Lewis
1963; Olson 1965), a union can p o ec union membe s
(e.g., aising he wage le el o union membe s) h ough
collec i e ba gaining wi h he employe ; he e o e, he e
is a posi i e union wage p emium (e.g., F eeman 1980;
Pa sley 1980; F eeman and Medo 1984; Hi sch and
Addison 1986; Ca d 1996; Fa be e al. 2021; Kulka ni
and Hi sch 2021; Masso e al. 2022). Thus, highe union
membe ship densi y is he endency o inc ease he
union wage p emium.
I he go e nmen p omo es he es ablishmen o
unions and expands union co e age in he public sec o ,
his may inc ease unions’ ba gaining powe signi ican ly,
which may lead o he union wage p emium in he pub-
lic sec o becoming g ea e han ha in he p i a e sec-
o . Blanch lowe and B yson (2010) and Rosen eld and
Denice (2019) epo ha he union densi y o collec i e
ag eemen co e age a e in he public sec o is highe
han ha in he p i a e sec o in he US and he UK,
and he union wage p emium in he public sec o is also
g ea e .
In China, union membe ship densi y in he public sec-
o is g ea e han in he p i a e sec o , which is simila
o he si ua ion in de eloped coun ies. Fo example,
based on da a om he CFPS, he a e age p opo ion
o union membe ship was 9.47% o he o e all sample,
27.7% o he public sec o , and only 4.4% o he p i-
a e sec o du ing he pe iod 2010–2018. The e o e, i is
p edic ed ha he ba gaining powe o he public sec o
will be g ea e han ha o he p i a e sec o , which may
signi ican ly inc ease he union wage p emium (ha e an
inc easing e ec ).
Second, compa ed wi h de eloped coun ies, wo ea-
sons may dec ease he union wage p emium o China’s
public sec o (ha e a dec easing e ec ). One eason is
he di e ence in he a ge o a union’s collec i e ag ee-
men be ween he public and p i a e sec o s. Speci i-
cally, in China, based on he egula ions men ioned
abo e, he ACFTU in he public sec o aims o imp o e
he wo k condi ions o all employees, including bo h
union membe s and non-membe s (You 2017; Guo and
Dai 2022), whe eas p i a e sec o unions a ge only
union membe s, as is he case in de eloped coun ies.
As he union’s in luence on non-membe s is g ea e in
he public sec o han in he p i a e sec o , he wage
gap be ween union membe s and non-membe s may be
na owe in he public sec o han in he p i a e sec o .
The second eason is he go e nmen (o he Com-
munis Pa y o China)’s in luence on human esou ce
managemen , including wage-se ing, which is g ea e
in he public sec o han in he p i a e sec o (Lin
e al.1994, 2020; Iwasaki e al. 2020; Ma and Iwasaki
2021; Jin e al. 2022). Consequen ly, he in luence o
unions’ collec i e ag eemen on wage-se ing may be
smalle in he public sec o , which means ha he posi-
i e union wage p emium may also be smalle in he
public sec o .
When he dec easing e ec is g ea e han he inc eas-
ing e ec , he union wage p emium in he public sec o
may be smalle han ha in he p i a e sec o . The e o e,
his s udy p oposes hypo hesis 1 (H1) as ollows:
H1 In China, he e is a posi i e union wage p emium
in bo h he public and p i a e sec o s. The p emium in
he p i a e sec o may be g ea e han ha in he public
sec o .
Rega ding he channels o unions’ impac on he wage
gap in he public and p i a e sec o s, his s udy es ima es
he e ec s o wo channels: he di e ence in union mem-
be ship densi y ( he endowmen e ec ) and he di e -
ence in he union wage p emium ( he p ice e ec ).
Fi s , a union wi h g ea e membe ship densi y may
ha e g ea e ba gaining powe , which may inc ease he
p obabili y o inc easing union membe s’ wage le el (e.g.,
Lewis 1963; F eeman 1980; Fa be e al. 2021). Blanch-
lowe and B yson (2010) and Rosen eld and Denice
(2019) epo ha he union wage p emium in he public
sec o is much mo e signi ican han ha in he p i a e
sec o in he US and he UK due o ha union membe -
ship densi y in he public sec o is g ea e han ha in he
p i a e sec o .
Al hough he collec i e ag eemen igh and powe a e
smalle o Chinese unions han hose in de eloped coun-
ies, he o me also aim o imp o e wo king condi ions,
including aising he wage le el. Du ing he planned-
economy pe iod, in he public sec o , he ACFTU was
es ablished, and he go e nmen en o ced wo ke s’ pa -
icipa ion in unions. Con e sely, in he p i a e sec o , he
es ablishmen o unions has been pe mi ed since he

Page 5 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
1990s (You 2017; Guo and Dai 2022). Thus, he public
sec o has g ea e union membe ship densi y han he
p i a e sec o , which may widen he wage gap be ween
bo h sec o s.
Second, he magni ude o he union wage p emium
may di e be ween he public and p i a e sec o s, which
a ec s he wage gap be ween hem. I he union wage
p emium in he public sec o is g ea e han ha in he
p i a e sec o , i may widen he wage gap, and ice e sa.
On he one hand, a highe union membe ship den-
si y may lead o g ea e ba gaining powe o he public
sec o , which may inc ease his sec o ’s union wage p e-
mium and widen he wage gap be ween he public and
p i a e sec o s (inc ease wage p emium e ec ). Empi i-
cal s udies ha e e ealed ha in de eloped coun ies, a
highe union membe ship densi y in he public sec o
inc eases he union wage p emium (Blanch lowe and
B yson 2010; Rosen eld and Denice 2019). I is an ici-
pa ed ha in China, a highe union membe ship densi y
in he public sec o may also inc ease he union wage
p emium in ha sec o . This, in u n, could con ibu e
o he widening wage gap be ween he public and p i a e
sec o s, simila o wha occu s in de eloped coun ies.
On he o he hand, in con as wi h de eloped coun-
ies, he union wage p emium may be smalle in China’s
public sec o . This could be a ibu ed o wo easons:
Fi s , he ACFTU aims o imp o e he wo k condi ions
o all employees, including union membe s and non-
membe s (Yuan 2015; You 2017; Guo and Dai 2022). Sec-
ond, he go e nmen ’s conside able in luence on wage
le el in he public sec o is g ea e han ha in he p i a e
sec o (Lin e al.1994, 2020; Iwasaki e al. 2020; Ma and
Iwasaki 2021; Jin e al. 2022).
Conside ing ha unions’ main unc ions in he public
sec o include imp o ing employees’ wo king condi ions
and p o ec ing hei igh s ( o bo h union membe s
and non-membe s), he dec ease in he wage p emium
e ec may be g ea e han he inc ease e ec , which may
dec ease he union wage p emium in he public sec o .
The e o e, gene ally, he di e ence in he union wage
p emium be ween he public and p i a e sec o s (whe e
he union wage p emium in he public sec o is lowe
han ha in he p i a e sec o ) may condi ionally na ow
he wage gap.
Consequen ly, I p opose hypo hesis 2 (H2):
H2 Bo h union membe ship densi y and union wage
p emium di e ences condi ionally a ec he wage gap
be ween he public and p i a e sec o s in China. The o -
me widens he wage gap be ween he public and p i a e
sec o s, while he la e na ows i .
I es H1 and H2 empi ically in he ollowing sec ion.
3 Me hodology
3.1 Models
3.1.1 Models o  es ing H1
This s udy uses a wage unc ion o calcula e he union
wage p emium in he o al sample (Eq.(1)) and in he
public and p i a e sec o s sepa a ely (Eq.(2)) o p o e
H1. I used he O dina y Leas Squa es (OLS) me hod as
he baseline es ima ions, which is exp essed by Eqs. (1)
and (2).
whe e subsc ip
i
is an indi idual;
ep esen s he ime
yea (2010, 2012, 2014, 2016, o 2018); supe sc ip
j
ep-
esen s he sec o (public o p i a e).
lnW
ep esen s he
loga i hm o he hou ly wage.
U
is a union-membe ship
dummy;
Pub
is a public-sec o dummy;
X
ep esen s
o he ac o s (e.g., educa ion, yea s o wo k expe ience,
gende , and occupa ion, and p o incial egional dum-
mies) ha may a ec wage le els;
n
ep esen s he ypes
o
X
.
βU
ep esen s he union wage p emium;
βpub
indi-
ca es he public sec o wage p emium;
βX
is he coe -
icien o
X
;
δ
deno es he ime yea ixed e ec ;
a
is a
cons an ; and
u
is an e o e m.
The e a e some conce ns on econome ic issues in
he OLS me hod based on he c oss-sec ional da a.
This s udy also uses a se o me hods o add ess hese
issues. Fi s , ega ding he omi ed a iable issue, se -
e al empi ical s udies ha e u ilized amily backg ound
a iables such as pa en al educa ion (e.g., a he ’s and
mo he ’s educa ion) as ins umen al a iables (IVs) in he
wage unc ion o educa ion (Lemke and Rischall 2003;
Mish a and Smy h 2013; Gong 2019). As pa en al edu-
ca ion a ainmen migh in luence hei adul child en’s
wages h ough a ious channels, such as in es men in
hei child en’s educa ion o o he unobse able cogni-
i e and non-cogni i e abili ies, I added he amily back-
g ound a iables (e.g., a he ’s and mo he ’s educa ion)
and e- an he OLS me hod o check he obus ness o
he esul s.
Second, wo sample selec ion biases may pe sis in
Eq.(1). The i s bias pe ains o wo ke s’ choice o wo k
in he public o p i a e sec o , while he second one
ela es o wo ke s’ choice o join a union. Fo ins ance,
(1)
lnW
i=a+βUUi+βPubPubi+βnX
n

1
X
i
+δ +ui
,
(2)
ln
Wj
i=aj+βj
UUj
i+βj
nX
n

1
Xj
i+δj
+uj
i
,
3 Page 6 o 25
X.Ma
wo ke s wi h pa en s in he public sec o a e mo e likely
o wo k in he public sec o and ha e a highe p obabili y
o joining unions. This may lead o a si ua ion in which
wo ke s in he public o p i a e sec o s and hose wi h
union membe ship a e no andomly selec ed. I used
a bi a ia e selec i i y app oach (Dubin and McFadden
1984) o add ess bo h ypes o selec ion bias. I es ablished
ou dummy a iables ( a he ’s CPC membe ship, mo h-
e ’s CPC membe ship, a he wo king in he public sec o ,
and mo he wo king in he public sec o ) as iden i ica ions
in he p obabili y unc ion o wo king in he public sec o
o joining a union in he i s -s age es ima ions.4 The a -
iables associa ed wi h amily backg ound (such as pa en-
al educa ion, employmen s a us, o membe ship in he
CPC) a e commonly employed as IVs in empi ical s udies
ocusing on indi iduals’ labo ma ke ou comes (Lemke
and Rischall 2003; Mish a and Smy h 2013; Gong 2019).
In his s udy, I disco e ed ha pa en al CPC membe ship
o wo kplace owne ship s a us (wo king in he public o
p i a e sec o ) signi ican ly impac ed he likelihood o an
indi idual acqui ing union membe ship o en y in o he
public sec o (see Addi ional ile1: TableS1). Howe e ,
hese a iables demons a ed insigni ican in luence on
an indi idual’s wage.5 Consequen ly, I u ilized hese ou
a iables as iden i ica ions in he selec ion bias adjus ed
models. Ne e heless, i is impo an o no e ha hese
ou iden i ica ion a iables canno be ega ded as
he sole comple e exogenous a iables used in he IV
me hod.
Nex , I calcula ed he co ec ion e ms o each sec-
o as
pub =φ(βN)/�(βN)
o he public sec o and
p i =φ(βN)/[1−�(βN)]
o he p i a e sec o . Fu -
he mo e, I calcula ed he co ec ion e m o he choice
o join a union as
u=φ(γ M)/�(γ M)
, and added hese
co ec ion e ms o he wage unc ion, as exp essed in
Eqs. (3) and (4).
Thi d, he e may be e e se causali y issue. Fo exam-
ple, union pa icipa ion may be a ec ed by a high union
(3)
lnW
i=a+βUUi+βUUi+βPubPubi+βnX
n

1
X
i
+γpubpubi +γuui +δ +ui
,
(4)
ln
Wj
i=aj+βj
UUj
i+βj
nX
n

1
Xj
i+γj
pubj
pubio γj
p ij
p ii+γj
uj
ui +δj
+uj
i
,
wage p emium. The lagged a iable o he union mem-
be ship dummy was used o add ess he e e se causali y
issue. I used da a om he su ey wa e da ing back wo
wa es p io o he su ey in ques ion (
−2
) as he lagged
pe iod in his s udy. Fo example, lagged union membe -
ship
U −2
(e.g., union membe ship in 2014) was used o
in es iga e union membe ship’s in luence on wages in he
cu en su ey yea (2018, as su eys a e conduc ed e e y
wo yea s). The LV model based on he OLS me hod is
exp essed by Eqs. (5) and (6).
Fou h,
ui
in Eq.(1) includes indi idual-speci ic and
ime-in a ian ac o s (
i
) and idiosync a ic e o s (
εi
).
An indi idual he e ogenei y issue (e.g., unobse able
pe sonali y, abili y, and p e e ence) may a ise in he es i-
ma ed esul s i
i
is main ained. This s udy u ilized he
longi udinal su ey da a and panel da a es ima ion me h-
ods (e.g., he FE/RE model) o add ess his p oblem. The
FE/RE model is exp essed in Eqs. (7) and (8):
whe e subsc ip
ep esen s ime yea , and
δ
deno es
he ime ixed e ec . Some s udies ha e ound ha indi-
idual he e ogenei y conside ably a ec s he union wage
p emium in de eloped coun ies (e.g., F eeman 1980),
whe eas empi ical s udies o China (Yao and Zhong
2013; Li and Xu 2014; Yuan 2015; Gunde son e al. 2016;
Li and Song 2017; Yang e al. 2018; Sun and Liu 2019)
(5)
lnW
i
=a+
βU
U
i −2
+
βPub
Pub
i
+βnX
n

1
X
i
+δ +ui ,
(6)
ln
Wj
i =aj+βj
UUj
i −2+βj
nX
n

1
Xj
i +δj
+uj
i
,
(7)
lnW
i
=a+
βU
U
i
+
βPub
Pub
i
+βnX
n

1
X
i
+δ + i+εi
,
(8)
ln
Wj
i =aj+βj
UUj
i +βj
nX
n

1
Xj
i + j
i+δj
+εj
i
,
4 The esul s o he p obabili y unc ion o wo king in he public sec o o
joining a union a e p esen ed in he Addi ional ile1: TableS1.
5 The esul s o wage unc ions using hese iden i ica ions a e a ailable
upon eques .
Page 7 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
ha e no add essed he he e ogenei y p oblem. This s udy
aims o add ess his issue in he Chinese con ex . How-
e e , i should be no ed ha al hough he FE/RE model
can add ess he p oblem due o unobse able ime-in a -
ian a iables o an indi idual, i s e ec i eness elies on
he a ia ion in each a iable. When he e’s minimal a i-
a ion in union membe ship du ing he analyzed pe iod
( om 2010 o 2018 in his s udy), he union e ec may
no be p ecisely es ima ed.6 Fu he mo e, samples show-
ing changes in union membe ship migh possess speci ic
cha ac e is ics ha may no accu a ely ep esen he
o e all union e ec s ac oss he en i e sample. I only used
he FE/RE model o conduc he obus ness checks.
Fi h, some s udies (e.g., Fa be e al. 2021; Goe ke and
Huang 2022) ha e highligh ed he exis ence o a so -
ing e ec wi hin union membe ship. Obse able ac o s
signi ican ly in luence he likelihood o joining a union,
leading o non- andom sample dis ibu ions be ween
union membe s and non-membe s. Fo ins ance,
esea ch by F eeman (1980), F eeman and Medo (1984),
and Fa be e al. (2021) demons a e ha union densi y is
highe among less-educa ed, Black, and low-wage wo k-
e s compa ed o hei coun e pa s— he well-educa ed,
Whi e, and high-wage wo ke s. Consequen ly, he g oups
o union membe s and non-membe s a e no andomly
selec ed. To mi iga e his issue, his s udy also employed
he PSM me hod o conduc he obus ness checks.
The PSM me hod a emp s o educe bias due
o obse able a iables in ea men e ec es ima es
ob ained by compa ing he ou comes o he uni s ha
ecei ed ea men o hose ha did no (Rosenbaum and
Rubin 1983). Using PSM b ings he obse ed da a close
o andomized expe imen al da a by ma ching and esa-
mpling o minimize selec i i y bias and coun e ac ual
s a es in he sample composi ion. To apply his me hod
o calcula e he a e age disposi ion e ec s, one mus (1)
selec app op ia e obse able a iables o esampling,
(2) un p obi o logi eg ession o es ima e he p open-
si y sco e, (3) ma ch he p opensi y sco e based on he
selec ed obse able a iables (co a ia es), and inally, (4)
calcula e he a e age ea ed e ec on he ea ed (ATT)
based on he ma ched samples, as in Eq.(9).
whe e
lnW j
i
(
1)
exp esses he loga i hm o union mem-
be s’ wage in sec o
j
and
lnW j
i
(
0)
ep esen s ha o he
(9)
ATT
=E

lnW
j
i (1)−lnW
j
i (0)|Union =1
,
non-membe s wi h a simila endowmen (e.g., educa ion,
yea s o wo k expe ience, and occupa ion) as he union
membe s in sec o
j
. This s udy epo s he ATT esul s
in he ollowing sec ion.
PSM assumes s ongly igno able ea men assign-
men , whe e ea men a iable assignmen s and la en
dependen a iables a e independen when condi ioning
obse ed co a ia es. Since co a ia es a e summa ized
in o one a iable (i.e., he p opensi y sco e), his a i-
able can be used e en when he co a ia e alues in wo
g oups do no o e lap (common suppo ). D awing om
p e ious s udies (F eeman 1980; F eeman and Medo
1984; Fa be e al. 2021, e c.) and speci ic cha ac e is ics
o he Chinese labo ma ke , such as segmen a ion by he
public and p i a e sec o s, as well as by he popula ion
egis a ion sys em (hukou) (Zhang e  al. 2016; Cheng
e al. 2020; Lin e al. 2020; Jin e al. 2022), his s udy uses
se e al a iables as obse ed co a ia es. These include
demog aphic ac o s (yea s o schooling, yea s o wo k
expe ience and expe ience squa ed alue, heal h s a us,
gende , e hnici y, hukou s a us, ma i al s a us, and CPC
membe ship), wo k- ela ed ac o s (occupa ion, indus y
sec o , wo kplace owne ship sec o ), amily backg ound
( a he ’s and mo he ’s CPC membe ship, a he ’s and
mo he ’s employmen s a us in he public sec o ), p o -
ince-le el egion, and su ey yea . Howe e , i should be
no ed ha he PSM me hod canno add ess he p oblem
due o unobse able a iables o an indi idual.
Las ly, conside ing ha he union wage p emium may
di e based on wage dis ibu ion, I also u ilized uncondi-
ional quan ile eg ession (QR) (Koenke and Base 1978)
and ixed e ec QR (Machado and San os Sil a 2019)
models. The QR model is exp essed by Eq.(10).
whe e
θ
ep esen s he wage dis ibu ion pe cen ile ha
should be be ween 0 and 1,
H
is he a iable including
he union dummy and o he con olled a iables includ-
ing ime-speci ic e ec and indi idual ixed e ec s. The
independen a iables a e he same as hose in Eq.(1).
ρθ(.)
ep esen s a check unc ion—a loss unc ion ha
e ie es he
θ
- h samplequan ile. I pe o med es ima-
ions o he public sec o , p i a e sec o , and he o al
sample.
Posi i e coe icien s o he union dummies (
βU
) o
he ATT ha a e s a is ically signi ican in he public
and p i a e sec o s in Eqs. (1)–(10) indica e ha H1 is
suppo ed.
(10)
min
θ


lnWi−β(θ)Hi


+(1−θ)


lnWi−β(θ)Hi

,
ρθ∈(0, 1)
6 The numbe o samples is 1048 o hose s aying as union membe , 5666
o hose s aying as non-union membe , 101 o hose ans e ing om
union membe o non-union membe (mo ing ou ), and 407 o hose
ans e ing om non-union o union membe (mo ing in) among he o al
samples used in his s udy.
3 Page 8 o 25
X.Ma
3.1.2 Models o  es ing H2
I used a wage decomposi ion me hod o es H2. Fi s , I
used a basic B-O decomposi ion me hod (Blinde 1973;
Oaxaca 1973), which is ep esen ed by Eqs. (11) and (12).
whe e
pub
and
p i
deno e he public and p i a e sec-
o s, espec i ely;
l
nW
pub
−lnW
p i
is he wage gap
be ween he wo sec o s;
H
ep esen s he ac o s ha
a ec wages, including union membe ship, demo-
g aphic ac o s, wo k- ela ed ac o s, p o incial le el
egions, and su ey yea , which a e simila o he a i-
ables used in Eq.(1);
β
is he coe icien o each ac o .

β
pub
(H
pub
−Hp i
)
o

βp i(H
pub
−Hp i
)
exp esses
he di e ences in endowmen e ec (e.g., di e -
ence in union co e age),

(β
pub
−βp i)H
p i
o

(β
pub
−βp i)H
pub
deno es he di e ences in p ice
e ec (e.g., union wage p emium).
I conduc ed a se o decomposi ion analyses as obus -
ness checks. Fi s , I used Oaxaca and Ransom’s (1994)
decomposi ion me hod ( he O-R me hod) based on he
wage unc ion using he RE model o add ess he index
numbe p oblem,7 as exp essed in Eq.(13).
whe e
β∗
is a sec o -neu al coe icien es ima ed based
on wage unc ions using he en i e sample. (
β∗
−β
p i
)
Hp i
ep esen s he gap caused by he oo-low endow-
men e u n o he p i a e sec o (known as “p i a e sec-
o loss”), and (
βpub
−β
∗
)
Hpub,
ep esen s he wage gap
gene a ed by he oo-high endowmen e u n o public
sec o (known as he “public sec o gain”). The sum o
hese wo decomposi ion alues ep esen s he wage gap
esul ing om he di e ences in he p ice e ec .
Second, unlike in de eloped coun ies, whe e only
non-manage employees can join unions, in China, bo h
manage s and non-manage employees can join unions.
Since mos union membe s a e non-manage employ-
ees who o en equi e mo e p o ec ion om unions han
manage s, I conduc ed he B-O decomposi ion using
(11)
lnW pub −lnW p i =

βpub

Hpub −Hp i

+
β
pub −
β
p iHp i,
(12)
lnW pub −lnW p i =

βp i

Hpub −Hp i

+
β
pub −
β
p iHpub,
(13)
l
nW
pub
−lnW
p i
=β
∗
H
pub
−H
p i
+(β
∗
−β
p i
)H
p i
+(β
pub
−β
∗
)H
pub
only non-manage employee samples (excluding manage
samples) o obus ness checks.
Thi d, while he undamen al Mince - ype wage unc-
ion (Mince 1958; 1974) adi ionally uses educa ion and
yea s o wo k expe ience as indica o s o human capi-
al, se e al s udies (e.g., Gus a sson and Li 2000; Mish a
and Smy h 2013) ha e used he occupa ion and indus y
sec o as b oade human capi al p oxies. Howe e , edu-
ca ion may in luence he choice o ypes o occupa ions
o indus ial sec o s. Fo ins ance, highly educa ed wo k-
e s a e mo e likely o become manage s o echnicians,
leading o he bad con ol a iable p oblem in eg es-
sion analysis. To add ess his issue and pe o m obus -
ness checks, ollowing Ma (2022), I also applied he B-O
decomposi ion me hod only using he basic human capi-
al a iables (excluding occupa ion and indus ial a i-
ables). The esul s o wage unc ions based on he OLS
me hod a e used in hese decomposi ion analyses.8
When he con ibu ion a e o he union in he endow-
men e ec has a posi i e alue and he con ibu ion
a e o he union in he p ice e ec has a nega i e alue
in hese decomposi ion esul s, i means ha he union
membe ship densi y widens he wage gap be ween he
public and p i a e sec o s in China, while and union
wage p emium di e ences na ows he wage gap. The e-
o e, H2 is suppo ed.
3.2 Da a and a iables
Longi udinal da a om he CFPS, a na ional longi udi-
nal su ey p ojec conduc ed by Peking Uni e si y since
2010, we e used (CFPS 2022) in his s udy. The baseline
su ey was o icially launched in 2010 in 25 p o inces,9
municipali ies, and au onomous egions, which ep e-
sen 95 pe cen o he o al popula ion in China. 14,960
households we e success ully in e iewed. Wi hin hese
households, 33,600 adul s (indi iduals aged 18yea s and
abo e) and 8,990 you hs (indi iduals aged 14yea s and
younge ) we e in e iewed du ing he i s wa e. Despi e
he a ailabili y o six wa es o CFPS da a, I op ed o use
7 The index numbe p oblem a ises because he B-O decomposi ion
me hod employs di e en e e ence g oups ( o example, Eq. (11) and
Eq. (12) a e es ima ed sepa a ely), leading o a ia ions in he esul s
be ween Eq.(11) and Eq.(12).
8 I also used he esul s o wage unc ions based on he FE and RE model
in he B-O decomposi ion analysis. The majo i y o hese esul s closely
esembled hose de i ed om he OLS me hod. De ailed esul s can be
ound in Addi ional ile1: Tables S3.
9 The CFPS did no include Xinjiang Uygu Au onomous Region, Tibe
Au onomous Region, Qinghai P o ince, Inne Mongolia Au onomous
Region, Ningxia Hui Au onomous Region, and Hainan P o ince in main-
land China and Hong Kong, Macau, Taiwan egions in 2010 baseline su ey.
The p opo ion o he popula ion in hese excluded egions is app oxima ely
5 pe cen o he o al popula ion in China.
Page 15 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
membe s. While he e may be a spillo e e ec ben-
e i ing non-membe s, he union wage p emium among
membe s ypically exceeds ha among non-membe s
wi hin co po a ions wi h a union. Con e sely, Chinese
unions, especially in he public sec o , ha e b oade a -
ge s ha encompass bo h union and non-membe s.
Fo ins ance, he 2021 T ade Union Law s ipula es ha
all b anch o ganiza ions o he ACFTU should p o ec
he legi ima e igh s and in e es s o o e all employees,
i espec i e o union membe ship. The e o e, he union
wage p emium ends o be smalle in China. Addi ionally,
in con as o collec i e ba gaining o wage inc eases
Table 5 Resul s o union e ec on wage o o al, public sec o and p i a e sec o based on he PSM me hod
The obse able a iables include demog aphic ac o s (yea s o schooling, yea o wo k expe ience, heal h s a us, sex, e hnici y, u ban hukou, ma ied, CPC
membe ship), wo k ac o s (occupa ion, indus y sec o ), pa en ’s wo k s a us, pa en ’ CPC membe ship, p o ince-le el egion and su ey yea (2012, 2014, 2016,
and 2018) dummies we e used in he p obi eg ession model o calcula e he p opensi y ma ching sco e. To al samples including public and p i a e sec o . Sou ce:
Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018
PSM he p opensi y sco es ma ching me hod. ATT a e age ea ed e ec on he ea ed
***p < 0.01
**p < 0.05
*p < 0.1
Sample T ea ed Con ols Di e ence SE - alue
(1) One o one Nea es -neighbo e ma ching (1:1)
Na ion Unma ched 2.581 2.205 0.376 0.020 18.93
ATT 2.580 2.463 0.117 0.031 3.78
Public Unma ched 2.625 2.328 0.297 0.024 12.40
ATT 2.624 2.548 0.077 0.037 2.10
P i a e Unma ched 2.479 2.142 0.337 0.036 9.43
ATT 2.479 2.357 0.122 0.052 2.33
(2) k-Nea es -neighbo ma ching (1:3)
Na ion Unma ched 2.581 2.205 0.376 0.020 18.93
ATT 2.580 2.460 0.120 0.026 4.62
Public Unma ched 2.625 2.328 0.297 0.024 12.40
ATT 2.624 2.530 0.094 0.031 3.04
P i a e Unma ched 2.479 2.142 0.337 0.036 9.43
ATT 2.479 2.373 0.106 0.044 2.40
(3) Ke nel ma ching
Na ion Unma ched 2.581 2.205 0.376 0.020 18.93
ATT 2.581 2.483 0.098 0.023 4.20
Public Unma ched 2.625 2.328 0.297 0.024 12.40
ATT 2.625 2.534 0.091 0.028 3.21
P i a e Unma ched 2.479 2.142 0.337 0.036 9.43
ATT 2.479 2.359 0.120 0.038 3.13
(4) Radius ma ching
Na ion Unma ched 2.581 2.205 0.376 0.020 18.93
ATT 2.580 2.484 0.096 0.024 4.05
Public Unma ched 2.625 2.328 0.297 0.024 12.40
ATT 2.625 2.530 0.095 0.029 3.27
P i a e Unma ched 2.479 2.142 0.337 0.036 9.43
ATT 2.474 2.377 0.097 0.039 2.46
(5) Local linea eg ession ma ching
Na ion Unma ched 2.581 2.205 0.376 0.020 18.93
ATT 2.580 2.484 0.096 0.031 3.12
Public Unma ched 2.625 2.328 0.297 0.024 12.40
ATT 2.624 2.529 0.096 0.037 2.62
P i a e Unma ched 2.479 2.142 0.337 0.036 9.43
ATT 2.479 2.373 0.106 0.052 2.02

3 Page 16 o 25
X.Ma
wi h employe s in de eloped coun ies, aising wage le -
els is no he p ima y unc ion o Chinese unions. A icle
1 o he 2021 T ade Union Law desc ibes he union’s ole
as a b idge o link connec ing he CPC o ganiza ions and
Chinese wo ke s. Gi en ha he basic wage le el in he
public sec o is con olled by he go e nmen , and he
scope o inc easing wages h ough collec i e ag eemen s
is limi ed in he p i a e sec o , he posi i e union wage
p emium is smalle in China.
Las ly, Table5 summa izes he esul s ob ained using
se e al PSM me hods. They a e he one- o-one nea es
neighbo ma ching (Model 1), k-nea es neighbo ma ch-
ing (1:3 ma ching in his s udy; Model 2), Ke nel ma ch-
ing (Model 3), adius ma ching (Model 4), and local
linea eg ession ma ching (Model 5) me hods.
I es ima ed he p obabili y unc ion o joining unions
o he o e all sample as well as sepa a ely o he pub-
lic and p i a e sec o s. These esul s we e used o de i e
p opensi y sco es.18 The esul s (see Appendix TableA2)
highligh signi ican in luences o demog aphic ac-
o s, wo k- ela ed ac o s, amily backg ound, egion,
and su ey yea s on he p obabili y o joining unions.
Fo ins ance, he esul s indica e highe p obabili ies o
joining unions among indi iduals in he public sec o ,
Han majo i y wo ke s, hose wi h highe educa ional
a ainmen , employees in manu ac u ing and cons uc-
ion sec o s, and hose whose a he s a e CPC membe s.
Mo eo e , he e was a no able inc ease in he likelihood
o joining unions in 2014, 2016, and 2018 compa ed o
he e e ence yea o 2010.
Unlike indings in ad anced capi alis coun ies (F ee-
man and Medo 1984; Fa be e al. 2021), whe e union
densi y is highe among disad an aged g oups—such as
less-educa ed indi iduals, Black indi iduals, and low-
wage wo ke s—compa ed o ad an aged wo ke s like
well-educa ed indi iduals and highe -wage ea ne s, in
China, he p obabili y o union membe ship is highe
among he ad an aged g oup han among he disad an-
aged g oup. This implies ha union co e age among
disad an aged wo ke s in China migh be lowe , sugges -
ing ha ade unions may no ex end as many bene i s
-50 0 50 100
S anda dized%biasac ossco a ia es
heal h
Occ_o he
Ind_sale
y2016
emale
Ind_ cm
Occ_go ma
mgo
Ind_o he
han
y2012
go
Ind_se
y2014
mcpc
Occ_cle k
exp
Occ_ ech
ma ied
pa y
cpc
y2018
yedu
own_s a e
Unma ched
Ma ched
Fig. 3 S anda dized bias ac oss con ol a iables among unma ched and ma ched g oup ( o al samples). Sou ce: Calcula ed using da a
om he CFPS o 2010, 2012, 2014, 2016 and 2018. The esul s o p o ince-le el egional dummy a iables we e no p esen ed in Fig. 3. These esul s
a e a ailable upon eques
18 Se en indi idual samples we e d opped due o he esampling. The num-
be o o al samples used in he PSM me hod is 14,076.
Page 17 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
o hese wo ke s, po en ially con ibu ing o a scena io
whe e ’ he ich ge iche while he poo ge poo e .
Figu e 3 displays he s anda d bias ac oss obse -
able a iables be o e and a e ma ching o he balance
check. The di e ences in he s anda d bias o each a i-
able be ween union membe s and non-membe s became
smalle (app oaching a alue nea “0”) a e ma ch-
ing. Figu e4 p esen s he esul s o he o e lap check.
These esul s indica e ha he ma ching p ocedu es a e
app op ia e.
The esul s om he i e ypes o ma ching me hods
we e simila . Fo example, he esul s based on one-
o-one nea es neighbo ma ching (Model1) indica ed
ha he esul o ATT is 0.117, which is simila o ha
(0.124) in Table 2, sugges ing he e exis s a posi i e
union wage p emium na ionwide. The esul s suppo
H1 again. Addi ionally, he alues o di e ences in he
ma ched g oup (0.117) became smalle han hose in he
unma ched g oup (0.376), sugges ing ha hese obse -
able ac o s may signi ican ly a ec he union wage p e-
mium h ough hei e ec s on he union densi y.
To compa e he esul s o he ATT be ween he pub-
lic and p i a e sec o s, all esul s indica e ha while
he ATT alues a e signi ican posi i e alues o bo h
he public and p i a e sec o s, he magni ude o ATT
alues o he public sec o is g ea e han ha o he
p i a e sec o . Fo ins ance, he esul s based on one-
o-one nea es neighbo ma ching (Model1) indica ed
ha he ATT alue is 0.077 o he public sec o and
0.122 o he p i a e sec o , indica ing ha while he e
exis posi i e union wage p emiums o bo h he public
and p i a e sec o s, he p emium o he p i a e sec o
is g ea e han ha o he public sec o . The esul s con-
i m hypo hesis H1 again.
0 .2 .4 .6 .8
P opensi ySco e
Un ea ed: O suppo Un ea ed: On suppo
T ea ed: On suppo T ea ed: O suppo
Fig. 4 Resul s o o e lap checks ( o al samples). Sou ce: Calcula ed
using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018
Table 6 Resul s o union wage p emium by wage dis ibu ion pe cen ile
The con olled a iables include demog aphic ac o s (yea s o schooling, yea o wo k expe ience, heal h s a us, sex, e hnici y, u ban hukou, CPC membe , ma ied),
wo k- ela ed ac o s (occupa ion, indus y sec o , wo kplace owne ship sec o ), p o ince-le el egion and su ey yea (2012, 2014, 2016, and 2018) dummies, which
a e no p esen ed in he able. These esul s a e a ailable upon eques . Sou ce: Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018
QR quan ile eg ession model; FE_QR ixed e ec s quan ile eg ession model. 5 ~ 95% indica e he wage dis ibu ion pe cen iles (5% is he lowes wage le el, 95% is
he highes wage le el)
***p < 0.01
**p < 0.05
*p < 0.1
(1) QR (2) QR_FE
Na ion Public P i a e Na ion Public P i a e
5% 0.135** 0.156* 0.119 0.039 0.002 0.104
10% 0.086*** 0.095** 0.106** 0.039 0.005 0.097
20% 0.067*** 0.078*** 0.057 0.038 0.010 0.095
30% 0.068*** 0.071*** 0.028 0.038 0.012 0.095
40% 0.076*** 0.059*** 0.048 0.037 0.016 0.094
50% 0.077*** 0.056** 0.073** 0.036 0.025 0.075
60% 0.079*** 0.037* 0.071** 0.034 0.036 0.059
70% 0.063*** 0.042* 0.044 0.033 0.040 0.058
80% 0.088*** 0.036 0.060 0.033 0.041 0.057
90% 0.091*** 0.062 0.015 0.033 0.047 0.056
95% 0.068*** 0.078 0.014 0.032 0.050 0.049
3 Page 18 o 25
X.Ma
5.1.4 Resul s bywage dis ibu ion
Conside ing unions’ a ying e ec s on he low-, mid-
dle-, and high-wage g oups, I also calcula ed he union
wage p emium by wage dis ibu ion pe cen iles o
bo h he public and p i a e sec o s using wo models:
(1) he QR model and (2) he ixed-e ec s QR model.
The esul s a e p esen ed in Table6.
The esul s ob ained om he QR model indica e
he p esence o a posi i e union wage p emium o he
o e all, public-sec o , and p i a e-sec o samples, espe-
cially o he low- and middle-wage g oups. Mo eo e ,
he magni ude o he union wage p emium in he low-
wage g oup in he public sec o exceeded ha in he
p i a e sec o . These esul s suppo H1.
Howe e , he esul s de i ed om he ixed-e ec s
QR model sugges ha , al hough he union wage p e-
miums ha e posi i e alues and hese p emiums a e
g ea e in he p i a e sec o ac oss low-, middle-, and
high-wage g oups o he o e all, public-sec o , and
p i a e-sec o samples, hey a e s a is ically insig-
ni ican . These esul s imply ha when add essing he
indi idual he e ogenei y issue, he na ion’s union wage
p emium is insigni ican .
5.2 Resul s o  es ing hypo hesis H2
5.2.1 Basic decomposi ion esul s based on heB‑O
decomposi ion me hod
The basic decomposi ion esul s based on he B-O
me hod a e p esen ed in Table7. The esul s o wage
unc ions based on he RE models we e used in he
decomposi ions.
Fi s , he union e ec on he endowmen componen
was 7.4% (Column [a]) and 10.8% (Column [c]) espec-
i ely, implying ha di e ences in union membe -
ship densi y con ibu ed o widening he wage gap. In
con as , he union e ec on he p ice componen was
−1.3% (Column [b]) and −4.7% (Column [d]) espec-
i ely, indica ing ha a ia ions in he union wage
p emium educed he wage gap. These esul s p o ide
suppo o hypo hesis H2.
Second, conce ning he o e all esul s o he endow-
men and p ice componen s, he con ibu ion a e
o he endowmen componen (96.0% in Column [a],
77.0% in Column [c]) exceeds ha o he p ice compo-
nen (4.0% in Column [b], 21.3% in Column [d]). This
sugges s ha di e ences in endowmen be ween he
wo sec o s a e he p ima y ac o s con ibu ing o he
wage gap.
Las ly, compa ing he con ibu ion a e o each ac o ,
he yea s o educa ion in he endowmen componen
and he e u n o educa ion and yea s o wo k expe i-
ence in he p ice componen a e g ea e han hose o he
o he ac o s. These esul s indica e ha he in luence o
human capi al is g ea e han ha o o he ac o s, which
is consis en wi h he human capi al heo y (Becke 1964;
Mince 1974).
5.2.2 Robus ness checks
In o al, I conduc ed h ee obus ness checks, and hei
esul s a e p esen ed in Table8 and 9. Fi s , I employed
he O-R decomposi ion me hod. The esul s a e dis-
played in Table8. In Column [a], he union e ec on he
endowmen componen is 10.4%, sugges ing ha di e -
ences in union membe ship densi y widen he wage gap.
In Column [d], he union e ec on he p ice componen
is -4.3%, indica ing ha a ia ions in he union wage
p emium na ow he wage gap. These indings sugges
ha bo h di e ences in union membe ship densi y and
Table 7 Basic decomposi ion esul s o wage gap be ween
public and p i a e sec o
Sou ce: Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018.
B-O decomposi ion me hod is used. Es ima ion 1 is based on Eq.(11), es ima ion
2 is based on Eq. (12) exp essed in he ex . The esul s o wage unc ions based
on he OLS me hod a e used in he decomposi ion analysis
(1) (2)
Endowmen
(a) P ice
(b) Endowmen
(c) P ice
(d)
Value:
To al wage
gap = 0.239
0.230 0.009 0.189 0.050
Con ibu ion a e (%)
To al 96.0% 4.0% 78.7% 21.3%
Each ac o
Union 7.4% −1.3% 10.8% −4.7%
Educa ion 79.0% 139.8% 41.8% 177.0%
Expe ience 1.1% 93.3% −0.6% 94.9%
Heal h −0.2% −2.8% −0.6% −2.5%
Gende 1.6% 35.9% 4.4% 33.1%
E hnici y −0.1% 3.3% −0.1% 3.2%
Ma ied −1.9% −34.0% 0.5% −36.4%
Pa y 5.5% 1.9% 0.2% 7.1%
Occupa ion 8.6% −34.7% 28.8% −54.8%
Indus y −1.1% 31.2% −4.9% 35.0%
Region −14.3% 0.9% −14.5% 1.1%
Yea 10.4% 15.8% 12.8% 13.5%
Cons an 0.0% −245.3% 0.0% −245.3%
Page 19 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
Table 8 Robus ness check (1): Decomposi ion using O-R decomposi ion me hods
Sou ce: Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018. O-R decomposi ion me hod (Oaxaca-Ransom 1994) is used. The esul s o wage
unc ions based on he OLS me hod a e used in he decomposi ion analysis
Endowmen P ice
(a) (b) loss o non-union (c) gain o union (d) o al (b + c)
Value: wage gap = 0.239 0.219 0.008 0.012 0.020
Con ibu ion a e (%)
To al 91.4% 3.4% 5.2% 8.6%
Each ac o
Union 10.4% −0.2% −4.1% −4.3%
Educa ion 54.0% 45.9% 118.8% 164.7%
Expe ience 0.1% 35.5% 58.8% 94.3%
Heal h − 0.4% − 0.9% − 1.7% − 2.6%
Gende 3.4% 13.1% 21.1% 34.2%
E hnici y 0.0% − 0.4% 3.7% 3.3%
Ma ied − 0.5% − 14.3% − 21.1% − 35.4%
Pa y 3.8% 1.2% 2.4% 3.6%
Occupa ion 23.1% − 12.4% − 36.7% − 49.1%
I ndus y 0.0% 9.3% 20.8% 30.0%
Region − 14.0% 1.4% − 0.8% 0.7%
Yea 11.7% 4.3% 10.2% 14.5%
Cons an 0.0% − 79.1% − 166.3% − 245.3%
Table 9 Robus ness check (2): Decomposi ion using di e en samples and a iables
Sou ce: Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018. B-O decomposi ion me hod (Blinde 1973; Oaxaca 1973) was used. Es ima ions (a)
and (c) we e based on Eq.(11), es ima ions (b) and (d) a e based on Eq.(12) exp essed in he ex . The esul s o wage unc ions based on he OLS me hod a e used in
he decomposi ion analysis
(1) Employees (2) Excluding occupa ion and indus y
Endowmen
(a) P ice
(b) Endowmen
(c) P ice
(d) Endowmen
(e) P ice
( ) Endowmen
(g) P ice
(h)
Value: wage gap 0.233 0.023 0.204 0.052 0.243 0.079 0.236 0.086
Con ibu ion a e (%)
To al 90.7% 9.3% 79.6% 20.4% 75.5% 24.5% 73.3% 26.7%
Each ac o
Union 6.1% −1.7% 11.0% −6.6% 5.7% −1.4% 10.0% −5.7%
Educa ion 69.0% 124.3% 34.9% 158.4% 70.1% 84.9% 45.5% 109.5%
Expe ience 0.4% 86.4% − 0.1% 86.9% 0.3% 67.9% 0.1% 68.1%
Heal h − 0.5% − 1.2% − 0.6% − 1.1% − 0.2% − 0.6% − 0.3% − 0.6%
Gende 1.3% 34.8% 3.7% 32.4% 1.2% 31.9% 4.1% 29.1%
E hnici y 0.0% − 0.4% 0.0% − 0.4% − 0.1% − 0.7% − 0.1% − 0.7%
Ma ied − 1.7% − 31.5% 0.4% − 33.6% − 1.4% − 31.7% 1.0% − 34.1%
Pa y 5.6% 0.8% 3.1% 3.3% 2.7% − 1.4% 6.7% − 5.4%
Occupa ion 8.5% -23.3% 28.3% − 43.0%
Indus y 4.5% 21.6% − 0.2% 26.3%
Region − 12.7% 1.1% − 12.9% 1.3% − 9.5% 4.9% − 10.9% 6.4%
Yea 10.2% 16.2% 12.2% 14.2% 6.6% 36.0% 17.2% 25.4%
Cons an 0.0% − 217.7% 0.0% − 217.7% 0.0% − 165.3% 0.0% − 165.3%
3 Page 20 o 25
X.Ma
di e ences in union wage p emiums in luence he o -
ma ion o he wage gap be ween he public and p i a e
sec o s. The o me widens he wage gap, while he la e
na ows i , he eby ea i ming hypo hesis H2.
Second, I u ilized he non-manage employee sam-
ples and conduc ed he B-O decomposi ion (Model 1 in
Table9). Columns [a] and [c] co espond o he esul s
based on Eq. (11), while Columns [b] and [d] co espond
o he esul s based on Eq. (12). The union e ec on he
endowmen componen is 6.1% (Columns [a]) and 11.0%
(Columns [c]) espec i ely, indica ing ha di e ences
in union membe ship densi y widen he wage gap. The
union e ec on he p ice componen is −1.7% (Columns
[b]) and -6.6% (Columns [d]) espec i ely, sugges ing ha
di e ences in union wage p emiums na ow he wage
gap. These esul s also suppo he no ion ha di e ences
in union membe ship densi y con ibu e o widening
he wage gap, while dispa i ies in union wage p emiums
na ow he wage gap, he eby con i ming hypo hesis H2
once again.
Thi d, I excluded occupa ion and indus ial a iables
om he con olled a iables and conduc ed he B-O
decomposi ion (Model 2 in Table9). Columns [e] and [ ]
co espond o he esul s based on Eq. (11), while Col-
umns [g] and [h] co espond o he esul s based on Eq.
(12). The union e ec on he endowmen componen is
5.7% (Columns [e]) and 10.0% (Columns [g]) espec i ely,
sugges ing ha di e ences in union membe ship densi y
widen he wage gap. The union e ec on he p ice com-
ponen is −1.4% (Columns [ ]) and -5.7% (Columns [h])
espec i ely, indica ing ha di e ences in union wage
p emiums na ow he wage gap. These esul s also p o-
ide u he suppo o hypo hesis H2.
Las ly, ollowing Jann (2008), I u ilized he h ee-
old decomposi ion me hod (Table 10). The endow-
men e ec is calcula ed a 96.2%, while he p ice e ec
s ands a 21.3%, and he in e ac ion be ween bo h e ec s
is −17.5%. These esul s indica e ha he impac o he
endowmen e ec on he wage gap su passes ha o he
p ice e ec , which aligns wi h he indings in Table7
(endowmen e ec : 85.2% and 77.0%; p ice e ec : 14.8%
and 23.0%). These esul s ea i m hypo hesis H2.
6 Conclusions
As ade unions a e ac i e in co po a ions wo ldwide,
he union e ec s on he labo ma ke ha e a ac ed
global a en ion. Howe e , empi ical e idence is sca ce
ega ding how ade unions’ e ec s on he wage gaps di -
e be ween he public and p i a e sec o s, especially o
China. Using na ional longi udinal su ey da a om i e
wa es (2010, 2012, 2014, 2016, and 2018) o he CFPS
and he Blinde -Oaxaca decomposi ion me hod, his
s udy i s es ima es he e ec s o union membe ship
densi y on he wage gap be ween he public and p i a e
sec o s in China.
This s udy p esen s wo key indings. Fi s , he esul s
based on he OLS me hod, selec ion bias adjus ed model,
LV model, and PSM me hod indica e he exis ence o a
signi ican posi i e union wage p emium na ionwide,
and he p emium in he public sec o is g ea e han ha
in he p i a e sec o . Howe e , he p emium becomes
insigni ican a e add essing he issue o indi idual
Table 10 Robus ness check (3): Decomposi ion using h ee- old B-O decomposi ion me hod
Sou ce: Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018. Th ee- old B-O decomposi ion me hod is used. The esul s o wage unc ions based
on he OLS me hod a e used in he decomposi ion analysis
Coe . z
(1) Di e en ial
P edic ion_1(Public) 2.408*** 226.34
P edic ion_2(P i a e) 2.169*** 233.7
Di e ence 0.239*** 16.96
(2) Decomposi ion Value Pe cen age (%)
Endowmen 0.230*** 15.87 96.2
P ice 0.051*** 2.85 21.3
In e ac ion − 0.042*** -2.27 − 17.5

Page 21 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
he e ogenei y. Second, he decomposi ion esul s indi-
ca e ha he union membe ship densi y di e ence ( he
endowmen e ec ) widens he wage gap be ween he
public and p i a e sec o s; con e sely, he union wage
p emium di e ence ( he p ice e ec ) na ows he wage
gap.
These indings ha e he ollowing p ac ical implica-
ions: he di e ence in union membe ship densi y wid-
ens he wage gap (Tables7, 8 and 9). The p opo ion o
union membe s in he p i a e sec o (2.8–5.7% in 2010
and 2018, espec i ely) was lowe han ha in he pub-
lic sec o ( ange: 13.9–40.9%) (Fig.2). The e o e, a pol-
icy expanding union co e age in he p i a e sec o may
e ec i ely na ow he wage gap be ween bo h sec o s.
Wi h he p og ess o ma ke -o ien ed e o ms, income
inequali y has expanded and become a se e e issue
(Sicula e al. 2021). Al hough Zhang and Cheng (2011)
a gued ha because unions inc ease adminis a ion and
bu eauc a iza ion p ocesses, hey canno e ec i ely
imp o e China’s income inequali y p oblems, ou esul s
indica ed ha expanding he union membe ship densi y
in he p i a e sec o can be expec ed o na ow he wage
gap be ween he public and p i a e sec o s, which would
con ibu e o educing income inequali y in China o a
le el compa able o ha in de eloped coun ies (Ca d
1996; Fa be e al. 2021).
Finally, his s udy had some limi a ions. Fi s , while I
examined union wage p emiums in he public and p i a e
sec o s and a emp ed o mi iga e po en ial endogenei y
issues by using he FE model as well as a selec ion-bias
adjus men , LV and FE models, and he PSM me hod, i
is impo an o no e ha he endogenei y issues could no
be ully esol ed. Fu he esea ch is needed o explo e
he causal ela ionship be ween unions’ e ec on he wage
gap be ween he public and p i a e sec o s. To add ess
he indi idual he e ogenei y issue, p e ious s udies ha e
p edominan ly u ilized he FE model (e.g., F eeman 1980;
Mellow 1981; Mince 1981; F eeman and Medo 1984;
Walsh 2013; N lhola e  al. 2019), a me hodology also
employed in his s udy. Apa om he FE model, se e al
al e na i e econome ic me hods wa an conside a ion.
Fi s ly, ce ain s udies (Ca d 1996; Kulka ni and Hi sch
2021; Kölling 2022) u ilize he in o ma ion on he ans-
o ma ion o union membe ship ollowing job changes.
Due o he limi ed samples o changing union membe -
ship in he CFPS, I could no employ his me hod; how-
e e , u u e esea ch could explo e i s applica ion wi h
mo e ex ensi e obse a ions om app op ia e su ey
da a. Secondly, in China, a ia ions in union es ablish-
men be ween he public and p i a e sec o s, coupled
wi h changes in unions in he p i a e sec o in esponse
o e ol ing egula ions and policies, sugges he po en-
ial applica ion o quasi-expe imen al me hods (e.g.,
di e ence-in-di e ences [DID], eg ession discon inui y
design [RDD]). Fo ins ance, he DID me hod has been
employed o in es iga e he causal e ec s o unions on
wages in DiNa do and Lee (2004), Lee and Mas (2012),
and B eda (2015) in de eloped coun ies.
Second, as Pa sley (1980) and Boo h and B yan (2004)
a gued, a union has spillo e and h ea e ec s on non-
membe s. These e ec s may di e be ween he public
and p i a e sec o s and, hence, could in luence he wage
gap be ween hem. S udying hese e ec s p esen s new
challenges o esea che s.
Thi d, due o he limi a ions o he wage- ela ed su -
ey i ems in he 2010–2018 CFPS wa es, his s udy solely
es ima ed unions’ e ec on o e all wage. Gi en ha he
composi ion o compensa ion componen s (e.g., basic
wage, bonuses, and allowances) may a y be ween he
public and p i a e sec o s, u u e esea ch should aim o
es ima e he union wage e ec by compensa ion ype.
Fou h, while I ha e u ilized a iables such as occupa-
ion, indus y sec o , and owne ship ype om he CFPS
o mi iga e he in luence o he wo kplace on wo ke s’
wages, i is no ewo hy ha o he wo kplace ac o s like
i m size, echnological le el, i m pe o mance (e.g.,
deb , p o i ), and openness (such as adable co po a-
ions) migh impac wo ke s’ wages. The esea ch o con-
ol hese wo kplace ac o s based on employe -employee
su ey da a has become a new issue (B eda 2015; Gü z-
gen 2016; Masso e al. 2022). Fu he mo e, some demo-
g aphic a iables, such as pe sonali y ai s and isk
a ibu es, may a ec he likelihood o choosing public o
p i a e sec o and wage le els (Kamal and Blacklow 2022;
Roe hlisbe ge e al. 2023). The empi ical s udy conside -
ing hese ac o s based on app op ia e su ey da a should
be examined in u u e s udies.
Despi e hese limi a ions, I belie e ha he cu en
s udy, which ook ad an age o longi udinal da a om
he CFPS, p o ides insigh s in o unions’ e ec on wages
in he wo ld’s la ges eme ging ma ke economy, while
also o e ing new empi ical e idence o unde s anding
how unions a ec he wage gap be ween he public and
p i a e sec o s based on he decomposi ion me hods.
3 Page 22 o 25
X.Ma
Appendix
See Tables11 and 12
Table 11 De ini ions and desc ip i e s a is ics o a iables
The samples we e limi ed o hose aged 16–60. Due o he limi ed space, he desc ip i e s a is ics o p o ince-le el egion dummy a iables we e no p esen ed in he
able, hey a e a ailable upon eques . Sou ce: Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018
De ini ions Mean SD
Log o hou ly wage loga i hm o he hou ly wage calcula ed based on o al wage and wo k hou s 2.310 0.870
Union 1 = ade union membe , 0 = non-membe 0.095 0.293
Employmen sec o
Public sec o 1 = wo king in he public sec o , 0 = wo king in he p i a e sec o 0.217 0.412
Educa ion Yea s o schooling 10.412 4.254
Expe ience Yea s o expe ience = age-6-yea s o schooling 22.078 12.763
E hnici y 1 = Han majo i y, 0 = e hnic mino i ies 0.943 0.232
Heal h 1 = heal h s a us is e y good o good, 0 = o he wise 0.401 0.490
U ban 1 = u ban hukou, 0 = u al hukou 0.420 0.494
Ma ied 1 = ha ing a spouse,0 = o he wise 0.771 0.420
Pa y 1 = a membe o Communis Pa y o China,0 = non-CPC membe 0.111 0.314
Occupa ion
Manage 1 = manage o echnician,0 = o he wise 0.095 0.294
Technician 1 = echnician,0 = o he wise 0.131 0.337
Cle k 1 = cle k,0 = o he wise 0.098 0.298
Ope a o 1 = ope a o ,0 = o he wise 0.383 0.486
O he occupa ion 1 = o he occupa ion,0 = o he wise 0.293 0.455
Indus y
Manu ac u ing and cons uc ion 1 = wo king in he manu ac u al o cons uc ion indus y sec o ,0 = o he wise 0.363 0.481
T a ic and communica ion 1 = wo king in he a ic o communica ion indus y sec o ,0 = o he wise 0.066 0.249
Sales 1 = wo king in he a ic o sale indus y sec o ,0 = o he wise 0.189 0.392
Se ice 1 = wo king in he a ic o se ice indus y sec o ,0 = o he wise 0.220 0.414
O he indus y 1 = o he indus y,0 = o he wise 0.162 0.368
Yea
y2010 1 = 2010 su ey yea , 0 = o he wise 0.145 0.352
y2012 1 = 2012 su ey yea , 0 = o he wise 0.154 0.361
y2014 1 = 2014 su ey yea , 0 = o he wise 0.223 0.416
y2016 1 = 2016 su ey yea , 0 = o he wise 0.225 0.417
y2018 1 = 2018 su ey yea , 0 = o he wise 0.254 0.435
Obse a ions 14,084
Page 23 o 25 3 Union membe ship and hewage gap be ween hepublic andp i a e sec o s: e idence omChina
Table 12 Resul s o p obabili y unc ion o joining union
The p obi eg ession model is used. Sou ce: Calcula ed using da a om he CFPS o 2010, 2012, 2014, 2016 and 2018.
***p < 0.01
**p < 0.05
*p < 0
(1) Na ion (2) Public (3) P i a e
Coe z Coe z Coe z
Employmen sec o (Re .: P i a e)
Public 0.647*** 18.96
Educa ion 0.115*** 17.14 0.116*** 12.22 0.117*** 12.36
Exp 0.047*** 7.98 0.047*** 6.04 0.054*** 5.98
Exp_sq 0.000*** − 3.68 0.000*** − 2.69 − 0.001*** − 3.13
Heal h − 0.213*** − 6.63 − 0.221*** − 5.18 − 0.184*** − 3.83
Female − 0.048 − 1.48 − 0.026 − 0.60 − 0.068 − 1.45
Han majo i y 0.194** 2.35 0.353*** 3.35 − 0.044 − 0.35
Ma ied 0.028 0.60 0.029 0.45 0.026 0.38
CPC membe ship 0.084** 2.11 0.138*** 2.92 0.119* 1.68
Occupa ion (Re .:
Ope a o )
O icial and manage − 0.169*** − 2.74 − 0.457*** − 5.37 0.105 1.23
Technician − 0.141** − 2.53 − 0.330*** − 4.49 0.090 1.04
Cle k − 0.039 − 0.78 − 0.276*** − 4.21 0.298*** 3.88
O he − 0.162*** − 3.35 − 0.342*** − 4.93 0.004 0.06
Indus y (Re . Manu.
and Cons.)
T a ic and commu-
nica ion − 0.173*** − 2.88 − 0.038 − 0.48 − 0.322*** − 3.35
Sale − 0.133** − 2.27 − 0.107 − 0.98 − 0.165** − 2.33
Se ice − 0.232*** − 4.95 − 0.124* − 1.92 − 0.340*** − 4.93
O he − 0.293*** − 6.15 − 0.189*** − 3.08 − 0.293*** − 3.70
Pa en backg ound
Fa he : CPC membe 0.170*** 4.89 0.167*** 3.75 0.161*** 2.94
Mo he : CPC membe 0.086 1.24 0.053 0.62 0.142 1.26
Fa he in public
sec o 0.150 1.05 0.286* 1.73 − 0.159 − 0.49
Mo he in public
sec o 0.164 0.55 0.052 0.15 0.451 0.83
Yea (Re .: y2010)
y2012 0.291*** 7.53 0.330*** 6.39 0.224*** 3.97
y2014 0.295*** 6.76 0.362*** 6.45 0.191*** 2.79
y2016 0.317*** 4.76 0.565*** 5.37 0.091 1.03
y2018 0.785*** 15.90 1.007*** 15.70 0.400*** 5.11
P o ince ixed e ec Yes Yes Yes
Cons an s − 3.647*** − 26.38 − 3.175*** − 16.58 − 3.534*** − 17.49
No. o sample 14,084 5,496 8,588
Log likelihood − 4641.980 − 2742.983 − 1977.211
Pseudo R2 0.202 0.160 0.134
P ob > chi2 0.000 0.145 0.000
3 Page 24 o 25
X.Ma
Supplemen a y In o ma ion
The online e sion con ains supplemen a y ma e ial a ailable a h ps:// doi.
o g/ 10. 1186/ s12651- 024- 00361-2.
Addi ional ile1: TableS1. Resul s o wage unc ion based on he OLS
me hod and FE model. TableS2. Resul s o p obabili y unc ions used in
selec ion bias adjus ed me hod. TableS3. Decomposi ion esul s based
on he B-O me hod and RE/FE model.
Acknowledgemen s
No applicable.
Au ho con ibu ions
XMA designed he s udy; collec ed he da a; pe o med he o mal analyses;
and w o e, ead, and app o ed he inal manusc ip .
Funding
This esea ch was suppo ed by he p ojec g an o he Join Usage and
Resea ch Cen e , Ins i u e o Economic Resea ch, Hi o subashi Uni e si y,
Japan (G an numbe : IERPK2323).
Da a a ailabili y
The da a used in his s udy can be ob ained om h ps:// opend a a. pku. edu.
cn/ da a e se/ CFPS? langu age= en. The da ase s cons uc ed by he au ho a e
a ailable upon eques .
Decla a ions
E hics app o al and consen o pa icipa e
The da ase used in his s udy, he China Family Panel S udies (CFPS), is pub-
licly a ailable (h p:// opend a a. pku. edu. cn/ en). The medical e hics commi ee
app o ed he CFPS s udy, and all in e iewees we e equi ed o sign in o med
consen . E hics app o al o he da a collec ion in CFPS was ob ained om
he Biomedical E hics Re iew Commi ee o Peking Uni e si y (IRB00001052-
14010). E hics app o al o he use o CFPS da a was ob ained om he Uni e -
si y o Newcas le Human Resea ch E hics Commi ee. All he pa icipan s ga e
hei w i en in o med consen be o e any s udy p ocedu es began.
Consen o publica ion
The au ho o his s udy decla es ha his manusc ip does no in inge
copy igh , mo al igh s, o o he in ellec ual p ope y igh s o any o he
pe son. The au ho o his s udy es i ies ha his pape is o iginal and has no
incu ed any ype o plagia ism. The au ho o his s udy decla es ha he a i-
cle is o iginal, has no al eady been published in a jou nal, and is no cu en ly
unde conside a ion by ano he jou nal. The au ho o his s udy ag ees o he
e ms o he Sp inge Open Copy igh and License Ag eemen .
Compe ing in e es s
The au ho decla es no con lic s o in e es associa ed wi h his s udy.
Recei ed: 20 June 2023 Accep ed: 7 Janua y 2024
Re e ences
Becke , G.S.: Human capi al: A heo e ical and empi ical analysis, wi h special
e e ence o educa ion, 1s edn. Columbia Uni e si y P ess, Columbia
(1964)
Blanch lowe , D.G., B yson, A.: The wage impac o ade unions in he UK
public and p i a e sec o s. Economica 77(305), 92–109 (2010)
Blinde , A.S.: Wage disc imina ion: Reduced o m and s uc u al es ima ion. J.
Hum. Res. 8, 436–455 (1973)
Boo h, A., B yan, M.: The union membe ship wage p emium puzzle: is he e a
ee ide p oblem? Ind. Labo Rela . Re . 57(3), 402–421 (2014)
B eda, T.: Fi ms’ en s, wo ke s’ ba gaining powe and he union wage p e-
mium. Econ. J. 125(589), 1616–1652 (2015)
Ca d, D.: The e ec o unions on he s uc u e o wages: a longi udinal analysis.
Econome ica 64(4), 957–979 (1996)
Ca dozo, D.P., da Cunha, M.S.: E idence o he public–p i a e wage gap o
B azil in he pe iod om 2008 o 2016. Ind. J. Labou Econ. 62, 577–594
(2019)
Casale, D., Posel, D.: Unions and he gende wage gap in Sou h A ica. J. A .
Econ. 20(1), 27–59 (2010)
Chang, J., Huang, H.: T ade unions, unemploymen , economic g ow h, and
income inequali y. Mac oecon. Dyn. 20(1), 404–428 (2016)
Chen, G., Demu ge , S., Fou nie , M.: Wage di e en ials and owne ship s uc-
u e o China’s en e p ises. Wo ld Econ. Pape 6, 11–31 (2005)
Cheng, H., Hu, D., Li, H.: Wage di e en ial be ween u al mig an and u ban
wo ke s in he People’s Republic o China. Asian De . Re . 37(1), 43–60
(2020)
China Family Panel S udies (CFPS): h ps:// opend a a. pku. edu. cn/ da a e se/
CFPS? langu age= en. Accessed on May 20, 2022.
Choi, H., Ramos, R.: The union wage p emium in a segmen ed labo ma ke :
new e idence om Ko ea. J. Asian Pac. Econ. 28, 1979352 (2021)
Demu ge , S., Li, S., Yang, J.: Ea ning di e en ials be ween he public and
p i a e sec o s in China: explo ing changes o u ban local esiden s in
he 2002s. China Econ. Re . 23, 138–153 (2012)
DiNa do, J., Lee, D.S.: Economic impac s o new unioniza ion on p i a e sec o
employe s: 1984–2001. Q. J. Econ. 119(4), 1383–1441 (2004)
Dubin, J., McFadden, D.: An econome ic analysis o esiden ial elec ic appli-
ance holdings and consump ion. Econome ica 52(2), 345–362 (1984)
Fa be , H.S., He bs , D., Kuziemko, I., Naidu, S.: Unions and inequali y o e he
wen ie h cen u y: new e idence om su ey da a. Q. J. Econ. 136(3),
1325–1385 (2021)
F eeman, R.B.: The exi - oice adeo in he labo ma ke : unionism, job enu e,
qui s and sepa a ions. Q. J. Econ. 94(4), 643–673 (1980)
F eeman, R.B., Medo , J. L.: Wha Do Unions Do? Basic Books (1984).
Gindling, T.H., Hasnain, Z., Newhouse, D., Shi, R.: A e public sec o wo ke s in
de eloping coun ies o e paid? E idence om a new global da ase .
Wo ld De . 126, 104737 (2020)
Gong, B.: Like a he like son? Re isi ing he ole o pa en al educa ion in es i-
ma ing e u ns o educa ion in China. Re . De . Econ. 23, 275–292 (2019)
Gunde son, M., Lee, B.Y., Wang, H.: Union pay p emium in China: an indi idual-
le el analysis. In . J. Manpow. 37(4), 606–627 (2016)
Guo, H., Dai, W.: O iginal aspi a ion and mission: his o ical e iew and expe i-
ence o Chinese ade unions pa icipa ing in na ional go e nance. J.
Beijing Fede a ion T ade Unions Cad e Coll. 37(4), 53–62 (2022)
Gü zgen, N.: Es ima ing he wage p emium o collec i e wage con ac s:
e idence om longi udinal linked employe –employee da a. Indus. Rela-
ions 55(2), 294–322 (2016)
Gus a sson, B., Li, S.: Economic ans o ma ion and he gende ea nings gap in
u ban China. J. Popul. Econ. 13(2), 305–329 (2000)
Hi sch, B. T., John T. A.: The Economic Analysis o Unions: New App oaches and
E idence. Allen and Unwin (1986).
Iwasaki, I., Ma, X., Mizoba a, S.: Co po a e owne ship and manage ial u no e
in China and Eas e n Eu ope: a compa a i e me a-analysis. J. Bus. Econ.
111, 105928 (2020)
Iwasaki, I., Ma, X., Mizoba a, S.: Owne ship s uc u e and i m pe o mance in
eme ging ma ke s: a compa a i e me a-analysis o Eas Eu opean EU
membe s a es, Russia, and China. Econ. Sys . 46(2), 100945 (2022)
Jin, X., Xu, L., Xin, Y., Adhika i, A.: Poli ical go e nance in China’s s a e-owned
en e p ises. China J. Accoun . Res. 13, 100236 (2022)
Kamal, M., Blacklow, P.: A i udes and pe sonali y in he Aus alian gende
wage gap. Appl. Econ. 54(47), 5442–5459 (2022)
Ke , A., Wi enbe g, M.: Union wage p emia and wage inequali y in Sou h
A ica. Econ. Model. 97(C), 255–271 (2021)
Koenke , R.W., Basse , G.J.: Reg ession quan ile. Econome ica 46(1), 33–50
(1978)
Kölling, A.: Sho age o skilled labo , unions and he wage p emium: a eg es-
sion analysis wi h es ablishmen panel da a o Ge many. J. Labo Res. 43,
239–259 (2022)
Kulka ni, A., Hi sch, B.T.: Re isi ing union wage and job loss e ec s using he
displaced wo ke su eys. Ind. Labo Rela . Re . 74(4), 948–976 (2021)
Lee, D., Mas, A.: Long- un impac s o unions on i ms: new e idence om
inancial ma ke s, 1961–1999. Q. J. Econ. 127(1), 333–378 (2012)
Lemke, R.J., Rischall, I.C.: Skill, pa en al income, and IV es ima ion o he e u ns
o schooling. Appl. Econ. Le . 10(5), 281–286 (2003)