Ji akom Si is isakulchai; Supanika Leu cha usmee
A icle
Re u ns o highe educa ion by ield o s udy in Thailand:
Compa a i e analysis o p e and pos -COVID-19 pandemic
Asian Jou nal o Economics and Banking (AJEB)
P o ided in Coope a ion wi h:
Ho Chi Minh Uni e si y o Banking (HUB), Ho Chi Minh Ci y
Sugges ed Ci a ion: Ji akom Si is isakulchai; Supanika Leu cha usmee (2025) : Re u ns o highe
educa ion by ield o s udy in Thailand: Compa a i e analysis o p e and pos -COVID-19 pandemic,
Asian Jou nal o Economics and Banking (AJEB), ISSN 2633-7991, Eme ald, Leeds, Vol. 9, Iss. 1, pp.
2-21,
h ps://doi.o g/10.1108/AJEB-09-2024-0106
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/334134
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/
Re u ns o highe educa ion by ield
o s udy in Thailand: compa a i e
analysis o p e and
pos -COVID-19 pandemic
Ji akom Si is isakulchai and Supanika Leu cha usmee
Facul y o Economics, Chiang Mai Uni e si y, Chiang Mai, Thailand
Abs ac
Pu pose –This s udy es ima es e u ns o highe educa ion ac oss di e en ields in Thailand o 2019 and 2021,
accoun ing o ield selec ion endogenei y. The compa ison o e s insigh s in o he impac o he pandemic and
o he economic shocks on he e u ns.
Design/me hodology/app oach –The s udy applies an econome ic causal amewo k, in eg a ing economic
heo y wi h empi ical analysis using da a om Thailand’s socioeconomic su eys in 2019 and 2021.
A mul inomial ea men e ec s model wi h sample selec ion co ec ion is used o es ima e he impac o
di e en ields o s udy on income, accoun ing o selec ion bo h in o highe educa ion in di e en ields and in o
employmen , add essing po en ial biases om abili y so ing and sample selec ion.
Findings –The s udy inds a ia ions in e u ns o educa ion ac oss ields. In 2019, eaching o e ed he highes
e u ns on a e age, ollowed by heal hca e. Social sciences, business and compu e - ela ed ields showed
mode a e e u ns, while he combined g oup o science, ag icul u e, enginee ing and a chi ec u e had
non-signi ican e u ns, indica ing a low weigh ed a e age ac oss hese di e se ields. In 2021, heal hca e
exhibi ed he highes e u n due o pandemic-d i en demand. Ac oss bo h yea s, con olling o occupa ion
educed he es ima ed e u ns by app oxima ely 50%, highligh ing he ole o occupa ional s a us in media ing
educa ional e u ns.
O iginali y/ alue –This s udy uniquely applies an econome ic causal amewo k o analyze e u ns o highe
educa ion by ield o s udy in Thailand. I o e s insigh s o policymake s o align educa ional p og ams wi h
labo ma ke demand and emphasizes he impo ance o da a-d i en decisions in esponding o dis up ions.
Keywo ds Re u ns o schooling, Field o s udy, Causal amewo k, Abili y so ing, Mul inomial ea men
Pape ype Resea ch pape
1. In oduc ion
Choosing a ield o s udy in highe educa ion is a pi o al decision ha signi ican ly impac s an
indi idual’s u u e ea nings. As mo e esou ces a e in es ed in highe educa ion wo ldwide,
unde s anding he bene i s o di e en s udy ields has become inc easingly i al. This
impo ance is e iden in he subs an ial g ow h o global highe educa ion pa icipa ion.
Acco ding o UNESCO (2022), he pe cen age o people en olled in highe educa ion doubled
om 19% in 2000 o abou 40% in 2020. This end is also seen in Thailand, whe e he
p opo ion o adul s wi h highe educa ion quali ica ions, including oca ional o college
deg ees, has s eadily inc eased, eaching 20% in 2021 (NSO, 2022).
The COVID-19 pandemic has signi ican ly al e ed labo ma ke s, inc easing demand in
sec o s like heal hca e while educing i in o he s such as ou ism and hospi ali y (ADB, 2021).
These shi s, coupled wi h he ongoing Fou h Indus ial Re olu ion, ha e led o a ying
e u ns on educa ion ac oss di e en ields o s udy (WEF, 2020). This s udy compa es labo
AJEB
9,1
2
JEL Classi ica ion — C21, J21, J24, J31
© Ji akom Si is isakulchai and Supanika Leu cha usmee. Published in Asian Jou nal o Economics
and Banking. Published by Eme ald Publishing Limi ed. This a icle is published unde he C ea i e
Commons A ibu ion (CC BY 4.0) licence. Anyone may ep oduce, dis ibu e, ansla e and c ea e
de i a i e wo ks o his a icle ( o bo h comme cial and non-comme cial pu poses), subjec o ull
a ibu ion o he o iginal publica ion and au ho s. The ull e ms o his licence may be seen a h p://
c ea i ecommons.o g/licences/by/4.0/legalcode
The cu en issue and ull ex a chi e o his jou nal is a ailable on Eme ald Insigh a :
h ps://www.eme ald.com/insigh /2615-9821.h m
Recei ed 8 Sep embe 2024
Re ised 18 No embe 2024
29 Decembe 2024
Accep ed 9 Janua y 2025
Asian Jou nal o Economics and Banking
Vol. 9 No. 1, 2025
pp. 2-21
Eme ald Publishing Limi ed
e-ISSN: 2633-7991
p-ISSN: 2615-9821
DOI 10.1108/AJEB-09-2024-0106
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
ma ke ou comes om 2019 o 2021, p o iding insigh s in o how he pandemic and Indus y
4.0 ha e collec i ely in luenced educa ional e u ns o highe educa ion ac oss di e en ields
o s udy in Thailand.
Re u ns o educa ion ha e been p ima ily es ima ed h ough wo me hods: he ull-
discoun ing me hod and he Mince ian ea nings unc ion, wi h he la e being mo e
commonly employed due o i s b oad applicabili y (Pa inos and Psacha opoulos, 2018).
Global esea ch e eals a consis en a e age p i a e e u n o app oxima ely 9% annually pe
addi ional yea o schooling, wi h minimal decline o e ime (Psacha opoulos and Pa inos,
2018). This end aligns wi h Tinbe gen (1975)’s concep o a “ ace be ween educa ion and
echnology,” whe e skill demand ou paces supply despi e inc eased educa ional a ainmen .
While gene al e u ns o educa ion ha e ma ginally dec eased, p i a e e u ns o highe
educa ion ha e isen, exace ba ing equi y conce ns. Goldin and Ka z (2009) a ibu e his o
skill-biased echnological p og ess, which has widened income inequali y and main ained
high p emiums o ad anced skills. This sugges s ha educa ional ad ancemen s lag behind
he g owing demand o highe -le el skills, po en ially indica ing echnology’s lead in his
ace. Consequen ly, analyzing e u ns by speci ic ields o s udy, a he han o e all highe
educa ion, becomes c ucial o iden i y which skills yield he highes e u ns in an e ol ing
labo ma ke .
Howe e , es ima ing e u ns by ield o s udy p esen s challenges beyond he ypical
conce ns o abili y and selec ion bias. Abili y so ing ac oss ields in oduces addi ional
po en ial biases o he es ima ion. Resea ch shows ha indi iduals wi h highe abili ies end o
selec ields ha yield highe e u ns, complica ing he assessmen o educa ional e u ns
(Psacha opoulos and Pa inos, 2018;A cidiacono, 2004). This abili y so ing can lead o
o e es ima ion o e u ns o ce ain ields, as adi ional me hods may ail o accoun o he
in luence o unobse ed cha ac e is ics on bo h educa ional a ainmen and labo ma ke
ou comes (Ebe ha d e al., 2017;Lemieux, 2014;Bol and Heisig, 2021).
To add ess biases s emming om unobse ed he e ogenei y, namely selec ion and abili y
so ing biases, esea che s ha e employed ad anced econome ic echniques. S uc u al
models, like hose used by A cidiacono (2004), cap u e indi iduals’ dynamic decisions
ega ding educa ion and occupa ion. Quasi-expe imen al app oaches, such as eg ession
discon inui y designs (Canaan and Mouganie, 2018), decomposi ion me hod (Bol and Heisig,
2021) and ins umen al a iable me hods (Ki keboen e al., 2016), ha e also been u ilized o
isola e he impac o ield-speci ic educa ion on ea nings.
To sys ema ically analyze he causal e ec o choosing di e en ields o s udy on wages,
while clea ly communica ing he limi a ions o he model and po en ial biases in es ima ing he
e u n o educa ion, i is c ucial o adop a s uc u ed app oach g ounded in he econome ic
causal amewo k. This amewo k p o ides a igo ous ounda ion o add essing he
complexi ies inhe en in causal in e ence, ensu ing ha he analysis is no only
me hodologically sound bu also anspa en ega ding he assump ions and po en ial sou ces
o bias. By explici ly aming he analysis wi hin his app oach, we o e a clea pa hway o
o he esea che s o eplica e and c i ique ou indings, while also highligh ing he c i ical
ac o s ha could lead o biased es ima es, such as unobse ed he e ogenei y o selec ion bias.
In doing so, we align ou esea ch wi h bes p ac ices in econome ic analysis, emphasizing he
impo ance o ca e ully conside ing model speci ica ions and he obus ness o causal
in e p e a ions when assessing he e u ns o educa ion ac oss di e en ields o s udy.
To add ess hese po en ial biases wi hin econome ic causal amewo k, his pape employs
a mul inomial ea men e ec s (MTE) model wi h sample selec ion co ec ion o analyze he
e u ns o educa ion ac oss a ious ields o s udy in Thailand, explici ly add essing bo h he
abili y so ing bias and selec ion in o paid employmen . The mul inomial ea men model
allows o he es ima ion o ea men e ec s when he ea men a iable is mul inomial in
na u e, meaning ha indi iduals can choose among mul iple educa ional ields (Deb and
T i edi, 2006). By compa ing educa ional e u ns be o e and a e he COVID-19 pandemic,
his s udy aims o p o ide a comp ehensi e unde s anding o how economic shocks in luence
Asian Jou nal o
Economics and
Banking
3
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
s uden choices and subsequen wage ou comes. This app oach enables he analysis o he
impac o di e en ields o s udy on wage ou comes while con olling o selec ion bias and
unobse ed he e ogenei y, he eby p o iding mo e accu a e es ima es o he economic e u ns
associa ed wi h each ield.
This esea ch con ibu es o he g owing body o li e a u e on educa ional e u ns,
pa icula ly in he con ex o de eloping economies like Thailand. By employing a anspa en
me hodology o analyze he e u ns o highe educa ion ac oss di e se ields o s udy, while
simul aneously accoun ing o selec ion bias and ield-speci ic abili y so ing, his esea ch
aims o p o ide mo e p ecise es ima es o he economic e u ns associa ed wi h each academic
ield. The insigh s gained can in o m policymake s and educa ional ins i u ions in aligning
educa ional p og ams wi h labo ma ke needs and suppo ing s uden s in making in o med
decisions abou hei ields o s udy, especially in he ace o economic dis up ion and
unce ain ies.
This pape is o ganized as ollows: Sec ion 2 p esen s he econome ic causal amewo k,
ollowed by Sec ion 3, which de elops he causal model, including he economic modeling o
e u ns o educa ion and ield o s udy decisions. Sec ion 4 discusses he iden i ica ion o
causal pa ame e s, ou lining he coun e ac ual app oach, hypo he ical models, and empi ical
s a egy. Sec ion 5 de ails he da a sou ces and sample cha ac e is ics. Sec ion 6 p esen s he
esul s, ocusing on he es ima ed e u ns o highe educa ion by ield o s udy and he ole o
occupa ional s a us as a media o . Finally, Sec ion 7 concludes he pape wi h policy
implica ions and sugges ions o u u e esea ch.
2. Econome ic causal amewo k
The concep o causal in e ence in econome ics da es back o Ragna F isch’s seminal lec u e
in 1930 (Bje khol and Qin, 2011). F isch concep ualized causali y as a hough expe imen
whe e economis s hypo he ically analyze how changes in inpu s would in luence ou pu s. This
pe spec i e laid he ounda ion o mode n causal analysis in economics, ocusing on wo
undamen al p inciples: au onomy and di ec ionali y (Heckman and Pin o, 2024).
Au onomy in econome ics e e s o he concep explained by F isch (1938). The e m
au onomy means au onomy o (causal) ela ionship and au onomy o a unc ion is de ined as
unc ions ha exhibi in a iance p ope y in he sense ha hey a e in a ian o any changes in
hei a gumen s (Heckman and Pin o, 2024). In econome ic causal model, his means ha
each causal ela ionship in a model is sel -con ained and no di ec ly a ec ed by changes in
un ela ed a iables. Fo ins ance, in a e u n- o-educa ion model, he ela ionship be ween
yea s o schooling and income should emain alid ega dless o ex e nal economic shi s,
p o ided he model is well-speci ied. This cha ac e is ic o au onomy is essen ial o isola ing
he ue e ec o educa ion on income wi hou con amina ion om ex aneous ac o s.
Fo he concep o di ec ionali y, Heckman and Pin o (2024) highligh ed ha causali y
unc ions in one di ec ion, lowing om he cause o he e ec . In gene al, i we change he
cause, i will a ec he ou come, bu changing he ou come won’ a ec he o iginal cause. In
he con ex o educa ion and ea nings, his implies ha changes in educa ional a ainmen
in luence income, no he o he way a ound. This concep helps a oid issues like ci cula
easoning o simul anei y bias, ensu ing ha he es ima ed e u n on educa ion e lec s a ue
causal e ec a he han jus a co ela ion.
Unlike adi ional s a is ical amewo ks, which ely on join dis ibu ions o desc ibe
ela ionships be ween a iables wi hou speci ying causal di ec ionali y, he econome ic
causal amewo k in oduces a c i ical laye o analysis. In s a is ics, he ela ionship be ween
wo a iables can be ully cap u ed by hei join dis ibu ion, ye his app oach lea es he
di ec ion o causali y ambiguous. Fo ins ance, while co ela ion o mu ual dependence
be ween educa ion and income can be quan i ied, i does no cla i y whe he educa ion di ec ly
causes highe income o i highe income leads o mo e educa ion. The econome ic causal
amewo k add esses his gap by es ablishing a s uc u ed app oach o de e mining causal
AJEB
9,1
4
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
ela ionships, emphasizing he di ec ionali y o in luence om cause o e ec . This s uc u ed
amewo k is pa icula ly aluable in applied econome ic analysis, whe e he goal is o en o
answe speci ic causal ques ions ele an o policy decisions o heo e ical inqui ies.
The econome ic causal amewo k (Heckman and Pin o, 2024) is buil a ound h ee co e
asks ha a e essen ial o causal in e ence: (1) cons uc ing a causal model, (2) iden i ying
causal pa ame e s, and (3) pe o ming s a is ical in e ence. These asks ensu e ha he analysis
emains aligned wi h he speci ic causal ques ions being in es iga ed by he analys . We will
explain he de ails o he h ee key asks in he ollowing sec ions o ensu e a comp ehensi e
unde s anding o he causal analysis p ocess in his esea ch.
3. Causal model
The i s ask, de eloping a causal model, in ol es de ining a sys em o s uc u al equa ions
ha map he ela ionships be ween a se o a iables and hei powe se . Each equa ion in his
sys em is an in a ian mapping om inpu s o ou pu s, whe e he inpu s a e unde s ood as he
di ec causes o he ou pu s.
Heckman and Pin o (2022) ma hema ically illus a e how causal models in economics can
be cons uc ed, using he Gene alized Roy model (Roy, 1951) as an example. They
demons a e how causal ela ionships can be ep esen ed wi h Di ec ed Acyclic G aphs
(DAGs), which cla i y he di ec ion o causali y. Addi ionally, hey highligh he Local
Ma ko Condi ion (LMC), which ully cha ac e izes causal models by speci ying ha a
a iable is condi ionally independen o i s non-descendan s gi en i s pa en s in he g aph.
The cons uc ion o he causal model in his s udy elies on an economic model o Field o
S udy Decisions and Wage Ou comes, which will be de ailed in he ollowing sec ions. This
amewo k unde pins he analysis by explici ly modeling he selec ion p ocess in o di e en
ields o s udy and how hese choices ansla e in o wage ou comes, allowing o a igo ous
examina ion o he causal ela ionships in ol ed.
3.1 Economic modeling o he e u ns o educa ion wi h endogenous decision on ield
o s udy
Indi iduals make educa ional decisions, including he le el o s udy and ield o s udy, o
maximize hei li e ime u ili y. This modeling amewo k is based on he wo k o Al onji e al.
(2016) in “The Analysis o Field Choice in College and G adua e School: De e minan s and
Wage E ec s.” Howe e , unlike he o iginal model, which allows o swi ching ields o s udy,
his pape assumes ha indi iduals selec hei ield o s udy once. This assump ion is made
due o limi ed obse a ions o ield-swi ching in Thailand, whe e mos indi iduals choose hei
ield o s udy only once, ei he in oca ional educa ion o a he bachelo ’s deg ee le el.
The decision-making p ocess is di ided in o wo dis inc pe iods: he educa ion pe iod and
he wo king pe iod. Du ing he educa ion pe iod (Pe iod 1), indi iduals ace a c i ical decision
ega ding whe he hey should in es in a highe educa ion and which ield o s udy o pu sue.
The model inco po a es Abilij, which is he belie o he indi idual i ega ding his abili y in he
ield o s udy j. The expec ed u ili y du ing his pe iod is exp essed as:
Uij1¼α0jþα1jAbilij;(Eq.1)
whe e Uij1 ep esen s he u ili y in he educa ion pe iod o indi idual iin ield j. F om his
speci ica ion, α0jcap u es he baseline u ili y associa ed wi h ield j, and α1j e lec s he
sensi i i y o u ili y o he indi idual’s abili y Abilij.
In he wo king pe iod (Pe iod 2 onwa d), he u ili y unc ion is ep esen ed as:
Uij ¼NPj�Abilij�þγ1jWagej�Abilij�;(Eq.2)
Asian Jou nal o
Economics and
Banking
5
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
whe e Uij deno es he u ili y o indi idual ide i ed om he chosen ield ja ime ¼2;. . . ;T
and Tis he e i emen pe iod. The e m NPjðAbilijÞcap u es he nonpecunia y bene i s, which
depends on he abili y Abili. The pa ame e γ1 e lec s he weigh gi en o he expec ed wage
WagejðAbilijÞ, which is also con ingen on he indi idual’s abili y.
Indi iduals make hei educa ional decisions o maximize he discoun ed sum o hei
expec ed payo s, exp essed as he ollowing expec ed li e ime u ili y unc ion:
Ui¼E"X
j
dij�Uij1þeij1�þX
T
¼2X
j
β −1dij�Uij þeij �#;(Eq.3)
whe e dij is he dummy a iable, which equals 1 i indi idual ichooses ield o s udy jand
0 o he wise, and eij is he idiosync a ic shock.
I an indi idual does no in es in highe educa ion, hen Uij ¼0, indica ing ha hei
u ili y is no malized o a baseline o only comple ing high school o lowe educa ion. In his
con ex , Uij e lec s he ela i e u ili y compa ed o no pu suing a speci ic ield o s udy.
I is c ucial o ecognize ha an indi idual’s abili y in a chosen ield o s udy, Abilij,
encompasses unobse ed cha ac e is ics ha may in luence he decision-making p ocess,
including p e e ences and as es (Al onji e al., 2016). This unobse ed ield-speci ic abili y
can cause “abili y so ing” bias in he e u ns o educa ion es ima ion. Speci ically, abili y
so ing can occu when indi iduals wi h highe abili y in a speci ic ield Abilij selec ields o
s udy ha yield highe e u ns, bo h du ing hei educa ion and in he labo ma ke . I abili y
Abilij in luences ou comes in he labo ma ke (i.e. γ1j>0), he educa ional ins i u ion’s ole is
o e eal o enhance his abili y, which hen impac s he indi idual’s p oduc i i y and
nonpecunia y bene i s in he labo ma ke . Con e sely, i abili y Abilij ma e s du ing he
educa ion pe iod (i.e. α1j>0), ields o s udy wi h highe po en ial e u ns migh be a oided
by indi iduals wi h lowe abili y due o he academic di icul y, leading o bo h abili y and sel -
selec ion bias. Thus, when es ima ing he ea men e ec o a ield o s udy on wage
ou comes, i is essen ial o accoun o po en ial abili y so ing, which could o he wise bias he
es ima es.
3.2 Causal model o ield o s udy decisions and wage
The causal model in his s udy is based on an economic model (desc ibed in Sec ion 3.1) ha
cap u es he ela ionship be ween ield o s udy decisions and wage ou comes. The model is
ep esen ed by he Di ec ed Acyclic G aph (DAG) in Figu e 1, whe e he ou come a iable is
wage (deno ed by Wage), he endogenous a iable is he le el o educa ion segmen ed by ield
o s udy (deno ed by Edu), and he con ounding ac o is abili y (deno ed by Abil), which
in luences bo h educa ional decisions and labo ma ke ou comes. The exogenous a iables
include indi idual cha ac e is ics (deno ed by IC) such as gende and age, and household
backg ound. Addi ionally, employmen s a us (deno ed by Emp) plays a key ole as an
in e media y a iable a ec ing bo h occupa ion and wage ou comes. The s uc u al equa ions
ha go e n he ela ionships among hese a iables a e as ollows:
Wage ¼ Wage�Edu;Abil;Exp;Occ;εWage�
Occ ¼ OccðEdu;εOccÞ
Exp ¼ Exp�εExp�
Edu ¼ EduðIC;Abil;εEduÞ
AJEB
9,1
6
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
Emp ¼ Emp�IC;Edu;Exp;εEmp�
IC ¼ ICðεICÞ
Abil ¼ AbilðεAbilÞ
In his model, εWage;εOcc;εExp;εEdu;εEmp;εIC and εAbil ep esen unobse ed ac o s a ec ing
he espec i e a iables. The Local Ma ko Condi ion (LMC) ully cha ac e izes he causal
model by asse ing ha each a iable is condi ionally independen o i s non-descendan s
gi en i s di ec pa en s in he DAG. Speci ically:
Wage ⊥ Emp;ICgj Edu;Exp;Occ;Abilg
Occ ⊥ IC;Exp;Emp;Abilgj Edug
Exp ⊥ Occ;Edu;IC;Abilg
Emp ⊥ Wage;Occ;Abilgj IC;Edu;Expg
Edu ⊥ Expgj IC;Abilg
IC ⊥ Exp;Abilg
Abil ⊥ Exp;ICg
Figu e 1. DAG ep esen s causal model o he e u ns o educa ion wi h endogenous decision on ield o s udy
Asian Jou nal o
Economics and
Banking
7
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
By cha ac e izing he causal model h ough he LMC, we ensu e ha he iden i ica ion o
causal e ec s is sys ema ically g ounded in he s uc u e o he ela ionships among a iables.
4. Iden i ica ion o causal pa ame e s
In his sec ion, we explain Task 2, which in ol es iden i ying he causal pa ame e using a
coun e ac ual app oach, and he hypo he ical model used o his pu pose is de ailed in
Sec ions 4.1–4.3. Task 3, ocusing on he es ima ion using he empi ical model, is co e ed in
Sec ion 4.4.
4.1 Coun e ac ual app oach
In causal in e ence, he coun e ac ual app oach is undamen al o es ima ing causal e ec s
de i ed om a causal model. The p ima y objec i e o a causal model is o es ima e he causal
e ec ha answe s a speci ic causal ques ion. A causal e ec , o causal pa ame e , is de ined
by he coun e ac ual concep , whe e he e ec is he di e ence in ou comes esul ing om
manipula ing o in e ening on he inpu a iable, while holding he con ounding ac o s
cons an . This implies ha he unde lying s uc u al equa ions emain unchanged e en when
he inpu a iable is al e ed.
Fo ins ance, conside a scena io whe e he ea men a iable Tcan ake wo alues, T¼1
( ea ed) and T¼0(un ea ed). The causal e ec is hen de ined as he di e ence be ween he
ou come unde ea men Yð1Þand he ou come wi hou ea men Yð0Þ,Yð1Þ−Yð0Þ. This
di e ence ep esen s he impac o he ea men on he ou come, assuming ha all o he
ac o s emain cons an . A key poin om his de ini ion is ha he causal e ec is de e mined
en i ely by he causal model, wi hou elying on any p obabili y model o s a is ical
assump ions. The causal e ec is a heo e ical idea ha is de ined wi hin he causal model and
does no depend on he likelihood o he ea men o he way he da a is dis ibu ed.
Howe e , when we mo e om heo y o empi ical es ima ion, p obabili y models become
essen ial. These models allow us o es ima e causal pa ame e s using obse ed da a. Fo
example, o es ima e he A e age T ea men E ec (ATE), which is he expec ed di e ence in
ou comes ac oss he popula ion, we would calcula e E½Yð1Þ−Yð0Þ�. He e, he p obabili y
model helps us u ilize he da a o es ima e he a e age causal e ec , inco po a ing s a is ical
assump ions and echniques o ensu e ha ou es ima es a e accu a e and obus .
4.2 Hypo he ical models
Haa elmo (1943) in oduced he concep o causal manipula ion using wha is now known as
he ix ope a o (Heckman and Pin o, 2015). The coun e ac ual ou come Yð Þcan be ob ained
by ixing he inpu Tin he ou come equa ion o a speci ic alue T¼ wi hin he suppo o T
(deno ed as ∈suppðTÞ). In adi ional s a is ics, when condi ioning on a a iable, he
dis ibu ion o all ela ed a iables can change, which di e s om he e ec o he ix ope a o
(o causal ope a o ). The ix ope a o speci ically impac s only he dis ibu ion o he
descendan s o he a iable being ixed, lea ing o he aspec s o he model unchanged. Since
s a is ical app oaches do no inhe en ly include manipula ion using he ix ope a o , Heckman
and Pin o (2015) de eloped he econome ic causal amewo k, which links he ix ope a o o
s a is ical condi ioning (o condi ional dis ibu ions). This in eg a ion allows o a mo e
p ecise and con olled analysis o causal e ec s in empi ical esea ch.
Heckman and Pin o (2015) p oposed he Hypo he ical Model, dis inguishing i om he
empi ical model ha gene a es obse able da a. The hypo he ical model is used o o mula e
hough expe imen s in ol ing he manipula ion o inpu s o de e mine causali y. In his
amewo k, a hypo he ical a iable e
Tis in oduced as an exogenous a iable ha causes he
ou come Y. When e
Tis manipula ed, i a ec s Ywi hou al e ing o he a iables in he s uc u al
model. The hypo he ical model, he e o e, ansla es he causal ope a ion o ixing X in o he
s a is ical ope a ion o condi ioning on e
T.
AJEB
9,1
8
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
Le Me ep esen he empi ical causal model used o es ima e causal pa ame e s and le Mh
ep esen he causal model o mula ed using he hypo he ical a iable e
T.Heckman and Pin o
(2015) in oduced c i e ia o sys ema ically connec he coun e ac ual and empi ical
dis ibu ions. Speci ically, o any disjoin se o a iables Y;Win he s uc u al causal
model de ined by e
T, and o any alues ; 0∈suppðTÞ, he hypo he ical model sa is ies:
Y⊥e
T���ðT;WÞ0Ph�Yje
T¼ ;T¼ 0;W�¼PhðYjT¼ 0;WÞ ¼ PeðYjT¼ 0;WÞ(Eq.4)
Y⊥T����e
T;W�0Ph�Yje
T¼ ;T¼ 0;W�¼Ph�Y���e
T¼ ;W�¼Pe�Y���T¼ ;W�(Eq.5)
Figu e 2 illus a es he hypo he ical model co esponding o he causal model depic ed in
Figu e 1. In his hypo he ical model, he Edu -inpu in he ou come equa ion is eplaced by he
ex e nal a iable g
Edu. As a esul , in he hypo he ical model, he e a e no a ows poin ing om
he ea men a iable Edu o he ou come. This eplacemen e ec i ely emo es he di ec
in luence o he ea men on he ou come, o i ixes he ea men (as ix ope a o ).
No ice ha he diag am in Figu e 2 speci ically illus a es a hypo he ical a iable se up
aimed a es ima ing he di ec e ec o educa ion (Edu) on wages (Wage). The o al and indi ec
e ec s diag ams a e no shown, as hey all ou side he scope o his s udy. Fo u he de ails
on modi ying he diag am o iden i y o al, di ec , and indi ec e ec s, eade s a e e e ed o
Heckman and Pin o (2015).
F om he LMC o he ou come a iable, Wage (Wage ⊥Edujðg
Edu;Exp;Occ;AbilÞ, we
ha e ha PhðWage���g
Edu ¼ ;Exp;Occ;AbilÞ ¼ PeðWage���Edu ¼ ;Exp;Occ;AbilÞ. Howe e ,
he a iable Abil is unobse able, which means ha PeðWagejEdu ¼ ;Exp;Occ;AbilÞcanno
be di ec ly es ima ed om he da a. The e o e, addi ional assump ions a e necessa y o enable
he es ima ion o causal pa ame e s om he obse ed da a. These assump ions ypically
in ol e ei he app oxima ing he e ec s o unobse ed a iable using obse able p oxies o
Figu e 2. DAG ep esen s hypo he ical model o he e u ns o educa ion wi h endogenous decision on ield
o s udy
Asian Jou nal o
Economics and
Banking
9
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
6.1.2 Re u ns o highe educa ion by ield o s udy in 2021: a compa ison. The compa ison o
he e u ns o highe educa ion be ween 2019 and 2021 is shown in Table 4. In 2021, he e u n
o educa ion in he heal hca e ield inc eased by 40.1%, becoming he highes among all ields
wi h an MTE es ima e o 1.943. This subs an ial inc ease likely e lec s he heigh ened demand
o heal hca e p o essionals du ing he COVID-19 pandemic. The eaching ield main ained a
high e u n, wi h an MTE es ima e o 1.571, a sligh and s a is ically insigni ican decline om
1.664 in 2019. O he ields exhibi ed some changes, bu he changes we e no s a is ically
signi ican . This lack o signi icance may indica e ei he no sys ema ic change o
he e ogeneous e ec s ac oss subg oups. O e all, e u ns emained ela i ely s able ac oss
mos ields, excep o he ma ked inc ease obse ed in heal hca e.
Table 3. Es ima ion o e u ns o highe educa ion by ield o s udy in 2019 and 2021 using Heckman sample-
selec ion model (HM) and he mul inomial ea men e ec s model wi h sample selec ion co ec ion (MTE)
(HM2019) (HM2021) (MTE2019) (MTE2021)
ln(income) ln(income) ln(income) ln(income)
Field o s udy
1. Teaching 1.321*** 1.335*** 1.664*** 1.571***
(0.045) (0.053) (0.205) (0.138)
2. Social science, law,
a , and humani ies
1.074*** 1.163*** 1.227*** 1.291***
(0.048) (0.049) (0.143) (0.137)
3. Business 1.019*** 1.136*** 1.127*** 1.432***
(0.035) (0.039) (0.328) (0.105)
4. Compu e ela ed 0.808*** 0.904*** 1.178*** 1.091***
(0.089) (0.051) (0.198) (0.101)
5. Science, ag icul u e, enginee , and a chi ec 0.932*** 1.016*** 0.066 0.112*
(0.044) (0.042) (0.088) (0.061)
6. Heal hca e 1.159*** 1.471*** 1.387*** 1.943***
(0.075) (0.047) (0.174) (0.089)
7. O he ields 0.963*** 1.005*** 1.304*** 1.158***
(0.074) (0.108) (0.093) (0.163)
Female 0.178*** 0.182*** 0.039 0.005
(0.033) (0.033) (0.121) (0.113)
Age 0.066*** 0.068*** 0.118*** 0.140***
(0.009) (0.010) (0.014) (0.014)
Age
2
�0.001*** �0.001*** �0.002*** �0.002***
(0.000) (0.000) (0.000) (0.000)
U ban 0.995*** 0.927*** 1.067*** 0.979***
(0.032) (0.031) (0.047) (0.047)
a h ho �2.534*** �2.562***
(0.029) (0.039)
lnsigma 1.151*** 1.146***
(0.012) (0.012)
IMR �2.839*** �2.792***
(0.475) (0.491)
Cons an 7.396*** 7.381*** 6.655*** 6.247***
(0.196) (0.196) (0.288) (0.279)
Obse a ions 64,207 65,885 53,337 54,828
No e(s): (1) Boo s apped s anda d e o s in pa en heses (***p< 0.01, **p< 0.05, *p< 0.1)
(2) Explana o y a iables used o he selec ion (in o employmen ) equa ion in he HM models and he
cons uc ion o he In e se Mill’s Ra ion (IMR) o he MTE models include age, and dummy a iables o
ma ied, u ban, college, and oca ional deg ee
(3) The MTE models we e es ima ed using he m ea eg command in S a a, de eloped by Deb (2009). Full
es ima ion esul s a e p o ided in Appendix A
Sou ce(s): Au ho s’ es ima ions using he SES 2019 and SES 2021 da ase s o he sample o indi iduals aged
25–60 yea s whose p ima y ac i i y is no a ending school (wi h popula ion weigh ing)
AJEB
9,1
16
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
In addi ion o ield o s udy, gende , age, and u ban esidence also in luence income le els.
Fo he emale a iable, he coe icien s in bo h MTE2019 and MTE2021 a e no s a is ically
signi ican , sugges ing no subs an ial gende wage gap a e accoun ing o selec ion e ec s.
The coe icien s o age and age squa ed indica e a nonlinea ela ionship wi h income. The
esul s sugges ha income ini ially inc eases wi h age, e lec ing e u ns o expe ience, bu
his g ow h diminishes and e en ually e e ses a la e s ages o he li e cycle. The u ning
poin occu s ea lie in MTE2019 compa ed o MTE2021, sugges ing ha income g ow h wi h
age was sligh ly mo e sus ained in 2021. U ban esidence consis en ly shows a subs an ial and
s a is ically signi ican wage p emium, wi h coe icien s o 1.067 in MTE2019 and 0.979 in
MTE2021, hough he sligh decline sugges s a ma ginal educ ion in u ban ad an ages.
6.2 Occupa ional s a us as a media o o he e ec s o educa ion o income
The ne e u ns o educa ion o each ield o s udy a e in luenced by di e ences in
occupa ional s a uses. This sec ion es ima es e u ns while con olling o occupa ion,
including (1) employe (business owne s wi h a leas one employee), (2) own accoun wo ke s
(ag icul u e), (3) own accoun wo ke s (non-ag icul u e), wi h (4) public o p i a e employee
as he baseline. The compa ison be ween models wi h and wi hou occupa ional s a us con ol
highligh s he indi ec e ec o ield o s udy on wage ou comes h ough occupa ional s a us
so ing.
The esul s om bo h 2019 and 2021 show a consis en pa e n o income ac oss di e en
occupa ional s a uses. Employe s and employees, who se e as he base g oup in he
eg essions, had he highes a e age mon hly income. In con as , own-accoun wo ke s—
business owne s wi hou employees o eelance s—ea ned lowe incomes, especially in he
ag icul u al sec o . Rega ding he e ec o COVID-19 in 2021, he income o employe s and
own-accoun wo ke s ou side he ag icul u al sec o d opped mo e sha ply compa ed o he
employee g oup.
When examining he e u ns o highe educa ion by ield o s udy while con olling o
occupa ion (Model MTE2019oc in Table 5), he 2019 es ima es we e app oxima ely 50%
lowe ac oss mos ields compa ed o he es ima es wi hou con olling o occupa ional s a us
(Model MTE2019 in Table 3). This subs an ial educ ion sugges s ha a signi ican po ion o
he obse ed e u ns o educa ion is media ed h ough occupa ional so ing, pa icula ly in o
oles such as employe and sala ied employee posi ions, which a e associa ed wi h highe
ea nings. In all ields, he e u ns declined when con olling o occupa ion, bu his e ec was
mos p onounced in he case o he science, ag icul u e, enginee ing, and a chi ec u e ield. Fo
his g oup, he es ima ed e u n changed om a sligh posi i e (0.066) o a nega i e alue
(�0.294), ep esen ing a mo e ex eme educ ion compa ed o o he ields. This sugges s ha
Table 4. Re u ns o highe educa ion in 2019 and 2021 by ield o s udy
Fields o s udy
Re u ns o
educa ion 2019
Re u ns o
educa ion 2021 Di e ence SE -s a p- alue
1. Teaching 1.664 1.571 �0.093 0.246 �0.378 0.706
2. Social science, law, a ,
and humani ies
1.227 1.291 0.064 0.199 0.322 0.747
3. Business 1.127 1.432 0.306 0.345 0.887 0.375
4. Compu e ela ed 1.178 1.091 �0.088 0.223 �0.395 0.693
5. Science, ag icul u e,
enginee , and a chi ec
0.066 0.112 0.045 0.106 0.424 0.672
6. Heal hca e 1.387 1.943 0.556 0.196 2.841 0.004
No e(s): Compa ison o MTE2019 and MTE2021 es ima es
Sou ce(s): Au ho s’ calcula ion using he SES 2019 and SES 2021 da ase s
Asian Jou nal o
Economics and
Banking
17
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
he highe ea nings seen among g adua es in hese ields a e la gely d i en by hei en y in o
highe -paying occupa ions a he han he in insic alue o he deg ee.
Con olling o occupa ion also al e s he es ima ed coe icien s o gende , age, and u ban
a iables, highligh ing he ole o occupa ional so ing in wage di e ences. Fo he emale
a iable, he coe icien inc eased om 0.039 o 0.274 when accoun ing o occupa ion,
indica ing ha he ini ially lowe ea nings o women a e la gely due o di e ences in
occupa ional oles. Women a e less likely o be employe s (1.50% o women s 3.35% o
Table 5. Es ima ion o e u ns o highe educa ion by ield o s udy con olled o occupa ional s a us in 2019
and 2021
(MTE2019oc) (MTE2021oc)
ln(income) ln(income)
Field o s udy
1. Teaching 0.826*** 0.729***
(0.314) (0.131)
2. Social science, law, a , and humani ies 0.615*** 0.717***
(0.232) (0.129)
3. Business 0.590 0.815***
(0.363) (0.119)
4. Compu e ela ed 0.660** 0.599***
(0.260) (0.107)
5. Science, ag icul u e, enginee , and a chi ec �0.294*** �0.196***
(0.073) (0.059)
6. Heal hca e 0.705*** 1.186***
(0.174) (0.081)
7. O he ields 0.736*** 0.636***
(0.089) (0.149)
Occupa ion
1. Employe 0.001 �0.016
(0.117) (0.091)
2. Own accoun wo ke s �2.722*** �2.735***
(Ag i) (0.048) (0.048)
3. Own accoun wo ke s �0.187*** �0.306***
(Non-ag i) (0.033) (0.032)
Female 0.274** 0.277***
(0.111) (0.102)
Age 0.103*** 0.112***
(0.013) (0.013)
Age
2
�0.001*** �0.001***
(0.000) (0.000)
U ban 0.480*** 0.468***
(0.041) (0.040)
IMR �3.355*** �3.440***
(0.429) (0.445)
Cons an 7.562*** 7.439***
(0.268) (0.253)
Obse a ions 53,337 54,828
No e(s): (1) Boo s apped s anda d e o s in pa en heses (***p< 0.01, **p< 0.05, *p< 0.1)
(2) Explana o y a iables used o he selec ion (in o employmen ) equa ion in he HM models and he
cons uc ion o he In e se Mill’s Ra ion (IMR) o he MTE models include age, and dummy a iables o
ma ied, u ban, college, and oca ional deg ee
(3) The MTE models we e es ima ed using he m ea eg command in S a a, de eloped by Deb (2009). Full
es ima ion esul s a e p o ided in Appendix A
Sou ce(s): Au ho s’ es ima ions using he SES 2019 and SES 2021 da ase s o a sample o indi iduals aged
25–60 yea s whose p ima y ac i i y is no a ending school (wi h popula ion weigh ing)
AJEB
9,1
18
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
men) and mo e likely o wo k as own-accoun wo ke s ou side ag icul u e (27.99 s 25.58%
o men), who ea n less income on he a e age [2]. Once adjus ed o occupa ional so ing, he
highe ea nings po en ial o women wi hin simila oles becomes e iden . The coe icien s o
age and age squa ed demons a e he nonlinea impac o expe ience on ea nings o bo h
models. Con olling o occupa ion educes he age coe icien ( om 0.118 o 0.103) and
sligh ly diminishes he absolu e alue o he age squa ed coe icien ( om �0.002 o �0.001),
sugges ing ha pa o he age- ela ed ea nings ad an age is linked o occupa ional so ing.
The ma ginal e ec o age emains posi i e un il la e s ages o he li e cycle, wi h he u ning
poin shi ing om age 29.5 in MTE2019 o age 51.5 when occupa ions a e con olled. This
indica es ha occupa ional so ing ampli ies e u ns o younge wo ke s while mi iga ing
declines o olde wo ke s [3]. Fo u ban esidence, he coe icien dec eases signi ican ly
( om 1.067 o 0.480) a e adjus ing o occupa ion, indica ing ha much o he u ban wage
p emium a ises om so ing in o highe -paying occupa ions. Rega dless, he emaining
posi i e e ec highligh s ad an ages ela ed o loca ion, such as be e access o high-paying
occupa ions, independen o occupa ional ac o s.
The esul s o 2021 show simila pa e ns o hose obse ed in 2019 when compa ing
es ima es wi h and wi hou con olling o occupa ion. Ac oss bo h yea s, con olling o
occupa ion consis en ly educes he es ima ed coe icien s o all ields o s udy, indica ing he
signi ican ole o occupa ional so ing in wage di e ences.
7. Conclusion
This s udy applies he econome ic causal amewo k de eloped by Heckman and Pin o (2022)
o es ima e he e u ns o highe educa ion by ield o s udy in Thailand, wi h a ocus on
compa ing he p e- and pos -COVID-19 pe iods. Using da a om Thailand’s socioeconomic
su eys (SES) collec ed in 2019 and 2021, we analyzed a sample o indi iduals aged 25–60
who a e no engaged in o mal educa ion, examining how he pandemic has in luenced
educa ional e u ns ac oss a ious ields.
Cen al o ou app oach is he applica ion o he economic causal amewo k, linking he
e u ns o educa ion wi h an endogenous ield o s udy model. This amewo k accoun s o
bo h obse ed and unobse ed abili ies a ec ing indi iduals’ choices o ield o s udy and
subsequen labo ma ke ou comes, as illus a ed in Figu es 1–3. We employ he con ol
unc ion app oach o adjus o unobse ed ac o s, using he mul inomial ea men e ec s
(MTE) model by Deb (2009) wi h sample selec ion co ec ion as he main empi ical ool. This
me hod enables accu a e es ima ion o he ea men e ec s o di e en ields o s udy on
income while add essing biases om abili y so ing and sample selec ion.
The esul s e eal signi ican a ia ion in he e u ns o highe educa ion by ield o s udy. In
2019, eaching and heal hca e had he highes e u ns (166.4 and 138.7% highe income,
espec i ely), while e u ns o science, ag icul u e, enginee ing, and a chi ec u e we e
minimal o no signi ican . By 2021, heal hca e exhibi ed he highes e u n (194.3%), likely
d i en by inc eased demand du ing he COVID-19 pandemic, while eaching main ained
s ong bu sligh ly lowe e u ns (157.1%). Ac oss bo h yea s, con olling o occupa ion
educed he es ima ed e u ns by abou 50%, indica ing ha a subs an ial pa o he obse ed
e u ns is due o occupa ional so ing in o highe -paying oles. This e ec was mos
p onounced o science, ag icul u e, enginee ing, and a chi ec u e, whe e posi i e e u ns
u ned signi ican ly nega i e, sugges ing ha he ea nings ad an age is p ima ily om access
o well-paid jobs a he han ield-speci ic skills. The analysis o demog aphic ac o s u he
highligh ed he ole o occupa ional so ing, pa icula ly o gende , age, and u ban esidence,
unde sco ing he con inued impo ance o expe ience and access o high-paying indus ies in
explaining wage di e ences.
The a ia ion in e u ns ac oss ields o s udy sugges s a need o policies ha be e align
educa ional o e ings wi h labo ma ke demand. High e u ns in ields like eaching and
heal hca e indica e s ong labo ma ke alue. The low o nega i e e u ns in science,
Asian Jou nal o
Economics and
Banking
19
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
ag icul u e, enginee ing, and a chi ec u e sugges po en ial issue and equi e u he s udy o
explo e speci ic causes and de elop app op ia e s a egies. Addi ionally, e o s should be
made o acili a e ansi ions in o highe -paying occupa ions, ensu ing ha he bene i s o
highe educa ion a e equi ably dis ibu ed. This can begin by sys ema ically es ima ing and
epo ing e u ns o a ious ypes o educa ion, enabling indi iduals o make in o med
decisions.
Howe e , he s udy has limi a ions. The eliance on c oss-sec ional da a and he lack o
household panel da a in Thailand wi h de ailed ield o s udy in o ma ion a e signi ican
cons ain s. Small sample sizes in ce ain ields may also limi he p ecision o he es ima es.
None heless, he da ase used emains he bes a ailable o his ype o analysis in Thailand.
Fu u e esea ch could build on hese indings by conside ing he ma ching be ween
educa ional ields and occupa ional ou comes, as highligh ed by Lemieux (2014), and by
inco po a ing ac o s like unemploymen isk and income a iabili y o be e unde s and wage
di e en ials ac oss ields.
No es
1. Acco ding o Thailand’s NSO, he wo king aged popula ion is 15 yea s old and abo e. Howe e , as
his s udy aims o examine he e u n o educa ion o hose who al eady inished hei o mal
educa ion, he age ange is selec ed o be 25–60 yea s old.
2. The dis ibu ion o male and emale wo ke s ac oss occupa ion was abula ed using he SES
2019 da a.
3. The ma ginal e ec o age is calcula ed as: dðln ðincomeÞ
dðAgeÞ¼βAge þ2$βAge2$Age. Unlike o he a iables
wi h cons an e ec s, he ma ginal e ec o age a ies ac oss he li e cycle due o he quad a ic
speci ica ion. In MTE2019 (wi hou occupa ion con ols), he ma ginal e ec u ns nega i e a age
29.5 yea s old. In he model wi h occupa ion con ols, he u ning poin shi s 51.5 yea s old.
Re e ences
Al onji, J.G., A cidiacono, P. and Mau el, A. (2016), “The analysis o ield choice in college and
g adua e school: De e minan s and wage e ec s”, in Hanushek, E., Machin, S. and Woessmann,
L. (Eds), Handbook o he Economics o Educa ion, Else ie , Vol. 5, pp. 305-396, doi: 10.1016/
b978-0-444-63459-7.00007-5.
A cidiacono, P. (2004), “Abili y so ing and he e u ns o college majo ”, Jou nal o Econome ics,
Vol. 121 Nos. 1-2, pp. 343-375, doi: 10.1016/j.jeconom.2003.10.010.
Asian De elopmen Bank (ADB) (2021), COVID-19 and Labo Ma ke s in Sou heas Asia: Impac s on
Indonesia, Malaysia, he Philippines, Thailand, and Vie Nam, Asian De elopmen Bank,
Manilla, a ailable a : h ps://www.adb.o g/publica ions/co id-19-labo -ma ke s-sou heas -asia
Bje khol , O. and Qin, D. (Eds). (2011), A Dynamic App oach o Economic Theo y: Lec u es by
Ragna F isch a Yale Uni e si y, Rou ledge.
Bol, T. and Heisig, J.P. (2021), “Explaining wage di e en ials by ield o s udy among highe educa ion
g adua es: e idence om a la ge-scale su ey o adul skills”, Social Science Resea ch, Vol. 99,
102594, doi: 10.1016/j.ss esea ch.2021.102594.
Canaan, S. and Mouganie, P. (2018), “Re u ns o educa ion quali y o low-skilled s uden s: e idence
om a discon inui y”, Jou nal o Labo Economics, Vol. 36 No. 2, pp. 395-436.
Deb, P. (2009), MTREATREG: S a a Module o Fi s Models wi h Mul inomial T ea men s and
Con inuous, Coun and Bina y Ou comes Using Maximum Simula ed Likelihood, S a is ical
So wa e Componen s, a ailable a : h ps://ideas. epec.o g/c/boc/bocode/s457064.h ml
Deb, P. and T i edi, P.K. (2006), “Maximum simula ed likelihood es ima ion o a nega i e binomial
eg ession model wi h mul inomial endogenous ea men ”, STATA Jou nal, Vol. 6 No. 2,
pp. 246-255, doi: 10.1177/1536867x0600600206.
AJEB
9,1
20
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
Ebe ha d, B., Podio, M., P�
e ez Alonso, A., Rado ica, E., A o ina, L., Peiseniece, L., Caama~
no Sendon,
M., Gonzales Lozano, A. and Sol�
e-Pla, J. (2017), “Sma wo k: he ans o ma ion o he labou
ma ke due o he ou h indus ial e olu ion”, In e na ional Jou nal o Business and Economic
Sciences Applied Resea ch, Vol. 10 No. 3, pp. 47-66.
Goldin, C. and Ka z, L.F. (2009), The Race be ween Educa ion and Technology, Ha a d
Uni e si y P ess.
F isch, R. (1938), “Au onomy o economic ela ions: S a is ical e sus heo e ical ela ions in
economic mac odynamics”, Pape gi en a League o Na ions Rep in ed in D.F. Hend y and
M.S. Mo gan (1995) The Founda ions o Econome ic Analysis, Camb idge Uni e si y P ess.
Haa elmo, T. (1943), “The s a is ical implica ions o a sys em o simul aneous equa ions”,
Econome ica, Vol. 11 No. 1, pp. 1-12, doi: 10.2307/1905714.
Heckman, J.J. (1979), “Sample selec ion bias as a speci ica ion e o ”, Econome ica, Vol. 47 No. 1,
pp. 153-161, doi: 10.2307/1912352.
Heckman, J.J. and Pin o, R. (2015), “Causal analysis a e Haa elmo”, Econome ic Theo y, Vol. 31
No. 1, pp. 115-151, doi: 10.1017/s026646661400022x.
Heckman, J.J. and Pin o, R. (2022), “The econome ic model o causal policy analysis”, Annual e iew
o economics, Vol. 14 No. 1, pp. 893-923, doi: 10.1146/annu e -economics-051520-015456.
Heckman, J.J. and Pin o, R. (2024), “Econome ic causali y: he cen al ole o hough expe imen s”,
Jou nal o Econome ics, Vol. 243 Nos. 1-2, 105719, doi: 10.1016/j.jeconom.2024.105719.
Heckman, J.J. and Robb, R. J (1985), “Al e na i e me hods o e alua ing he impac o in e en ions:
an o e iew”, Jou nal o Econome ics, Vol. 30 Nos. 1-2, pp. 239-267, doi: 10.1016/0304-4076
(85)90139-3.
Ki keboen, L.J., Leu en, E. and Mogs ad, M. (2016), “Field o s udy, ea nings, and sel -selec ion”,
Qua e ly Jou nal o Economics, Vol. 131 No. 3, pp. 1057-1111, doi: 10.1093/qje/qjw019.
Lemieux, T. (2014), “Occupa ions, ields o s udy and e u ns o educa ion”, Canadian Jou nal o
Economics/Re ue canadienne d’�
economique, Vol. 47 No. 4, pp. 1047-1077, doi: 10.1111/
caje.12116.
Na ional S a is ical O ice (NSO) (2022), S a is ical Yea book Thailand 2022, Na ional S a is ical
O ice, Bangkok.
Psacha opoulos, G. and Pa inos, H.A. (2018), “Re u ns o in es men in educa ion: a decennial e iew
o he global li e a u e”, Educa ion Economics, Vol. 26 No. 5, pp. 445-458, doi: 10.1080/
09645292.2018.1484426.
Roy, A.D. (1951), “Some hough s on he dis ibu ion o ea nings”, Ox o d Economic Pape s, Vol. 3
No. 2, pp. 135-146, doi: 10.1093/ox o djou nals.oep.a041827.
Si is isakulchai, J. and Leu cha usmee, S. (2023), “Re isi ing e u ns o educa ion in Thailand:
s uc u al causal model amewo k”, In e na ional Jou nal o Unce ain y, Fuzziness and
Knowledge-Based Sys ems, Vol. 31 No. 2, pp. 223-241, doi: 10.1142/s0218488523400135.
Tinbe gen, J. (1975), Income Dis ibu ion: Analysis and Policies, No h- Holland.
UNESCO (2022), “Highe educa ion global da a epo ”, The 3 d UNESCO Wo ld Highe Educa ion
Con e ence, a ailable a : h ps://www. igh - o-educa ion.o g/ esou ce/highe -educa ion-global-
da a- epo -con ibu ion-wo ld-highe -educa ion-con e ence
Wo ld Economic Fo um (WEF) (2020), The Fu u e o Jobs Repo 2020, Wo ld Economic Fo um,
Gene a, a ailable a : h ps://www.we o um.o g/publica ions/ he- u u e-o -jobs- epo -2020/
Supplemen a y ma e ial
The supplemen a y ma e ial o his a icle can be ound online.
Co esponding au ho
Supanika Leu cha usmee can be con ac ed a : [email p o ec ed]
Fo ins uc ions on how o o de ep in s o his a icle, please isi ou websi e:
www.eme aldg ouppublishing.com/licensing/ ep in s.h m
O con ac us o u he de ails: [email p o ec ed]
Asian Jou nal o
Economics and
Banking
21
Downloaded om h p://www.eme ald.com/ajeb/a icle-pd /9/1/2/9700190/ajeb-09-2024-0106.pd by ZBW Ge man Na ional Lib a y o Economics use on 16 Decembe 2025
