Mülle , Ge i
Wo king Pape
Panel e idence on wi hin-occupa ion change in job asks
and indi idual wages
IAB-Discussion Pape , No. 02/2024
P o ided in Coope a ion wi h:
Ins i u e o Employmen Resea ch (IAB)
Sugges ed Ci a ion: Mülle , Ge i (2024) : Panel e idence on wi hin-occupa ion change in job
asks and indi idual wages, IAB-Discussion Pape , No. 02/2024, Ins i u ü A bei sma k - und
Be u s o schung (IAB), Nü nbe g,
h ps://doi.o g/10.48720/IAB.DP.2402
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/283020
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-sa/4.0/deed.de
IAB-DISCUSSION PAPER
A icles on labou ma ke issues
2|2024 Panel E idence on Wi hin-Occupa ion Change in
Job Tasks and Indi idual Wages
Ge i Mülle
ISSN 2195-2663
Panel E idence on Wi hin-Occupa ion
Change in Job Tasks and Indi idual Wages
Ge i Mülle (IAB)
Mi de Reihe „IAB-Discussion Pape “ will das Fo schungsins i u de Bundesagen u ü A bei
den Dialog mi de ex e nen Wissenscha in ensi ie en. Du ch die asche Ve b ei ung on
Fo schungse gebnissen übe das In e ne soll noch o D ucklegung K i ik ange eg und Quali ä
gesiche we den.
The “IAB Discussion Pape ” is published by he esea ch ins i u e o he Ge man Fede al
Employmen Agency in o de o in ensi y he dialogue wi h he scien i ic communi y. The p omp
publica ion o he la es esea ch esul s ia he in e ne in ends o s imula e c i icism and o
ensu e esea ch quali y a an ea ly s age be o e p in ing.
Con en s
Con en s ................................................................................................................................... 3
Abs ac .................................................................................................................................... 4
Zusammen assung ................................................................................................................... 4
JEL classi ica ion ...................................................................................................................... 4
Keywo ds ................................................................................................................................. 4
Acknowledgemen s .................................................................................................................. 4
1 In oduc ion ....................................................................................................................... 5
2 Empi ical Implemen a ion .................................................................................................. 9
3 Da a .................................................................................................................................. 11
3.1 Measu emen o Job Tasks ................................................................................................ 11
3.2 Classi ica ion o Occupa ions ............................................................................................ 12
3.3 Wages, Human Capi al and Demog aphic Con ols ......................................................... 13
3.4 Sample Selec ion ............................................................................................................... 14
4 Desc ip i e S a is ics ........................................................................................................ 15
5 Job Tasks and Wages ........................................................................................................ 17
6 Wage E ec s o Rou ine In ensi y by a Job’s O he Task Requi emen s ............................ 20
7 Robus ness Checks ........................................................................................................... 22
8 Concluding Rema ks ......................................................................................................... 24
Re e ences .............................................................................................................................. 26
Appendix ................................................................................................................................ 29
IAB-Discussion Pape 2|2024
4
Abs ac
D awing on newly a ailable panel da a, his pape p esen s an empi ical analysis o he wage
e ec s o changing job asks, assessed o indi iduals a hei wo kplace. I am he e o e able o
exploi wi hin-occupa ion wi hin-indi idual a ia ion, o e ime, o s udy wage e u ns o
cogni i e, in e pe sonal, physical and ou ine ask in ensi y. The indings o Au o and Handel’s
(2013) pionee ing wo k on he signi icance o such wi hin-occupa ion (“in ensi e ma gin”) ask
a ia ion a e eassessed. Unobse ed wo ke a ibu es and ongoing sel -selec ion in o
occupa ions can be accoun ed o in a much mo e comp ehensi e way han possible wi h pu ely
c oss-sec ional da a, as in he o iginal wo k.
Zusammen assung
Au G undlage neu e ügba e Panelda en p äsen ie dieses Papie eine empi ische Analyse de
Lohne ek e sich e ände nde Tä igkei en am indi iduellen A bei spla z. Un e Ausnu zung de
Da en a ia ion inne halb on Be u en und Indi iduen, übe die Zei , we den Lohn endi en ü die
Ausübung on kogni i en, in e pe sonalen, physischen und Rou ine-Tä igkei en geschä z . Die
E kenn nisse de ein luss eichen S udie on Au o und Handel (2013) übe die Bedeu ung des
Tä igkei swandels inne halb on Be u en („in ensi e Ma gin“) we den e neu un e such .
Insbesonde e kann indi iduelle unbeobach e e He e ogeni ä und Selbs selek ion in Be u e
besse Rechnung ge agen we den, als dies in de u sp ünglichen Que schni ss udie möglich
wa .
JEL classi ica ion
J24, J31
Keywo ds
Job asks, occupa ions, ou ine-biased echnical change, wages, wo kplaces
Acknowledgemen s
I am g a e ul o B i a Ma hes and Johannes Luds eck o help ul discussions. I would also like o
hank pa icipan s a he TASKS VI In e na ional BIBB/IAB/ZEW Con e ence 2022 o hei
commen s and sugges ions. All opinions, omissions, and e o s emain my own.
IAB-Discussion Pape 2|2024
5
1 In oduc ion
This pape complemen s he exis ing empi ical li e a u e on occupa ional asks and wo ke s’
wages wi h an explici ocus on ha pa o ask change happening a he job-le el, ha is, wi hin
and independen o occupa ion. While nume ous in luen ial s udies, s a ing wi h he wo k by
Au o /Le y/Mu nane (2003), ha e conside ed occupa ional ask equi emen s, employmen and
wage ou comes, e idence on he wi hin-occupa ion dimension (“in ensi e ma gin”) is sca ce.
The ew pionee ing s udies ha do exis on his opic, he mos p ominen one being by
Au o /Handel (2013), a e all c oss-sec ional in na u e, which poses clea limi a ions on he
analyses. This poin s ou a gap in he li e a u e ha he p esen wo k seeks o add ess o he
i s ime.
Since he pape by Au o /Le y/Mu nane (2003, hence o h ALM), he ask-based app oach o he
labo ma ke has spa ked o a la ge body o empi ical li e a u e conce ned wi h he s udy o
echnology-induced change in ask equi emen s and associa ed shi s in employmen and
wages. Much o his wo k ollows he ou iniza ion hypo hesis (o ou ine-biased echnical
change, RBTC) acco ding o which p og essing compu e iza ion and au oma ion subs i u e o
ou ine asks he e o o e done by humans (e.g. Le y/Mu nane 1996; Le y 2010). A he cu en
s a e o echnology, “ ou ine” comp ises any ask ha is ule-based and codi iable in o m o
p og am ou ines o be ollowed and execu ed by a machine o an algo i hm. Wi h p ices o
compu ing powe and au oma ion e e alling, i ms shi hei demand o asks ca ied ou by
human labo away om ou ine, owa ds non ou ine; which may induce wo ke s o adap and
eloca e (e.g. ac oss occupa ions).1
A he hea o he ask-based app oach lies a concep ual dis inc ion be ween human p oduc i e
a ibu es (inna e abili y, acqui ed skills) and he ba ches o main wo k ac i i y ( asks) o which
hese a ibu es a e applied, alongside conjec u es on he shi ing bounda y be ween asks done
by humans and hose done by machines. In his ega d, he o iginal ounda ion o he ALM
app oach is a a he ine-g ained, quali a i e, conside a ion o wha i is ha indi iduals do a
wo k, and wha he eo machines a e inc easingly able o ake o e (ALM: 1282−86, and he wo ks
ci ed he ein). This ini ial mic o-pe spec i e on wo ke s’ job con en and ask shi s expe ienced
a hei wo kplaces has somewha been los , howe e . One cen al eason o his s a e o a ai s
is measu emen , as e.g. discussed in Au o /Handel (2013: S61−62), o he li e a u e o e iew on
he ask app oach by Au o (2013: 189−192). Mos la ge-scale labo ma ke su eys commonly
used in he ield simply lack de ailed in o ma ion on indi iduals’ main wo k ac i i ies. Much o
he ea ly pola iza ion li e a u e he e o e s udies (change in) ask con en in e ms o
occupa ional a ilia ion o wo ke s, and conside s occupa ion-le el employmen and wage
ou comes; mos o en d awing on agg ega e da a. E en when wo ke samples a e a ailable and
indi idual-le el ou comes can be conside ed (e.g. in Fi po/Fo in/Lemieux 2011; Co es 2016;
Ca aglia/E he idge 2020; Böhm 2020), he key explana o y a iables cap u ing wo ke s’ job ask
con en emain occupa ion-le el.
1 This line o easoning o e s one explana ion o “pola iza ion” phenomena in occupa ional employmen and wages obse ed
bo h in he Uni ed S a es and Eu ope (e.g. Au o /Ka z/Kea ney 2006; Goos/Manning 2007; Au o /Do n 2009;
Goos/Manning/Salomons 2009, 2014).
IAB-Discussion Pape 2|2024
6
The auxilia y assump ion ha each occupa ion equa es o a cha ac e is ic bundle o main wo k
ac i i ies deli e s he missing link, and has become s anda d p ac ice in empi ical
implemen a ions o a ask-based app oach. Codes om s a is ical classi ica ion sys ems o
occupa ions (SOC, ISCO, e c.) a e eadily a ailable in mos da a sou ces on he labo ma ke and
he e o e used as p oxies. These codes a e commonly agg ega ed up in o (o en only less han a
hand ul o ) b oade ca ego ies assumed o ep esen simila con en s, and in ensi ies, in e ms o
he asks demanded om wo ke s (e.g. ou ine–non ou ine). This g ouping o occupa ions ei he
ollows he hie a chical logic o he unde lying classi ica ion sys em, o happens manually,
judgemen -based. Some imes, expe a ings o he ypical (a e age) ask con en o jobs wi hin
a gi en occupa ion a e me ged om da abases such as he U.S. Dic iona y o Occupa ional Ti les
(DOT), o he Occupa ional In o ma ion Ne wo k (O*NET), o ei he quali a i ely in o m a
pa icula g ouping decision (as in e.g. Au o /Do n 2013), o o cons uc quan i a i e (me ic)
scales o occupa ion-le el ask con en (as in e.g. Ross 2017).
Wha is he eby necessa ily dis ega ded is ha pa o a ia ion in asks a ec ing wo ke s wi hin,
and independen o , hei occupa ion. I is easily concei able ha di e en wo ke s holding he
same occupa ional i le (o assigned by esea che s o he same agg ega e s a is ical g oup) may
be acing qui e di e en ask equi emen s a hei espec i e wo kplace. E en “iden ical”
wo ke s, in e ms o hei educa ion and o he indi idual p oduc i e a ibu es, holding he same
occupa ion, a e no necessa ily engaged in he same main ac i i ies; shaped by he speci ic
goods o se ices p o ided, o ganiza ional p ocesses and wo k lows wi hin a i m, and so o h.
Likewise, he de elopmen o e ime in jobs’ ask composi ion may be as he e ogeneous. Fo
some indi iduals, he kind o ask change expe ienced a hei wo kplace could e en un coun e
o any “global” end, such as he ubiqui ous shi in equi emen s away om ou ine owa ds
non ou ine. Ye , so a , ask a ia ion a his le el o g anula i y has a ely been conside ed
empi ically.
The mic o-le el panel da a used in his pape opens up new possibili ies o analysis. A
s anda dized se o ask measu es adop ed om he U.S. STAMP su ey (Handel 2007, 2008) and
adminis e ed o pa icipan s in he Adul Coho o he Ge man Na ional Educa ional Panel S udy
(NEPS) assesses he main asks pe o med by indi iduals a hei wo kplace(s) o e ime. This
allows o a ollow-up on he pionee ing wo k by Au o /Handel (2013, hence o h AH), who we e
i s o in es iga e he alue added o pe son-le el job ask measu es in explaining wages based
on one c oss-sec ion o da a om he P ince on Da a Imp o emen Ini ia i e (PDII). Decomposing
he a iance o each one o hei co e job ask measu es (abs ac , manual, and ou ine
equi emen s) in o wi hin- and be ween-occupa ion componen s, AH ind ha he e is as much
a ia ion wi hin occupa ions as he e is be ween hem.2 While his con i ms he impo ance
a ibu ed o wo ke s’ occupa ional a ilia ion in de e mining hei asks, i also poin s owa ds a
e y sizeable “ esidual” a ia ion, ypically assumed away and disca ded. Fu he , AH p esen
e idence ha his wi hin-occupa ion (“in ensi e ma gin”) a ia ion is an impo an p edic o o
indi idual wages.
As i s ands, hei pape is s ill he e e ence wo k agains which o assess one’s own indings. O
cou se, AH a e well awa e o he limi a ions he c oss-sec ional da a impose on hei analyses.
2 Wi h some sligh a ia ion by he ask domain conside ed (see AH: S80, Table 4).
IAB-Discussion Pape 2|2024
7
They clea ly poin ou ha he wage p emia and penal ies associa ed wi h hei job ask
measu es a e o be iewed as sugges i e (o causal e ec s), and likely be con ounded by a
numbe o known selec ion issues. Fi s , hey canno exclude ha he wi hin-occupa ion
a ia ion in job asks is no me ely e lec ing unobse ed a ia ion a he indi idual le el.
Second, omi ed indi idual a ibu es may no only be co ela ed wi h job asks wi hin he
chosen occupa ion, bu also be a key d i e o he occupa ional choice in he i s place. AH’s
concep ual discussion o he canonical Roy model o occupa ional sel -selec ion (Roy 1951), in
combina ion wi h some o he empi ical es s p esen ed, may he e o e be iewed as a call o e-
app oaching he ma e in a way ha was una ailable o hem. Wha ollows om hei
discussion, in pa icula , is ha he p oduc i e alue o an indi idual’s ec o o unobse ed
abili ies (e iciency in exe cising ce ain asks) is dependen on he occupa ion, and may hence
be ewa ded di e en ially ac oss al e na i e choices. In o de o es ima e he wi hin-occupa ion
wage e ec s o job asks as unbiased as possible, one he e o e no only needs o keep cons an
he unobse ed abili ies o indi iduals, alongside wi h hei occupa ion, bu also ake in o
accoun he occupa ion-speci ici y o he ( e u ns o hese) unobse ables.
Gi en he a ailable panel da a, an empi ical s a egy ha nes s such a Roy- ype model is o
exploi a ia ion wi hin unique wo ke -occupa ion combina ions. He eby, no only unobse ed
indi idual and occupa ion cha ac e is ics a e accoun ed o sepa a ely and addi i ely (as e.g. in
Ross 2017). The condi ioning is on he in e ac ion be ween indi idual and occupa ion. Since
occupa ion-speci ic e u ns o indi idual unobse ables a e cha ac e is ics o a gi en “pai ”,
ime-demeaning wi hin each wo ke -occupa ion combina ion sweeps he unobse ed selec ion
ac o s a he desi ed le el. The same iden i ica ion s a egy has ecen ly been used in ela ed
wo k by Co es (2016) and Ca aglia/E he idge (2020). In line wi h hese au ho s, he unique
indi idual-occupa ion combina ions a e called spells and, hence o h, spell- ixed e ec s
es ima ion o he wage equa ions.
In an icipa ion o some o my esul s, le me b ie ly no e he e ha when I eplica e AH’s c oss-
sec ional analyses using my da a in a simila way (i.e. by pooling he panel wa es, and using a
simila a ie y o con ols), I ind qui e compa able coe icien s on he job ask a iables in e ms
o sign, s a is ical signi icance, and magni ude (change in log wages due o 1SD change in a ask
measu e). This is qui e ema kable as he e a e some no able di e ences be ween he wo da a
sou ces in e ms o he su ey and ques ionnai e design, coun y/labo ma ke and ime
pe iod(s) conside ed. When compa ing hese c oss-sec ional es ima es wi h hose ob ained om
he spell- ixed e ec s speci ica ion, he changes in e u n es ima es a e consis en wi h s ong
wo ke so ing based on unobse able abili y di e ences. Fo example, he la ge and s a is ically
signi ican wage p emium o exe cising abs ac asks (a combina ion o cogni i e and
in e pe sonal equi emen s) ound be o e, anishes. Likewise, he la ge and signi ican wage
penal y o exe cising manual asks is educed by abou h ee-qua e s. The ask domain ha
u ns ou o be mos obus ly ela ed o wages, nega i ely so, is ou ineness. I shows he
smalles ela i e change in coe icien when mo ing om he c oss-sec ional o he spell ixed
e ec s e e ence speci ica ion (a educ ion by abou hal ). In es iga ing some e ec
he e ogenei ies, I ind ha he ela ion be ween ou ineness and wages is mode a ed by wha
o he asks indi iduals a e equi ed o pe o m concu en ly a hei job. Some o hese asks
appea o lend p o ec ion agains , while o he s ampli y, ou ineness’ ad e se e ec s on wo ke s’
IAB-Discussion Pape 2|2024
8
wages. Al hough my indings clea ly con i m AH’s conce n abou abili y- ela ed so ing issues,
hey do no ejec and a he suppo hei claim ha wi hin-occupa ion job ask a ia ion is a
ele an componen o o e all ask change, wa an ing sepa a e and equal conside a ion.
The e ha e been ea lie a emp s o ollow-up on AH, no ably by Cassidy (2017), Roh bach-
Schmid (2019) and S o m (2020), who all d aw on indi idual-le el job ask da a a ailable om
he Ge man Quali ica ion and Ca ee Su ey (GQCS). Nex o he B i ish Skills Su ey (BSS)
(Fels ead e al. 2007; Fels ead/G een 2008), he GQCS is one o he e y ew da a sou ces
in e na ionally ha allows o a wi hin-occupa ion pe spec i e (e.g. Spi z-Oene 2006;
Black/Spi z-Oene 2010). Howe e , as he BSS, he GQCS is designed as a epea ed c oss-sec ion.
This places clea limi s on any empi ical analysis in accoun ing o indi idual unobse ables and
hei ole in occupa ional sel -selec ion in much he same way as in AH’s o iginal pape . S a ing
back in 1979, he GQCS has also no been speci ically designed wi h a ask-based app oach in
mind. Mo eo e , he e a e only ew ques ionnai e i ems on job asks wi h a consis en , unal e ed,
ques ion wo ding ac oss consecu i e c oss-sec ions o e he las decades (see Roh bach-
Schmid /Tiemann 2013).
As al eady men ioned be o e, he e is a ela ed li e a u e ha d aws on indi idual-le el panel
da a o conside wage e u ns o occupa ional asks and a ilia ion (e.g. Co es 2016,
Ca aglia/E he idge 2020). The panel dimension o hei da a allows hese au ho s o con ol o
ime, indi idual and occupa ion ixed e ec s, as well as he in e ac ion among he la e wo (i.e.
spell ixed e ec s). Selec ion issues a e he eby add essed in a much mo e comp ehensi e way
han in AH o o he p io wo k. Co es’ pape , based on he U.S. Panel S udy o Income Dynamics
(PSID), has been he i s o use indi idual-le el panel da a o es ima e occupa ional ask p emia,
applying spell ixed e ec s es ima ion. Wo ke s’ asks, howe e , a e again p oxied o by an
agg ega ion o de ailed occupa ion codes in o h ee b oad ca ego ies ( ou ine, non ou ine
cogni i e, non ou ine manual). Mo eo e , any de ailed occupa ion code’s assignmen o one o
hese agg ega e ca ego ies is ixed a he beginning o he obse a ion pe iod, and kep cons an
o e ime. In a e y simila ein, Ca aglia and E he idge use indi idual-le el panel da a o he
B i ish Household Panel Su ey (BHPS) and Ge man Socio-Economic Panel (GSOEP) o es ima e
occupa ional ask e u ns o wo majo Eu opean economies.
The implici assump ion in all hese s udies is ha wo ke s employed in he same agg ega e
occupa ional g oup a e exposed o iden ical ask equi emen s a hei wo kplaces, and change
he ein. This iew is clea ly challenged by AH’s key inding o a la ge and independen sha e o
job-le el ask a ia ion wi hin occupa ions. Fu he mo e, i one assigns occupa ions o b oad
ask ca ego ies once and o all, nei he he a e age occupa ional ask con en is allowed o
change o e ime, no he indi idual job con en wi hin. In ligh o he long ime pe iods spanned
by he PSID, BHPS, GSOEP wi h ongoing echnical change and changing equi emen s, his
una oidable es ic ion appea s pa icula ly un a o able.
Pe haps closes in spi i o he p esen wo k is Ross (2017). Following Co es (2016), he d aws on
PSID da a and employs a ixed e ec s es ima ion app oach. Ins ead o g ouping de ailed
occupa ion codes in o e y b oad, ime-in a ian ca ego ies, he me ges in o ma ion om he
O*NET expe da abase. By using he occupa ional a ilia ion o he indi idual wo ke s in his
sample (a any gi en poin in ime) as a link, he cons uc s sco es o an occupa ion’s ou ine and
abs ac ask in ensi y. This s a egy allows him o conside ime- a ia ion in occupa ional ask
IAB-Discussion Pape 2|2024
15
(Panel B). The agg ega ion o occupa ions shown he ein is o exposi ional easons only,
showcasing he o e all sample composi ion a a compa a i e le el o de ail as in he ci ed
e e ence s udies. As explained in Sec ion III, he key eg ession models accoun o spell ixed
e ec s de i ed om he wo-digi le el KldB-2010 codes, wi h wen y-eigh dis inc occupa ions
being popula ed in each panel wa e (see Panel A o Table A2).
4 Desc ip i e S a is ics
I s a ou by p esen ing some desc ip i e e idence on he ou ask measu es and hei a ia ion
o e ime, wi hin occupa ions, and wi hin job spells. Table 1, below, depic s sample a e ages by
su ey wa e (Columns 1–3), mean changes o e he ull obse a ion window (Column 4), as well
as a iabili y o asks in e ms o s anda d de ia ions (Column 5). All s a is ics in he able a e
based on he o iginal sco es on a 0−1 scale, as desc ibed in Ma hes e al. (2014).
Table 1: Sample Va ia ion o NEPS Job Task Measu es
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Job Task Measu es
2011/12
mean
2015/16
mean
2019/20
mean
change in
means,
2011/12 -
2019/20
s d. de .
pooled
sample
sha e o ask
a ia ion wi hin
occupa ions
sha e o ask
a ia ion wi hin
job spells
Cogni i e
.633
.631
.626
-.007
.209
.756
.504
In e pe sonal
.467
.458
.447
-.020
.202
.706
.434
Physical
.306
.304
.292
-.014
.270
.593
.469
Rou ine
.387
.400
.412
.025
.172
.865
.429
Obse a ions
944
858
688
2,490
No e: The able shows a ia ion in NEPS job ask measu es: means by panel wa e, o e all a ia ion (s d. de .) and mean change
o e he ull obse a ion pe iod. Depic ed a e he aw sco es on a 0-1 scale as de eloped in Ma hes e al. (2014). The wo
igh mos columns show he (p opo ion o ) a iance in job ask sco es le a e demeaning he da a wi h espec o
occupa ions (KldB-2010 wo-digi codes) and job spells (unique indi idual-occupa ion combina ions).
Sou ce: Tex . © IAB
I no e only minu e o e all changes om one wa e o ano he , wi h no clea ly disce nable end
o e he gi en pe iod o any single measu e. Cogni i e, in e pe sonal, and physical mean ask
change is signed nega i ely, and posi i ely o ou ine. Ye , hese di e ences a e always
indis inguishable om ze o, s a is ically, wi h he espec i e s anda d de ia ion being la ge by
mul iple o de s o magni ude o each measu e.14
Gi en he ocus o his pape , and he compa a i e ad an age o he NEPS da a (which is no
long- e m end analysis), a mo e ele an examina ion conce ns he ex en o ask a ia ion
p esen wi hin occupa ions. Fo his pu pose, I eg ess ou any ixed occupa ion e ec s om each
one o he ou job ask measu es and hen di ide he a iance o he ( espec i e) esidual sco es
by hei o e all a iance. Ve y much in line p io c oss-sec ional s udies on job asks, also he
14 The wa e-speci ic s anda d de ia ions o each ask domain a e i ually iden ical o hei pooled coun e pa s, which is why I
do no epo hem sepa a ely.
IAB-Discussion Pape 2|2024
16
NEPS da a e idence a sizeable sha e o wi hin-occupa ion a ia ion (Column 6); wi h some sligh
di e ences by domain.15 O cou se, he exac magni udes always hinge on he kind and le el o
de ail o he occupa ional ca ego iza ion. The mo e (less) agg ega ed he selec ed coding
scheme, he mo e (less) a ia ion he e is wi hin occupa ions, by cons uc ion. In ha ega d, he
KldB wo-digi le el used h oughou his pape co e s a middle g ound in e ms o de ail.16
As a inal s ep, column 7 conside s he sha e o esidual ask a ia ion p esen wi hin spells, ha
is, a e ha ing ne ed ou all cha ac e is ics o indi iduals, occupa ions, and hei unique
combina ions. E en a e applying his da a ans o ma ion, a subs an ial sha e o he ini ial job
ask a ia ion is s ill emaining: abou hal , o somewha less han hal , depending on he
domain. This implies ha no only do asks a y subs an ially ac oss indi iduals holding he
same occupa ion, bu also o a gi en indi idual in a gi en occupa ion: ha is, o e ime, wi hin
he same spell. Bo h ypes o “wi hin” a ia ion o asks a e usually no conside ed and assumed
away. Columns 6 and 7 clea ly unde sco e AH’s poin ha we miss a nonnegligible po ion o ask
change, i we do no ex end he scope o ou empi ical analyses o include he pe son-job le el.
In he nex able, Table 2, I look a he way job asks a e in e ela ed wi h one ano he . S a ing
wi h he ull c oss-sec ional a ia ion in Panel A, we no e a ce ain co-occu ence o in e pe sonal
and cogni i e ask domains, and nega i e associa ions o hese wo wi h ou ineness.17 Physical
asks in u n co ela e nega i ely wi h bo h, cogni i e and in e pe sonal asks, albei much
weake so wi h he la e . A he same ime, physical asks a e sligh ly posi i ely co ela ed wi h
ou ine job asks. These associa ion pa e ns i well wi h wha AH ha e ound using hei c oss-
sec ional job asks da a (ibid.: S79, Table 3). To a deg ee, howe e , hey e lec di e ences in he
ypical (a e age) ask con en s o he occupa ions held by wo ke s.
In o de o conside in e co ela ions o job asks wi h such occupa ion-le el e ec s emo ed,
Panel B ecalcula es all associa ions among job ask measu es d awing on hei espec i e
wi hin-occupa ion a ia ion only. Compa ed o Panel A, we see a gene al weakening o all c oss-
co ela ions, which is exac ly wha one would expec o happen i asks a e sys ema ically
bundled oge he a he occupa ion le el. Ye , all associa ions o a leas mode a e s eng h
be o e (say, >|.3|) do pe sis and a e also ound again wi hin occupa ions.
To comple e he discussion o Table 2, Panel C desc ibes how he a ious job ask domains co-
de elop o e ime ( he las decade), again, ne o any occupa ion e ec s. No e, by conside ing
he in e co ela ions o i s -di e enced e sions o he job ask measu es, any indi idual wo ke
he e ogenei y is kep cons an , oo. Le us i s u n o change o e ime in jobs’ ou ine in ensi y
and how his ela es o changes in he o he domains, as his has ecei ed by a he mos
a en ion in he li e a u e.
15 Physical asks exhibi a somewha highe le el o be ween occupa ion a ia ion.
16 When using he mos ine-g ained i e-digi KldB-2010 o ou -digi ISCO-08 classi ica ions p o ided in he NEPS da a, abou
hal o he a ia ion in job asks is s ill le unexplained by occupa ion indica o s (318 and 189 dis inc codes in my sample,
espec i ely), again wi h some a ia ion by domain, and well in line wi h he magni udes epo ed by AH. De ailed esul s
a ailable upon eques . So, he gene al conclusion ha he e exis s a non-negligible amoun o wi hin-occupa ion ask
a ia ion clea ly holds h ough. As a gued in he da a sec ion, he ull i e-digi KldB-2010 o ou -digi ISCO-08 classi ica ions
a e oo de ailed o se e as e e ence ames in he gi en sample, wi h qui e many occupa ional ca ego ies con aining less han
a hand ul o obse a ions, no seldom only a single on pe code.
17 The co ela ion be ween cogni i e and in e pe sonal asks is clea ly posi i e (+.369) bu a om pe ec (1.00), suppo ing he
iew ha hese can be conside ed dis inc ac i i ies, o equi emen s. No seldom, hese wo domains a e amalgama ed o a
single measu e o he “abs ac ness” o asks (c . AH, in pa icula S91-92; and Ross 2017).
IAB-Discussion Pape 2|2024
17
Table 2: In e co ela ions o Job Task Measu es
Cogni i e
In e pe sonal
Physical
Rou ine
A. (pooled) C oss-sec ional
Cogni i e
1.000
In e pe sonal
+.369
1.000
Physical
−.354
−.079
1.000
Rou ine
−.507
−.357
+.174
1.000
B. (pooled) C oss sec ional, occupa ion-le el e ec s emo ed
Cogni i e
1.000
In e pe sonal
+.279
1.000
Physical
−.193
+.066
1.000
Rou ine
−.438
−.338
+.005
1.000
C. Fi s -di e enced, occupa io-le el e ec s emo ed
Cogni i e
1.000
In e pe sonal
+.126
1.000
Physical
−.021
+.052
1.000
Rou ine
−.181
−.200
−.073
1.000
No e: All co ela ions a e based on he ull sample o 2,490 obse a ions as summa ized in Table A2.
Sou ce: Tex . © IAB
The bo om ow o Panel C shows ha declining ou ine ask equi emen s go along wi h
inc easing cogni i e and in e pe sonal equi emen s ( he co ela ions in FDs being mode a ely
s ong wi h −.181 and −.200, espec i ely). The ac ha hese co ela ions a e a om pe ec ,
howe e , indica es ha he e is a lo o he e ogenei y p esen in he way jobs (wi hin he same
occupa ion, and o s a is ically iden ical wo ke s) change in e ms o hei ask composi ion.
S ill, o e all, hey a e commensu a e wi h RBTC explana ions o ask change, sugges ing ha he
kind o shi s ypically ound be ween occupa ions (i.e. away om ou ine and owa ds
non ou ine-cogni i e/in e pe sonal) also wo k wi hin occupa ions. Finally, he co ela ion
be ween i s -di e enced ou ine and physical ask measu es is signed posi i ely bu appea s a
bi weak (<|.1|) o d aw any clea conclusions. As a as he in e ela ions among changes o e
ime along o he domains a e conce ned, he e is some weak e idence poin ing owa ds
inc easing complemen a i y be ween cogni i e and in e pe sonal job ask con en s (co el. in
FDs abou +.13). This is a phenomenon ha has been no ed ea lie in e ms o skills, as opposed
o asks, by Weinbe ge (2014).
5 Job Tasks and Wages
Tu ning o he analysis o job asks and wages, displayed in Table 3 below, I pick up he h ead
whe e AH le o and begin by es ima ing pooled c oss-sec ion wage eg essions in columns 1
and 2. I con ol o a simila s anda d a ie y o human capi al and demog aphic co a ia es as
hey do: a se o indica o s o wo ke s’ highes le el o ( oca ional) educa ion a ained, o eign
IAB-Discussion Pape 2|2024
18
bo n and ma i al s a us. I also always include quad a ics in age and job spell enu e.18 Column 2
addi ionally accoun s o occupa ion-speci ic e ec s. Column 3 is he i s eg ession model
exploi ing he indi idual-le el panel s uc u e o my da a, he eby abso bing he ec o o
(p ac ically ime-in a ian ) human capi al and demog aphic con ols in o he se o wo ke - ixed
e ec s. Column 4 ex ends he model u he by ime-demeaning he da a wi hin each unique
wo ke -occupa ion spell, as explained abo e. All speci ica ions include ime ixed e ec s and a
dummy a iable o an indi idual’s cu en wo kplace being loca ed in o me Eas Ge many.
Each column o Table 3 epo s es ima ion esul s wi h (log) eal hou ly wages as he ou come
a iable and pe son-yea as he uni o obse a ion; s anda d e o s being clus e ed a he
indi idual le el h oughou . The explana o y job ask a iables a e no malized o ha e mean ze o
and s anda d de ia ion one.
Table 3: Wage Re u ns o Cogni i e, In e pe sonal, Physical and Rou ine Job Tasks
Ou come is (log) eal hou ly wages
(1)
(2)
(3)
(4)
Cogni i e
+.080***
+.083***
−.001
−.005
[s.e.]
[.012]
[.012]
[.010]
[.010]
In e pe sonal
−.013
+.007
−.000
−.003
[s.e.]
[.009]
[.010]
[.008]
[.009]
Physical
−.106***
−.090***
−.034***
−.027**
[s.e.]
[.010]
[.011]
[.011]
[.012]
Rou ine
−.053***
−.033***
−.024***
−.025***
[s.e.]
[.009]
[.009]
[.007]
[.007]
Time ixed e ec s
X
X
X
X
Human capi al and demog aphic con ols
X
X
Occupa ion ixed e ec s
X
X
Indi idual ixed e ec s
X
Indi idual-occupa ion spell ixed e ec s
X
Obse a ions
2,490
2,490
2,490
2,235
No e: The da a sou ce is he Adul Coho o he Ge man NEPS co e ing he pooled wa es 2011/12, 2015/16 and 2019/20 wi h
wo kplace ask measu es. The analysis sample is es ic ed o ull- ime wo king males wi h nonmissing in o ma ion. Each
column epo s es ima ion esul s wi h (log) eal hou ly wages as he ou come a iable and pe son-yea as he uni o
obse a ion. The cogni i e, in e pe sonal, manual and ou ine sco es a e no malized o ha e mean ze o and a s anda d
de ia ion o one. Occupa ion codes a e in acco dance wi h he 2010 issue o he Ge man Classi ica ion o Occupa ions
(Klassi ika ion de Be u e, KldB-2010) a he wo-digi le el. S anda d e o s a e epo ed in b acke s and clus e ed a he
indi idual le el.
***p<.01, **p<.05, *p<.10
Sou ce: Tex . © IAB
S a ing wi h he c oss-sec ional model shown in column 1, we see a s a is ically signi ican
posi i e coe icien o cogni i e job asks, and signi ican nega i e coe icien s o physical and
ou ine asks. The magni udes a e qui e sizeable, wi h p edic ed wage p emia and penal ies o
abou +8, −11 and −5 log poin s due o a one s anda d de ia ion change in he espec i e ask
scale. Adding occupa ion ixed e ec s in column 2 educes he es ima es ob ained o physical
18 No e, ha o all membe s o my sample age g ows collinea ly wi h gene al expe ience (age − # yea s o educa ion − 6), by
su ey design.
IAB-Discussion Pape 2|2024
19
and ou ine asks somewha , abou 2 log poin s each. In e pe sonal asks do no show up
signi ican ly. Al hough an immedia e compa ison wi h AH’s c oss-sec ional es ima es is impeded
by di e ences in ask measu es, samples, o he su ey design ea u es, coun ies and ime
poin (s) conside ed, he o e all pa e ns and magni udes a e su p isingly simila (c . AH: S82,
Table 5, in pa icula columns 4 and 7).19
Column 3 is he i s speci ica ion ha goes beyond AH and es ima es ask e u ns based on
indi idual-le el ixed e ec s eg essions. The la ge and signi ican coe icien o cogni i e asks
ound be o e is d i en down o ze o. Likewise, we see a la ge educ ion in he nega i e coe icien
o physical asks by almos 6 log poin s. These changes in es ima es a e consis en wi h (s ong)
wo ke so ing based on unobse able abili y di e ences. The es ima e ob ained o ou ine
asks is educed by only abou 1 log poin and in e pe sonal job asks again do no show up
signi ican ly.
Column 4 shows es ima ion esul s o he mos comp ehensi e spell ixed e ec s speci ica ion.
This s ep educes he coe icien o physical asks somewha u he , while ou ine asks’ wage
impac emains i ually iden ical a abou −2.5 log poin s ( o one s anda d de ia ion inc ease in
he sco e).
Appendix Table A3 o e s a numbe o sensi i i y checks o al e na i e ways o occupa ional
coding, by using ine le els o g anula i y (no. o digi s) and di e en classi ica ion schemes
(Ge man KldB s. ISCO). The o e all conclusion om hese checks is ha my pa icula choice is
a he innocuous and no he main d i e behind he esul s epo ed abo e. Any o he way o
accoun ing o occupa ion yields almos iden ical es ima es. Wi h an eye o sample size I p e e
o s ick o he mo e agg ega e KldB-2010 wo-digi le el h oughou he pape .
Appendix Table A4 p esen s some u he sensi i i y checks wi h espec o ou lie s (indi idual
wages, occupa ion size), sample de ini ion, and he inclusion o ime- a ying occupa ion ixed
e ec s; all es ima es being be highly compa able o hose ob ained om he e e ence
speci ica ion in Table 3, Column 4.
As an in e im conclusion, I no e ha bo h physical and ou ine asks happen o ecei e signi ican
wage penal ies; ne o wo ke s’ occupa ional a ilia ion, hei inhe en unobse able p oduc i e
a ibu es, and any occupa ion-speci ici y hese unobse ables may ha e. In con as , he wage
p emia o cogni i e asks p e iously ound in c oss-sec ional speci ica ions can be explained
en i ely by abili y- ela ed so ing. This p ecisely con i ms he conce n AH a icula ed wi h espec
o hei own c oss-sec ional es ima es, bu we e unable o add ess gi en he da a a ailable o
hem.
19 The es ima es do no change by much when condi ioning on occupa ion a he mo e de ailed h ee-digi le el, as in AH (using
he Ge man KldB in my case).
IAB-Discussion Pape 2|2024
20
6 Wage E ec s o Rou ine In ensi y by a
Job’s O he Task Requi emen s
As b ie ly no ed in he da a sec ion on measu emen , ye wo h highligh ing again he e, he NEPS
su eys cogni i e, in e pe sonal and physical job ask equi emen s as sepa a e domains and
wi hou men ion o he “ ou ineness” concep in he unde lying i em o mula ions. Analogously,
ou ine in ensi y o job asks is assessed wi hou e e ence o he o he domains. By example,
he e is no a emp o quali y whe he a ce ain le el o epo ed ou ine in ensi y ep esen s,
say, cogni i e- ou ine a he han manual- ou ine. F om he da a we only know whe he ou ine
in ense asks a e exe cised in a job en i onmen ha is a he same ime cogni i ely demanding
(o no ), physically demanding (o no ), and so o h. In a way, he ques ionnai e design akes he
ALM amewo k a ace alue: i is “pu e” ou ineness o human job asks - hei suscep ibili y o
au oma ion and subs i u ion by machines and algo i hms h ough codi iabili y- ha is
conside ed mos ele an and o ha e all kinds o implica ions o wages and he di ision o
labo .20
Commensu a e wi h his iew and su ey implemen a ion, i appea s na u al o s a ou wi h
wage eg ession models ha conside he a ious ask domains side by side; ha is, holding he
espec i e o he domain(s) ixed. When e-es ima ing he mos comp ehensi e e e ence
speci ica ion (Column 4 o Table 3) epea edly, ye each ime en e ing only a single ask measu e,
I ob ain i ually iden ical pa ame e es ima es. So, in e ms o i s es ima ed a e age wage e ec ,
ou ineness indeed appea s o be ope a ing la gely o hogonal o he o he domains conside ed.
While he ALM amewo k assumes a a he agnos ic s ance wi h ega d o whe he ou ineness
mani es s i sel in o m o −o in conjunc ion wi h− cogni i e, in e pe sonal, o physical asks, i
ne e heless has o eme ge in some way. I he e o e appea s wo hwhile o examine in some
mo e de ail whe he he p ima y and pe sis en nega i e ou ineness e ec is mode a ed by a
job’s ypical o he equi emen s. By unmasking some o he unde lying e ec he e ogenei ies
connec ed o he ocal ou ine ask domain, we can po en ially iden i y job ask p o iles ha a e
pa icula ly (un) a o able om a wo ke ’s pe spec i e.
Mo e conc e ely, I ake he job spell ixed e ec s speci ica ion om Table 3 as my poin o
depa u e and ask whe he (wi hin-job spell) changes in ou ine ask in ensi y ha e a di e en ial
wage impac depending on whe he a wo ke ’s cu en spell anks low, in e media e, o high in
e ms o i s’ o he ask equi emen s. I ake a p agma ic app oach and spli spells in o h ee
e iles, anking om low, in e media e o high in e ms o cogni i e, in e pe sonal, and physical
equi emen s, consecu i ely. These ca ego ical indica o s a e hen in e ac ed wi h he ou ine
ask sco es, as de ined be o e.21 Es ima ion esul s a e depic ed in Table 4.
20 As Au o (2013: 186) w i es, i is all abou he bounda y be ween “labo ” and “capi al” asks, and which happens o shi o e
ime depending on changes in echnological easibili y and economic cos (o au oma ion). Au o /Do n (2013), o example, also
cons uc a “pu e” ou ineness index, pu ged o any abs ac and/o manual occupa ion ask equi emen s, and pu i a he
cen e o hei empi ical analysis.
21 Using ine ca ego ies (qua iles, quin iles) o using ully con inuous in e ac ions (linea , squa ed, cubed, e c.) does no
unea h any eliable addi ional insigh s; and asks oo much o he da a, in my iew. The a ailable sample size and panel
a ia ion simply does no lend i sel well o go u he in o he ask dis ibu ions’ ails o iden i y complex nonlinea pa e ns.
IAB-Discussion Pape 2|2024
21
Table 4: Wage E ec s o Rou ine Task In ensi y by Job's O he Task Requi emen s
Ou come is (log) eal hou ly wages
A. by le el o cogni i e ask in ensi y
low
in e media e
high
Rou ine
−.028**
−.029**
−.015
0.47
F(high−low)
[s.e.]
[.013]
[.012]
[.013]
.492
p- alue
B. by le el o in e pe sonal ask in ensi y
low
in e media e
high
Rou ine
−.027**
−.020*
−.033**
0.11
F(high−low)
[s.e.]
[.013]
[.012]
[.014]
.736
p- alue
C. by le el o physical ask equi emen s
low
in e media e
high
Rou ine
−.008
−.017
−.048***
4.42
F(high−low)
[s.e.]
[.012]
[.012]
[.014]
.036
p- alue
No e: Model speci ica ion and analysis sample a e he same as in he mos comp ehensi e speci ica ion o Table 3 (Column 4).
Job spells a e classi ied in o e iles acco ding o hei le el o cogni i e, in e pe sonal and physical ask equi emen s. The
esul ing ca ego ical a iables a e hen ully in e ac ed, consecu i ely, wi h he s anda dized ou ine ask measu e as used
be o e. S anda d e o s a e epo ed in b acke s and clus e ed a he indi idual le el.
***p<.01, **p<.05, *p<.10
Sou ce: Tex . © IAB
Panel A conside s he e ogeneous ou ineness e ec s by le el o cogni i e ask in ensi y. I ind
ha he nega i e wage e ec o ou ine ask in ensi y is concen a ed in job spells ha ank low
o in e media e in e ms o hei cogni i e equi emen s. Ye , all h ee es ima ed coe icien s a e
signed nega i ely and he di e ence in coe icien s be ween he highes and lowes e ile is no
s a is ically signi ican , as he F- es in he igh mos column shows.
Panel B sugges s ha he le el o in e pe sonal ask equi emen s on he job is no a s ong
mode a o o he nega i e ou ineness e ec , albei some sligh a ia ion in coe icien s ac oss
e iles. The di e ence be ween any wo ca ego ies ne e exceeds one s anda d e o in
magni ude, and ne e mo e han hal a s anda d e o when compa ing each indi idual
coe icien o he es ima ed o e all a e age e ec om he e e ence speci ica ion.
This s ands in s a k con as o he indings o he le el o physical ask equi emen s, displayed
in Panel C. The wage penal y o ou ine in ense asks appea s o be con ined o wo kplace
con ex s ha a e a he same ime physically demanding. The poin es ima e o abou −5 log
poin s o jobs anking in he highes e ile o physical ask in ensi y is oughly wice as la ge as
he o e all a e age e ec . An equal change in (wi hin-job spell) ou ineness does no ha e any
signi ican nega i e e ec when un olding in wo kplace con ex s ha a e physically less
demanding.
In sum, ou ineness may be conside ed a cha ac e is ic o human job asks ha cu s h ough
o he domains and is a ec ing wages independen ly, ye i s e ec s eng h is clea ly mode a ed
by he kind and le el o o he asks equi ed concu en ly.22 High cogni i e demands appea o
22 Table 4 conside s le els o cogni i e, in e pe sonal and physical ask equi emen s consecu i ely. One could also hink o
e alua ing he ou ineness e ec a a ious combina ions o hese o he domains, ha is in a mul idimensional way. Absen any
IAB-Discussion Pape 2|2024
22
shield wo ke s agains he ad e se e ec s o ou ineness, while hese ad e se e ec s a e
pa icula ly p onounced in jobs whe e physical ask equi emen s a e also high.
To comple e he discussion, I ha e also examined e ec he e ogenei ies when in e ac ing asks
in a domain-by-domain way (see Appendix Table A5). Tha is, I ask how he wage e ec o any
pa icula domain a ies o e he e iles o i s own in ensi y le el dis ibu ion. By example, a
(po en ial) payo acc uing o cogni i e asks may be di e en depending on whe he a job a es
low, in e media e, o high in e ms o i s cogni i e equi emen s. Repea ing his o each domain,
sepa a ely, and always holding he o he domains cons an in he backg ound, he ollowing
esul s eme ge: Fi s , I do no de ec any impo an nonlinea i ies in cogni i e o in e pe sonal
asks ha migh ha e been masked by he linea panel speci ica ion.23 Second, he nega i e
a e age e ec o physical asks on wages ound be o e, is d i en by jobs wi h high physical
equi emen s, ce e is pa ibus. Thi d, in con as , he nega i e a e age wage e ec o ou ineness
eme ges pe sis en ly ac oss all e iles o he ou ine ask in ensi y dis ibu ion, and in abou he
same s eng h. I he e o e en a i ely conclude ha he p o ec ion agains , o ampli ica ion o ,
ou ineness’ ad e se e ec s on wages is ela ed o he (bundle o ) o he asks ha wo ke s a e
equi ed o exe cise; pe haps mo e so han o he le el o ou ine in ensi y i sel .
7 Robus ness Checks
The chosen spell ixed e ec s app oach exploi s ime- a ia ion wi hin unique indi idual-
occupa ion combina ions. Toge he wi h an a ay o addi ional ixed e ec s and con ols, his
emo es a hos o in luencing ac o s ypically no accoun ed o by p io s udies on job asks.
S ill, in o de o in e p e he es ima ed coe icien s as “selec ion- ee” wage p emia and
penal ies, one mus be willing o assume he absence o any o he non andom e o componen s
ela ed o bo h asks and wages. In ligh o he non-expe imen al na u e o he app oach, i is a
necessa y assump ion ha i sha es wi h all o he ecen panel-da a based s udies employing
spell ixed e ec s es ima ion, no ably Co es (2016) and Ca aglia/E he idge (2020).
Wha appea s possible, beyond ha , is o hink mo e conc e ely abou ma gins o he e ogenei y
and in e media e selec ion s eps ha ha e been omi ed so a , and a guably pose a ele an
conce n. One po en ial candida e ha may cause spu ious co ela ion be ween asks and wages
–e en when holding he indi idual-occupa ion spell ixed– is employe he e ogenei y, and he
ela ed ac ha he e is non andom assignmen o indi iduals o employe s. In con as o he
issue o occupa ional so ing and mobili y (e.g. wo ke s deselec ing ou ine occupa ions) his
conce n has ecei ed only scan a en ion in he empi ical ask li e a u e. Howe e , wo ke s may
well eac o changing ask demands and e u ns by a change o employe . I may e en be hei
i s line o adjus men , be o e swi ching he occupa ion. The mechanism is en i ely analogous o
ha discussed in e ms o occupa ional sel -selec ion. Indi iduals choose (and a e being chosen
by) employe s based on whe e hei pa icula se o abili ies is expec ed o be mos ewa ding
(o p oduc i e). By ex ending he spell de ini ion acco dingly, ha is, o unique indi idual-
clea heo e ical p edic ions o es , howe e , sea ching ac oss all so s o ask p o iles would in ol e a ai amoun o da a
mining in which I would no like o engage.
23 The igh mos column o Table A5 displays es s a is ics and p- alues, based on he co esponding F- es o each domain.
IAB-Discussion Pape 2|2024
23
occupa ion-employe combina ions, he spell ixed e ec s app oach can be eadily adap ed.
De ining job spells in his manne also sweeps any employe he e ogenei y pe se, e.g.
di e ences in pay policies (“high wage i ms” s. “low wage i ms”), con en s o collec i e
ag eemen s, and many o he ac o s.
I am he e o e able o con ol o unobse ed employe cha ac e is ics (and hei in e ac ion wi h
indi idual unobse ables) wi hou ha ing explici ly a ailable he nume ical iden i ie s, such as
employe IDs, which is a ely possibly. Adding employe ixed e ec s o panel wage eg essions is
usually no an op ion i based on su ey da a samples. A bes , a ew obse able cha ac e is ics
(e.g. indus y, numbe o employees) a e a ailable and linked o he pe son-le el esponses.
Howe e , since he NEPS depa s om a “pu e” panel s udy when i comes o he collec ion o
employmen in o ma ion om i s esponden s, i allows one o in e employe changes o
wo ke s.
Employmen is su eyed by asking o cohe en episodes, wi h begin and end da e, and no as a
s a us a iable (applying a he in e iew da e). By design, any epo ed episode e e s o an
unin e up ed pe iod o ime ha an indi idual wo ks o he same employe , wi hin he same
occupa ion. I ei he changes, he employe o he occupa ional a ilia ion, a new episode is being
eco ded. This ques ionnai e design and episode da a s uc u e implies, con e sely: whene e an
occupa ion code emains unchanged be ween wo adjacen employmen episodes o he same
indi idual, hen, an employe change mus ha e occu ed. This p ocedu e allows me o modi y
and ex end he spell de ini ion o unique indi idual-occupa ion-employe combina ions. I
he eby accoun s o employe -speci ic e u ns o indi idual unobse ables and also emo es
any ixed employe e ec s (in one go) in he eg ession analyses, as discussed.
Table 5: Robus ness o Es ima es w. . . Employe Cha ac e is ics and Changes
Cogni i e
In e pe sonal
Physical
Rou ine
(i) Adding " i m ype" ixed e ec s ( i m size Χ indus y cells)
(n=2,235)
−.002
−.005
−.025**
−.024***
[.011]
[.009]
[.012]
[.008]
(ii) Using only wi hin-indi idual-occupa ion-employe spell a ia ion in job asks
(n=2,100)
+.001
−.001
−.024**
−.020***
[.011]
[.009]
[.012]
[.008]
No e: Model speci ica ion and analysis sample a e he same as in he mos comp ehensi e speci ica ion o Table 3 (Column 4).
Fi m ypes a e de ined by he ull in e ac ion o ca ego ical i m size and indus y a iables. I agg ega e he de ailed wo-digi
indus y codes a ailable in he da a (acco ding o he Ge man Klassi ka ion de Wi scha szweige, WZ 2008) in o b oade
agg ega es (mo e in line wi h ISIC/NACE) and dis inguish he ollowing 9 ca ego ies: (1) Mining, elec ici y, gas, s eam, wa e , (2)
Manu ac u ing, (3) Cons uc ion, (4) Wholesale, e ail, anspo a ion, s o age, (5) In o ma ion and communica ion, (6) Financial
and insu ance ac i i ies, (7) P o essional, scien i ic, echnical se ices, (8) Public adminis a ion, educa ion, e c., (9) O he
se ice ac i i ies. The i m size a iable dis inguishes 5 ca ego ies: (a) less han 5 employees, (b) 5 o less han 20, (c) 20 o less
han 200, (d) 200 o less han 2.000, (e) mo e han 2.000 employees. Since one cell is emp y, I am able o con ol o 44 dis inc
i m ypes. Fo comple eness, no e, ha I allow he wage impac o each i m ype o di e by ime pe iod. S anda d e o s a e
epo ed in b acke s and clus e ed a he indi idual le el.
***p<.01, **p<.05, *p<.10
Sou ce: Tex . © IAB
Table 5 displays he es ima es ob ained when employe cha ac e is ics and employe ela ed
sel -selec ion o wo ke s a e addi ionally con olled. As a i s ake, o compa ison, obse able
IAB-Discussion Pape 2|2024
24
employe cha ac e is ics a e used o gene a e a ec o o indica o a iables cap u ing “ i m
ypes”, an in e ac ion o ca ego ical i m size in o ma ion and indus y codes. This would
usually be he only way o wa d wi h su ey panel da a in e ms o accoun ing o employe
he e ogenei y in p oduc i i y, pay policies, e c., as discussed abo e. Row (i) o Table 5 shows ha
he es ima ed coe icien s on physical and ou ine ask in ensi y a e only minimally a ec ed
when compa ed o he e e ence speci ica ion in Table 3 (Column 4). In ow (ii) o Table 5 he
es ima es ob ained om a spell ixed e ec s model using he augmen ed spell de ini ion,
including he employe dimension, a e p esen ed. Compa ed o ow (i), only he es ima ed
coe icien on he ou ine in ensi y domain shows some sensi i i y (change om −.024 o −.020).
Quali a i ely, his con i ms ha employe s’ cha ac e is ics, and sys ema ic wo ke so ing among
hem, may indeed be explaining pa o he ela ion be ween job asks and wages, as suspec ed.
Quan i a i ely, hey do no appea o be dominan (he e).
8 Concluding Rema ks
This pape p esen s empi ical e idence ha measu es o job-le el ask con en ha e a signi ican ,
independen , impac on indi idual wages. I base my analyses on newly a ailable panel da a o
Ge many ha allows me o apply indi idual-occupa ion(-employe ) spell ixed e ec s es ima ion
in his con ex , o he i s ime. My indings sugges ha RBTC explana ions o occupa ional ask
change and economic ou comes also hold wi hin occupa ions. Fo example, a pe sis en
nega i e wage e ec o ou ine ask in ensi y −in he sense o inc easing subs i u abili y o
human asks h ough machines and algo i hms− is p esen a he job-le el, oo. Wha also
eme ges is ha his nega i e wage impac is mode a ed by wha o he asks wo ke s a e equi ed
o pe o m concu en ly. Complex cogni i e ac i i ies lend p o ec ion, while manual o physical
asks exace ba e he wage penal y. By con olling o unobse ed wo ke , occupa ion, employe
and ma ch-speci ic he e ogenei y, I ind selec ion issues o play a majo ole, and o explain la ge
pa s o he c oss-sec ional associa ion be ween job asks and wages documen ed in p e ious
s udies; mos p ominen ly by AH. S ill, despi e he sizeable educ ions in he poin es ima es o
he a e age wage e u ns o se e al ask domains, I conside my esul s suppo i e o AH’s
gene al conclusion: in a-occupa ional a ia ion a he job-le el is an impo an componen ha
wa an s sepa a e conside a ion, complemen a y o he ask change happening be ween
occupa ions. This holds e en mo e so, as indi iduals do no , o canno , always easily lea e hei
occupa ion o o igin. While occupa ional mobili y has ecei ed a lo o a en ion in he li e a u e,
s aying is he de aul s a e o many, and examining how ask change a ec s hose who (a e
con ined o) s ay is equally ele an .
A inal ema k ela es o he ease wi h which occupa ions a e o en equa ed o asks, and only so.
The e is usually no much discussion abou wha occupa ions a e, and why hey should ma e
o indi idual economic ou comes, besides he idea ha hey e lec cha ac e is ic ask bundles.
As poin ed ou ea lie by Bi a ello/K ama z/Mai e (2018: 1), an occupa ion is no a well-de ined
economic concep . Wo ke s’ job asks a e some hing else han hei occupa ion and, con e sely,
occupa ions a e mo e han cha ac e is ic ask bundles (alone). A possible al e na i e, o
complemen a y, iew on occupa ions is ha o labo ma ke ins i u ions. Ce i ica es and
IAB-Discussion Pape 2|2024
31
Table A2: Summa y S a is ics
A. Panel S uc u e
(i)
Numbe o obse a ions
2,490
P opo ions by pa icipa ion pa e n ac oss panel wa es:
"111" ( ully balanced)
.725
"1.1"
.069
"11."
.206
(ii)
Numbe o indi iduals
944
(iii)
Numbe o occupa ions (KldB-2010 wo-digi codes)
28
Mean # o obs. pe occupa ion code
89
Median # o obs. pe occupa ion code
65
Min−Max
8−243
(i )
Numbe o job spells / unique indi idual-occupa ion combina ions
1,128
B. Sample Desc ip i es (Means and p opo ions)
( )
Mean log hou ly eal wages
3.16
S d. de .
.447
( i)
Highes le el o educa ion
≤ Lowe seconda y schooling + app en iceship
.182
Uppe seconda y schooling + app en iceship
.439
College, uni e si y
.379
( ii)
Mean age (in yea s)
45.3
S d. de .
7.8
Min−Max
25−58
( iii)
Fo eign bo n
.060
(ix)
Ma ied
.754
(x)
Mean job spell enu e (in yea s)
11.3
S d. de .
8.8
Min−Max
0−39
(xi)
Occupa ional ca ego ies (agg ega ed)
Manual p o essions
.243
Se ices
.086
Comme cial and Adminis a i e
.215
Technicians and Enginee s
.188
P o essionals and Manage s
.269
(xii)
Wo kplace loca ed in ( o me ) Eas Ge many
.158
No e: The able summa izes he panel s uc u e in e ms o indi iduals and occupa ions, and p o ides sample
means/p opo ions o he dependen a iable as well as some selec ed human capi al and basic demog aphic a iables. The
agg ega ed occupa ional ca ego ies a e based on Bloss eld’s Occupa ional Classi ica ion as p o ided in he NEPS da a Files. On
he o iginal concep ion, see Bloss eld (1985: 68), and on upda ing i o he KldB-2010 see Schimpl-Neimanns (2020).
Sou ce: Tex . © IAB [FV: Abbildung/Tabelle_Quelle]
IAB-Discussion Pape 2|2024
32
Table A3: Sensi i i y o Es ima es o Al e na i e Ways o Occupa ional Coding
Ou come is (log) eal hou ly wages
(1)
(2)
(3)
(4)
(5)
Cogni i e
−.007
−.009
−.010
−.011
−.008
[s.e.]
[.011]
[.010]
[.011]
[.010]
[.010]
In e pe sonal
−.000
+.002
+.003
−.000
+.003
[s.e.]
[.009]
[.009]
[.009]
[.009]
[.009]
Physical
−.026**
−.027**
−.025*(*)
−.025**
−.026**
[s.e.]
[.012]
[.012]
[.013]
[.012]
[.013]
Rou ine
−.025***
−.026***
−.024***
−.025***
−.023***
[s.e.]
[.008]
[.007]
[.008]
[.007]
[.007]
Indi idual-occupa ion spell ixed e ec s
based on:
KldB-2010: wo-digi + i h
X
KldB-2010: h ee-digi + i h
X
KldB-2010: i e-digi
X
ISCO-08: h ee-digi
X
ISCO-08: ou -digi
X
Time ixed e ec s
X
X
X
X
X
Obse a ions
2,182
2,157
2,126
2,169
2,143
No e: Da a and analysis sample a e as desc ibed in Sec ion 3. Each column uses he maximum numbe o obse a ions,
excluding single ons. Na u ally, he highe he le el o occupa ional de ail accoun ed o , he mo e single on obse a ions
d opped and he smalle he espec i e es ima ion sample. Occupa ion codes a e in acco dance wi h he 2010 issue o he
Ge man Classi ica ion o Occupa ions (Klassi ika ion de Be u e, KldB-2010) o he 2008 issue o he In e na ional S anda d
Classi ica ions o Occupa ions (ISCO-08), a a ying le els o g anula i y. In addi ion o con olling o ime e ec s and
indi idual-occupa ion spell ixed e ec s, all models include quad a ics in age and job spell enu e, and an indica o o whe he
he cu en wo kplace is loca ed in ( o me ) Eas Ge many. S anda d e o s a e epo ed in b acke s and clus e ed a he
indi idual le el.
***p<.01, **p<.05, *p<.10
Sou ce: Tex . © IAB [FV: Abbildung/Tabelle_Quelle]
IAB-Discussion Pape 2|2024
33
Table A4: Fu he Sensi i i y Checks o Job Task Re u n Es ima es
Cogni i e
In e pe sonal
Physical
Rou ine
(i) Dele ing wage ou lie s ( op and bo om 1%)
(n=2,199)
+.001
+.001
−.023**
−.024***
[.010]
[.008]
[.011]
[.007]
(ii) Dele ing small occupa ions (wi h less han 20 obse a ions)
(n=2,175)
−.003
−.000
−.026**
−.024***
[.011]
[.009]
[.012]
[.008]
(iii) Fo cing sample o be ully balanced
(n=1,671)
−.001
−.004
−.026**
−.023***
[.011]
[.009]
[.012]
[.008]
(i ) Con olling o ime- a ying occupa ion ixed e ec s
(n=2,235)
−.004
−.005
−.024**
−.022***
[.010]
[.009]
[.012]
[.007]
No e: Model speci ica ion and analysis sample a e he same as in he mos comp ehensi e speci ica ion o Table 3 (Column 4).
S anda d e o s a e epo ed in b acke s and clus e ed a he indi idual le el.
***p<.01, **p<.05, *p<.10
Sou ce: Tex . © IAB [FV: Abbildung/Tabelle_Quelle]
Table A5: He e ogeneous Job Task E ec s: domain × domain
Ou come is (log) eal hou ly wages
low
in e media e
high
Cogni i e
+.011
−.023
−.009
0.69
F(high s. low)
[s.e.]
[.017]
[.019]
[.018]
.406
p- alue
In e pe sonal
+.003
−.019
+.013
0.14
F(high s. low)
[s.e.]
[.018]
[.013]
[.017]
.707
p- alue
Physical
−.007
−.020
−.041**
0.93
F(high s. low)
[s.e.]
[.031]
[.019]
[.017]
.336
p- alue
Rou ine
−.024*
−.024**
−.028**
0.05
F(high s. low)
[s.e.]
[.014]
[.012]
[.013]
.823
p- alue
No e: Model speci ica ion and analysis sample a e he same as in he mos comp ehensi e speci ica ion o Table 3 (Column 4).
Job spells a e classi ied in o e iles acco ding o hei le el o cogni i e, in e pe sonal, physical and ou ine ask equi emen s.
The esul ing ca ego ical a iables a e hen ully in e ac ed, domain by domain, wi h he espec i e s anda dized ask measu e
as used be o e. S anda d e o s a e epo ed in b acke s and clus e ed a he indi idual le el.
***p<.01, **p<.05, *p<.10
Sou ce: Tex . © IAB [FV: Abbildung/Tabelle_Quelle]
Imp in
IAB-Discussion Pape 2|2024
Da e o publica ion
30 Janua y 2024
Publishe
Ins i u e o Employmen Resea ch
o he Fede al Employmen Agency
Regensbu ge S . 104
90478 Nü nbe g Ge many
Righ s o use
This publica ion is published unde he ollowing C ea i e Commons Licence:
A ibu ion – Sha eAlike 4.0 In e na ional (CC BY-SA 4.0)
h ps://c ea i ecommons.o g/licenses/by-sa/4.0/deed.de
Download o his IAB-Discussion Pape
h ps://doku.iab.de/discussionpape s/2024/dp0224.pd
All publica ions in he se ies “IAB-Discussion Pape ” can be downloaded om
h ps://iab.de/en/publica ions/iab-publica ions/iab-discussion-pape -en/
Websi e
h ps://iab.de/en/
ISSN
2195-2663
DOI
10.48720/IAB.DP.2402
Co esponding au ho
Ge i Mülle
Email: Ge i .Muelle @iab.de
