Job creation and trade in manufactures: Industry-level analysis across countries

Shi e aw, Admasu; Hailu, Degol A icle Job c ea ion and ade in manu ac u es: Indus y-le el analysis ac oss coun ies IZA Jou nal o Labo & De elopmen P o ided in Coope a ion wi h: IZA – Ins i u e o Labo Economics Sugges ed Ci a ion: Shi e aw, Admasu; Hailu, Degol (2016) : Job c ea ion and ade in manu ac u es: Indus y-le el analysis ac oss coun ies, IZA Jou nal o Labo & De elopmen , ISSN 2193-9020, Sp inge , Heidelbe g, Vol. 5, Iss. 3, pp. 1-36, h ps://doi.o g/10.1186/s40175-016-0052-z This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/152422 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 p://c ea i ecommons.o g/licenses/by/4.0/ ORIGINAL ARTICLE Open Access Job c ea ion and ade in manu ac u es: indus y-le el analysis ac oss coun ies Admasu Shi e aw 1* and Degol Hailu 2 * Co espondence: ashi e aw@wm. edu 1 The College o William and Ma y Williamsbu g, Vi ginia, USA Full lis o au ho in o ma ion is a ailable a he end o he a icle Abs ac This pape examines indus y-le el esponses o manu ac u ing employmen in he con ex o globaliza ion using a la ge sample o de eloped, de eloping, and ansi ion economies. We ind ha de eloping coun ies need a ypically high a es o alue-added g ow h (abou 10 %) o inc ease manu ac u ing employmen app eciably (abou 4 %). The employmen bene i s o expo o ien a ion a e also modes e en in “compa a i e ad an age”indus ies o de eloping coun ies. Howe e , di e si ying he expo baske con ibu es signi ican ly o employmen g ow h, pa icula ly in he medium- and high- echnology indus ies. Impo compe i ion does no unde mine employmen g ow h in low- echnology indus ies o de eloping coun ies while i displaces jobs in he same indus ies in O ganisa ion o Economic Co-ope a ion and De elopmen (OECD) and ansi ion economies. Fo de eloping coun ies, impo -induced job losses a e highe in he mo e capi al- in ensi e medium- echnology indus ies. Jobs in high- echnology indus ies a e less sensi i e o impo s wi h posi i e ela ionships obse ed in he OECD. In es men also complemen s job c ea ion in low- echnology indus ies o de eloping coun ies ha ha e ye o indus ialize. JEL codes: J21, L60, O14, O25 Keywo ds: Labo demand, Employmen elas ici y, Manu ac u ing, Expo o ien a ion, Impo compe i ion 1 In oduc ion Economic de elopmen and po e y educ ion in de eloping coun ies depend c i ically on access o gain ul job oppo uni ies. The Wo ld Bank’s2013Wo ld De elopmen Re- po highligh ed he b oade socioeconomic signi icance o jobs ha anscend he p i a e e u ns o employmen (Wo ld Bank 2012). Employmen in o mal sec o manu ac u ing, he ocus o his pape , ecei es special a en ion om policymake s and esea che s pa ly because i p o ides ela i ely s able and be e paying jobs. Such desi able a i- bu es o jobs dese e emphasis as he bulk o employmen oppo uni ies in de eloping coun ies a e c ea ed by small i ms in he un egis e ed sec o whe e ea nings a e ypic- ally uns eady and job- ela ed bene i s a e nonexis en (Wo ld Bank 2012; Goldbe g and Pa cnik 2003). While a numbe o s udies unde sco e he c i ical ole o indus ial p og ess and di e - si ica ion o economic g ow h (Hausmann e al. 2007; Jones and Olken 2005; Imbs and Waczia g 2003), he labo ma ke implica ions o indus ializa ion and di e si ica ion a e © 2016 Shi e aw and Hailu. Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons license, and indica e i changes we e made. Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 DOI 10.1186/s40175-016-0052-z no en i ely clea pa ly because such p ocesses a e aking place in he con ex o inc eas- ingly globalized economies. Unce ain y abou employmen p ospec s also a ises om e- cen episodes o “job-less”economic g ow h expe ienced by bo h de eloped and de eloping coun ies a leas in he sho o medium uns (Caballe o and Hammou 1997; Kapsos 2005). In ac , many de eloping coun ies g apple wi h pe sis en ly high un- employmen a es, especially in u ban a eas, in spi e o mo e open and g owing econ- omies. O e coming cons ain s o economic g ow h and expo s he e o e does no seem o gua an ee sa is ac o y labo ma ke ou comes. A ecen e iew o empi ical s udies om semi-indus ialized de eloping coun ies by Goldbe g and Pa cnik (2007) shows ha ade libe aliza ion has been accompanied by ising wage inequali y con a y o he ex- pec a ion o mo e a o able ou comes o low-skilled wo ke s. E idence is s ill lacking on he mo e p essing issue o unemploymen and global in eg a ion pa icula ly in de elop- ing coun ies. The limi ed empi ical li e a u e coming mainly om de eloped coun ies p o ides a mixed pic u e on he employmen e ec s o ade libe aliza ion as discussed la e in his pape . Labo economis s asse ha s ic labo ma ke egula ions a e among he main cul- p i s o slow employmen g ow h (Bo e o e al. 2004; Hal iwange e al. 2008). The as- sump ion is ha employe s espond o es ic i e hi ing and i ing egula ions wi h lacklus e job c ea ion, in es ing ins ead on p oduc i i y enhancing ac i i ies (Caballe o and Hammou 1997). Howe e , F eeman (2010) and Wo ld Bank (2012) epo only mod- es , i any, impac s o labo ma ke egula ions in de eloping coun ies pa ly because o weak en o cemen o egula ions. 1 Concu en wi h ade libe aliza ion, mos de eloping coun ies ha e also been elaxing labo laws du ing he 1990s (Goldbe g and Pa cnik 2007) u he a enua ing he po en ial explana o y powe o labo ma ke ins i u ions. O e he las decade o so, economis s began o explo e i m he e ogenei y in size, p oduc i i y, and expo o ien a ion o be e unde s and he mechanisms h ough which ade openness may a ec labo ma ke ou comes. Mo e ecen ade heo ies ex end he Mel i z (2003) model wi h he e ogeneous i ms by in oducing labo ma ke impe ec ions a ising no as such om hi ing and i ing egula ions bu om sea ch ic ions, e iciency wages, o i m-speci ic wage ba gaining. Leading con ibu ions in his ega d include Egge and K eickemeie (2009), Helpman e al. (2010), Felbe may e al. (2011a), and Da is and Ha igan (2011). These new models highligh he condi ions unde which ade openness could inc ease wage inequali y and he unemploymen a e. This ma ks a sig- ni ican imp o emen o e compa a i e ad an age and in a-indus y ade models whose implica ions o en do no ma ch obse ed labo ma ke ou comes. While he abo emen ioned heo e ical con ibu ions inspi e he empi ical analysis in his pape , some o hei es ic i e ea u es need o be add essed. Fo ins ance, po en ial di e ences in he labo ma ke e ec s o ade o de eloped and de eloping coun ies a e o en igno ed mainly because o he ocus on in a-indus y ade among coun ies wi h simila economic s uc u es. 2 Simila ly, a ia ion in indus y-le el esponses o ade e- o ms canno be in e ed di ec ly om he la es ade models gi en hei emphasis on equilib ium unemploymen a e and wage inequali y. Howe e , he con o e sies su - ounding ade libe aliza ion a ise p ima ily om di e gences in economic s uc u e and pa e ns o specializa ion ac oss coun ies. The empi ical app oach in his pape he e o e add esses such di e ences by using indus y-le el analysis o employmen o coun ies wi h dissimila economic s uc u es. As compa ed o he la ge empi ical li e a u e on he Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 2 o 36 p oduc i i y and wage inequali y e ec s o globaliza ion, s udies ha add ess i s employ- men e ec s a e a e and di e ema kably on a numbe o dimensions. These include di - e ences in esponse a iables, le els o agg ega ion, measu es o ade openness, and da a sou ces and quali y, which complica e meaning ul compa ison o he exis ing e idence and he lessons o be d awn. This pape con ibu es o his li e a u e by es ima ing a labo demand model o a la ge sample o de eloping, ansi ion, and de eloped coun ies obse ed o e he pe iod 1990– 2009. The model examines indus y-le el esponses o manu ac u ing employmen o changes in ma ke demand, ac o p ices, and ade openness as well as in es men and expo baske di e si ica ion. We use ou -digi In e na ional S anda d Indus ial Classi i- ca ion (ISIC) indus ies ha a e ma ched wi h ade lows om 72 coun ies. The analysis allows model pa ame e s o a y ac oss egional labo ma ke s and indus y ca ego ies a di e en le el o echnological ad ancemen . We also use coun y- and indus y-speci ic ins umen al a iables o add ess endogenei y p oblems ha o en unde mine es ima ion o labo demand models. As a p e iew o ou esul s, we ind a nega i e wage elas ici y o labo demand ha ends o decline in indus ies wi h ela i ely high labo sha e o ou pu . G ow h in de- mand inc eases manu ac u ing employmen al hough he elas ici y is a less p opo ion- a e such ha de eloping coun ies may need o achie e a ypically high a es o alue- added g ow h (abou 10 %) o inc ease employmen by abou 4 %. While demand and own-p ice elas ici ies o labo demand a e ema kably simila ac oss all coun ies and in- dus ies, sys ema ic egional di e ences eme ge in he labo ma ke implica ions o ade openness. The employmen esponse o ade openness also depends c ucially on he echnological composi ion o indus ies. Expo o ien a ion has a bes modes employ- men bene i s e en in “compa a i e-ad an age”indus ies o de eloping coun ies, while i ends o slowdown job des uc ion in low- echnology indus ies o O ganisa ion o Eco- nomic Co-ope a ion and De elopmen (OECD) coun ies. Howe e , we did no ind em- ploymen educing e ec s o expo s as implied by ecen ade heo ies. Impo pene a ion does no educe employmen g ow h in low- echnology indus ies o de elop- ing coun ies while i leads o subs an ial job displacemen in he same indus ies in he OECD. Fo de eloping coun ies, impo -induced job losses a e ela i ely high in he mo e capi al-in ensi e medium- echnology indus ies. Di e si ica ion o he expo baske boos s job c ea ion pa icula ly in skill-in ensi e indus ies o bo h de eloped and de el- oping coun ies. Fo coun ies ha ha e ye o indus ialize and become accomplished ex- po e s o manu ac u es, in es men has a s ong complemen a y e ec on employmen . The es o he pape is o ganized as ollows: Sec ion 2 ou lines he heo e ical pe spec- i es ha inspi e ou labo demand model and p o ides a e iew o exis ing empi ical e i- dence. Sec ion 3 discusses he da ase s and desc ibes he dis ibu ion o key a iables as well as he indus ial composi ion o employmen and ade. Es ima ion issues and ins u- men s a e add essed in Sec ion 4 while Sec ion 5 discusses he esul s. Sec ion 6 p o ides es ima es o an ex ended labo demand model while Sec ion 7 concludes he pape wi h some policy implica ions. 2 Theo e ical amewo k We d aw bo h on heo ies o labo demand and in e na ional ade o assess manu ac- u ing employmen in he con ex o globaliza ion. Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 3 o 36 2.1 De i ed demand o labo Following Hame mesh (1993), he own-p ice elas ici y o labo demand holding ou pu cons an can be exp essed as: ηLL ¼∂lnL ∂lnw¼−1−SL ½σð1Þ whe e Lis labo and wis wage a e, σis he elas ici y o subs i u ion be ween labo and o he inpu s, and S L is he e enue sha e o labo . Since wage shocks a ec p oduc p ices, he esul ing adjus men in he scale o p oduc ion will also a ec labo demand. Hame - mesh (1993) shows his e ec o depend on he p ice elas ici y o consume demand η. The e o e, he o al own-p ice elas ici y o labo demand can be exp essed as: ηLL ¼−1−SL ½σ−SLηð2Þ Equa ion (2) shows a nega i e wage elas ici y o labo demand wi h wo componen s: he “subs i u ion e ec ” ep esen ed by [1−S L ]σand he “scale e ec ” ep esen ed by S L η (Slaugh e 2001). In an indus y wi h low labo sha e, wage spikes a ec labo demand p ima ily h ough he ac o subs i u ion e ec while he scale e ec domina es in indus- ies wi h ela i ely high labo sha e. Ou basic labo demand unc ion, ollowing Hame mesh’s log-linea speci ica ion can hus be exp essed as: ln Lj  ¼δþβln Vj  þαln wj  þεj ð3Þ whe e Ls ands o uni s o labo , Vis alue added, wis wage a e, jindexes indus y, and indexes ime. In assessing he e ec o ade openness on labo demand, Rod ik (1997) a gues ha impo s no only push he p ice index down bu also aise he p ice elas ici y o demand (η) as access o a wide a ie y o consume goods inc eases. The own-p ice elas ici y o labo demand will hus inc ease in impo compe ing indus ies. He also unde sco es ha be e access o impo ed in e media e inpu s and pa s may gi e domes ic i ms u he lexibili y in o ganizing p oduc ion which implies an inc ease in σand hus highe wage elas ici y o labo demand ollowing a educ ion in impo es ic ions. Slaugh e (2001) p o ides u he de ails on hese mechanisms. To be e unde s and he ne e ec in a gi en indus y wi h a known labo sha e, Eq. (2) can be ea anged as: η LL =−σ−S L (η−σ). The e o e, in indus ies whe e η>σ, heown- p ice elas ici y o labo demand inc eases wi h S L . Since he p ice elas ici y o p oduc de- mand is equi alen o he elas ici y o subs i u ions be ween inal goods (Slaugh e 2001), an inc ease in η LL wi h he labo sha e o ou pu implies ha consume s’elas ici y o subs i u- ion be ween p oduc s is g ea e han i ms’elas ici y o subs i u ion be ween inpu s. T ade libe aliza ion can in ensi y his by inc easing consume s’choices mo e han he inpu choices o domes ic i ms. Howe e , i consume demand is ela i ely p ice inelas ic and impo s p ima ily inc eases σ, hen he o al p ice elas ici y o labo demand will be lowe o indus ies wi h ela i ely high labo sha e. 2.2 T ade heo ies and labo demand Recen ade heo ies wi h he e ogeneous i ms highligh labo ma ke e ec s o ade openness o he han h ough g ea e own-p ice elas ici y o labo demand. The basis o Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 4 o 36 his a gumen is he ecogni ion ha i ms pa icipa ing in expo ma ke s a e ypically la ge and mo e e icien han i ms p oducing only o domes ic ma ke s (Be na d e al. 2003; Be na d and Jensen 1999). Fo malizing hese obse a ions, he Meli z (2003) ade model shows ha he selec ion o expo e s om he uppe end o he i m size and p od- uc i i y dis ibu ions in ensi ies he ealloca ion o jobs owa d mo e p oduc i e i ms. Such job chu ning, howe e , lea es equilib ium wages and unemploymen in ac because o he assump ion o pe ec ly compe i i e labo ma ke s in he Meli z (2003) model. The inno a ion o ecen ade heo ies is hus o in oduce labo ma ke impe ec ion in Meli z- ype models o assess ade- ela ed wage inequali y and unemploymen . Helpman e al. (2010) p opose a ade model wi h he e ogeneous i ms and sea ch ic- ions. Fi ms in his model sc een wo ke s o hei abili ies and la ge i ms ha e economies o scale in sc eening allowing hem o hi e wo ke s wi h abo e a e age abili ies. Opening ade inc eases p o i s o la ge i ms en e ing expo ma ke s and aises hei incen i e o sc een wo ke s e e mo e s ic ly. Since wo ke s wi h supe io abili ies a e di icul o e- place, wages in expo i ms will be highe on a e age as compa ed o wages in non- expo ing i ms who sc een less, hence inc easing wage inequali y a e ade libe aliza ion. S ic e sc eening o wo ke s in he Helpman e al. (2010) model also educes he hi ing a e among expo i ms, which can inc ease he equilib ium unemploymen a e i he acancy- o-unemploymen a io is low. Felbe may e al. (2011a) p opose a simila model wi h sea ch ic ions whe e ade e- duces he unemploymen a e as long as i leads o a e age p oduc i i y g ow h. These au- ho s a gue ha he alue o a new job o an employe inc eases wi h a e age i m p oduc i i y (hence inc easing he job c ea ion a e), bu p oduc i i y g ow h equi es a e- duc ion in a iable ade cos s o an inc ease in he numbe o ading pa ne s. 3 A ade model by Egge and K eckmei e (2009) elies on wo ke s’p e e ence o a “ ai ” wage, which is ied o i m pe o mance. T ade openness unde his assump ion leads o wage inequali y as wages ise signi ican ly in highly p oduc i e i ms ha en e in o expo ma ke s. Non-expo i ms ace s i compe i ion om impo ed inal goods while expe i- encing p essu e o keep up wi h high wages in expo i ms. Unemploymen in he Egge and K eckmei e (2009) model inc eases because he combined e ec educes p o i ma - gins o non-expo i ms o cing hem o ei he cu jobs o exi he ma ke . Da is and Ha inga (2011) add an e iciency wage a gumen o a Meli z- ype model whe e la ge i ms pay highe han he ma ke -clea ing wage o elici e o om hei wo ke s as i is pa icula ly ha de o moni o e o in la ge i ms. While Da is and Ha inga (2011) p e- dic a ise in unemploymen a e ollowing ade openness, i is expec ed o be mode a e as compa ed o he olume o job ealloca ion ac oss i ms. O e all, he abo emen ioned ade models wi h impe ec labo ma ke s sugges ha ee ade could ha e undesi able consequences in e ms o inequali y and/o unemploy- men despi e clea wel a e bene i s. These p edic ions a e qui e di e en om bo h he adi ional Hecksche -Ohlin-Samuelson model and he in a-indus y ade models, and hey seem o be mo e consis en wi h ac ual da a and public opinion on ade. In he compa a i e ad an age ade models, o ins ance, ade openness would lead o in e - indus y ealloca ion o labo ( om impo compe ing indus ies o expo o ien ed ones) and educes unemploymen in a labo -abundan coun y. In he in a-indus y ade models, coun ies wi h simila endowmen s can gain om ade wi hou job ealloca ion ac oss i ms, as hey assume no i m he e ogenei y. Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 5 o 36 The p eceding discussion sugges s ha a labo demand model needs o accoun o ex- po o ien a ion and impo compe i ion. An inc ease in he e enue sha e o expo i ms could educe employmen g ow h as in he Helpman e al. (2010) and Da is and Ha inga (2011) models, o i may lead o employmen g ow h as in Felbe may e al. (2011a). Since he isk o i m exi declines wi h i m size and p oduc i i y (Be na d and Jensen 1999), one would also expec jobs in expo e i ms o be ela i ely mo e s able. I is he e o e possible ha as an indus y ge s inc easingly expo o ien ed, employmen becomes mo e secu e while job c ea ion a es may slowdown. Al hough ade heo ies wi h he e oge- neous i ms ypically assume i m p oduc i i y o be a andom d aw om a Pa e o dis i- bu ion, in p ac ice, expo e i ms would engage in long- e m p oduc i i y enhancing ac i i ies such as inno a ion, aining o wo ke s, and in es men in he la es machine y and equipmen —in e en ions ha a e mo e equen among la ge i ms (Re enga 1997). The esul ing p oduc i i y gains, including hose om lea ning- h ough-expo ing, may allow expo e s o inc ease ou pu wi h limi ed job c ea ion. The e ec on ne employ- men hus becomes an empi ical ques ion. Impo compe i ion, apa om educing wo ke s’ba gaining powe (Rod ik 1997), may lead o job des uc ion as in Egge and K eckmeie (2009). Howe e , he employmen e - ec a guably a ies ac oss indus ies wi hin a coun y depending on sou ces o compe i- i eness. I access o in e media e inpu s is a majo cons ain o domes ic i ms, educing impo es ic ions may inc ease a e age p oduc i i y. Compe i i e p essu e could also lead o p oduc i i y g ow h by educing ei he x-ine iciency o i m/wo ke en s (Re enga 1992). P oduc i i y g ow h may hus allow domes ic i ms o compe e e - ec i ely wi h impo s and a oid apid job des uc ion. Simila o he e ec s o expo s on labo demand, he ac ual employmen esponse o impo compe i ion is ul ima ely an empi ical ques ion. The labo demand model wi h expo o ien a ion and impo pene a ion a he in- dus y le el can be exp essed as: ln Lj  ¼δþβln Vj  þαln wj  þϕX Y  j þφM MþY  j þεj ð4Þ whe e Xs ands o expo s, Yis ou pu , and Mis impo s. A de ailed discussion o es ima ion issues will be picked up in Sec ion 4, while we now u n o a e iew o ela ed empi ical s udies. 2.3 Exis ing empi ical e idence The mul iplici y o new ade models wi h di e en assump ions and labo ma ke impli- ca ions sugges s ha he unemploymen e ec s o ade need o be assessed empi ically. The empi ical li e a u e unsu p isingly shows mixed esul s. Fo a sample o indus ialized coun ies in he OECD, Felbe may e al. (2011b) ind ha ade openness, measu ed as he ade sha e o GDP, educes he long- un unemploymen a e signi ican ly. Using a la ge sample o de eloped and de eloping coun ies, Du e al. (2009) also ind ha ade libe aliza ion educes he na ional unemploymen a e in he long un al hough un- employmen may inc ease in he sho un. While hese pape s assess na ional unemploy- men a es, hey say e y li le abou indus y-speci ic e ec s, which can be impo an gi en he ac oss-coun y di e ences in indus y s uc u e. Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 6 o 36 S udies ha examine indus y-le el employmen and ade o en use da a om a single coun y and p o ide in e es ing esul s. Ea lie s udies by G eenaway e al. (1999) o he UK, Re enga (1992) and Sachs e al. (1994) o he USA, and Re enga (1997) o Mexico all ind employmen educing e ec s o ade openness. 4 Mo e ecen ly, Au o e al. (2013) ind ha impo compe i ion om China has a signi ican nega i e e ec on local labo ma ke s in he USA. Dau h e al. (2014) ind ha impo s om China and Eas Eu ope ha e a mild ad e se employmen e ec on employmen while he ne employmen e ec o ade expos- u e is posi i e because o s ong Ge man expo s o hese coun ies. Hasan e al. (2007) in- es iga e he e ec s o India’s 1991 ade e o m on indus y-le el labo demand. They ind signi ican pos - e o m inc eases in he own-p ice elas ici y o labo demand, which end o be highe o Indian s a es wi h lexible labo ma ke s. Howe e , he coe icien s on indica- o s o ade policy shi s a e by and la ge insigni ican . Also, using indus y- and s a e-le el da a om India, Hasan e al. (2012) ind ha ade p o ec ion inc eases unemploymen in s a es wi h lexible labo ma ke s pa icula ly in u ban a eas wi h expo -o ien ed indus ies. The e idence om La in Ame ican coun ies is no as encou aging. Acco ding o A anasio e al. (2004), ade libe aliza ion in Colombia du ing he la e 1980s and ea ly 1990s signi ican ly inc eased wage inequali y and he likelihood o in o mal sec o em- ploymen whe e wages and bene i s a e minimal. The a gumen is ha o mal sec o i ms exposed o inc eased impo compe i ion cu pe manen employmen posi ions and ou - sou ce some ac i i ies o low-wage wo ke s in he in o mal sec o . In B azil, Menezes- Filho and Muendle (2011) ind la ge ansi ions o wo ke s ou o he labo o ce and in o unemploymen ollowing ade libe aliza ion. The au ho s a gue ha ealloca ion o labo o expo e s and compa a i e-ad an age sec o s is no enough o accommoda e impo - induced job displacemen s om he o mal sec o . Bo h A anasio e al. (2004) and Menezes-Filho and Muendle (2011) use household su eys o ma ch wo ke s wi h hei indus ial a ilia ion, and i will be in e es ing o examine he e idence based on ac ual indus y-le el da a as we do la e in his pape . The e is now a la ge li e a u e on he p oduc i i y e ec s o ade libe aliza ion in de eloping coun ies. 5 S udies on A ican coun ies a e a he sca ce and ocus on Sou h A ica. Using indus y-le el da a, Jenkins (2008) shows ha bo h impo s and expo s ha e nega i e e - ec s on labo demand in Sou h A ican manu ac u ing while Johnson and Sub amanian (2001) show ha a i educ ions inc ease p oduc i i y g ow h. Fi m-le el e idence by Söde bom and Teal (2000) show ha expo e i ms in A ica a e la ge , mo e p oduc i e, and capi al in ensi e on a e age han i ms supplying only o domes ic ma ke s. The empi ical e idence on unemploymen e ec s o ade libe aliza ion is hus qui e mixed al hough he weigh o e idence leans sligh ly owa d he unemploymen inc easing e ec . Such mixed esul s a e unsu p ising gi en he wide a ia ion ac oss s udies in e ms o dependen a iables, indica o s o ade openness, model speci ica ion, es ima- ion me hods, and ime ho izon. Some s udies use na ional unemploymen a es while o he use s a e- and indus y-le el da a. T ade measu es also include indices o ade ol- ume as well as ac ual ade policy changes. How much o he dispa i y in he exis ing em- pi ical e idence e lec s di e ences in da a and me hodologies is qui e unknown. One o he con ibu ions o his pape is hus o p o ide a consis en se o empi ical e idence on he employmen - ade ela ionship using he same da a sou ces, le el o disagg ega ion, model speci ica ion, and es ima ion me hod o a la ge sample o de eloping, ansi ion, and de eloped coun ies o e a compa able ime ho izon. The analysis also akes in o Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 7 o 36 accoun po en ial di e ences in he employmen esponse o ade openness based on he echnological ad ancemen o manu ac u ing indus ies. This app oach allows a be e compa ison o pe o mances ac oss coun ies while add essing di e ences in indus y s uc u e and deg ee o ade in eg a ion. 3 Da a and desc ip i e s a is ics 3.1 Da a The pape combines wo da ase s on manu ac u ing indus ies. The i s one is UNIDO’s Indus ial S a is ics da abase (INDSTAT4) ha p o ides disagg ega ed da a on indus ial ac i i ies a he ou -digi ISIC le el. INDSTAT4 epo s, among o he a iables, he num- be o wo ke s (L) and cu en US dolla alues o o al ou pu (Y), alue added (V), he wage bill (W), and in es men (I) in an indus y. The o he da a sou ce is he Uni ed Na ions Commodi y T ade da abase (COMTRADE), which p o ides highly disagg ega ed da a on expo s (X) and impo s (M) a he le el o six-digi Ha monized Sys em ade classi ica ion codes (HS codes). The ade da a is es ic ed o manu ac u ing indus ies. The wo da ase s a e combined a he ou -digi ISIC le el using a conco dance p o ided by he UN s a is ics di ision. The esponse a iable o in e es , ln(L ij ), is he loga i hm o o al numbe o wo ke s in coun y iand indus y ja ime . The wage a e (w)inanindus y ep esen s hea e age wage pe wo ke ob ained by di iding he o al wage bill by he o al numbe o wo ke s. Pa ial labo p oduc i i y ( ) is calcula ed as alue added pe wo ke in a ou -digi ISIC indus y. Cons an p ice alues o ou pu , alue added, and wages a e calcula ed using p oduce p ice indices (PPI) o wo-digi ISIC indus ies om INDSTAT2 da abase. This da abase p o ides indus y-le el indices o olume o ou pu om which we we e able o eco e he unde lying p oduce p ice index. Fo indus ies wi h missing alues o PPI, we use he manu ac u ing sec o PPI. The expo sha e o ou pu (X/Y) and impo pene a ion a es [M/(M+Y)] a e calcu- la ed by using ade da a om COMTRADE and o al ou pu om INDSTAT4. The numbe o expo i ems (HS) e e s o he numbe o six-digi p oduc s in a ou -digi ISIC indus y o which a coun y has nonze o expo s. The numbe o compe ing coun- ies (CX) is he numbe o expo ing coun ies a ound he wo ld; o each HS code, a coun y is an ac i e expo e . We ha e da a on he abo emen ioned a iables o 72 coun ies: eigh om Sub- Saha an A ica (SSA), 11 om Asia and he Paci ic (ASIA), en om La in Ame ican and he Ca ibbean (LAC), eigh om Middle Eas and No h A ica (MENA), 16 om Cen al and Eas Eu ope (CEE), and 19 om he OECD. The sample pe iod ex ends om 1990 o 2009 wi h unbalanced panel a he indus y le el. See Appendix C o he lis o coun ies. 3.2 G ow h a es Table 1 p o ides summa y s a is ics on manu ac u ing ou pu and employmen by egion. Real manu ac u ing alue added g ew a abou 3 % pe annum du ing 1990–2008 6 o he en i e sample wi h g ow h a es in de eloping coun ies exceeding ha o OECD coun ies. A abou 3 % pe annum, SSA and MENA a e he only egions wi h ela i ely s ong employmen g ow h du ing he sample pe iod. To pu g ow h a es in pe spec i e, i Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 8 o 36 cha ac e is ics ha a e indus y speci ic and ime in a ian . This es ima o will also ad- d ess he in luence o coun y ixed e ec s. The ime ixed e ec s will cap u e a ia ions o e ime in in e es a es and o he mac oeconomic shocks ha a e common o all in- dus ies. Since sample coun ies a e a di e en s ages o indus ializa ion and success in one indus y could ha e spillo e e ec s on ela ed indus ies, i is un ealis ic o assume independence o ε ij ac oss indus ies wi hin a coun y. Robus s anda d e o s will hus be used h oughou he analysis. Despi e hese sa egua ds, he panel ixed e ec s model is unlikely o add ess all he endogenei y p oblems sa is ac o ily. Job losses in speci ic manu ac u ing indus ies may igge selec i e ade p o ec ion measu es ha educe impo s, while inc eased emale labo o ce pa icipa ion and/o mig a ion may uel expo compe i i eness. The e o e, we use ins umen al a iables o accoun o p oblems o e e se causa ion and ime a ying un-obse ables and ob ain exogenous a ia ion in impo pene a ion a e and expo o ien a ion as well as eal wages and eal alue added. The ins umen s we cons uc a e coun y and indus y speci ic and a y o e ime. One o ou ins umen s o impo pene a ion a e is a weigh ed a e age exchange a e index o majo impo supplie coun ies o a pa icula indus y o a gi en coun y. Since he US Dolla is he cu ency o choice o mos in e na ional ansac ion, we use US Dolla ex- change a es o impo pa ne coun ies. Exchange a e indices om he IMF Financial S a- is ics da abase a e weigh ed by he sou ce coun y’s sha e in o al impo s o a coun y in a gi en indus y using ade lows om he COMTRADE da abase. We also cons uc an a e age implici GDP de la o o impo supplie coun ies weigh ed in a simila ashion as an addi ional ins umen o impo pene a ion a e. The assump ion is ha exchange a es and p ice indices o key impo supplie coun ies a e exogenous o economic condi ions in a speci ic indus y o he impo ing coun y while a ec ing impo cos s. The ins umen s o impo pene a ion a e can be exp essed as: EXIij ¼X k Mijk X k Mijk 0 B B @ 1 C C A EXUS k ð6Þ PPIij ¼X k Mijk X k Mijk 0 B B @ 1 C C A PPIk ð7Þ whe e EXI ij is a weigh ed a e age exchange a e index o impo supplie coun ies, M ijk ep esen s impo by coun y i om supplie coun y kin indus y ja ime ,andEX US k is US Dolla exchange a es o impo supplie coun ies. PPI k ep esen s implici GDP de- la o s o sou ce coun ies as a p oxy o hei p oduce p ice indices while PPI ij is a weigh ed a e age indus y-speci ic index o impo sou ce coun ies. In cons uc ing he ins umen s o impo pene a ion, we allow he composi ion o pa - ne coun ies o a y ac oss indus ies and o e ime, which oge he wi h he annual da a on exchange a es and GDP de la o s p o ide us wi h indus y-speci ic ins umen s o each samplecoun y.Weusea2.5%impo sha easacu o poin ode e mine hese o im- po pa ne coun ies. The selec ed sou ce coun ies accoun on a e age o 89 % o indus y-le el impo s in ou sample. As muchasda aallow,we ake h eeda apoin seach Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 15 o 36 o he 1990s and 2000s o allow changes in he composi ion o impo sou ce coun ies. In a ew cases, da a limi a ions es ic us o only wo se s o pa ne coun ies, i.e., one se o he 1990s and ano he se o he 2000s. Ou ins umen s o impo compe i ion di e om ha o Au o e al. (2013) whe e hey used impo s by o he de eloped coun y om China o ins umen US impo s om China. Dau h e al. (2014) ollow a simila app oach o Ge many. Unlike hese au ho s, we examine he employmen impac o impo in ensi y in gene al a he han ocusing on compe i ion om eme ging economies—China in he case o he US and China and Eas Eu ope in he case o Ge many. Ou app oach measu es he changes in impo supplie coun ies di ec ly a he han he indi ec app oach which elies on he beha io o impo ing coun ies a a simila le el o de elopmen . To ins umen he expo -ou pu a io, we cons uc indus y-speci ic weigh ed a e age eal exchange a e index o expo ing coun ies. We use he o icial US dolla exchange a e index adjus ed o indus y-speci ic p ice indices. The p oduce p ice indices o majo ex- po des ina ion coun ies a e weigh ed by hei espec i e expo sha es om a gi en in- dus y using ade lows om COMTRADE. The o icial exchange a e is assumed o be exogenous o i ms in a ou -digi ISIC indus y and so also a e p oduce p ices in he ex- po des ina ion coun ies. In addi ion o his, we also use a weigh ed GDP index o expo des ina ion coun ies o cap u e demand shocks in pa ne coun ies using he expo sha es as weigh s. To be included in he calcula ion o he eal exchange a e index and he GDP index, a des ina ion coun y should accoun o a leas 2.5 % o an indus y’sexpo s om a sample coun y. The selec ed des ina ion coun ies accoun on a e age o 90 % o indus y-le el expo s om ou sample. The ins umen s o expo -ou pu a io can be exp essed as: REXIUS ij ¼EXUS i PPIij X k0 Xijk0 X k0 Xijk0 0 @1 APPIk0 0 B B B B B B @ 1 C C C C C C A ð8Þ GDPIij ¼X k0 Xijk0 X k0 Xijk0 0 B B @ 1 C C A GDPIk0 ð9Þ whe e REXIUS ij is weigh ed a e age eal exchange a e index o expo des ina ion coun ies, X ijk ' is o al expo om indus y jo coun y i o des ina ion coun y k′a ime ,EX US i is he o icial exchange a e o expo ing coun y i, PPI ij is indus y-le el p oduce p ice index ( om INDSTAT2) o expo ing coun y, and PPI k' is implici GDP de la o o an expo des ina ion coun y. GDPI k' is he GDP index o an expo des ina ion coun y while GDPI ij is he agg ega e GDP index o expo des ina ion coun ies applicable o indus y jo an expo ing coun y. 12 Assuming ha wo ke s in a ou -digi indus y will conside he a e age wage in he e- spec i e wo-digi indus y as an al e na i e wage, we use he la e o ins umen eal wage a es a he ou -digi le el. The ac ha we a e using eal wages al eady akes in o accoun he wo-digi p oduce p ice index in ha indus y. Real alue added o each Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 16 o 36 ou -digi indus y is ins umen ed by he eal GDP o he same coun y o cap u e de- mand shocks om he en i e domes ic economy. Following he discussion in Sec ion 3, he labo demand model will be es ima ed sepa - a ely o he six egions. By c ea ing sub-samples o coun ies wi h compa able geo- g aphic, his o ical, and ins i u ional backg ounds, we hope o u he mi iga e he e ec s o ime a ying coun y-le el he e ogenei y ha ou model does no con ol o . These in- clude di e ences in echnological capabili ies, human capi al, and ins i u ional quali ies ha could a ec he employmen esponse o exogenous shocks in ou co a ia es. The as egional di e ence in indus ial s uc u e and dynamics shown ea lie in Sec ion 3 seem o suppo his s a egy a he han imposing he same coe icien o all de eloped and de eloping coun ies. Wi hin each egion, he employmen model will also be es ima ed o he en i e manu- ac u ing sec o as well as o high-, medium-, and low- echnology indus ies. The idea is o allow o he e ogeneous employmen esponses o exogenous shock in ou co a ia es ac oss g oups o indus ies which may a ise om unobse ed echnological and ma ke cha ac e is ics. The obse ed s uc u al di e ences in Sec ion 3 suppo his classi ica ion o indus ies, which hope ully enhances he s a is ical p ecision and policy ele ance o he econome ic analysis. 5 Es ima ion esul s 13 Table 4 p esen s es ima es o he labo demand model o g oups o de eloping, ansi- ion, and de eloped coun ies. As would be expec ed, we ind a nega i e and s a is ically signi ican own-p ice elas ici y o labo demand ac oss all egions. In de eloping coun- ies, he coe icien on eal wage anges om 0.44 in MENA o 0.54 in SSA wi h an a e - age o 0.48. In e es ingly, Table 4 shows posi i e and s a is ically signi ican coe icien s on he in e ac ion e m be ween eal wages and he labo sha e o ou pu o all egions. This sugges s ha he elas ici y o employmen wi h espec o eal wages declines as he labo sha e o ou pu inc eases. A a e age alues o labo sha e, which is abou 42 % in de el- oping coun ies, he esul s indica e abou 5 % educ ion in he own-p ice elas ici y o labo demand. This is consis en wi h he neoclassical ac o demand heo y whe e he a e o ac o subs i u ion declines as i ms use mo e o one inpu keeping ou pu con- s an . The sign o he in e ac ion e m also sugges s ha he elas ici y o subs i u ion o inpu s o manu ac u ing i ms is g ea e han consume s’elas ici y o subs i u ion be- ween inal p oduc s, al hough in heo y he la e has no uppe bound (Slaugh e 2001). Fo mal compa ison o coe icien s in Appendix A: Table 10 in he Appendix shows ha he coe icien on wage in OECD coun ies is signi ican ly la ge han coe icien s in each de eloping egion. Among de eloping coun ies, s a is ically signi ican di e ences exis mainly on he in e ac ion e m be ween labo sha e and log wages. 14 Demand o manu ac u ed goods has he expec ed posi i e and s a is ically signi ican e ec on labo demand. Ac oss he de eloping wo ld, he a e age demand elas ici y o employmen is abou 0.43, i.e., a 1 % g ow h in eal alue added leads o a 0.43 % inc ease in employmen . While coe icien s on alue added in Table 4 show s a is ically insigni i- can di e ences among de eloping egions, Appendix A: Table 10 shows ha hey a e sig- ni ican ly lowe han ha o OECD coun ies. Gi en he ac ual con ac ion o manu ac u ing employmen in he OECD, he posi i e demand elas ici y indica es he Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 17 o 36 ealloca ion o jobs om he declining LT indus ies owa d be e pe o ming MT and HT indus ies. The coe icien s on eal alue added in Table 4 sugges ha he employmen bene i s o indus ializa ion in de eloping coun ies a e a he mode a e han wha policymake s may ha e wished o . A de eloping coun y needs o main ain abou 10 % g ow h in eal alue added o achie e a 4.3 % annual g ow h in manu ac u ing employmen . Howe e , pe o mance indica o s in Tables 1 and 2 sugges ha such spec acula a es o alue- added g ow h a e a he in equen . In coun ies whe e popula ion and labo o ce con- inue o g ow a 2–3 % pe annum, inc easing he employmen sha e o manu ac u ing may equi e e en highe a es o indus ial expansion. While demand and wage elas ici ies o labo demand a e ema kably simila ac oss de- eloping egions in e ms o coe icien size and signi icance, Table 4 shows s iking di e - ences in he employmen esponse o pa icipa ion in in e na ional ade (see also Appendix A: Table 10). LAC and MENA a e he only de eloping egions whe e impo pene a ion signi ican ly dampens domes ic labo demand. This is no su p ising in he case o LAC gi en i s high impo pene a ion a e ela i e o o he de eloping egions. This inding is also consis en wi h exis ing coun y-le el s udies in La in Ame ica as dis- cussed ea lie (Re enga 1997; A anasio e al. 2004; Menezes-Filho and Muendle , 2011). Impo pene a ion causes no ad e se employmen e ec s in ASIA, while i signi ican ly inc eases employmen g ow h in SSA. The la e is likely he esul o be e access o impo ed in e media e inpu s mo e han o se ing he nega i e employmen e ec s o compe i ion om impo ed inal p oduc s. While impo compe i ion has s a is ically sig- ni ican job des uc ion e ec s in CEE and OECD coun ies, he employmen e ec in he OECD is a he ma ginal as compa ed o CEE. La ge egional di e ences also eme ge in he employmen e ec s o expo o ien a ion. O he hings being equal, g ow h in manu ac u ed expo s inc eases labo demand signi i- can ly in LAC and ASIA. Posi i e albei weake employmen e ec s o expo s a e also Table 4 Es ima ed labo demand model by egion: IV-panel ixed e ec s SSA MENA LAC ASIA DVPG CEE OECD Ln (V) 0.3597*** 0.3724*** 0.3935*** 0.4614*** 0.4337*** 0.2770*** 0.6402*** (0.0336) (0.0206) (0.0227) (0.0149) (0.0096) (0.0109) (0.0082) M/(M+Y) 0.2437** −0.3097*** −0.3961*** 0.0214 0.0679 −0.3188*** −0.0600** (0.0967) (0.0922) (0.0836) (0.0987) (0.0443) (0.0688) (0.0251) X/Y −0.0382 −0.0045 0.0355*** 0.0212*** 0.0268*** 0.0172* 0.0141*** (0.0305) (0.0358) (0.0082) (0.0079) (0.0061) (0.0096) (0.0037) Ln (w)−0.5421*** −0.4406*** −0.4635*** −0.4785*** −0.4804*** −0.3649*** −0.6440*** (0.0214) (0.0136) (0.0172) (0.0117) (0.0071) (0.0092) (0.0062) Ln (w)*s 0.0743*** 0.0337*** 0.1215*** 0.0823*** 0.0577*** 0.0020*** 0.0674*** (0.0068) (0.0029) (0.0065) (0.0045) (0.0021) (0.0002) (0.0016) Cons an 5.3675*** 4.7048*** 4.7174*** 4.0743*** 4.2927*** 5.8563*** 2.0429*** (0.4848) (0.3097) (0.3828) (0.2517) (0.1581) (0.1741) (0.1382) Obse a ions 1367 3619 2789 5235 13010 7950 16258 Numbe o indus ies 162 397 449 448 1456 1068 1501 No e: All a iables a e lagged by one pe iod. Vis eal alue added, w is eal wage, sis he sha e o labo in ou pu , X/Y is he expo sha e o ou pu , and M/(M + Y) is impo pene a ion a e. Robus s anda d e o s indica ed in pa en heses.*, **, and *** ep esen s a is ical signi icance a 10, 5, and 1 %, espec i ely. DVPG e e s o de eloping coun ies Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 18 o 36 obse ed in CEE and OECD coun ies. Owing pa ly o hei limi ed pa icipa ion in ex- po ma ke s, employmen in SSA and MENA does no a y wi h he expo -ou pu a io o an indus y. O e all, he e idence does no suppo he nega i e ela ionship be ween employmen and expo g ow h p edic ed by some o he new ade heo ies wi h he e o- geneous i ms. Bu ou indings a e mo e consis en wi h Felbe may e al. (2011a) whe e expo o ien a ion inc eases job c ea ion. None heless, he coe icien s on he expo a io a e ypically small e en o egions whe e hey a e s a is ically signi ican and di e ences ac oss egions a e la gely insigni ican as shown in Appendix A: Table 10. Such unimp es- si e employmen bene i s sugges ha he size and p oduc i i y ad an ages o expo ing i ms may ha e allowed hem o inc ease expo s wi h limi ed adjus men o labo . The esul s in Table 4 e lec a e age employmen esponses ac oss all manu ac u ing indus ies igno ing po en ial he e ogenei y ac oss g oups o indus ies. By es ima ing he labo demand models o echnology-based g oups o indus ies, we allow employmen esponses o a y wi hin each egion. Table 5 examines employmen in low- echnology in- dus ies while Tables 6 and 7 epo he esul s o medium- and high- echnology indus- ies, espec i ely. Because o limi ed numbe o obse a ions o he SSA sample, Table 6 combines he MT and HT indus ies in A ica. Table 5 indica es ha o mos coun ies, he elas ici y o employmen wi h espec o alue added is la ge in LT indus ies (by abou en pe cen age poin s) ela i e o he manu- ac u ing sec o a e ages epo ed in Table 4 ( ha o LAC and CEE a e close o he a e - age). This is no he case o MT indus ies in Table 6, whe e he coe icien s on alue added a e ypically lowe han he espec i e egional a e age in Table 4 o all egions. These esul s a e consis en wi h he desc ip i e s a is ics ha LT indus ies a e ypically labo -in ensi e while MT indus ies end o be capi al-in ensi e. The employmen in ensi y o alue-added g ow h in HT indus ies (Table 7) is o en be e han MT indus ies bu less han LT indus ies. LAC and CEE a e excep ions whe e demand elas ici ies o employmen in HT indus ies a e s onge ela i e o LT indus ies. As discussed ea lie , HT indus ies a e he only indus ies in CEE and OECD coun ies ha exhibi ed employmen g ow h. Table 5 Es ima ed labo demand model o low- echnology indus ies by egion: IV-panel ixed e ec s SSA MENA LAC Asia DVPG CEE OECD Ln (V) 0.4403*** 0.5452*** 0.3779*** 0.6026*** 0.5489*** 0.2782*** 0.7516*** (0.0480) (0.0345) (0.0408) (0.0241) (0.0159) (0.0168) (0.0141) M/(M+Y)−0.0290 −0.5260*** −0.0830 0.1866 −0.0586 −0.6591*** −0.2830*** (0.1435) (0.1187) (0.1897) (0.1470) (0.0660) (0.1084) (0.0442) X/Y −0.0193 0.1211*** −0.0163 0.0401** 0.0346** 0.0128 0.1343*** (0.0299) (0.0425) (0.0452) (0.0157) (0.0135) (0.0176) (0.0096) Ln (w)−0.4931*** −0.5793*** −0.4370*** −0.6230*** −0.5629*** −0.3172*** −0.7399*** (0.0342) (0.0275) (0.0288) (0.0215) (0.0134) (0.0157) (0.0111) Ln (w)*s 0.1110*** 0.0775*** 0.1401*** 0.1784*** 0.1284*** 0.0013*** 0.0507*** (0.0138) (0.0100) (0.0124) (0.0090) (0.0054) (0.0002) (0.0021) Cons an 3.8088*** 2.9508*** 4.8977*** 2.3120*** 2.7883*** 5.6677*** 0.9704*** (0.6850) (0.4734) (0.7177) (0.3922) (0.2464) (0.2737) (0.2545) Obse a ions 671 1143 944 1610 4368 2683 4692 Numbe o indus ies 80 127 145 139 491 347 435 No e: See no es unde Table 4 Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 19 o 36 Al hough impo pene a ion a es in LT indus ies a e highe han o he indus ies, his does no seem o induce signi ican educ ion in labo demand in de eloping coun ies (Table 5). The only excep ion is he esou ce ich MENA egion whe e he coe icien on impo pene a ion is la ge, nega i e, and s a is ically signi ican . As de eloping coun ies ha e become he main sou ces o LT expo s (accoun ing o abou wo hi ds o o al LT expo s o coun ies in ou sample), he absence o a s ong ad e se employmen e ec o impo s in his sec o could be he esul o ade among de eloping coun ies. Fo de- eloped and ansi ion economies, howe e , LT impo s cause signi ican displacemen o low-skill jobs. This indica es he ela i e compe i i eness o de eloping coun ies in LT Table 6 Es ima ed labo demand model o medium- echnology indus ies by egion: IV-panel ixed e ec s SSA MENA LAC Asia DVPG CEE OECD Ln (V) 0.3261*** 0.3141*** 0.3973*** 0.2838*** 0.3607*** 0.4240*** 0.5806*** (0.0511) (0.0365) (0.0444) (0.0292) (0.0176) (0.0215) (0.0166) M/(M+Y) 0.3232** −0.5156** −0.5473*** −0.4580** 0.1886** −0.2788** −0.2495*** (0.1323) (0.2418) (0.1523) (0.1977) (0.0757) (0.1288) (0.0626) X/Y −0.0092 −0.0226 0.0353 0.0351*** 0.0233** 0.0144 0.0761*** (0.1455) (0.0766) (0.0773) (0.0113) (0.0117) (0.0109) (0.0232) Ln(w) −0.5740*** −0.3979*** −0.4930*** −0.3936*** −0.4630*** −0.4886*** −0.5816*** (0.0302) (0.0214) (0.0311) (0.0213) (0.0117) (0.0180) (0.0119) Ln (w)*s 0.0646*** 0.0345*** 0.1546*** 0.0732*** 0.0548*** 0.0519*** 0.0809*** (0.0081) (0.0051) (0.0130) (0.0092) (0.0035) (0.0031) (0.0034) Cons an 5.9938*** 5.4613*** 4.9503*** 6.9532*** 5.4960*** 4.3036*** 2.5997*** (0.7348) (0.5998) (0.7560) (0.5174) (0.3044) (0.3289) (0.2906) Obse a ions 696 1380 893 1600 4437 2539 5366 Numbe o indus ies 82 149 143 137 491 346 497 No e: See no es unde Table 4 Table 7 Es ima ed labo demand model o high- echnology indus ies by egion: IV-panel ixed e ec s MENA LAC ASIA DVPG CEE OECD Ln (V) 0.3715*** 0.4394*** 0.3771*** 0.4335*** 0.4098*** 0.6365*** (0.0506) (0.0361) (0.0315) (0.0209) (0.0243) (0.0144) M/(M+Y)−0.1984 −0.4408** 0.1953 −0.0517 −0.1176 0.1222** (0.1926) (0.1821) (0.1734) (0.1059) (0.1241) (0.0505) X/Y −0.1101 0.0480*** −0.0059 0.0294*** 0.0160 −0.0020 (0.0974) (0.0078) (0.0148) (0.0089) (0.0318) (0.0043) Ln(w) −0.5269*** −0.5007*** −0.4396*** −0.5072*** −0.5031*** −0.6869*** (0.0302) (0.0323) (0.0195) (0.0138) (0.0182) (0.0111) Ln (w)*s 0.0953*** 0.0831*** 0.0457*** 0.0717*** 0.0232*** 0.0953*** (0.0110) (0.0095) (0.0069) (0.0049) (0.0017) (0.0036) Cons an 5.0214*** 4.1416*** 5.4001*** 4.4400*** 4.4876*** 2.3626*** (0.7198) (0.5628) (0.5494) (0.3390) (0.3540) (0.2277) Obse a ions 983 870 1821 3797 2449 5527 Numbe o indus ies 107 144 155 424 337 506 No e: See no es unde Table 4 Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 20 o 36 indus ies and ou indings suppo he iew ha ade wi h de eloping coun ies educe jobs in high-income coun ies (Au o e al. 2013). As shown in Table 6, jobs in medium- echnology indus ies a e pa icula ly suscep ible o impo compe i ion ac oss all egions excep SSA. Unlike he LT indus ies, he employmen - educing e ec s o impo compe i ion in MT indus ies a e a bi s onge o de eloping coun ies as compa ed o CEE and OECD coun ies. Un il e y ecen ly, mo e han 50 % o wo ld expo s o MT p oduc s a e supplied by OECD coun ies show- ing hei compe i i e ad an age. MT indus ies in A ica seem o bene i s om inc eased impo pene a ion, which could be explained by be e access o in e media e inpu s. Jobs in high- echnology indus ies a e by a he leas sensi i e o impo compe i ion ela i e o LT and MT indus ies. The only excep ion is La in Ame ica whe e s a is ically signi i- can employmen con ac ion occu s as a esul o impo pene a ion. On he lip side, he wo egions ha domina e wo ld expo s in HT indus ies, i.e., OECD and ASIA, ha e posi i e employmen elas ici ies wi h espec o impo pene a ion, al hough he coe i- cien o ASIA is no s a is ically signi ican . This sugges s ha g ea e ade openness in high- echnology indus ies could be bene icial o mos coun ies. No all de eloping coun ies bene i om expo -o ien ed low- echnology manu ac u - ing. Table 5 shows ha only MENA and ASIA ha e s a is ically signi ican employmen bene i s om LT expo s. In e es ingly, he coe icien on expo -ou pu a io is posi i e and signi ican o LT indus ies o he OECD sugges ing ha expo o ien a ion helps slowdown he a e o job des uc ion in his sec o . This is consis en wi h he declining employmen sha e and inc easing expo a io o his sec o in he OECD (Fig. 2). Expo o ien a ion in MT indus ies has employmen bene i s only in ASIA and OECD egions sugges ing hese a e he dominan expo e s in his indus y. High- echnology expo s a e no impo an sou ces o employmen gain wi h he excep- ion o LAC whe e he coe icien is posi i e and signi ican . The size o he coe icien on expo a io is qui e low e en when s a is ically signi ican ea i ming he limi ed con i- bu ion o expo -o ien ed manu ac u ing o job c ea ion. Once again, he ac ha ex- po e s a e ela i ely la ge and p oduc i e seems o allow hem o inc ease expo s wi h limi ed job c ea ion. Equally impo an is he obse a ion ha expo o ien a ion is no endange ing exis ing jobs as some o he la es ade heo ies p edic ed. Summing up, low- echnology indus ies s ill seem o ha e signi ican employmen po- en ials o de eloping coun ies. A shi owa d MT manu ac u ing may inc ease p od- uc i i y and alue-added g ow h in de eloping coun ies, as shown ea lie in Fig. 2, bu such a ansi ion is less likely o gene a e mo e employmen because o he capi al- in ensi e na u e o he sec o and i s limi ed expo o ien a ion. F om an employmen pe spec i e, a s uc u al change ha a o s HT o e MT indus ies may se e de eloping coun ies be e as hei human capi al and pe capi a income con inue o ise. This is pa ly because he ou pu sha e o HT indus ies in o al manu ac u ing is s ill e y low in de eloping coun ies and employmen in his sec o is a less suscep ible o impo com- pe i ion. The la e is pe haps an indica ion o he ole o in e media e inpu s as well as g owing FDI lows in his sec o om OECD o de eloping coun ies. 6 Ex ended model In his sec ion, we ex end he labo demand model by conside ing addi ional dimensions o expo ac i i ies. Coun ies and indus ies wi h compa able expo -ou pu a ios may Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 21 o 36 di e in he ange o expo i ems, i.e., he ex ensi e ma gin. Hausmann e al. (2007) a gue ha he disco e y o new expo i ems is bound o be sub-op imal because o sunk cos s ha a e p i a e and he public-good na u e o success ul new expo p oduc s. The au- ho s also show di e si ica ion as an impo an p edic o o bo h he le el and u u e g ow h a e o pe capi a income. In a ela ed li e a u e, Be na d e al. (2010) and Goldbe g e al. (2010) ind ha mul ip oduc i ms a e signi ican ly la ge , mo e p oduc - i e, and mo e expo o ien ed han single p oduc i ms. Whe he di e si ica ion a ec s employmen oppo uni ies, holding expo - a io cons an , is hus wo h explo ing. Ou conjec u e is ha he employmen bene i s o adding a new expo i em may de- pend on he cu en di e si y o he expo baske . Fo a de eloped coun y wi h a b oad ange o manu ac u ed expo i ems, adding one mo e p oduc may no ha e a no iceable employmen e ec as compa ed o a de eloping coun y wi h only a ew expo i ems. We a emp o explo e his ela ionship by including in he labo demand model he o al numbe o six-digi expo i ems (HS codes) wi hin a ou -digi ISIC indus y and com- pa e he ou come o de eloped and de eloping coun ies. Once a new expo i em is in oduced, subsequen employmen g ow h will be d i en by he in ensi e ma gin, which in u n depends pa ly on he compe i i eness o he global ma ke in ha pa icula indus y. P oduc s wi h ela i ely low disco e y cos s a e likely o be expo ed by i ms om mul iple coun ies. Wi h ewe buye s and many supplie s, impo e s may exe cise ma ke powe and use he h ea o swi ching supplie coun ies/ i ms o dic a e he e ms o exchange in hei ad an age. The esul ing unce ain y o de- mand and con ac e ms may educe expo e s’incen i e and abili y o c ea e and e ain jobs. To cap u e his e ec , he ex ended labo demand model includes a measu e o he numbe o ac i e expo e coun ies (CX) o each ou -digi indus y. Finally, cons ain s o job c ea ion may a ise om inadequa e in es men s in machin- e y and equipmen . Since Eq. (5) al eady con ols o alue added, he ocus he e is on he ole o in es men on labo p oduc i i y o domes ic i ms acing compe i i e p es- su e. One example is he educ ion in main enance cos and down ime by eplacing old machines by new ones. The p oduc i i y e ec could also be associa ed wi h he in- c ease in he scale o p oduc ion as in es men in ensi ies. We expec such comple- men a y employmen e ec s o be impo an o coun ies a ea ly s ages o indus ializa ion. Fo o he s, high in es men a es may dampen labo demand as p o- duc ion becomes mo e capi al-in ensi e. The ex ended model includes he in es men a e as one o he explana o y a iables, and i is calcula ed as he pe cen age o manu- ac u ed ou pu dedica ed o in es men pu poses. The ex ended model is exp essed as ollows: ln LðÞ ij ¼δþβln VðÞ ij; −1þαln wðÞ ij; −1þϕX Y 0 @1 Aij; −1 þφM MþY 0 @1 Aij; −1 þ ψln wðÞ ij; −1sij; −1þγln HSðÞ ij; −1þπln CXðÞ ij; −1þij þu þεij ð10Þ The esul s om he ex ended model a e p esen ed in Table 8. Because o he subs an- ial educ ion in he numbe o obse a ions due o missing alues o he new a iables, he ex ended model is i s es ima ed o he en i e manu ac u ing sec o wi hou del ing in o echnology-based indus y ca ego ies. The coe icien s on p e iously discussed co a - ia es emain simila o hose in Table 4 wi h mino di e ences in magni ude and a ew Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 22 o 36 losses o p ecision due o sample size educ ion. The e o e, he emaining discussion o- cuses on he new a iables in he ex ended model. A key obse a ion is he posi i e and s a is ically signi ican coe icien on he numbe o expo i ems (HS) o de eloping coun ies. This is pa icula ly ue in SSA and LAC coun ies and o a ce ain ex en in MENA oo. This sugges s ha a e con olling o expo -ou pu a io, adding new expo i ems has signi ican employmen bene i s o coun ies ha ha e ye o become accomplished expo e s o manu ac u es. No such em- ploymen bene i s accompany expo di e si ica ion in ASIA and OECD coun ies, which cu en ly domina e wo ld expo s o manu ac u es. The ma ginal employmen bene i s om p oduc adding seem o decline wi h he scale o expo ac i i ies. This appea s o esona e wi h he indings o Imbs and Waczia g (2003) ha as coun ies de elop, hei economies become inc easingly less (mo e) concen a ed (di e si ied) un il a h eshold le el o pe capi a income is eached, beyond which specializa ion kicks in. Pa icipa ing in expo ma ke s in which he e a e a la ge numbe o compe i o s does no ha e any no iceable employmen e ec s. None heless, he e is some e idence ha expo ing wha mos o he coun ies expo inc eases job oppo uni ies in SSA. This sug- ges s ha coun ies wi h limi ed echnological capabili ies can ge a oo hold in in e - na ional ma ke s by expo ing i ems wi h low disco e y cos s. Simila ly, he ac ha he coe icien on CX is nega i e o he Asian sample, despi e lacking p ecision, sugges s ha he egion is app oaching a poin whe e indus ial expansion and job c ea ion h ough low- echnology expo s may no longe be a iable op ion. This oppo uni y is abou o Table 8 Ex ended labo demand model by egion: IV-panel ixed e ec s SSA MENA LAC ASIA DVPG CEE OECD Ln (V) 0.3352*** 0.3631*** 0.2503*** 0.4066*** 0.3084*** 0.3115*** 0.6764*** (0.0502) (0.0350) (0.0335) (0.0286) (0.0158) (0.0220) (0.0139) M/(M+Y) 0.0784 −0.6395*** −0.6473*** −0.2628** −0.3702*** −0.4983*** 0.0200 (0.1441) (0.1099) (0.1161) (0.1159) (0.0599) (0.0836) (0.0494) X/Y 0.0002 0.0338 0.0430*** −0.0012 0.0188*** 0.0068 0.0126*** (0.0308) (0.0335) (0.0088) (0.0084) (0.0060) (0.0102) (0.0033) Ln (w)−0.4880*** −0.4128*** −0.3239*** −0.4079*** −0.3674*** −0.2919*** −0.6771*** (0.0344) (0.0240) (0.0253) (0.0245) (0.0124) (0.0194) (0.0122) Ln (w)*s 0.0731*** 0.0209*** 0.0747*** 0.1174*** 0.0378*** 0.0203*** 0.0989*** (0.0107) (0.0033) (0.0087) (0.0102) (0.0027) (0.0017) (0.0032) Ln (HS) 1.0269*** 0.1124 b 2.2233*** 0.2012 0.2922*** 0.7878*** 0.1813 d (0.2831) (0.0762) (0.4782) (0.3500) (0.0565) (0.1338) (0.1194) Ln (CX) 0.4071 a 0.0289 0.1368 −0.1497 0.0705 0.1752 0.0307 (0.2527) (0.1928) (0.1635) (0.1641) (0.0964) (0.1902) (0.0542) I/Y 0.4089** 0.0925** −0.0805 0.0234 0.0315 c −0.0067 0.0810 b (0.1927) (0.0418) (0.0641) (0.0246) (0.0201) (0.0189) (0.0547) Cons an −2.9992 4.1023*** −10.0472*** 4.0980* 3.5978*** −1.3813 0.1106 (2.1329) (1.1365) (3.3685) (2.3353) (0.6381) (1.3317) (0.9039) Obse a ions 534 1104 1322 1551 4511 2211 6357 Numbe o indus ies 57 131 201 148 537 352 716 No e: See no es unde Table 4.HS is he numbe o six-digi expo i ems in a 4-digi indus y, CX is he numbe o compe ing coun ies, and I/Y is he in es men a e. Supe sc ip s a,b,c, and d ep esen s a is ical signi icance a 11, 13, 14, and 12 %, espec i ely Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 23 o 36 a ail i sel o low-cos coun ies in A ica. The o he in e es ing esul in Table 8 is he posi i e and signi ican ela ionship be ween in es men and employmen in de eloping coun ies as a g oup wi h he excep ion o LAC. I is s iking ha he employmen e ec s o in es men a e la ge and highly signi ican o manu ac u ing indus ies in SSA. In Table 9, we allow he coe icien s o he new a iables o a y ac oss echnology- based indus y ca ego ies. Medium- and high- echnology indus ies o de eloping coun- ies a e me ged oge he o economize on obse a ions. The esul s show ha labo de- mand in de eloping coun ies inc eases signi ican ly as new expo i ems (HS) a e added o bo h LT and he combined MT/HT indus ies. Howe e , he employmen bene i s a e subs an ially la ge in he MT/HT indus ies e lec ing pe haps g ea e possibili ies o di- e si ica ion in mo e ad anced indus ies. This is consis en wi h Hausmann e al. (2007) whe e g ow h in pe capi a income inc eases wi h he in oduc ion o new p oduc s in echnologically ad anced indus ies. P oduc adding also helps de eloped coun ies c ea e manu ac u ing jobs bu only in he HT indus ies; e en hen, he coe icien on HS is ba ely signi ican . The coe icien on CX ha p oxies he compe i i eness o expo ma - ke s a he indus y le el is nega i e o he mos pa as would be expec ed bu s a is i- cally insigni ican . Fo LT indus ies in de eloping coun ies, howe e , an inc ease in he numbe o compe ing coun ies seems o inc ease employmen . While in es men inc eases job c ea ion in de eloping coun ies as shown in Table 8, only in LT indus ies does his complemen a i y u n ou o be s a is ically signi ican . This seems o be consis en wi h he obse ed inc ease in he employmen sha e o LT Table 9 Ex ended labo demand model by egion and indus y g oup: IV-panel ixed e ec s De eloping coun ies OECD coun ies Low- ech Med/high- ech Low- ech Med- ech High- ech Ln (V) 0.4288*** 0.2261*** 0.7620*** 0.6067*** 0.5924*** (0.0214) (0.0247) (0.0208) (0.0237) (0.0426) M/(M+Y)−0.2558*** −0.4861*** −0.3644*** −0.4502*** 0.2714*** (0.0811) (0.0859) (0.0877) (0.1002) (0.0898) X/Y 0.0347*** 0.0172** 0.1199*** 0.1186*** −0.0025 (0.0132) (0.0069) (0.0102) (0.0225) (0.0044) Ln (w)−0.4815*** −0.3286*** −0.7149*** −0.6477*** −0.6969*** (0.0181) (0.0178) (0.0200) (0.0214) (0.0322) Ln (w)*s0.1161*** 0.0231*** 0.1374*** 0.0945*** 0.0772*** (0.0066) (0.0031) (0.0061) (0.0053) (0.0074) Ln (HS) 0.1911*** 0.3515*** 0.0403 −0.1885 0.6140 a (0.0720) (0.0864) (0.1689) (0.1675) (0.3980) Ln (CX) 0.2196* −0.1575 −0.0085 0.0970 −0.1306 (0.1235) (0.1465) (0.1118) (0.0679) (0.1206) I/Y 0.0433* 0.0297 −0.0684 −0.1055 0.2332** (0.0249) (0.0317) (0.1273) (0.0831) (0.0995) Cons an 2.1329*** 5.4557*** −0.1446 3.7549*** −0.5908 (0.8238) (0.9565) (1.3057) (1.2790) (3.0254) Obse a ions 2430 2081 2811 2234 994 Numbe o indus ies 287 250 314 254 114 No e: See no es unde Table 4. HS is he numbe o six-digi expo i ems in a ou -digi indus y, CX is he numbe o compe ing coun ies, and I/Y is he in es men a e. Supe sc ip s a ep esen s s a is ical signi ican a 13 % Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 24 o 36 Table 13 Regional compa ison o labo demand models: high- echnology in e ac ions Compa ison wi h ASIA Compa ison wi h OECD ASIA-MENA ASIA-LAC OECD-ASIA OECD-MENA OECD-LAC Ln (V) 0.5958 0.3924 0.4011 0.6880 0.4369 (0.0356)*** (0.0529)*** (0.0205)*** (0.0220)*** (0.0362)*** M/(M+Y)−0.1863 −0.4486 0.2039 −0.1864 −0.3298 (0.1886) (0.2695)* (0.1278) (0.1300) (0.1859)* X/Y −0.1293 0.0825 −0.0057 −0.1251 0.0814 (0.0953) (0.0706) (0.0110) (0.0657)* (0.0487)* Ln (w)−0.6018 −0.4708 −0.4437 −0.6356 −0.4827 (0.0271)*** (0.0485)*** (0.0142)*** (0.0183)*** (0.0334)*** Ln (w)*s 0.1207 0.0720 0.0488 0.1300 0.0816 (0.0101)*** (0.0140)*** (0.0051)*** (0.0069)*** (0.0097)*** *ln (V)−0.2587 −0.0296 0.2277 −0.0771 0.1865 (0.0434)*** (0.0592) (0.0257)*** (0.0265)*** (0.0389)*** *M/(M+Y) 0.3819 0.6262 −0.0517 0.3236 −0.0824 (0.2601) (0.3116)** (0.1430) (0.1440)** (0.0492)* *X/Y 0.1283 −0.0842 0.0040 0.1243 0.4767 (0.0966) (0.0719) (0.0138) (0.0662)* (0.1936)** *ln (w) 0.1534 0.0303 −0.2416 −0.0446 −0.1990 (0.0338)*** (0.0515) (0.0195)*** (0.0224)** (0.0353)*** *ln (w)*s −0.0767 −0.0266 0.0459 −0.0370 0.0121 (0.0122)*** (0.0153)* (0.0067)*** (0.0080)*** (0.0103) _cons 4.7270 5.3752 3.2307 2.5950 2.8839 (0.4310)*** (0.4209)*** (0.2288)*** (0.2335)*** (0.2131)*** N2800 2682 7327 6493 6375 No e: See Appendix A: Table 10 o no es Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 31 o 36 Table 14 Fi s S age Reg ession Resul s o ASIA Co esponding o Table 4 Value added Wage a e Impo pene a ion Expo -ou pu EXI ij −0.2344 0.1321 0.0771 0.0329 0.0468 0.0376 0.0075 0.0254 PPI ij −0.0024 −0.0073 −0.0001 −0.0001 0.0006 0.0005 0.0001 0.0003 REXIUS ij −0.3890 −0.8549 −0.0297 −0.0513 0.0175 0.0140 0.0028 0.0095 REXIUS ij -squa ed 0.0057 0.0048 0.0007 −0.0003 0.0012 0.0010 0.0002 0.0007 GDPI ij −0.0073 0.0012 −0.0007 −0.0009 0.0008 0.0006 0.0001 0.0004 Ln (w)- wo-digi SIC 0.2787 0.0824 −0.0005 −0.0189 0.0181 0.0145 0.0029 0.0098 GDPI-domes ic 0.0133 0.0117 0.0006 0.0008 0.0007 0.0006 0.0001 0.0004 F s a is ic 108.7 716.34 18.81 8.46 No e: EXI ij is a weigh ed a e age exchange a e index o impo supplie coun ies, PPI ij is a weigh ed a e age indus y- speci ic p ice index o impo sou ce coun ies, REXIUS ij is weigh ed a e age eal exchange a e index o expo des ina ion coun ies, GDPI ij is he agg ega e GDP index o expo des ina ion coun ies applicable o indus y jo an expo ing coun y, ln (w)- wo-digi SIC is he log o wage a e a he wo-digi SIC le el, and GDPI-domes ic is GDP index a he coun y le el. Robus s anda d e o s a e in pa en hesis Table 15 Fi s s age eg ession esul s o SSA co esponding o Table 4 Value added Wage a e Impo pene a ion Expo -ou pu EXI ij 0.3426 0.0741 −0.0026 −0.0182 (0.0500) (0.0676) (0.0166) (0.0394) PPI ij −0.0058 −0.0073 −0.0006 −0.0002 (0.0011) (0.0015) (0.0004) (0.0008) REXIUS ij −0.1949 −0.7543 −0.0315 −0.0574 (0.0272) (0.0368) (0.0090) (0.0214) REXIUS ij -squa ed −0.0301 0.0044 0.0008 0.0010 (0.0039) (0.0053) (0.0013) (0.0031) GDPI ij −0.0028 0.0004 0.0014 −0.0005 (0.0010) (0.0014) (0.0003) (0.0008) Ln (w)- wo-digi SIC 0.0779 −0.1010 0.0346 −0.0061 (0.0371) (0.0502) (0.0123) (0.0292) GDPI-domes ic 0.0098 −0.0082 0.0013 0.0013 (0.0011) (0.0015) (0.0004) (0.0009) F s a is ic 53.76 89.23 17.98 2.09 No e: See no es o Appendix B: Table 14 Appendix B: Fi s s age eg ession esul s Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 32 o 36 Table 16 Fi s s age eg ession esul s o LAC co esponding o Table 4 Value added Wage a e Impo pene a ion Expo -ou pu EXI ij −0.0815 −0.0554 −0.0028 0.0362 0.0334 0.0330 0.0085 0.0260 PPI ij −0.0002 −0.0086 −0.0003 −0.0005 0.0005 0.0005 0.0001 0.0004 REXIUS ij −0.0780 −0.6938 −0.0640 −0.0523 0.0124 0.0123 0.0031 0.0097 REXIUS ij -squa ed −0.0177 −0.0154 0.0030 0.0024 0.0017 0.0017 0.0004 0.0013 GDPI ij −0.0016 −0.0011 0.0003 0.0017 0.0010 0.0009 0.0002 0.0007 Ln (w)- wo-digi SIC −0.1089 −0.0095 −0.0149 −0.0138 0.0242 0.0239 0.0061 0.0188 GDPI-domes ic 0.0210 0.0378 0.0011 0.0025 0.0022 0.0022 0.0006 0.0017 F s a is ic 53.98 293.27 31.95 3.98 No e: See no es o Appendix B: Table 14 Table 17 Fi s s age eg ession esul s o OECD co esponding o Table 4 Value added Wage a e Impo pene a ion Expo -ou pu EXI ij −0.1835 −0.0197 0.0155 0.0443 0.0195 0.0142 0.0066 0.0206 PPI ij −0.0056 −0.0118 0.0000 −0.0005 0.0002 0.0002 0.0001 0.0002 REXIUS ij −0.3381 −0.8440 −0.0367 −0.1030 0.0054 0.0039 0.0018 0.0057 REXIUS ij -squa ed 0.0063 0.0193 0.0011 0.0040 0.0010 0.0007 0.0003 0.0010 GDPI ij −0.0057 −0.0009 0.0005 0.0005 0.0003 0.0002 0.0001 0.0004 Ln (w)- wo-digi SIC 0.0842 0.0272 −0.0197 −0.0333 0.0079 0.0058 0.0027 0.0084 GDPI-domes ic 0.0140 0.0140 0.0004 −0.0003 0.0003 0.0002 0.0001 0.0004 F s a is ic 366.11 2974 85.45 40.67 No e: See no es o Appendix B: Table 14 Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 33 o 36 Table 18 Fi s s age eg ession esul s o CEE co esponding o Table 4 Value added Wage a e Impo pene a ion Expo -ou pu EXI ij 0.1079 0.2663 0.0136 −0.0074 0.0314 0.0285 0.0060 0.0239 PPI ij −0.0016 −0.0045 −0.0003 −0.0006 0.0004 0.0004 0.0001 0.0003 REXIUS ij −0.2813 −0.6260 −0.0306 −0.0670 0.0066 0.0060 0.0013 0.0050 REXIUS ij -squa ed −0.0228 −0.0454 −0.0037 −0.0058 0.0020 0.0018 0.0004 0.0015 GDPI ij 0.0016 0.0039 0.0005 0.0000 0.0007 0.0007 0.0001 0.0005 Ln (w)- wo-digi SIC 0.0521 0.0043 −0.0122 −0.0223 0.0171 0.0155 0.0033 0.0130 GDPI-domes ic 0.0047 0.0060 −0.0001 −0.0002 0.0003 0.0003 0.0001 0.0002 F s a is ic 210.77 774.19 36.84 10.7 No e: See no es o Appendix B: Table 14 Table 19 Fi s s age eg ession esul s o MENA co esponding o Table 4 Value added Wage a e Impo pene a ion Expo -ou pu EXI ij −0.1983 −0.0764 0.0321 0.0317 0.0329 0.0311 0.0076 0.0181 PPI ij −0.0028 −0.0079 −0.0002 −0.0007 0.0006 0.0006 0.0001 0.0003 REXIUS ij −0.2730 −0.7793 −0.0182 −0.0407 0.0106 0.0101 0.0024 0.0059 REXIUS ij -squa ed 0.0074 0.0231 0.0007 0.0011 0.0008 0.0008 0.0002 0.0004 GDPI ij 0.0026 0.0026 0.0001 −0.0004 0.0009 0.0009 0.0002 0.0005 Ln (w)- wo-digi SIC 0.0452 0.0030 −0.0111 0.0080 0.0254 0.0240 0.0058 0.0140 GDPI-domes ic 0.0007 −0.0002 0.0006 0.0027 0.0013 0.0013 0.0003 0.0007 F s a is ic 154.17 456.14 11.36 4.73 No e: See no es o Appendix B: Table 14 Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 34 o 36 Appendix C Compe ing in e es s The IZA Jou nal o Labo & De elopmen is commi ed o he IZA Guiding P inciples o Resea ch In eg i y. The au ho s decla e ha hey ha e obse ed hese p inciples. Acknowledgemen s We hank Be hanu Abegaz and Pe e McHen y o hei commen s and sugges ions. We a e also g a e ul o he impo an commen s p o ided by a e e ee and he edi o o IZA-JOLD ha helped us imp o e he pape . Responsible edi o : Da id Lam Au ho de ails 1 The College o William and Ma y Williamsbu g, Vi ginia, USA. 2 UNDP (Uni ed Na ion De elopmen P og amme), New Yo k, USA. Recei ed: 25 Augus 2015 Accep ed: 15 Janua y 2016 Re e ences A anasio O, Goldbe g PK, Pa cnik N (2004) T ade e o ms and wage inequali y in Colombia. J De Econ 74:331–366 Au o D, D on D, Hanson G (2013) The China synd ome: local labo ma ke e ec s o impo compe i ion in he Uni ed S a es. Am Econ Re 103(6):2121–68 Be na d B, Jensen JB (1999) Excep ional expo pe o mance: cause, e ec s, o bo h? J In Econ 47:1,1–25 Be na d A, Ea on J, Jensen JB, Ko um S (2003) Plan s and p oduc i i y in in e na ional ade. Am Econ Re 94:1265–1290 Be na d BA, Redding SJ, Scho PK (2010) Mul iple-p oduc i ms and p oduc swi ching. Am Econ Re 100(1):70–97 Bhalo a S (1998) The puzzle o jobless g ow h in Indian manu ac u ing. Ox Bull Econ S a 60:1,5–32 Bo e o CJ, Djanko S, La Po a R, Lopez-De-Silanes F, Shleie e A (2004) The egula ion o labo . Q J Econ 119(4):1339–1382 Caballe o R, Hammou M (1997) Jobless g ow h: app op iabili y, ac o -subs i u ion and unemploymen , NBER Wo king Pape No. 6221, Na ional Bu eau o Economic Resea ch. Dau h W, Suedekum J, Findeisen S (2014) The ise o he Eas and he Fa Eas : Ge man labo ma ke s and ade in eg a ion. J Eu Econ Assoc 12(6):1643–75 Da idson C, Ma usz SJ, She chenko A (2008) Globaliza ion and i m le el adjus men wi h impe ec labo ma ke s. J In Econ 75:295–309 Da is D, Ha igan J (2011) Good jobs, bad jobs, and ade libe aliza ion. J In Econ 84:26–36 Table 20 Lis o sample coun ies by egion Asia and Paci ic Cen al and Eas e n Eu ope Middle Eas and No h A ica Sub-Saha an A ica La in Ame ica and Ca ibbean OECD China Albania Egyp Bo swana A gen ina Aus alia Fiji Aze baijan I an E i ea Bahamas Aus ia Hong Kong Bela us Jo dan E hiopia Boli ia Belgium Indonesia Bulga ia Kuwai Madagasca B azil Denma k Sou h Ko ea Cyp us Mo occo Malawi Chile Finland Macao, China Es onia Oman Mau i ius Colombia Ge many Malaysia Geo gia Qa a Senegal Ecuado G eece Mongolia Ky gyz Republic Is ael Sou h A ica Mexico Hunga y Philippines La ia Pa aguay I eland Singapo e Li huania Pe u Japan S i Lanka Moldo a Luxembou g Poland Mal a Romania Ne he lands Russian Fede a ion New Zealand Slo ak Republic Po ugal Slo enia Spain Sweden Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 35 o 36 Du P, Mi a D, Ranjan P (2009) In e na ional ade and unemploymen : heo y and c oss-na ional e idence. J In Econ 78:32–44 Egge H, K eickemeie U (2009) Fi m he e ogenei y and he labo ma ke e ec s o ade libe aliza ion. In Econ Re 50(1):187–216 Fallon P, Lucas R (1993) Job secu i y egula ions and he dynamic demand o indus ial labo in India and Zimbabwe. J De Econ 40:241–275 Felbe may G, P a J, Schme e H-J (2011a) Globaliza ion and labo ma ke ou comes: wage ba gaining, sea ch ic ions, and i m he e ogenei y. J Econ Theo y 146:39–73 Felbe may G, P a J, Schme e H-J (2011b) T ade and unemploymen : wha do he da a say. Eu Econ Re 55:741–758 F eeman R (2010) Labo egula ions, unions, and social p o ec ion in de eloping coun ies: ma ke dis o ions o e icien ins i u ions? ol 5, Handbook o de elopmen economics. Else ie , Ams e dam, pp 4657–4702 Goldbe g P, Pa cnik N (2003) The esponse o he in o mal sec o o ade libe aliza ion. J De Econ 72:463–496 Goldbe g P, Pa cnik N (2007) Dis ibu ional e ec s o globaliza ion in de eloping coun ies. J Econ Li XLV:39–82 Goldbe g KP, Khandelwall AK, Pa cnik N, Topalo a P (2010) Mul ip oduc i ms and p oduc u no e in he de eloping wo ld: e idence om India. Re Econ S a 92(4):1042–1049 G eenaway D, Hine R, W igh P (1999) An empi ical assessmen o he impac o ade on employmen in he Uni ed Kingdom. Eu J Poli Econ 15:485–500 Hal iwange J, Sca pe a S, Schweige H (2008) Assessing job lows ac oss coun ies: he ole o indus y, i m size and egula ions, NBER Wo king Pape No. 13920, Na ional Bu eau o Economic Resea ch. Hame mesh DS (1993) Labo demand. P ince on Uni e si y P ess, P ince on Hasan R, Mi a D, Ramaswamy KV (2007) T ade e o ms, labo egula ions, and labo -demand elas ici ies: empi ical e idence om India. Re Econ S a 89(3):466–481 Hasan R, Mi a D, Ranjan P, Ahsan R (2012) T ade libe aliza ion and unemploymen : heo y and e idence om India. J De Econ 97:269–280 Ha zich onoglou T (1997) Re ision o he high- echnology sec o and p oduc classi ica ion. OECD Science, Technology and Indus y Wo king Pape 1997/02, Pa is Hausmann R, Hwang J, Rod ik D (2007) Wha you expo ma e s. J Econ G ow h 12:1,1–25 Helpman E, I skhoki O, Redding S (2010) Inequali y and unemploymen in a global economy. Econome ica 78(4):1239–1283 He endo B, Roge son R, Valen inyi A (2013) G ow h and s uc u al ans o ma ion, NBER Wo king Pape No. 18996, Na ional Bu eau o Economic Resea ch. Imbs J, Waczia g R (2003) S ages o di e si ica ion. Am Econ Re 93:1,63–86 İȘcan T (1998) T ade libe aliza ion and p oduc i i y: a panel s udy o he Mexican manu ac u ing indus y. J De S ud 34(5):123–148 James B, and Olken B (2005) The ana omy o s a -s op g ow h, NBER Wo king Pape No. 11528, Na ional Bu eau o Economic Resea ch. Jenkins R (2008) T ade, echnology and employmen in Sou h A ica. J De S ud 44:1,60–79 Jonsson G, Sub amanian A (2001) Dynamic gains om ade: e idence om Sou h A ica. IMF S a Pap 48(1):197–224 Kapsos S (2005) The employmen in ensi y o g ow h: ends and mac oeconomic de e minan s, Employmen S a egy Pape s 2005/12, In e na ional Labo O ganiza ion. Kim E (2000) T ade libe aliza ion and p oduc i i y g ow h in Ko ean manu ac u ing indus ies: p ices, p o ec ion, ma ke powe , and scale e iciency. J De Econ 62:55–83 Meli z M (2003) The impac o ade on in a-indus y ealloca ion and agg ega e indus y p oduc i i y. Econome ica 71(6):1695–1725 Menezes-Filho NA, Muendle M (2011) Labo ealloca ion in esponse o ade e o m, NBER Wo king Pape No. 17372, Na ional Bu eau o Economic Resea ch. Pa cnik N (2002) T ade libe aliza ion, exi , and p oduc i i y imp o emen s: e idence om Chilean plan s. Re Econ S ud 69:245–276 Re enga A (1992) Expo ing jobs?: he impac o impo compe i ion on employmen and wages in US manu ac u ing. Q J Econ 107(1):255–284 Re enga A (1997) Employmen and wage e ec s o ade libe aliza ion: he case o Mexican manu ac u ing. J Labo Econ 15(S3):S20–S43 Rod ik D (1997) Has globaliza ion gone oo a ? Ins i u e o In e na ional Economics, Washing on, DC Sachs DJ, Sha z HJ, Dea do A, Hall RE (1994) T ade and jobs in U.S. manu ac u ing. B ook Pap Econ Ac 1994(1):1–84 Scho A (2004) He e ogeneous p oduc i i y eponse o a i educ ion. E idence om B azilian manu ac u ing i ms. J De Econ 75:373–396 Slaugh e M (2001) In e na ional ade and labo -demand elas ici ies. J In Econ 54:27–56 Söde bom M, Teal F (2000) Skills, in es men and expo s om manu ac u ing i ms in A ica. J De S ud 37(2):13–43 Waczia g R, Wallack JS (2004) T ade libe aliza ion and in e sec o al labo mo emen s. J In Econ 64:411–439 Wo ld Bank (2012) Jobs, Wo ld De elopmen Repo 2013. In e na ional Bank o Recons uc ion and De elopmen , Washing on, DC Shi e aw and Hailu IZA Jou nal o Labo & De elopmen (2016) 5:3 Page 36 o 36