INTERNATIONALIZATION OF FIRMS AND SKILL STRUCTURE

Transcription

1 UNVERSTY OF LJUBLJANA FACULTY OF ECONOMCS MOJCA LNDČ NTERNATONALZATON OF FRMS AND SKLL STRUCTURE DOCTORAL DSSERTATON Ljubjana, 206

2

3 AUTHORSHP STATEMENT Te undersigned Mojca Lindič, a student at te University of Ljubjana, Facuty of Economics, (ereafter: FELU), decare tat am te autor of te doctora dissertation entited nternationaization of firms and ski structure, written under supervision of Fu Professor Jože Damijan, PD. n accordance wit te Copyrigt and Reated Rigts Act (Officia Gazette of te Repubic of Sovenia, Nr. 2/995 wit canges and amendments) aow te text of my doctora dissertation to be pubised on te FELU website. furter decare te text of my doctora dissertation to be based on te resuts of my own researc; te text of my doctora dissertation to be anguage-edited and tecnicay in aderence wit te FELU s Tecnica Guideines for Written Works wic means tat o cited and / or quoted works and opinions of oter autors in my doctora dissertation in accordance wit te FELU s Tecnica Guideines for Written Works and o obtained (and referred to in my doctora dissertation) a te necessary permits to use te works of oter autors wic are entirey (in written or grapica form) used in my text; to be aware of te fact tat pagiarism (in written or grapica form) is a crimina offence and can be prosecuted in accordance wit te Copyrigt and Reated Rigts Act (Officia Gazette of te Repubic of Sovenia, Nr. 55/2008 wit canges and amendments); to be aware of te consequences a proven pagiarism carge based on te submitted doctora dissertation coud ave for my status at te FELU in accordance wit te reevant FELU Rues on Doctora Dissertation. Date of pubic defence: 5 Apri 206 Committee Cair: prof. dr. Črt Kostevc Supervisor: prof. dr. Jože Damijan Member: prof. dr. Katja Zajc Kejžar Member: Professor Jozef Konings Ljubjana, 5 Apri 206 Autor s signature:

4

5 Acknowedgements Tis doctora dissertation woud not be possibe to compete witout te ep of many peope. Foremost, woud ike to express my gratitude to my mentor, prof. dr. Jože Damijan, for is guidance, vauabe comments and suggestions, wic benefited tremendousy during te course of writing tis doctora dissertation and during doctora studies in genera. woud ike to extend my gratefuness to te committee members, prof. dr. Katja Zajc Kejžar, Professor Jozef Konings, PD, and prof. dr. Črt Kostevc, wose recommendations for improvements and constructive feedback improved tis dissertation substantiay. n addition, am indebted to prof. dr. Sašo Poanec, wose ep as been invauabe. Moreover, gratefuy acknowedge caenging questions and suggestions for improvements, received from feow students, referees and conference discussants. woud aso ike to tank my current and former co-workers at te Bank of Sovenia, Facuty of Economics and UMAR for teir support and insigtfu discussions, wic contributed to tis dissertation. Many tanks aso to organizations and individuas, wo enabed an access to te data used in tis dissertation; te Statistica office of te Repubic of Sovenia, te Tax Autorities of Sovenia, te Bank of Sovenia, and te Agency of te Repubic of Sovenia for Pubic Lega Records and Reated Services. Finay and incomparaby, woud ike to express my sincere gratitude to Matej, my parents, Luka and is famiy, my inimitabe friends and te rest of my famiy, wo ave been te irrepaceabe cornerstones on my doctora study s pat. consider mysef extremey fortunate to be surrounded by so many individuas, wo supported and encouraged me unconditionay during te times of ardsips. dedicate tis doctora dissertation to you and to my itte anges.

6

7 NTERNATONALZATON OF FRMS AND SKLL STRUCTURE SUMMARY Te motivation for writing tis doctora dissertation is among oters derived from statistica findings and forecasts on increasing internationa activities of firms, and te increasing empoyment of tertiary educated empoyees. Te main aim of te dissertation is to study te inkages and causaities between internationa trade and te ski structure of firms. Te dissertation consists of tree parts. Te first part studies te mecanisms of ski upgrading in trading firms by deveoping a teoretica mode tat reates te individua s incentives for acquiring iger skis to te profit-maximizing beaviour of trading firms. Taking into account te beaviour of individuas and teir decisions for ski upgrading, te first part of te mode sows tat ony te ig abiity individuas ave incentives for acquiring iger skis, as ong as tey are ater compensated wit iger wages after entering empoyment. Tese findings are incorporated in te second part of te mode, wic studies te beaviour of firms, in te way tat iger wage eves signa a iger empoyment of ig abiity, ig skied empoyees. Te second part of te mode concudes tat ig-productive firms ave incentives for investing in iger tecnoogy, to empoy ig-skied abour, and to engage in internationa trade. Te decisions for tecnoogy dress-up and ski upgrading coincide wit te firm s decision to start importing and exporting as te atter requires iger tecnoogy and ig-skied abour. Te second part of te dissertation confirms some of te key impications of te teoretica mode, by empiricay studying te inkages between firms ski structure, importing and exporting. Te anaysis empasises te importance of importing, serving as a prerequisite for increasing te tecnoogy eve troug importing intermediate goods and/or tecnoogy before te start of exporting. n order to anayse tese inkages, a ric empoyer-empoyee pane dataset for Sovenian manufacturing firms for te period was used. Te dataset encompasses severa databases, covering te information on te baance seet data and te income statements of firms, teir import and export activities, and te caracteristics of empoyees. Te propensity score matcing and severa matcing tecniques were appied in order to increase te robustness of resuts. Te resuts sow tat firms wit a better ski structure sef-seect into importing and ater aso sustain a iger sare of skied empoyees in te first and second year after te start of importing, compared to non-importing firms. Aso, import starters furter increase teir ski sare in te second year after te start of importing. Studying te impact of importing on te start of exporting sows tat te start of importing intermediate goods positivey infuences te start of exporting in te subsequent year after te start of importing, wereas te impact in te second year is uncommon. n contrast, importing capita goods as a positive impact on te start of exporting ony in te second year after te start of importing. Terefore, te

8 resuts point toward amost an immediate impact of importing intermediate goods on te start of exporting, wie te impact of capita goods takes onger to sow effect. Te fina part of te dissertation studies te effect of outsourcing and offsoring on te ski structure of firms. Te study anayses weter controing for bot activities in one mode aters te previous empirica studies, wic controed ony for one factor; weter controing for a destination country of outsourcing and offsoring brings new insigts; and weter controing for te occupationa eve of workers brings additiona contributions. Specificay, besides te conventiona approac for defining skis, i.e. te educationa eve, skis are aso defined by tree major occupationa groups; Managers, Professionas and Tecnicians. To estimate te abovementioned ypoteses, a matced empoyerempoyee dataset for Sovenian manufacturing and service firms between 997 and 200, and te metods for pane data anaysis were used (i.e. pooed OLS, fixed effects, and random effects). Te resuts of te mode on average sow a positive impact of offsoring on te ski sare of firms, wie te impact of outsourcing is uncommon. Wen controing for ig- and ow-income countries, te resuts for manufacturing firms sow a positive and simiar impact on te sare of skied empoyees for offsoring to bot groups of countries. n service firms, resuts sow a weaker impact of offsoring to ig-income countries on te reative empoyment of skied empoyees, compared to offsoring to owincome countries. Wen taking into account aso te occupationa eves for defining skis, offsoring to ig-income countries sows a stronger impact on te reative empoyment of Professionas in manufacturing firms, as compared to offsoring to ow-income countries. Wie te resuts for manufacturing firms do not vary significanty wen using different definitions of skied abour, tis does not od for te resuts of service firms. Wen defining skis ony by te occupationa eve, offsoring to ig-income countries sows a weaker impact on te reative empoyment of Tecnicians in comparison to te impact of offsoring to ow-income countries. Wen defining skis by occupationa and educationa eve, offsoring to ig- and ow-income countries sows a positive and simiar impact on te reative empoyment of tertiary educated Managers. Te impact of education terefore differs between occupationa groups, indicating tat firms differentiate between more and ess educated individuas witin te same occupationa group. Tis doctora dissertation contributes to te fied of knowedge in severa ways. First, besides anaysing te beaviour of eterogeneous firms, te teoretica mode aso studies te beaviour of individuas and teir decision for ski upgrading. Tis is important since individuas decisions ater infuence te abour demand of firms. Second, wen studying te sequencing between importing, ski upgrading and exporting, different types of importing were taken into account to furter deepen te expanatory power of te anaysis. Finay, te anaysis introduces a nove approac for defining skis, by controing aso for te occupationa eve of workers. Tis is important since individuas do not obtain skis ony wie acquiring education, but aso during te course of empoyment. Keywords: importing, exporting, ski structure of firms, offsoring, outsourcing.

9 NTERNACONALZACJA PODJETJ TER STRUKTURA ZNANJA ZAPOSLENH POVZETEK Motivacija za pisanje doktorske disertacije izaja iz statistični podatkov za zadnje obdobje, ki z izjemo obdobja gospodarske recesije kažejo rast uvoza in izvoza ter rast deežev in stopenj zaposenosti terciarno izobraženi. Namen doktorske disertacije je anaiza povezav in vzročnosti med mednarodno trgovino in strukturo zaposeni. Prvi de doktorske disertacije v teoretičnem modeu anaizira strukturo znanja zaposeni v mednarodno deujoči podjetji. Mode najprej anaizira posameznikove odočitve gede nadgrajevanja znanja, kjer rezutati kažejo, da se za dodatno izobraževanje odočijo e najboj sposobni posamezniki, če so za to kasneje v času zapositve kompenzirani z višjim pačiom. Ti zakjučki so kasneje vkjučeni v drugi de modea, ki anaizira odočitve podjetij, in sicer v predpostavki, da višji stroški pač signaizirajo višje stopnje zaposenosti visoko usposobjeni in visoko izobraženi posameznikov. Zakjučki modea kažejo pozitivne učinke na visoko produktivna podjetja, če ta vagajo v visoko tenoogijo, če izbojšajo strukturo zaposeni in če pričnejo z mednarodno aktivnostjo. Odočitev podjetja za izbojšanje tenoogije se skada z odočitvijo za začetek uvažanja in izvažanja, saj mednarodne aktivnosti podjetij zatevajo višjo raven tenoogije in bojšo strukturo znanja zaposeni. zsedki teoretičnega modea so bii v nadajevanju anaizirani v empiričnem modeu, katerega namen je bi predvsem preverba, ai se podjetja z bojšo izobrazbeno strukturo zaposeni odočijo za začetek uvažanja; ai imajo uvozniki bojšo izobrazbeno strukturo kot podjetja, ki ne uvažajo; ter ai uvoz preko dostopa do vmesni proizvodov in/ai do tenoogije suži kot sredstvo povišanja ravni tenoogije podjetja pred začetkom izvažanja. Za potrebe anaize so bii uporabjeni paneni podatki na ravni podjetij in zaposeni v predeovani dejavnosti v Soveniji v obdobju od eta 996 do 200. Podatki združujejo baze z informacijami o bianca stanja in izkazi uspea podjetij, uvozu in izvozu, ter astnosti zaposeni. Zgoraj omenjene ipoteze so bie anaizirane z metodo propensity score matcing, z namenom večje robustnosti rezutatov pa je bio uporabjeni več razičic matcing tenik. Rezutati kažejo, da začnejo uvažati podjetja z reativno bojšo strukturo znanja zaposeni ter da podjetja po začetku uvažanja tudi oranijo bojšo strukturo znanja zaposeni v primerjavi s podjetji, ki ne uvažajo. Poeg tega podjetja v drugem etu po začetku uvažanja še izbojšajo strukturo znanja zaposeni. V nadajevanju sem z anaizo proučevaa vpiv uvažanja na začetek izvažanja. Rezutati nakazujejo, da je eno eto po začetku uvažanja vmesni proizvodov vpiv na začetek izvažanja pozitiven, medtem ko je vpiv v drugem etu neznačien. Po drugi strani pa ima začetek uvažanja kapitaski dobrin pozitiven vpiv na začetek izvažanja šee v drugem etu po začetku uvažanja. Ti rezutati nakazujejo na skoraj takojšen vpiv začetka uvažanja

10 vmesni dobrin na začetek izvažanja, medtem ko se vpiv začetka uvažanja kapitaski dobrin zgodi z zamikom. Motivacija za pisanje tretje študije, ki proučuje vpiv outsourcinga in offsoringa na strukturo znanja v podjetji, izvira iz žeje po doprinosu k obstoječim raziskavam preko vkjučitve obe mer offsoringa in outsourcinga v en mode. Anaiza med drugim kontroira tudi za visoko in nizko razvite države, definicija znanja posameznika pa vkjučuje tudi vrsto pokica posameznika. Tri gavne skupine pokicev definirajo posameznike z višjo stopnjo znanja: vodje (Managers), profesionaci (Professionas), in teniki (Tecnicians). Za potrebe anaize so bii ponovno uporabjeni sovenski paneni podatki na ravni zaposeni in podjetij. Anaiza je obravnavaa obdobje med eti 997 in 200 in podjetja v predeovani ter storitveni dejavnosti, uporabjene pa so bie ocenjevane metode za panene podatke (pooed OLS oziroma metoda najmanjši kvadratov, metoda fiksni učinkov ter metoda sučajni učinkov). Rezutati modeov v povprečju nakazujejo pozitiven vpiv offsoringa na strukturo znanja podjetij. Anaiza v predeovani dejavnosti kaže, da ima offsoring v visoko razvite države pozitiven in podoben vpiv na strukturo znanja podjetij kot offsoring v nizko razvite države. V storitveni podjetji pa rezutati nakazujejo na šibkejši vpiv offsoringa v visoko razvite države v primerjavi z offsoringom v nizko razvite države. Ko se v definicijo znanja zaposeni vkjuči tudi pokic posameznika, rezutati za predeovana podjetja kažejo na močnejši vpiv offsoringa v visoko razvite države na reativno stopnjo zaposenosti profesionacev, v primerjavi z offsoringom v nizko razvite države. Rezutati za podjetja v predeovani dejavnosti se ne spremenijo bistveno, ko so uporabjene razične definicije ravni znanja v podjetji. Nasprotno pa se rezutati ob uporabi razični definicij bistveno razikujejo v storitveni podjetji. Ko definicija znanja zajema e pokic posameznika, ima offsoring v visoko razvite države šibkejši vpiv na reativno zaposenost tenikov kot offsoring v nizko razvite države. Ko definicija znanja zajema tako pokic posameznika kot tudi stopnjo izobrazbe, pa offsoring v visoko razvite in nizko razvite države kaže primerjiv in pozitiven vpiv na reativno stopnjo zaposenosti terciarno izobraženi vodij. Stopnja izobrazbe ima tako razičen vpiv na zaposenost razični skupin pokicev. Gavni prispevki doktorske disertacije so nasednji. Teoretični mode poeg proučevanja vedenja podjetij proučuje tudi vedenje posameznikov in njiove odočitve gede vaganja v nadajnje izobraževanje. Proučevanje odnosa posameznikov je pomembno z vidika kasnejšega vpiva te odočitev na povpraševanje po deovni sii podjetij. Anaiza vpiva uvoza na začetek izvažanja kontroira tudi vrsto uvoza z namenom povečanja razagane moči anaize. Zadnja prednost pa je vkjučitev stopnje pokica v definicijo znanja zaposeni. Sednje je pomembno zaradi dejstva, da je vseživjenjsko učenje vse boj nujno in da zaposeni ne preneajo z nadgrajevanjem svojega znanja, ko preneajo z rednim izobraževanjem. Kjučne besede: uvoz, izvoz, struktura znanja zaposeni, offsoring, outsourcing.

11 TABLE OF CONTENTS NTRODUCTON... Description of te dissertation topic area... Researc purpose and contributions of te dissertation... 5 Researc metods and data... 8 Structure and contents of te dissertation... 9 HOW TRADNG FRMS UPGRADE SKLLS AND TECHNOLOGY: THEORETCAL MODEL.... ntroduction....2 Literature review Te Mode Setup of te Mode ndividuas Firms Preferences Firm entry and exit Tecnoogy and factor eterogeneity nternationa activities of firms Firm beaviour Concusion UNTANGLNG THE RELATONSHP BETWEEN SKLL STRUCTURE, MPORTS AND EXPORTS ntroduction Literature review Metodoogy and data Metodoogy Hypotesis : Correation between importing and a better ski structure of firms is positive Hypotesis 2: Firms wit a better ski structure start importing Hypotesis 3: Ski structure canges in favour of te skied abour after te start of importing Hypoteses 4 and 5: mporting capita or intermediate goods as a positive impact on te start of exporting Data and descriptive statistics Resuts and discussion Base resuts Extensions of te mode Concusions THE EFFECTS OF OUTSOURCNG AND OFFSHORNG ON SKLL STRUCTURE: EVDENCE FROM MATCHED FRM-EMPLOYEE DATA i

12 3. ntroduction Literature review Metodoogy Framework and specification of te basic mode Extensions of te mode Data and descriptive statistics Empirica anaysis Basic mode Extensions of te mode Differentiation between ig- and ow-income countries Aternative definition of skis Concusion CONCLUSONS Summary of te main findings Limitations of te doctora dissertation... 9 Scientific contributions and future researc suggestions REFERENCES APPENDCES ii

13 LST OF FGURES Figure. mport and export growt in Sovenia and in EU Figure 2. Empoyment sare by educationa eve in Sovenia and in EU Figure 3. Empoyment rate by educationa eve in Sovenia and in EU Figure 4. Te fow-cart of events in te two-stage mode... 2 Figure 5. Occupationa structure of manufacturing and service firms in Sovenia iii

14 LST OB TABLES Tabe. Caracteristics of Sovenian manufacturing firms, broken-down by importing and exporting activities (mean vaues) Tabe 2. Resuts of testing Hypotesis : Correation between importing and a better ski structure of firms is positive Tabe 3. Resuts of testing Hypotesis 2: Firms wit a better ski structure start importing Tabe 4. Resuts of testing Hypotesis 3: Ski structure canges in favour of te skied abour after te start of importing... 5 Tabe 5. Resuts of testing Hypotesis 4: mporting capita goods as a positive impact on te start of exporting Tabe 6. Resuts of testing Hypotesis 5: mporting intermediate goods as a positive impact on te start of exporting Tabe 7. Mode extensions of testing Hypotesis 2: Firms wit a better ski structure start importing Tabe 8. Mode extensions of testing Hypotesis 3: Ski structure canges in favour of te skied abour after te start of importing Tabe 9. Mode extensions of testing Hypotesis 4: mporting capita goods as a positive impact on te start of exporting Tabe 0. Mode extensions of testing Hypotesis 5: mporting intermediate goods as a positive impact on te start of exporting Tabe. Caracteristics of Sovenian manufacturing and service firms Tabe 2. Caracteristics of Sovenian firms wic offsore and outsource Tabe 3. Caracteristics of empoyees in skied occupations Tabe 4. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms (observation period: ) Tabe 5. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, differentiating between ig- and ow-income countries (observation period: ) Tabe 6. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, using occupationa cassification for defining skis (observation period: )... 8 Tabe 7. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, for te major skied occupationa groups (observation period: ) Tabe 8. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, using educationa eve and occupationa cassification for defining skis (observation period: , ony tertiary educated) Tabe 9. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, for te major skied occupationa groups (observation period: , ony tertiary educated) iv

15 NTRODUCTON Every man [tus] ives by excanging. (Adam Smit) Description of te dissertation topic area Trade as payed an important roe in te economic deveopment, were tree reativey recent waves ave been especiay important for internationa trade and te word in genera; ndustria Revoution, post-war period after te bot Word wars, and te current wave after te 980s. Tese andmarks improved transportation inks and induced revoution in communication tecnoogies, wic in turn reduced distances and enabed te expansion of internationa trade (WTO, 204). Aong wit te canges in te society came aso te canges in teoretica modes of trade, were te excange of goods was te economists matter of interest for centuries. Teoretica trade modes first focused mainy on exporing te effects of countries different capabiities to produce certain products wit te same amount of input. Before Smit s Te Weat of Nations, te Scottis piosoper David Hume benefited te trade teory by offering opposition to mercantiists by arguing tat net exporting does not increase weat as it eads to iger domestic prices, wic eventuay increases imports. Later, in te second af of te 8 t century, Adam Smit deveoped a teory of absoute advantage in wic countries trade, wen teir absoute abour productivities differ. n Smit s mode, wic uses abour as te ony input, countries compare abour productivities in te production of a particuar good and te country wit te igest abour productivity becomes an exporter of tat good. Eac country terefore speciaises in te production of te good wit te absoute advantage and imports te ess efficienty produced good from abroad. n te case wen a country as no absoute advantage, trade does not occur (Krugman, Obstfed & Meitz, 202). Haf of a century ater, David Ricardo presented a teory of reative advantage. n tis mode, a country exports a particuar good if it as a reative advantage in te production of tis good; tat is if te good is produced wit a ower eve of te reative opportunity costs. Eac country ater speciaises in te production of te comparative-advantage product. Regarding tis teory, internationa trade occurs aso wen one of te countries does not ave an absoute advantage in any production of goods. Using Ricardo s comparative advantage mode as a bencmark, te Heckscer-Oin mode (te H-O mode) predicts tat countries wi adjust teir production and trading on beaf of teir factor endowments. Tis is a genera equiibrium mode of internationa trade and was deveoped by Ei Heckscer and Berti Oin. Contrary to te previous two modes, te H- O mode assumes two factors of production, abour and capita, and compares te reative abundances of tese factors between countries. A country wit a iger capita-to-abour ratio (i.e. te capita abundant country) speciaises in te production of te capita intensive

16 product and ater exports it. Meanwie, a abour abundant country speciaises in producing and exporting a abour intensive product and imports a capita intensive product. Accordingy, te prices of te abundant factor rise as a consequence of trade. Wie tere were no factor gains in Smit s or Ricardo s mode due to te usage of ony one factor in te mode, te Stoper-Samueson teorem in te H-O mode indicates tat te rea returns of te factor-abundant owners wi increase, and te rea returns of te owners of te oter factor wi decrease as a consequence of trade (Krugman, Obstfed & Meitz, 202). More recent modes contro aso for different types of returns to scae and competition. Krugman (979) deveoped a mode, wic uses a monopoistic competition as a framework and instead of using internationa differences in tecnoogy (as Smit and Ricardo) or differences in factor endowments (as in te H-O mode), it uses te economies of scae as a driver of trade. As a resut, trade deveops aso between countries tat ave simiar eves of tastes, tecnoogy eves and factor endowments. Panagariya (98) deveoped a mode, wic studies production patterns in a two-sector, a two-commodity mode, were one commodity is produced in a sector wit increasing returns to scae (RS) and te oter commodity is produced in a sector wit decreasing returns to scae (DRS). Te mode assumes two open economies tat differ in size and concudes tat a arge country wi export an RS commodity, wie a sma country wi export a DRS commodity. n is ater study, Krugman (99) deveoped an economic geograpy mode wit te purpose of studying te reasons for te concentration of manufacturing in some regions, rater tan oters, by taking into account te economies of scae and te transportation costs. Te resut of te mode is two-sided on te one and, wit ig transportation costs and weak economies of scae, te manufacturing production is concentrated around iger demand. On te oter and, ow transportation costs and strong economies of scae increase te manufacturing production in te areas tat aready ave a arge concentration of manufacturing to start wit. Te abovementioned papers concentrated mainy on studying te transnationa excange of goods between different countries on a eve of a country or industry. n tis regard, te Meitz (2003) mode presents an important miestone in te teoretica modes of trade by taking into account eterogeneous firms. n a cosed economy, ony te firms wit a sufficient productivity eve stay on te market and produce. Wen comparing incumbents and new entrants, te atter on average ave a ower eve of productivity and a iger probabiity to exit te market. n an open economy ony te most productive firms export since tey are abe to endure te additiona exporting costs. Again, te east productive firms exit te market. Te exit of east productive firms and te increasing saes of exporting firms a consequence of suppementary export saes cause te reative productivity eve in te economy to increase and transfer market sares and profits towards most productive firms. Te Meitz (2003) mode presents groundwork for many teoretica trade modes, wic expand te mode in severa ways. Bustos (20a) introduces an option to upgrade te 2

17 tecnoogy eve. n te mode, firms ave an option to upgrade teir eve of tecnoogy by paying an additiona fixed cost, wic resuts in reducing teir eve of margina costs of production. Te mode concudes tat ony te most productive firms upgrade tecnoogy and export, te firms wit intermediate productivity eves export but do not upgrade teir tecnoogy, wie te east productive firms serve te domestic market witout upgrading teir tecnoogy eves. n er foowing paper, Bustos (20b) enances te previous mode by controing for ski-upgrading witin firms. Firms ave an option to upgrade teir tecnoogy eve by investing in ski-intensive tecnoogy. Simiary as in te previous mode, te adoption of a ig-tecnoogy eve entais iger fixed costs and ower margina costs of production, wie taking into account aso te eve of abour costs in te ow- and ig-tecnoogy firms. Te wages of skied empoyees present a iger sare in tose firms tat adopt te ski intensive ig-tecnoogy. Before trade iberaisation, appointed wit ig trade costs, firms again form tree groups; te east productive firms do not export and continue using ow-tecnoogy, te intermediate productive firms export but aso use ow tecnoogy, wie te most productive firms export and upgrade teir eve of tecnoogy. After trade iberaisation, represented wit a reduction in variabe trade costs, te east productive firms arrange four different groups; te east productive exit te market, wie te increasing productivity eves enabe te firms to, first, stay on te domestic market, use ow tecnoogy and downgrade skis, second, start exporting but downgrade skis and continue using ow tecnoogy, and tird, start exporting and upgrade te tecnoogy eve and skis. Firms wit te medium productivity eve continue exporting, switc to ig tecnoogy and upgrade skis, wie te most productive firms continue using ig tecnoogy and exporting but downgrade skis. Finay, Amiti and Davis (20) upgraded te Meitz (2003) mode by controing aso for imports in te mode, wie evauating te effect of trade iberaisation on wages. Te mode contros for imports by introducing a iger fixed cost for importing intermediates from a particuar foreign market. Among oters, te concusions of te teoretica mode point to te exit of te east productive firms, to a wage decrease of firms, serving ony te domestic market, and to a wage increase of sufficienty arge importers and exporters. Finay, te mode aso proposes tat a firm wit iger profits and wages wi import a arger sare of its inputs or export a arger sare of its outputs. Findings from te abovementioned teoretica modes, indicating tat importing and exporting firms ave superior caracteristics in comparison to non-trading firms, were empiricay tested in severa papers. Tis introductory part presents ony a few empirica studies, wereas additiona studies wi be presented more in dept in te upcoming capters. At first, te empirica studies mainy concentrated on anaysing te caracteristics of exporters. Studies confirm tat exporters ave a iger eve of empoyment, wages and productivity, compared to non-exporters. One reason for tis is te sef-seection of firms into exporting as future exporters sow superior caracteristics 3

18 severa years before tey start exporting (see for exampe Yang & Maick, 200; and Bernard & Jensen, 999). Bernard and Jensen (999) find tat before te start of exporting, te increase in firm size and wages, wic in turn increases te prospects of becoming an exporter, is especiay important. Bernard and Jensen (999) empasise tat exporting brings severa benefits to firms and individuas in te form of iger wages, better future empoyment opportunities, increased growt of sipments, innovation and productivity, increases in empoyment growt, and enanced surviva probabiity. n addition to te sef-seection of firms into exporting, studies aso confirm earning-byexporting. Taking into account a subsampe of Sovenian manufacturing firms, Kostevc (2009) confirms previous studies on te sef-seection of firms into exporting, but aso finds tat more productive exporters enter more competitive exporting markets, wic in turn additionay increases teir productivity, in comparison to exporters, serving ess competitive markets. By estabising evidence for a iger growt of saes, productivity, and empoyment of new exporters, compared to non-exporting Cinese firms, Yang and Maick (200) aso confirm earning-by-exporting of firms. Wie te studies on anaysing te caracteristics of exporters are abundant, te importance of importing as been exposed ony recenty. Studying te Spanis manufacturing firms, Damijan and Kostevc (205) confirm tat more productive firms sef-seect into importing. By importing, firms get an access to new, ceaper and/or betterquaity products, wic decrease firms variabe costs and enabe greater investing in innovations. An increase in innovations, as a consequence of importing, ater as a positive impact on exporting, resuting again in additiona innovations. mporting is terefore a prerequisite for furter innovations and exporting. Studying te Begian firm-eve data, Amiti, tskoki and Konings (204) among oters confirm tat more productive firms import a arger sare of teir inputs from abroad, wic additionay increases teir productivity. A positive impact of imports on productivity was confirmed aso by Hapern, Koren and Szeid (20), studying te Hungarian micro-eve data. Te study finds ampe productivity gains in te manufacturing sector due to importing inputs internationay, were te vast sare of tis productivity increase can be attributed to an increased voume and vaue of imported inputs. Even more, imports ad an important impact on te Hungarian economic growt. Anoter section of te trade iterature, important for tis dissertation topic, anayses te connectedness between internationa trade and abour demand. n tis area of researc, studies usuay take into account te impact of outsourcing and offsoring on te abour demand. Anaysing te Sovenian manufacturing sector, Zajc Kejžar and Ponikvar (2004) find tat investment iberaisation infuenced firms from different parts of te tota factor productivity distribution differenty. More precisey, te most efficient firms face productivity increase, te east efficient incumbent firms face job destruction, wie te intermediate efficient firms experience bot penomena. Taking into account te taian manufacturing firms, Lo Turco and Maggioni (202) anayse te impact of offsoring on 4

19 te abour demand. Te important contribution of te study is differentiating between offsoring to ig- and ow-income countries. Resuts indicate tat offsoring to owincome countries as a negative impact on te abour demand. Wie te drawback of te abovementioned studies is not differentiating between skied and unskied workers, tere are many papers, wic contro for skis as we. n teir two papers, Feenstra and Hanson (996, 999) study te impact of outsourcing on te reative demand for skied empoyees (Feenstra & Hanson, 996) and wages of skied empoyees (Feenstra & Hanson, 999) in te United States. Resuts indicate tat outsourcing on average increases te reative demand for skied abour and teir reative wages. Te fear of job destruction in deveoped countries as increased in te recent years due to te enanced gobaisation and trade iberaisation in deveoping countries. Neverteess, te majority of studies find te treat not being arge and commony infuencing ony te owskied empoyees (see for exampe Mion & Zu, 203; Hijzen et a., 2005; Egger & Egger, 2003; and Strauss-Kan, 2003). Researc purpose and contributions of te dissertation Te abovementioned trend of increasing import and export growt is pronounced aso in Figure, wic sows import and export growt for Sovenia, te country of interest in tis dissertation, and in te EU. Bot, import and export, grew by a iger margin in Sovenia before te crisis, compared to te EU-28 average. Figure. mport and export growt in Sovenia and in EU mport growt 20 Export growt Growt, in % 0-5 Growt, in % Year Year Sovenia EU-28 Source: SORS, own cacuations As te iterature review exposed te interconnectedness of internationa trade and te ski structure of firms, te foowing two figures take into account te trends of te 5

20 empoyment sares and rates, divided by te eve of education, for te EU-28 and Sovenia. Te igest empoyment sare appertains to te empoyees wit te secondary eve of education, foowed by te tertiary educated empoyees and empoyees wit te primary educationa eve (Figure 2). n Sovenia, te empoyment sares of empoyees wit te secondary and primary eve of education decreased in te observation period, wie te empoyment sare of empoyees wit tertiary educationa eve increased. Simiar trends are evident aso in te EU data, wit te exception tat te empoyment sare of empoyees wit te secondary eve of education remained rougy te same troug te observation period. t is important to mention tat te cange in Sovenian empoyment sare by educationa eve is aso a consequence of ig participation rate of te younger popuation in tertiary education, wic does not necessariy mean te quaitative sift in te abour force structure. Terefore, te causaity is in tis case at east partiay running from te abour suppy and not demand. Figure 2. Empoyment sare by educationa eve in Sovenia and in EU-28 Empoyment sare (EU-28) Empoyment sare (Sovenia) Sare, in % Sare, in % Year Year Primary Tertiary Secondary No response Source: SORS, own cacuations Simiar trends are sared aso wen comparing te empoyment rates by educationa eve in Sovenia and in te EU-28, were te above-average empoyment rates beong to empoyees wit te tertiary and secondary eve of education, wie te empoyment rates of empoyees wit te primary eve of education are beow-average (Figure 3). 6

21 Figure 3. Empoyment rate by educationa eve in Sovenia and in EU-28 Empoyment rate (EU-28) Empoyment rate (Sovenia) Rate, in % Rate, in % Year Year Tota Secondary Primary Tertiary Source: SORS, own cacuations As a sma and open economy, Sovenia is igy dependent on te trends in te goba environment and internationa trade. Aso, its structure of empoyees as canged greaty in favour of skied individuas. After observing tese statistica findings, was motivated to furter expore te inkages and causaities between internationa trade and skis, in order to add to te existing iterature. n addition, according to te abovementioned statistics, beieve Sovenia is a suitabe country for making in-dept anayses. Te aim of te teoretica paper in te first capter is to add to te existing teoretica modes on eterogeneous firms by controing aso for imports and taking into account an individua s decision for ski upgrading. Te main objective of te mode is to consider te foowing two researc questions: (i) How does an individua s decision for ski upgrading ater affect te firm s abour demand?; and (ii) Do ow- and ig-tecnoogy firms take different decisions in empoying skied individuas, and starting to import and/or export after trade iberaisation? Contributions of te teoretica mode in te first capter are twofod. First, te mode fis te gap in te internationa trade teory by accounting aso for imports and tus broadening te modes of Meitz (2003) and Bustos (20a, 20b). Second, since an individua s decision for ski upgrading ater as an important impact on te firm s abour demand and productivity, te mode is spit into two parts. Te first part of te mode anayses te resoutions of individuas wit iger and ower eves of abiity for ski upgrading. Findings of te first part of te mode are ater incorporated in te second part, wic anayses firms decisions for tecnoogy upgrading and te start of importing and/or exporting. 7

22 Te main goa of te anaysis in te second capter is to empiricay test some of te concusions, set by te teoretica mode in te first capter. Te mode in tis capter considers answering te ensuing researc questions: (i) Do firms wit a better ski structure aso start importing?; (ii) Do importing firms furter increase teir ski eve after te start of importing?; and (iii) Do firms use imports for suppying intermediate goods and/or increasing teir eve of tecnoogy before te start of exporting? Since previous studies on te sequencing between internationa activities of firms usuay considered ony imports, exports and tecnoogy upgrading (see for exampe Damijan & Kostevc, 205; and Yang & Maick, 200), tis study fis te void in te iterature by incuding aso te ski structure of firms and different types of imports into te mode. n addition, te mode aso brings insigts for poicy impications wit regard to stimuating te internationa cooperation of firms and creating stimuus for iger empoyment of skied abour. Finay, te aim of te empirica mode in te tird capter is to study te impact of outsourcing and offsoring on te ski structure of firms. Te anaysis addresses te foowing researc questions: (i) Does controing for bot penomena offsoring and outsourcing in one mode add to te findings of previous empirica studies, wic contro ony for one factor in teir mode?; (ii) Does controing for a destination country of outsourcing and offsoring bring new insigts to te resuts?; and (iii) Does controing for an occupationa eve of workers wen defining skis bring additiona contribution to te resuts of te paper? Tis anaysis contributes to te fied of knowedge in severa ways. Firsty, by taking into account severa measures wen defining skis. Besides te usua eve of education, tis study considers aso te occupationa cassification of empoyees. Since skis can be acquired aso troug empoyment and experience, beieve it is important to contro for bot, te occupationa and te educationa eve wen defining skis. Secondy, since previous studies, controing ony for one of te two factors in teir modes, concuded tat bot, offsoring and outsourcing, on average ave a positive impact on te reative empoyment of skied empoyees (see for exampe Feenstra & Hanson, 996; Egger & Egger, 2003; and Mion & Zu, 203), tis anaysis contros for bot penomena in one mode in order to avoid te missing variabe bias probem. Taking into account bot factors in one mode is important aso due to te fact tat bot are expected to increase in te future, according to forecasts of UNCTAD (203). Researc metods and data Researc metods in tis dissertation are used in accordance wit te proposed aims of a particuar capter. Te teoretica mode in capter one extends te teoretica modes on eterogeneous firms by Meitz (2003), Bustos (20a, 20b), and Amiti and Davis (20), by incuding aso te anaysis of individuas beaviour into te mode. Te atter part refers to te work of Stark and oters (see for exampe Stark & Wang, 200; Stark, Hemenstein & Prskawetz, 998; Stark & Cau, 998; and Stark, Hemenstein & Prskawetz, 997 for reference). Te outine of te second capter is based on te empirica 8

23 metods used in Damijan and Kostevc (205), and Yang and Maick (200), wo appy propensity score matcing. extend teir anayses by studying more torougy te sequencing between importing and exporting, by means of controing for te type of imports, and te ski structure of firms in te mode. For robustness cecks, various types and variations of te matcing tecniques are used. Finay, te tird capter uses different metods for te pane data anaysis; pooed OLS, fixed effects and random effects. Again, severa different variations of te modes are used in order to enance te robustness of resuts. Te empirica anaysis of te second and tird capter combines severa different databases, wic form a ric firm-eve and empoyee-eve pane dataset for Sovenian firms, covering te period from 996 to 200. Te dataset comprises information on te baance seet data and income statements of Sovenian firms (i.e. number of empoyees, capita per empoyee, vaue added per empoyee), teir export and import activities (i.e. voume and vaue of exports and imports, type of exported and imported goods, destination of exports and imports), caracteristics of empoyees (i.e. gender, age, gross wage, educationa eve, occupationa eve), and te information on te foreign direct investments (FD) fows for a particuar Sovenian firm. Te dataset combines te foowing databases: persona income-tax data, transaction-eve data on exports and imports of goods, Statistica Registry of Empoyees, firm-eve accounting data, and FD. Te data was provided by te Statistica office of te Repubic of Sovenia (SORS), te Tax Autorities of Sovenia (TARS), te Bank of Sovenia, and te Agency of te Repubic of Sovenia for Pubic Lega Records and Reated Services (AJPES). Structure and contents of te dissertation Te core of tis doctora dissertation consists of tree capters, anaysing inkages and causaities between te trade status and te ski structure of Sovenian firms, were Capter introduces a teoretica mode, wic is ater tested empiricay in Capter 2. Finay, Capter 3 studies te effects of offsoring and outsourcing on te ski structure of firms. Te first capter introduces a teoretica mode, wic studies te decisions of individuas for investing in iger education, and te decisions of firms for tecnoogy and ski upgrading, and te decision for starting importing and exporting. Te first section introduces te motivation and goas of te anaysis, and is ater foowed by te second section wit a review of te reevant iterature. Te tird section, wic introduces te mode, is spit in tree subsections. Te first subsection gives a sort description of te mode, and is foowed by te first part of te teoretica mode, wic studies te beaviour of individuas. Te fina subsection anayses te beaviour of firms, were it first presents additiona background and te presumptions of te mode regarding firm preferences, te entry and exit of firms, tecnoogy and factor eterogeneity, and te 9

24 decision to begin wit internationa activities. Te ast section presents te main findings and te concusion. Te findings of te teoretica mode from te first capter are ater anaysed empiricay in Capter 2, wic untanges te reationsip between te ski structure of firms, imports and exports. Te first section introduces te topic, presents te motivation and contributions of te anaysis, wie te subsequent section introduces te reevant iterature from te fied of researc. Te metodoogy and data are torougy presented in te foowing section. n its first subsection, te modes, wic are ater estimated by propensity score matcing, and matcing tecniques tat are ater appied to te modes, are introduced. Te second subsection overviews te data used and te main descriptive statistics. Te foowing section firsty provides te tests on te quaity of matcing, before presenting te resuts in te next subsections. Resuts of te basic mode are incuded in subsection 2.4. and present te resuts on te inkages between importing, exporting and te ski structure of firms. Te extensions of te mode are incuded in subsection 2.4.2, wic take into account additiona ags of te outcomes of interest. Te fina section of Capter 2 gives a summary and a brief discussion of te resuts. Capter 3 presents te empirica anaysis of te effects of outsourcing and offsoring on te ski structure in Sovenian firms. As in te previous two capters, Capter 3 aso begins wit te introduction and te iterature review in order to present te main aims and motivation of te anaysis, and te reevant background of te topic. Te foowing section presents te metodoogica framework and specification of te basic mode, and its extensions. Te description of te data and te descriptive statistics are presented in section 3.4. Te resuts in te subsequent section are spit in two parts; te first part summarises te resuts of te basic mode wit te variabes of interest being outsourcing and offsoring. Te second subsection of te resuts presents te robustness cecks, wic contro for offsoring and outsourcing from ig- and ow-income countries, and introduce additiona measures of skis. Te fina section presents a summary of te main findings. Atoug eac of te capters incudes a pertinent introduction and concusion, te dissertation aso incudes a genera introduction and concusion. Te atter is incuded in Capter 4 and presents te main findings and te discussion of te resuts. Te foowing capter incudes te ist of references, wie te ast capter consists of appendices. Te appendices contain additiona reaisations of te equations from te main text, a suppementary expanation of particuar metodoogies used, and compementary tabes wit compete resuts and robustness cecks. Te fina appendix presents a toroug abstract of te dissertation in te Sovenian anguage. 0

25 HOW TRADNG FRMS UPGRADE SKLLS AND TECHNOLOGY: THEORETCAL MODEL. ntroduction Te iberaisation of internationa trade increases firm s productivity for two reasons; one is due to easier access to a better seection of advanced tecnoogies and anoter is due to a better aocation of production factors. Te atter canne was among oters empasized in te Meitz (2003) mode, wie te former was for exampe stressed in Bustos (20b). Te Meitz (2003) mode expores te effects of trade on intra-industry reaocations and aggregate industry productivity by taking into account eterogeneous firms tat differ regarding teir eve of productivity. Te mode concudes tat ony te most productive firms engage in exporting activities. Te Meitz (2003) mode represents groundwork in te recent trade iterature and was used as a basis aso in te Bustos (20b) mode, wic expores te effects of trade iberaisation on ski upgrading in exporting firms, were te mode aso differentiates between ig- and ow-tecnoogy firms. Tis paper aims to fi te void in te internationa trade teory by broadening te teoretica modes of Meitz (2003) and Bustos (20b), and correspondingy incuding imports to te mode. By doing tis, te mode aso expains recent empirica findings on te importance of importing as one of te drivers of firm s productivity gains. Evauating trade iberaisation after Cina s entry to te Word Trade Organization, Boom, Draca and Van Reenen (20) find tat te increased Cinese import competition increased te innovations and adoption of new tecnoogies, wic in turn increased te productivity witin firms, wie between firms it transferred empoyment toward innovative and tecnoogicay advanced firms. Te positive impact of importing on te firm s productivity was confirmed aso by Hapern, Koren and Szeid (20), studying te Hungarian data, Kasaara and Rodrigue (2008), studying te Ciean data, and Amiti and Konings (2007), studying te ndonesian data. Te atter study points out tat tese productivity increases are a consequence of importing ig-quaity intermediates, te enanced diversification of inputs, and iger earning opportunities (Amiti & Konings, 2007). Taking into account importers and exporters, Smeets and Warzynski (200) confirm tat bot, exporting and importing, increase te firm s productivity, wie firms wit te igest eve of productivity are engaged in bot trading activities. n reation to tese findings, empirica papers aso certify te positive impact of importing on exporting. Bas and Strauss-Kan (204) empasize tree cannes troug wic importing affects exporting positivey. First is te indirect productivity canne of increased productivity after importing, wic can in turn ave a positive effect on overcoming export costs. Second is a direct cost canne due to canging te input structure towards more cost-effective importing intermediates. Finay, troug te quaity/tecnoogy transfer, imported intermediate inputs can enabe exporting products to be of suc quaity and tecnoogy eves, as desired in te export markets.

26 Positive effects of importing on exporting were for exampe confirmed aso by Feng, Li and Swenson (202), studying te Cinese data. n addition, since te individua s decisions for acquiring iger skis ater ave an important impact on te beaviour of profit-maximizing firms, anoter motivation for writing tis paper was to combine specific individua s and firm s decisions. Since te existing trade modes are based on broader, firm-eve decisions, te impetus of te present paper is to expore more in dept aso te beaviour of individuas and teir decision for ski upgrading, as tese decisions ave in turn te effect on ski upgrading witin a firm. Te mode in tis paper bases its framework on te modes of Bustos (20a, 20b) and Meitz (2003), and on te work of Stark and oters (see for exampe Stark & Wang, 200; Stark, Hemenstein & Prskawetz, 998; Stark & Cau, 998; and Stark, Hemenstein & Prskawetz, 997 for reference), wo deveoped modes on uman capita formation. Te mode first expores te beaviour of individuas, wo decide weter to invest in acquiring iger skis or not. n tis part, te mode differentiates between ig abiity and ow abiity individuas, were te individua s abiity eve defines te cost eve for acquiring skis. ndividua s abiity is discoverabe ony to te individua. Upon te eve of tese costs, individuas decide weter to invest in obtaining te skis or not, were tis decision depends aso upon teir future wage eve. Resuts suggest tat ony ig abiity individuas find it profitabe to invest in acquiring additiona skis, wie tey in turn demand iger wages after entering empoyment. Te findings tat iger firm s abour costs signa iger empoyment of more skied workers are ten incorporated in te second part of te mode. Tis part focuses on exporing te beaviour of eterogeneous firms tat decide on wen to start investing in iger tecnoogy, and wen to start engaging in trading activities. n tis part of te mode, profit-maximizing firms differ upon teir eve of abour productivity, were te proxy for iger abour productivity are iger abour costs, indicating a iger empoyment eve of skied empoyees. Once more, te atter judgement is backed up by te resuts from te first part of te mode. nvesting in iger tecnoogy and starting to import and export brings iger fixed costs, but decreases te eve of firm s margina costs, and/or increases te empoyment of skied workers, and/or increases revenues. Findings from te second part suggest tat te tecnoogicay advanced firms empoy a iger number of skied workers and tat ony te most productive firms find it profitabe to start trading, investing in iger tecnoogy and ski upgrading. Tis paper contributes to te iterature in two ways. Firsty, since te mentioned empirica papers empasized te importance of differentiating between importing and exporting, tis mode accounts for bot. Terefore, te mode broadens te content of te papers of Bustos (20a, 20b) and Meitz (2003), wo take into account ony exporters. Secondy, wie oter teoretica trade modes ony anaysed decisions from a firm s point of view, tis paper s contribution is to combine beaviour of individuas and firms in one mode of trade. Te mode terefore broadens te existing trade modes by anaysing te beaviour 2

27 of individuas and teir decision for ski upgrading. Tis is ater incorporated in te firmeve decisions, by taking into account te firm s abour demand and productivity. Te remainder of te paper is organised in te foowing manner: te next section presents a brief introduction of te teoretica background, wic is furter on used as a reference point to te teoretica mode, incuded in te tird section. Te ast section summarises te main findings and incudes a concusion..2 Literature review Meitz (2003) deveoped an important teoretica mode, wic expores te effects of trade on intra-industry reaocations and aggregate industry productivity. Te mode uses eterogeneous firms tat differ regarding te eve of productivity, were firms wit iger eves of productivity produce te same amount of products at ower margina costs. After observing teir eve of productivity, firms decide to exit or enter te market, were new entrants ave a ower eve of productivity and a iger probabiity to exit tan firms tat are aready on te market. Wen exporing te effects of trade, te autor ony focuses on exports. After firms start exporting, tey are faced wit iger costs for two reasons; one reason is iger per-unit trade costs, and te oter reason is iger fixed costs. Te atter can be expained as a consequence of estabising new networks, adapting te product to te new market, setting up new distribution cannes, etc. After introducing te possibiity to export to te mode, firms again observe teir eve of productivity. Once more, te east productive firms decide to exit te market, te firms wit medium-eve of productivity decide to serve te domestic market, wie te most productive firms serve te domestic market and export (Meitz, 2003). Te Meitz (2003) mode presents te groundwork for many subsequent teoretica modes on trade. Bustos upgraded te Meitz (2003) mode by incuding tecnoogy upgrading (Bustos, 20a) and ski upgrading (Bustos, 20b) into te mode. n te first mode, Bustos (20a) takes into account profit maximizing firms wic decide weter to start exporting and weter to invest in iger tecnoogy. By adopting iger tecnoogy, firms pay iger fixed production costs, wie teir margina costs are reduced. After proving tat using ig tecnoogy and serving te domestic market is aways dominated by some oter coice, firms form four different groups: te east productive firms exit, te ow productive firms use ow tecnoogy and serve te domestic market, te medium productive firms sti use ow tecnoogy but aso export, wie ony te most productive firms upgrade teir tecnoogy eve and export (Bustos, 20a). Te gains of different production factors, abour and capita to be precise, were incuded aready in te Heckscer-Oin mode (te H-O mode), wic predicts tat countries adjust teir production and trading on beaf of teir factor endowments. Te Stoper- Samueson teorem in te H-O mode indicates tat te rea returns of te factor-abundant owners increase, and te rea returns of te owners of te oter factor decrease as a 3

28 consequence of trade (Krugman, Obstfed & Meitz, 202). Reating to te concusions of te H-O mode, te reative demand for skied workers a scarce factor in deveoping countries soud decrease after trade iberaisation. However, te empirica findings sow te opposite (see for exampe Godberg & Pavcnik, 2007). Bustos (20b) as fied te gap in trade iterature, by exporing te effects of trade iberaisation on ski upgrading in exporting firms. Te mode accounts for two categories of workers, skied and unskied. As in te previous mode (Bustos, 20a), firms form four different groups before trade iberaisation, wereas after iberaisation, tey form six groups in tota. Te east productive firms exit. Among te firms tat did not export before trade iberaisation, a fraction of tese firms continue serving te domestic market, use ow tecnoogy and downgrade skis; anoter fraction of tese firms sti uses ow tecnoogy, but tey start exporting and downgrade skis, wie te most productive of tese firms start to export, upgrade teir tecnoogy and skis. Firms tat were aready exporting before trade iberaisation and used ow tecnoogy continue to export, switc to ig tecnoogy and upgrade skis. Finay, te most productive firms tat were exporting and using ig tecnoogy before trade iberaisation continue exporting and using ig tecnoogy, but tey downgrade skis. Te concusions of te teoretica mode were ater tested aso empiricay, by studying te effect of Brazi s tariff reduction on Argentinian firms. Te mode s predictions tat ow-tecnoogy firms downgrade skis and tat firms in te upper-midde range of productivity distribution upgrade skis after trade iberaisation are consistent wit te empirica findings. On te oter and, te prediction tat te most productive ig-tecnoogy firms downgrade skis after trade iberaisation is not consistent wit te empirica findings (Bustos, 20b). Finay, as presented in te introduction, it is important to contro for te imports in trade modes, as imports usuay serve as a prerequisite to exporting activities (see for exampe empirica studies of Damijan & Kostevc, 205; and Atomonte & Békés, 200). To be precise, by studying te connections between importing, exporting and innovation in Spanis firms, Damijan and Kostevc (205) find tat importing enabes firms to first start wit process and product innovation, and ater aso wit exporting. n addition, exporting stimuates furter innovation. Atoug empirica studies sow te importance of importing, te atter is infrequenty incuded in te teoretica modes of trade. One of te modes tat does account for importing is te teoretica mode by Amiti and Davis (20), wo base teir teoretica mode on te Meitz (2003) mode and contro for imports, by incuding additiona costs of importing in te mode. Te teoretica mode in tis paper combines different aspects of te modes, presented in te iterature review and adds aso a toroug anaysis of individuas beaviour and teir decision for ski upgrading. t is necessary to study tese decisions, as tey ater ave an important impact on te firm s productivity eve, abour demand and abour costs. For tis purpose, severa papers of Stark and oters were taken into account (see for exampe Stark & Wang, 200; Stark, Hemenstein & Prskawetz, 998; Stark & Cau, 998; and Stark, 4

29 Hemenstein & Prskawetz, 997 for reference). Te primary focus is on te paper by Stark and Wang (200), wic deveoped a mode of uman capita formation in an environment wit and witout migration. bring te mode into use as a bencmark and use it for expaining te individua s coice for ski upgrading..3 Te Mode Tis section presents a simpe teoretica mode, te first part of wic studies te decision of individuas to invest in acquiring additiona skis. Te findings of te first part of te mode are ater incorporated in te second part, wic anayses te decision of eterogeneous firms to start trading and investing in iger tecnoogy..3. Setup of te Mode Te mode takes into account te country, endowed wit eterogeneous workforce and eterogeneous firms. ndividuas differ according to teir abiity eves, wic are exogenousy determined. Firms on te oter and differ according to te different productivity eves, wic are te end resut of different tecnoogies used, and in regards to firms being incuded in internationa trade. Concerning te atter, te mode differentiates between importers, exporters and importing-exporting firms, wereas concerning te former it differentiates between ig-tecnoogy and ow-tecnoogy firms..3.2 ndividuas Tis part of te teoretica mode foows te work of Stark and oters (see for exampe Stark & Wang, 200; Stark, Hemenstein & Prskawetz, 998; Stark & Cau, 998; and Stark, Hemenstein & Prskawetz, 997 for reference). Eac individua in te economy is endowed wit a certain amount of efficiency units (θ), wic represents te abiity of a worker. f te average abiity of workers in te economy is θ, and te abiities of ig abiity and ow abiity workers are θ S and θ U, respectivey, te foowing appies: θ U < θ < θ S. Te mode aso assumes tat te individua s abiity is discoverabe ony to te individua. For brevity, te mode denotes a individuas wit above-average abiities by θ S, and individuas wit beow-average abiities by θ U. Derivations of te mode terefore assume two abiity eves. After individuas evauate teir eve of abiity, tey decide weter to invest in acquiring iger skis or not. t is assumed tat te costs for acquiring iger ski eves are different for individuas wit different abiities. To be precise, costs for acquiring uman capita for ig abiity individuas (k S ) are ower tan te costs of ow abiity individuas (k U ); i.e. k U > k S >. A individuas ave an opportunity to acieve iger eves of education and become skied. However, since it is aso assumed tat te costs for acquiring te igest eves of uman capita are too ig for ow abiity individuas, tey wi be abe to obtain te uman capita ony up to a certain eve and wi not be abe to acieve above-average ski eves. Te drawback of tis assumption is in measuring te 5

30 costs for acquiring iger eves of education ony by te eve of individua s abiity and not taking into account oter aspects, as for exampe individua s financia capabiity. n reaity, due to reasons suc as individua s financia constraints or te ack of stamina, not a igy capabe individuas decide to acquire iger eves of education. n order to empasise te difference between individuas abiity and skis, and te period after individuas acquire skis, te mode denotes ig abiity, ig skied individuas wit Θ S, and ow abiity, ow skied individuas wit Θ U. Wie te abiity of individuas is not discoverabe and observabe by oters, te mode assumes tat firms can discover and observe individuas skis. Te abiity of individuas is terefore refected in teir skis. n reaity, firms can for exampe detect individuas skis in te form of teir educationa eve. However, as it was argued before, te ski eve of individuas cannot be fuy refected in teir educationa eve. Neverteess, since individuas persona caracteristics woud be ard to measure objectivey, since many empirica studies measure ski eve wit te eve of education, and since tis measure embraces some important aspects of individuas skis, as for exampe teir abiity, stamina, diigence, etc., beieve it is a sufficient proxy for measuring individuas skis. Yet, it woud be usefu to incude te fact tat individuas skis cannot be fuy discoverabe and observabe in te future mode extensions. Te mode assumes tat iger opportunity costs of education are ater refected in individuas iger wages. More precisey, individuas wit iger abiities wi ave incentives for acquiring above-average eves of uman capita, if teir costs for acquiring ig ski eves wi be ater compensated wit iger gross earnings wen tey are empoyed. By tis, te mode contros for te different financia background of individuas. Te gross earnings of ig abiity, ig skied workers (w S ) soud terefore be iger tan te gross earnings of ow abiity, ow skied workers (w U ); i.e. 0 < w U < w S. Tus, eac individua initiay bears te costs of acquiring uman capita. However, te costs are ater transmitted onto firms in te form of iger expected gross earnings of ig abiity, ig skied individuas. Te function of gross earnings for unskied workers is te foowing: w U (Θ U ) = λ[n(θ U + )] k U Θ U, () were te first term on te rigt and side (λ[n(θ U + )]) represents persona returns to uman capita, and te ast term represents costs of acquiring uman capita. Te parameter λ is assumed to be positive. Furtermore, for convenience, te foowing is assumed as we: λ > k U > k S >. Te atter assumption is important in order for individuas to ave incentives for acquiring uman capita. Oterwise, persona returns to uman capita woud be too ow, compared to costs of acquiring uman capita. 6

31 Simiary, te function of gross earnings of skied workers can be written as: w S (Θ S ) = λ [n(θ S + )] k S Θ S. (2) Te succeeding caim proves tat te optima ski eve of workers wit ow abiity and ow skis is ower tan te optima ski eve of workers wit ig abiity and ig skis. t is important to prove tat in order to make furter inferences on te wage eve of skied workers. Caim : Te optima ski eve of individuas wit ow abiity is ower tan te optima ski eve of individuas wit ig abiity. Proof: To get te optima ski eve of ig and ow abiity individuas, first order conditions of gross earnings for eac eve of skis are derived. wu ( U ) k U U U (3) ws ( S ) k S S S (4) Wen cecking te maxima, te foowing optima ski eves of workers are cacuated. Optima ski eve of te ow abiity workers (Θ U * ) is: Optima ski eve of te ig abiity workers (Θ S * ) is: Θ U * = λ(k U ) -. (5) Θ S * = λ(k S ) -. (6) Wen comparing bot optima eves and taking into account tat k S < k U, it is confirmed tat Θ U * < Θ S *. Atoug te previous caim confirms tat te ig abiity workers wi ave iger optima ski eves tan te ow abiity workers, it aso as to be proven tat te ig abiity workers wi ave incentives to invest in teir educationa attainment and make te best of teir potentia. As mentioned before, ig abiity workers wi ave incentives to invest in teir educationa attainment and become skied, if teir future income woud increase because of tat investment. By inserting optima ski eves of ig abiity and ow abiity individuas (expressions (5) and (6)) in te functions of gross earnings (expressions () and (2)), te foowing can be derived: w U (Θ U * ) < w S (Θ S * ) (7) λ[n(λ / k U )] k U [(λ / k U ) ] < λ[n(λ / k S )] k S [(λ / k S ) ]. (8) 7

32 Taking into account te assumption λ > k U > k S >, it can be confirmed tat te gross earnings of workers wit ow optima abiity (w U (Θ U * )) are ower, compared to te gross earnings of workers wit ig optima abiity (w S (Θ S * )). For consistency purposes it was aso confirmed tat w U (Θ U * ) > 0. Te proof for tis caim can be found in Appendix A. As ony te ig abiity individuas ave incentives to invest in acquiring iger skis, tota workforce (L) in te country comprises ig abiity, ig skied workers (L S ) and ow abiity, ow skied workers (L U ). Workforce in te country as a woe is terefore te foowing: L = L S + L U..3.3 Firms Tis part of te mode takes into account eterogeneous profit-maximizing firms tat differ in teir eve of abour productivity and decide weter to adopt a ski-intensive tecnoogy, and weter to start exporting and importing. Te previous part of te mode concuded tat skied workers ave a iger eve of abiity and can ence be empoyed in a more productive way. Tis finding wi be accounted for in te current part of te mode, wen taking into account te eve of firm s abour productivity. Te subsequent part of te teoretica mode foows te work of Meitz and Redding (204), Amiti and Davis (20), Bustos (20a and 20b), and Meitz (2003) Preferences Foowing Meitz and Redding (204), and Bustos (20a, 20b), tis part considers two symmetric countries tat engage in biatera trade after trade iberaisation. Consumer preferences are described by a continuum of orizontay-differentiated varieties and are assumed to take te Constant Easticity of Substitution (CES) form: M Q q( ) d, (9) 0 were defines a particuar variety of a product, M is te number of existing varieties, and is a constant easticity of substitution. Te foowing ods: /( ), were is a parameter wic determines te constant easticity of substitution, so tat appies. Tese preferences define te foowing demand function for eac variety ω: q ( ) XP p( ). Here, X represents te aggregate spending eve of consumers, p(ω) te price of eac variety, and P te price index, equa to: P M p( ) d. (0) 0 8

33 Firm entry and exit Foowing Meitz and Redding (204), and Bustos (20b), firms pay a sunk fixed entry cost f X to enter an industry. After tat, firms draw te eve of teir productivity φ from a cumuative distribution G(φ) and wit regard to tis eve tey decide weter to exit te market or to produce Tecnoogy and factor eterogeneity Products are produced by using a composite factor of production, L, wic is composed of skied abour (L S ) and unskied abour (L U ). From te previous subcapter, it foows tat skied workers ave a iger eve of abiity, wic is refected in teir iger wage eve w U < w S. Furtermore, foowing Meitz and Redding (204), and Bustos (20b), by paying an additiona fixed cost, firms can upgrade to a ig-tecnoogy eve, wic is aso more ski-intensive and reduces te firm s margina costs of production. On te oter and, te ow-tecnoogy eve is ess ski-intensive and demands ower fixed costs for producing goods. As in te case of distribution of skis, te distribution of ig-tec and ow-tec firms is discrete. Tota costs for ow-tecnoogy firms are as foows: q TC f ws wu, () were f denotes fixed costs, w S and w U are wages of skied and unskied workers, respectivey, q is te eve of firm s output, φ is productivity eve, and (0, ) denotes ski intensity. On te oter and, firms can invest in iger ski-intensive tecnoogy. Tota costs for te atter can be defined by: q TC f ws wu, (2) were η >, γ >, (0, ), and α > ß. Te mode assumes tat due to a smaer reative sare of skied empoyees in ow-tecnoogy firms, wo use ow-tecnoogy equipment, te abour productivity in ow-tecnoogy firms is ower tan te abour productivity in ig-tecnoogy firms. On te oter and, as a resut of investing in ski-intensive tecnoogy, ig-tecnoogy firms cange teir ski structure by empoying a iger number of ig abiity, ig skied empoyees. Accordingy, te mode assumes tat skiintensive tecnoogy is brougt into use more productivey wen empoying reativey more skied individuas wit ig abiities. Reating to te findings from te first part, wic studied te incentives for individua s ski upgrading, te mode aso assumes tat firms wit iger abour productivity ave iger abour costs, as a consequence of a 9

34 iger empoyment of skied workers, wo earn iger wages; w S > w U. Higer abour costs can terefore be considered as a proxy for iger empoyment of skied workers. Tese assumptions are consistent wit te findings of empirica studies, wic confirm tat bigger firms use more tecnoogy-advanced equipment, pay iger wages and empoy more productive workers (dson & Oi, 999). Simiar caracteristics ave aso been confirmed in trading firms, wic are arger in size and more productive (Atomonte & Békés, 200). Tese concusions are refected in te assumption tat α > ß (expressions () and (2)), wen defining te tota costs of ow- and ig-tecnoogy firms. Te parameters α and ß, and consequenty aso firms wages, are determined exogenousy. Since defining market-cearing wage and abour market equiibrium woud furter increase te compexity of te mode, tis extension to te mode woud be out of te scope of te current paper. Neverteess, it woud be usefu to expand te mode by incuding tese concepts in te future mode extensions nternationa activities of firms Te mode is buit as a 2-stage mode, were costs of trade decrease significanty ony in te second stage, as a consequence of trade iberaisation. n te first stage, firms decide weter to invest in ski-intensive ig-tecnoogy, wereas in te second stage, firms decide weter to engage in trading activities. Simiary as in Meitz and Redding (204), and Bustos (20b), firms decide to start exporting after reaising teir eve of productivity, φ, and taking into account te iger costs of exporting. On te one and, additiona fixed costs of exporting, f E, arise from estabising new saes cannes, advertising, adapting to new aws and rues, etc., wie on te oter, firms aso ave to pay additiona iceberg variabe trade costs τ, meaning tat τ number of units ave to be sipped abroad in order for one unit to arrive, were τ > (Meitz & Redding, 204). For very simiar reasons as in te case of exports, importing aso entais iger fixed costs, denoted by f (Amiti & Davis, 20). Additiona costs of exporting and importing make an assortment of te most productive firms tat can afford to endure iger costs. n te current setup of te mode, te costs of importing and exporting are defined exogenousy and are not compared by teir eigt. n te future mode extensions, it woud be interesting to compare aso te eigt of te costs, in order to make a connection between importing and exporting, and te sequencing pattern between te two. mporting can for instance ead to reduction of fixed exporting costs as an importing firm aready knows te foreign market, as estabised its networks, etc. As a resut, starting to import can increase te probabiity of a firm to start exporting Firm beaviour Some additiona assumptions concerning costs and te cange in productivity eves are made beow. As introduced earier, tis mode is of a two-stage type, were in te first stage, firms decide weter to invest in ig-tecnoogy or not and in te second stage, 20

35 after trade iberaisation, firms decidee weter to engage in internationa activities or not. Wen firms start importing, tey ave access to ceaperr tecnoogy and/or access to ceaper intermediates. t iss terefore anticipated tat importing increases te productivity of firms for two different reasons. First, importing intermediates aows oter factors of production to be used more productivey. Second, importing more affordabe tecnoogy equipment in turn increases te firm s productivity. Accordingy, te eve of abour productivity in firms increases after importing. Te mode aso a assumes te fixed costs of acquiring ig-tecnoogy are igerr tan te fixed costs of importing; i.e. f < fη. Te reason beind tis assumption is tat wen comparing te t tecnoogy eve of ow- teir tecnoogy importers and ig-tecno oogy firms, wic invest in tecnoogy witin own R&D departments, it is i assumed tat te increase in tee productivity eve wi not be as big in ow-tecnoogy firms tat start importing, compared to te productivity increase in firms tat start investing in ig-tecnoogy. Atoug ow-tecnoogy firms sti ave a more affordabe option forr increasingg teir eve of productivity troug importing, te benefits are not as ig, compared to investing into deveoping custom-made ig- to a tecnoogy equipment witin firms. Finay, due to exporting, firms se teir products iger number of customers and terefore reac iger revenues. Te foowing paragraps describe a two-stage mode, were in eac of te steps, firms decide between severa options and coose te most profitabe one. tt is assumed tat in te first stage (before trade iberaisation), importing and exporting is beyond te reac due to ig costs, so firms can ony coosee weter to invest in iger tecnoogy or not. n te second step, after trade iberaisation, firms ave an option to t start importing, exporting or bot. Te foowing diagram summarises te steps of te mode. Figure 4. Te fow-cart of events in te two-stage mode FRMS Stay on te market Exit Low-tecnoogy Hig-tecnoogy Domestic ony mport Export mport and export Domestic ony mport Export mport and export n te isted steps, firms compare severa different profit options, wic are described next. Foowing Meitz and Redding (204), and Bustos (20b), te market structure is of monopoistic competition type, t weree eac firm cooses its price in order to maximise its 2

36 profits. Te profit maximising price is a constant mark-up over margina costs. n te first stage, ow-tecnoogy firms () carge te price wie ig-tecnoogy firms (2) carge te price Firms compare te foowing two options: a) No trade, use ow tecnoogy (): ws wu p, (3) ws wu p. (4) r ( ) ( ) fws wu, (5) were ( ) are te tota profits of firms wit ow-tecnoogy eves, is te eve of abour productivity, and r ( ) are revenues, wit w S wu r XP. b) No trade, use ig tecnoogy (2): r ( ) fw w ( ) S U, (6) were ( ) are te tota profits of firms wit ig-tecnoogy eves, and r ( ) are te revenues, wit w S wu ( ) XP r. According to Meitz and Redding (204), firms first assess teir eve of productivity and upon tat decide weter to stay and produce or weter to exit te market. f tey stay, tey maximise te eve of teir profits wit regard to te eve of teir productivity. Tis * generates a surviva bound productivity, returning zero profits: ( * ) 0. Wen comparing zero-profit bounds of ow- and ig-tecnoogy firms in te first stage of te mode; i.e. r ( ) r ( ) ( ) ( ) fw fw, (7) 22

37 it foows tat due to te iger fixed costs of adopting new tecnoogy, ony te most productive firms wi be abe to afford investing in ig-tecnoogy. For convenience, is denoted as tota abour costs in ow-tecnoogy firms ( W ws wu ), and W as tota abour costs in ig-tecnoogy firms ( W w w ). Least productive firms wi terefore * use ow-tecnoogy. Furtermore, te exit bound productivity,, is defined by: S U W * * ( ) 0 W Af, (8) were A. X P To get te eve of productivity, above wic a firm finds it profitabe to invest in igtecnoogy,, te subsequent two expressions are compared: ( ) ( ), yieding te foowing: W f W W ) A ( W. (9) Now, it must appy tat *, wic is true as ong as W ( W ) ( W W ) W. Te atter expression stands wen te wages in igtecnoogy firms ( W ) are significanty iger tan te wages in ow-tecnoogy firms ( W ). Tis is consistent wit te findings from te first part of te paper, wic concudes tat iger wages signa a iger empoyment of skied workers. beieve tis assumption is vaid as it confirms previous empirica findings tat firms, wic use more tecnoogyadvanced equipment, aso pay iger wages, and empoy more productive workers (see for exampe dson & Oi, 999). Terefore, ony te most productive firms use ski-intensive tecnoogy and upgrade skis. n addition, taking into account te ast term in te upper expression W W, te reative increase in wages due to investing in iger tecnoogy as to be iger tan te reative decrease in margina costs; i.e. W W, wic additionay empasises te importance of iger empoyment of skied workers in igtecnoogy firms. After trade iberaisation in te second stage, ow-tecnoogy firms compare te foowing four options: 23

38 24 a) No trade, use ow tecnoogy (): ) ( ) ( U S w fw r, (20) were ) ( are te tota profits of firms wit ow-tecnoogy eves, is te eve of abour productivity, and ) ( r are te revenues, wit ) ( U S w w XP r. b) Start importing, use ow tecnoogy (): Wen ow-tecnoogy firms start importing, teir costs and productivity eve increase and add up to: U S w w q f f TC, were, and. ntroducing te factors and enabes controing for te decrease in margina costs and te canges of te ski structure in favour of te skied empoyees after ow-tecnoogy firms start importing. However, as expained above, te increase in te productivity eve is not as big as it woud be if te firms invested in deveoping te custom-made tecnoogy witin teir own R&D departments. Firms carge te price: U S w w p. Taking tese facts into account, profit is as foows: U S w w f f r ) ( ) ( ) (, (2) were ) ( are te tota profits of ow-tecnoogy firms tat start importing, and () r are te revenues, wit ) ( U S w w XP r. c) Start exporting, use ow tecnoogy (): Wen ow-tecnoogy firms start exporting, teir costs add up to: U S E E w w q f f TC. Consequenty, firms carge te price: U S E w w p. Taking tese facts into account, te profit is:

39 25 ) ( ) ( ) ( U S E E E w w f f r, (22) were ) E ( are te tota profits of ow-tecnoogy firms tat start exporting, and () E r are te revenues, wit ) ( U S E w w XP r. d) Start importing and exporting, use ow tecnoogy (): Wen ow-tecnoogy firms start importing and exporting, teir costs add up to: U S E E w w q f f f TC. Consequenty, firms carge te price: U S E w w p. Taking tese facts into account, te profit is: U S E E E w w f f f r ) ( ) ( ) (, (23) were ) E ( are te tota profits of ow-tecnoogy firms tat start importing and exporting, and ) ( E r are te revenues, wit ) ( U S E w w XP r. Wen comparing te zero-profit bounds in tis stage of te mode, te assumption of identica countries is considered (Bustos, 20a), from wic it foows tat te price index (P) and te expenditure eve (X) are te same at ome and abroad. First, te zero-profit bounds of ow-tecnoogy firms tat do not engage in internationa activities are compared to te bounds of tose wic start importing in te second stage of te mode: W f f r fw r ) ( ) ( ) ( ) ( ) (. (24) For convenience, again used te abbreviation for te tota abour costs in ow-tecnoogy firms ( W ) and denoted te tota abour costs of importing ow-tecnoogy firms by U S w w W. t foows tat ony te most productive ow-tecnoogy firms wi be abe to afford paying iger fixed costs of importing, wie te east productive ow-tecnoogy firms wi continue serving te domestic market. To get te eve of productivity, above wic a ow-tecnoogy firm finds it profitabe to start importing,, one compares te subsequent two expressions: ) ( ) (, and gets te foowing:

40 Te expression A W f W W ) fw W (. (25) W (. * appies, as ong as f W ) f ( W W ) f W W Tis is true wen te wages in importing ow-tecnoogy firms ( W ) are significanty iger tan te wages in ow-tecnoogy firms ( W ), wic again signas a iger empoyment eve of skied workers, as foows from te first part of te mode. Tis assumption is aso vaid, since te empirica data confirms tat importing firms are on average arger and pay iger wages (see for exampe Atomonte & Békés, 200). n W addition, taking into account te ast term in te upper expression W, te reative increase in wages due to importing as to be iger tan te reative decrease in margina W costs after te start of importing; i.e.. Tis statement corresponds to te initia W assumption tat te decrease in margina costs due to imports is ower tan it woud be, soud te firms invest in deveoping custom-made tecnoogy witin teir own R&D departments. Furtermore, wen comparing te zero-profit bounds of ow-tecnoogy firms tat do not engage in internationa activities and of tose wic start exporting in te second stage of te mode: E E r ( ) r ( ) ( ) ( ) fw ( f f E ) W, (26) it foows tat exporting ow-tecnoogy firms do not invest in upgrading teir ski structure nor do tey invest in acquiring ower margina costs. Terefore, since te productivity eve of ow-productive firms stays te same after tey start exporting, owtecnoogy firms wi export ony if te costs of exporting are ower tan te increase in revenues after te start of exporting. However, foowing Meitz and Redding (204), it is assumed tat te fixed costs of exporting are too ig for ow-tecnoogy firms and terefore present a seection, so tat ony te most productive firms start exporting. As a resut, firms tat do not invest in acquiring a iger eve of productivity eiter troug importing or troug investing in iger tecnoogy cannot start exporting since teir productivity eve is too ow. n addition, te zero-profit bounds of ow-tecnoogy firms wic do not engage in internationa activities and of tose tat start importing and exporting in te second stage of te mode, are compared wit te foowing expressions: 26

41 E E r ( ) r ( ) ( ) ( ) fw ( f f fe ) W. (27) n reation to te upper comparison, ow-tecnoogy firms wi find engaging in importing and exporting activities profitabe ony if te increase in revenues and productivity eve is bigger tan te increase in costs of exporting and importing. To get te eve of productivity, above wic a ow-tecnoogy firm finds it profitabe to start importing and E E E E exporting,, te subsequent two expressions are compared: ( ) ( ), yieding te foowing: A f W W ) ( f fe) W ) W W (. (28) E Tis aows us to ceck wen te productivity eve of ow-tecnoogy firms tat import ( ) is ower tan te productivity eve of ow-tecnoogy firms tat export and import E ( ): E f W W ) fw W W f( W W ) ( f fe) W W W (. (29) Since f E > 0 and τ >, it foows tat, wen te wages in importing owtecnoogy firms ( W ) are significanty iger tan te wages in ow-tecnoogy firms ( W ), wic was aready assumed. Terefore, ony te most productive ow-tecnoogy firms tat wi be abe to compensate for iger exporting costs wi start exporting and importing. To sum up, after trade iberaisation in te second stage of te mode, ony te most productive ow-tecnoogy firms coose to upgrade skis and to start exporting and importing, ess productive ow-tecnoogy firms ony import, and te east productive owtecnoogy firms continue serving te domestic market. On te oter and, ow-tecnoogy firms wi not decide to engage in exporting activities witout increasing teir eve of productivity by importing, as teir productivity eve woud be too ow to bear exporting costs. Te mode now focuses on evauating te foowing four options of ig-tecnoogy firms after trade iberaisation in te second stage: a) No trade, use ig tecnoogy (2): r ( ) fw w E ( ) S U, (30) 27

42 28 were ) ( are te tota profits of firms wit ig-tecnoogy eves, and ) ( r are te revenues, wit ) ( U S w w XP r. b) Start importing, use ig tecnoogy (2): Wen ig-tecnoogy firms start importing, teir costs and productivity eve increase and add up to: U S w w q f f TC, were, and. ntroducing factors and enabes controing for te increase in productivity eve and te canges in ski structure in favour of te skied empoyees after ig-tecnoogy firms start importing. n addition, firms carge te price: U S w w p. Taking tese facts into account, te profit is: U S w w f f r ) ( ) ( ) (, (3) were ) ( are te tota profits of ig-tecnoogy firms tat start importing, and () r are te revenues, wit ) ( U S w w XP r. c) Start exporting, use ig tecnoogy (2): Wen ig-tecnoogy firms start exporting, teir costs add up to: U S E E w w q f f TC. Consequenty, firms carge te price: U S E w w p. Taking tese facts into account, profit is as foows: ) ( ) ( ) ( U S E E E w w f f r, (32) were ) ( E are te tota profits of ig-tecnoogy firms tat start exporting, and () E r are te revenues, wit ) ( U S E w w XP r.

43 29 d) Start importing and exporting, use ig tecnoogy (2): Wen ig-tecnoogy firms start importing and exporting, teir costs add up to: U S E E w w q f f f TC. Consequenty, firms carge te price: U S E w w p. Taking tese facts into account, te profit is: U S E E E w w f f f r ) ( ) ( ) (, (33) were ) E ( are te tota profits of ig-tecnoogy firms tat start importing and exporting, and ) ( E r are te revenues, wit ) ( U S E w w XP r. Te foowing two expressions are considered wen comparing te zero-profit bounds of ig-tecnoogy firms tat do not engage in internationa activities and of tose wic start importing in te second stage of te mode: W f f r W f r ) ( ) ( ) ( ) ( ) (. (34) For convenience, te abbreviation for te tota abour costs in ig-tecnoogy firms ( W ) is appied, wie tota abour costs of importing ig-tecnoogy firms are denoted by U S w w W. To cacuate te eve of productivity in importing ig-tecnoogy firms,, te subsequent two expressions are compared: ) ( ) (, yieding te foowing: W W W f W W f A ) (. (35) n order for tis expression to be positive, 0, it is important for te foowing expression to od: W W. Since W W, it foows tat te margina cost reduction of ig-tecnoogy firms tat do not engage in internationa activities ( ) and of tose wic start importing in te second stage of te mode ( ), soud not differ substantiay. Tis coincides wit te assumption from te previous part of te paper, stating tat importing brings ower margina cost reduction, compared to te margina cost reduction due to investment into ig-tecnoogy.

44 n addition, te eve of productivity of ig-tecnoogy domestic firms,, and te eve of productivity of ig-tecnoogy importing firms,, is compared as we. Te expression appies, as ong as W W W (. Te atter f W W ) W f ( W W ) fw expression is vaid wen te wages in ig-tecnoogy firms ( W ) are significanty iger tan te wages in ow-tecnoogy firms ( W ). Aso, te wage eve in ig-tecnoogy firms soud increase substantiay as a consequence of importing ( W ). Again, foowing te concusions made wen studying te ski upgrading at te eve of individuas, bot presumptions signa a iger empoyment eve of skied workers and were aready assumed in te previous part of te paper. Next, te foowing two expressions are considered wen comparing te zero-profit bounds of ig-tecnoogy firms tat do not engage in internationa activities and of tose wic start exporting in te second stage of te mode: E E r ( ) r ( ) ( ) ( ) fw ( f f E ) W. (36) To get te eve of productivity, above wic a ig-tecnoogy firm finds it profitabe to E E E E start exporting,, te subsequent two expressions are compared: ( ) ( ), yieding te foowing: E W A f EW. (37) Since it was aready assumed tat, te productivity eve of ig-tecnoogy exporting E firms wi be positive; 0. n addition, te eve of productivity of ig-tecnoogy domestic firms,, and te eve of productivity of ig-tecnoogy exporting firms, is compared as we. Te expression W f ( W W ) W f W E W E, E appies, as ong as. Te atter expression confirms tat ony te most productive ig-tecnoogy firms, wic wi be abe to compensate for iger exporting costs, start exporting. By confirming tat te most productive ig-tecnoogy firms engage in trading activities after trade iberaisation in te second stage due to teir initia iger eve of productivity, it is necessary to compare te zero-profit bounds of ig-tecnoogy firms tat start importing and of tose wic start exporting in te second stage of te mode: 30

45 E E r ( ) r ( ) ( ) ( ) ( f f ) W ( f fe ) W. (38) Hig-tecnoogy firms coose between te start of importing and exporting on beaf of teir productivity eve; ig-tecnoogy firms decide to import if teir productivity eve is not yet ig enoug to start exporting, wereas more productive ig-tecnoogy firms start exporting in order to increase teir revenues. Tis makes it possibe to compare te productivity eves of ig-tecnoogy firms tat start importing ( ) and ig-tecnoogy E firms tat start exporting ( ) and see tat ig-tecnoogy firms start importing, wen te eve of bound productivity is iger; i.e. E. Te atter f EW W f( W W ) f W W W expression appies if te wage eve in ig-tecnoogy firms ( W ) is significanty ower tan te wage eve in ig-tecnoogy importing firms ( W ): W W, wic is again a sign of a iger empoyment eve of skied workers. Moreover, te decision between te start of importing and exporting wi depend on externa factors; i.e. te cost eve of importing and exporting. f te costs of importing are significanty iger tan te costs of exporting, ony te most productive ig-tecnoogy firms wi be abe to afford importing. n contrast, wen te opposite ods, ony te most productive ig-tecnoogy firms wi be abe to afford exporting. Te next step compares te zero-profit bounds of importing ig-tecnoogy firms and of ig-tecnoogy firms tat start importing and exporting in te second stage of te mode: E E r ( ) r ( ) ( ) ( ) ( f f ) W ( f f f E ) W. (39) t foows tat ig-tecnoogy firms wi find exporting and importing profitabe ony if te increase in revenues wi be bigger tan te increase in costs of exporting. To get te eve of productivity, above wic a ig-tecnoogy firm finds te start of importing and exporting profitabe,, te subsequent two expressions are compared: E E E ( ) ( ), obtaining te foowing: E E W f W A E. (40) Tis sows wen te productivity eve of ig-tecnoogy firms tat import ( ) is ower tan te productivity eve of ig-tecnoogy firms tat export and import ( E ): 3

46 ( ) W W f W W W f W W E f E. Again, te atter expression appies if te wage eve in ig-tecnoogy firms ( W ) is significanty ower tan te wage eve in ig-tecnoogy importing firms ( W ): W W. Findings from te part of te mode, studying te ski upgrading at te eve of individuas, again indicate iger wages being a signa of a iger empoyment eve of skied workers. n addition, if te costs of importing are significanty iger, compared to te costs of exporting, ony te most productive firms wi be abe to afford te start of importing. Finay, since te decision of ig-tecnoogy firms on wen to start exporting and importing depends aso on externa factors; i.e. te cost eve of exporting and importing, te anaysis from te previous paragrap as to be repeated for ig-tecnoogy firms tat decide between starting to export, and starting to export and import. Terefore, te zeroprofit bounds of exporting ig-tecnoogy firms and of ig-tecnoogy firms tat start importing and exporting in te second stage of te mode are compared wit te foowing expressions: E E E E r ( ) r ( ) ( ) ( ) ( f fe ) W ( f f f E ) W. (4) From tis it foows tat ig-tecnoogy firms find exporting and importing profitabe ony if te increase in te eve of productivity is bigger tan te increase in costs of importing. To get te eve of productivity, above wic a ig-tecnoogy firm finds it profitabe to start importing and exporting, E E E E compared: ( ) ( ), yieding te foowing: E, te subsequent two expressions are ( )( ) f f W W f W ( W W ) E A E. (42) One can now ceck wen te productivity eve of ig-tecnoogy firms tat export ( ) is ower tan te eve of ig-tecnoogy firms tat export and import ( ): E E f EW W ( f f E )( W W ) f W ) W W. Te atter expression appies if te wage eve in ig-tecnoogy firms ( W ) is significanty ower tan te wage eve in ig-tecnoogy importing firms ( W ), wic again signas a iger empoyment eve of skied workers after importing. Concerning externa factors, if te costs of importing are significanty ig, ony te most productive ig-tecnoogy firms wi be abe to engage in bot, exporting and importing. E E 32

47 To sum up, after trade iberaisation in te second stage of te mode, ony te east productive ig-tecnoogy firms serve ony te domestic market, were te decision on weter to start importing, exporting or bot depends on te eve of wages before and after importing, on te firm s productivity eve and on externa factors; i.e. te eve of export and import costs. nterestingy, wen ig-tecnoogy firms decide weter to start exporting or not, te fina decision is not based on te wage eve of ig-tecnoogy nontrading firms and ig-tecnoogy exporting firms. Making inferences from te first part of te paper wic studied te ski upgrading at te eve of individuas, tis woud be a sign of a iger empoyment eve of skied empoyees. Terefore, ski upgrading occurs ony in importing firms or firms tat engage in bot; importing and exporting..4 Concusion Te teoretica modes of trade ave been evoving troug istory in a desire of a toroug interpretation of internationa fows. Recent teoretica trade modes account for firm eterogeneity, and aso for tecnoogy and ski upgrading. Guided by tese teories, deveoped a teoretica mode, wic expores te individua s decisions for investing in ski upgrading and te firm s decisions to start tecnoogy upgrading and trading. Te mode in tis paper is divided in two parts. First part expores te beaviour of individuas and teir decisions on weter to invest in acquiring iger ski eves. ndividuas ave a coice to acquire iger skis, were te decision depends on teir abiity eve. Te findings suggest tat since te education costs of ow abiity workers for acquiring iger skis are excessive, ony ig abiity workers acieve iger ski eves. n addition, as a resut of ig education costs, ig abiity, ig skied workers demand iger wages after entering empoyment. Higer wages terefore signa iger empoyment of skied workers. Tis finding is brougt into use in te second part of te mode, wic takes into account te firm s decisions on weter to invest in iger tecnoogy and weter to engage in internationa activities. Te mode suggests tat before trade iberaisation, ony te most productive firms invest in acquiring iger tecnoogy eves. Higer abour costs of tese firms signa a iger empoyment of skied workers. After trade iberaisation, costs of importing and exporting diminis and firms ave an option to start engaging in internationa activities. Tis part again takes into account findings from te first part of te mode, tat te iger abour costs of firms signa iger empoyment of skied workers. Taking into account ow-tecnoogy firms first, te most productive ow-tecnoogy firms coose to ski upgrade and to start exporting and importing, ess productive ow-tecnoogy firms aso upgrade skis but start ony importing, and te east productive ow-tecnoogy firms continue serving ony te domestic market. Low-tecnoogy firms terefore use importing as means of increasing teir productivity eve before te start of exporting. Tis finding was confirmed aso in empirica studies (see for exampe Damijan & Kostevc, 205; and Atomonte & Békés, 200). Furtermore, ow-tecnoogy firms do not engage excusivey in exporting, as teir productivity eve is too ow to cover exporting costs. Wit regards to ig-tecnoogy 33

48 firms, ony te east productive ig-tecnoogy firms do not start importing and/or exporting after trade iberaisation, were te decision on weter to import, export, or bot, depends on te firm s productivity eve, te ski upgrading before and after importing, and on externa factors; te eve of export and import costs. Ski upgrading in ig-tecnoogy firms after trade iberaisation takes pace ony in firms tat start importing, or tat start engaging in bot, importing and exporting. Te mode igigts severa facts, wic woud be noteworty of furter empirica testing. One coud empiricay anayse te foowing findings of te teoretica mode: (i) firms wit better ski structure aso start importing; (ii) importing firms ave a better ski structure tan non-importing firms; and (iii) by aving an access to tecnoogy and/or to intermediates, imports serve for increasing te tecnoogy eve and/or cost reductions before te start of exporting. Te key contributions of tis mode are a differentiation between importers and exporters, and a toroug anaysis of te beaviour of individuas and firms, were te connection between te two as been made by inking fragments of modes on te individua s and te firm s beaviour. Te possibe imitations of te mode present additiona assumptions, wic ad to be made wen deveoping te mode; e.g. te increase in te wage eve of skied workers after investing in ig tecnoogy 34 W W, and after te start of W importing, compared to te decrease in te margina costs in tese firms. W Furtermore, te mode aso assumes tat te productivity eve increase after importing is ower compared to te productivity eve increase after investing in ig-tecnoogy. Tese additiona assumptions to some extent imit te vaue of te mode, as it woud be ard to test tem empiricay. Moreover, foowing Bustos (20b), and Stark and oters, te mode assumes a discrete distribution of skied, wie it woud be more reaistic to assume a continuous distribution of skied and to take into account aso te semi-skied workers. Even toug tis mode extension woud greaty increase te compexity of te mode and considering tat te empirica studies aso use discrete variabes for defining workers skis, it woud be usefu to take tis imitation into account in te future mode extensions. n addition, atoug te mode considers tree dynamic pase sifts; i.e. te individua s decision to acquire skis, te firm s decision to opt for ig tecnoogy, and te firm s decision to start importing and/or exporting, it is imited in discussing ony two firm s decisions simutaneousy (e.g. ig-tecnoogy vs. ow-tecnoogy, no trade vs. importing, etc.). Since nowadays firms face te canging environment wic demands compex decision-making on a daiy basis, tis structure of te mode woud be imited to transform in everyday environment. Atoug osing a more static structure of te mode woud greaty increase its compexity, tis imitation woud be usefu to be taken into account in furter studies. Neverteess, despite te aforementioned sortcomings, beieve te mode s concusions bring contributions to te fied of knowedge, since te concusions

49 are aso consistent wit previous empirica findings and open severa possibiities for furter empirica anayses. 35

50 2 UNTANGLNG THE RELATONSHP BETWEEN SKLL STRUCTURE, MPORTS AND EXPORTS 2. ntroduction Exporing te sources of iger productivity in firms as been of great interest in various fieds of researc wic sow tat firms iger productivity can be attributed to touger market competition, tecnoogica spiovers, uman capita and internationa trade (Syverson, 20), to name ony a few. Firms, engaged in trade, tend to be more productive due to cost reductions and tecnoogica transfers, wic can be acieved by offsoring, outsourcing and suppy cain management (Onodera, 2008). Anoter reason for observed superior performance of trading firms is te sef-seection of more productive firms into trading activities (Aw et a., 20; Aw et a., 2008; Greenaway & Kneer, 2004; Meitz, 2003; Voge & Wagner, 2008; Wagner, 2007). An aternative source of iger productivity in firms is te empoyment of skied empoyees, wo use given resources more efficienty and can adopt and start using new tecnoogies more quicky (Corvers, 997). Tese determinants of firm performance invovement in trading activities and te engagement of skied empoyees ave aso a positive impact on one anoter. More precisey, empirica studies prove tat imports ave an important impact on te demand for skied workers (Burstein et a., 203; Parro, 203; and Rave & Resef, 203). Furtermore, imports aso ave a positive impact on te start of exporting activities (Damijan & Kostevc, 205; and Wagner, 202b). Te aim of te anaysis is to test some of te impications made in te teoretica part of te previous capter. An additiona aim is to add to te empirica studies wic exposed te impact of trade status on te firms ski structure, and te interaction between te firms ski structure, importing and exporting. Te foowing ypoteses, wic were, among oters, discussed in te teoretica mode, were cosen to be empiricay anaysed in tis study: (i) firms wit better ski structure start importing; (ii) importing firms increase teir ski sare after te start of importing; and (iii) aving access to tecnoogy or to intermediates troug importing increases te probabiity of exporting. Te motivation for coosing tese particuar ypoteses was to contribute to te findings of empirica studies, wic took into account te sequencing between importing and exporting (for instance Damijan & Kostevc, 205), and empoyment and exporting (for instance Yang & Maick, 200). n addition, wen coosing wic concusions of te teoretica mode to test empiricay, some data imitations were considered as we. Te empirica anaysis terefore examines te differences in te eves of te ski structures of importers and non-importers, and import starters and non-importers. n addition, te anaysis aso examines te impact of aving access to ceaper intermediates via imports (measured by imports of intermediate goods), and te impact of aving access to tecnoogies via imports (measured by imports of capita goods) on te start of exporting activities. To te best of my knowedge, no suc anaysis of te casua inks and 36

51 sequencing between te ski structure of firms, imports, and exports as been done yet. Te purpose of tis study is to fi tis void in te iterature and to provide some poicy impications. Anoter motivation for studying te inkages between imports, exports and te ski structure of firms is in te increasing importance of internationa trade and skied workforce in te ast decade in most of te deveoped word. According to Eurostat, te sare of empoyees wit attained tertiary education as been steadiy increasing in te EU, wereas te increase of te sare in Sovenia was above te EU average. n te ast decade, te sare of empoyees wit attained tertiary education as increased by 8.2 percentage points in te EU-28 and by.8 percentage points in Sovenia. Additionay, te empoyment rates of persons wit attained tertiary education ave aways been te igest, reacing 82. % in te EU-28 in 204 (82.0 % in Sovenia), wie te tota empoyment rate in te EU-28 was 64.9 % (63.9 % in Sovenia). On te oter and, te increase of imports and exports measured as te vaue of EUR as been even iger. Atoug internationa trade decined during te crisis, te overa increase of imports reaced 49.2 % in te EU-28 (79.7 % in Sovenia) in te ast decade, wie te increase of exports was even iger, reacing 53.0 % in te EU-28 (06.7 % in Sovenia) (Eurostat). n order to study te above ypoteses tat are based on te impications of te forma mode outined in te previous capter, a inked empoyer-empoyee pane dataset for Sovenian manufacturing firms is used, covering te period from 996 to 200. Wit te aim of considering evident differences between importing and non-importing firms, te propensity score matcing approac is appied. Te dataset is constructed from severa data sources wit information on te financia figures of firms, te vaues of imports and exports, and te caracteristics of empoyees. Resuts confirm some previous findings and deiver new insigts to understanding te inkages between imports, exports and ski upgrading. Resuts confirm previous findings tat importing activities ave a positive impact on te demand for skis (see for exampe Crino, 202; and Mesci et a., 2008). n addition, firms wit a iger ski sare sef-seect into importing and continue to ave a iger ski sare, compared to non-importing firms. Wen anaysing te effect of starting to import intermediate or capita goods on te start of exporting, find tat importing intermediate goods as an immediate positive impact on te start of exporting in te year after te start importing. On te oter and, importing of capita goods as a positive impact on te start of exporting not earier tan in te second year after te start of importing capita goods. Tese resuts point to a different roe of te capita and intermediate goods in te production process. Since intermediate goods usuay ave a reativey sort expiration date and require additiona manufacturing processing or are used for resae, te impact of importing tese goods migt be instantaneous but sortterm. n contrast, since capita goods generay present firms fixed assets and are empoyed in te production process, te impact of importing tese goods migt appear graduay but ast onger time. 37

52 Besides contributing to te fied, te paper aso adds insigts into poicy impications. Based on te resut of tis empirica study, it is important tat poicy-makers focus on estabising an environment tat stimuates internationa cooperation. Next, creating incentives for a iger empoyment of skied empoyees aso seems to be important for te start of importing, wic consequenty as a positive impact on te start of exporting. n turn, importing pays an important roe in boosting empoyment of skied individuas. Te remainder of te paper is organised as foows. n te next section, a brief summary of te reevant iterature is given. Section tree introduces te empirica mode, describes te data and presents te descriptive statistics. Te basic resuts, extensions of te mode and a sort discussion are incuded in section four. Te ast section summarises and concudes. 2.2 Literature review Existing studies sow tat greater trade openness is one of te main reasons for increases in te demand and suppy of more educated abour, wic as been evident trougout te deveoped word (Foster et a., 202; Mesci et a., 2008; Attanasio et a., 2003; Tokarick, 2002; Muender, 2004; Feenstra & Hanson, 999). However, unti recenty, te majority of abour economists agreed tat tecnoogica improvements were te eading reason for a steady increase in te suppy and demand for skied abour, wie a positive impact of trade as been widey downpayed. Neverteess, recent studies, wic incude onger anaysis periods, expand teir researc aso on emerging markets and use up-to-date estimation metods, confirm te roe of trade in sifting te demand towards more educated abour. Te crucia finding is tat trade is an important driver of tecnoogica cange and consequenty as a great impact on upgrading te ski structure of firms and teir innovation activities (Boom et a., 20; Crino, 202; Mesci et a., 2008). Moreover, severa empirica studies confirm tat trading firms empoy more educated abour (Brambia et a., 200) and pay iger wages (Hepman et a., 20; and Scank et a., 2007). Tis is due to te fact tat exporting firms are on average arger and more productive tan teir non-exporting counterparts and due to te screening of potentia empoyees more intensivey. Accordingy, te workforce of exporting firms as an aboveaverage abiity and wages (Hepman et a., 20). nitiay, empirica studies focused mainy on exporing exporting status and oter determinants, wereas current studies empasise te importance of incuding aso importing status in te anaysis (Damijan & Kostevc, 205; and Muus & Pisu, 2007). Productivity gains due to greater access to imports were for exampe proven by Amiti and Konings (2007), wo make a distinction between productivity gains, wic are foowed by ower tariffs on fina goods and te ones tat are foowed by ower tariffs on intermediate goods. Wie ower output tariffs increase productivity by increasing import competition, ower input tariffs increase productivity due to access to ceaper imported inputs (Amiti & Konings, 2007). Simiary, trade iberaisation in ndia granted access to a greater number of imported inputs for firms, wic resuted to an increased number of 38

53 products introduced by domestic firms (Godberg et a., 200). Bas and Strauss-Kan (20) expicate te importance of imported inputs by confirming a strong impact of imported inputs on firms productivity and export performance in France, were te impact on productivity is more pronounced for imports from deveoped countries, compared to imports from deveoping countries. Aso, a iger number of imported inputs and/or a more diverse spectrum of imported inputs increase te probabiity to survive in export markets as tis enabes firms to cover te fixed costs of exports (Bas & Strauss-Kan, 20). A positive impact of imports on firms future exporting activities was confirmed aso by Wagner (202a). Furtermore, te composition of imports as been recognised as being important as we. n particuar, a reduction of trade costs increases trade in capita goods, wic in turn eads to an increase in te ski premium i.e. te wage of skied abour, reative to te wage of unskied abour and wefare gains for skied abour. Te mecanism at work ere is te capita-ski compementarity, wic creates te ski-biased trade (Burstein et a., 203; and Parro, 203). Rave and Resef (203) aso confirm tat te composition of imports infuences te demand for skied abour and te ski premium. Autors concude tat R&D-intensive capita equipment is compementary to skied abour, wie ess innovative capita is compementary to non-skied abour. Consequenty, te imports of R&Dintensive capita equipment raise te ski premium, wereas te imports of ess innovative capita ower te ski premium (Rave & Resef, 203). Wie te majority of papers focus teir researc on finding te types of correations between te determinants of firms productivity, more recent papers aso focus on exporing te directions of causaities between tese determinants. Once more, studies mainy focus teir attention on exporting firms (see for exampe Damijan et a., 200; and Yang & Maick, 200), wie te evidence on importing firms is scarce. n teir recent paper, Damijan and Kostevc (205) incude aso importers in teir researc and confirm te ink between imports, exports and innovation, using te Spanis microdata. Teir concusion is tat firms predominanty earn from importing, wic in turn as a positive effect on innovation and finay on exports. Te atter ten paves te way for furter innovation. Te resuts are prominent especiay for smaer firms, were imports enabe tese firms to introduce new production processes and to improve te caracteristics of teir produced goods. Damijan and Kostevc (205), owever, did not contro for te ski structure of firms in teir researc. Empirica studies tat expore te reationsip between imports and te ski structure of firms usuay find a positive impact of imports on te ski structure of firms. Mesci et a. (2008) demonstrate tat sectors wit te igest increase of imported inputs, reative to tota inputs, aso ave te igest reative increase of skied workers abour costs. Autors expain tis increase as a consequence of transferring te ski-intensive tecnoogies wit imports, wic contributes to a ski-biased increase of abour demand in favour of te skied workers. Moreover, teir resuts are backed up by te fact tat ony 39

54 imports from industriaised countries indicate te transfer of new and ski-intensive tecnoogies (Mesci et a., 2008). Focusing on te effects of service offsoring on empoyment, were service offsoring is defined as te sare of service inputs in te tota non-energy inputs, Crino (202) aso confirms te ski-biased effects of offsoring. To be precise, offsoring increases te demand for ig- and medium-skied abour, wie inders te demand of ow-skied abour. Te issue of trade and ski structure of firms was aso covered by recent papers on te effect of trade iberaisation in deveoping countries. Focusing especiay on Cina, tese studies sow tat reducing te obstaces of importing from deveoping countries as on average a positive impact on te ski structure of firms in deveoped countries, wie it adversey affects te empoyment of ess-educated abour (Pierce & Scott, 202; Boom et a., 20; and Mion et a., 200). By studying te effects of trade iberaisation witin a deveoping country, Fieer et a. (204) confirm increases in te ski premium, ski intensity and quaity upgrading among exporters. As a spiover effect, tis in turn induces quaity upgrading among oter domestic firms, wie te east productive firms cange for te worse and become ess ski intensive. Boom et a. (20) give severa expanations for te positive infuence of trade iberaisation on firms ski structure. Firsty, trade iberaisation increases te opportunities for empoying abour and capita, wic in turn reduces te costs of innovation and te production of new goods. n te wake of te aternative expanation, te iberaisation of internationa trade increases competition, wic in turn fosters innovation. Moreover, owering trade barriers aso enarges te market size, wic in sequence reaocates te fixed costs of innovation to a iger number of agents and enabes firms to sare te knowedge more easiy. Finay, as a consequence of trade iberaisation and ence te decreasing of trade costs in emerging markets, firms in deveoped countries sift teir product mix towards more tecnoogicay advanced products and consequenty benefit wen using intermediates from te emerging markets (Boom et a., 20). Motivated by te above-mentioned study of Damijan and Kostevc (205), wic studied te sequencing patterns between importing, exporting and innovation, and defined te importance of importing for te start of oter activities, decided to expore te sequencing between importing and exporting more in dept, by incuding aso te type of imports into te anaysis. As aready presented in te introductory part, intermediate and capita goods ave different roes in firms production processes. Terefore, it is important to test weter te start of importing tese goods as a significant impact on te start of exporting and weter tere are differences in te impact on te start of exporting wen controing for te types of imported goods. Te mecanism beind tis, i.e. te earning-byimporting, was introduced in te teoretica mode in te previous capter. Te mode concudes tat troug continuous process of importing, firms get an access to tecnoogy and/or an access to intermediates, wic in turn increases firms productivity before te start of exporting. Furtermore, since one of te concusions of te teoretica mode aso 40

55 points to a iger empoyment of skied abour in more productive firms and since te presented statistica data point to an increasing sare of tertiary educated empoyees in te country of interest in tis study, Sovenia, aso introduce te ski structure of firms in te anaysis. Distinguising between te imports of intermediate and capita goods, and introducing te ski structure of firms in te anaysis are terefore te main contributions to te existing iterature. 2.3 Metodoogy and data As stated before, te teoretica mode from te previous capter provides a framework for te empirica mode in te present capter. Te mode tests severa ypoteses, derived from te teoretica mode: (i) correation between importing and a better ski structure of firms is positive; (ii) firms wit a better ski structure start importing; (iii) ski structure canges in favour of te skied after te start of importing; (iv) importing capita goods (wic acts as a proxy for access to tecnoogy via imports) as a positive impact on te start of exporting; and (v) importing intermediate goods (wic is a proxy for access to ceaper intermediates via imports) as a positive impact on te start of exporting. Te ast two points refer to te finding of te teoretica mode, wic concudes tat firms use importing in order to increase teir eve of productivity before te start of exporting. mporting increases te firms productivity by making a production process more cost effective. Firms can acieve tis in two ways. Te first is by importing iger quantities of more affordabe intermediates, or importing intermediates of iger quaity from abroad. Te second is by importing capita goods tat are more affordabe or of iger quaity, wic can in turn increase te tecnoogica advancement and cost-effectiveness of firms production processes. Te atter migt in turn aso demand more sopisticated intermediate inputs, meaning te ast two points migt occur and in and. Te foowing subsections first present te metodoogy and ten continue by describing te data and presenting te main descriptive statistics Metodoogy Wit te aim of anaysing te presented ypoteses, appy propensity score matcing, wic was aso used wen studying sequencing patterns between importing, exporting and innovation (Damijan & Kostevc, 205), and exporting and firm performance (Yang & Maick, 200). Te abovementioned papers serve as a bencmark to tis study. Propensity score matcing is used in order to expore different beavioura patterns of firms tat sare simiar caracteristics. Moreover, te statistica iterature suggests aso oter advantages of propensity score anaysis; e.g. te propensity score does not rey on te correct specification of te functiona form of te reationsip, it makes a more unambiguous comparison between treated and contro units and is more objective, as modeing and te outcome anaysis are done separatey (Zanutto 2006; Hi et a., 2004; Becker & cino, 2002; and Rubin, 997). Te estimation of te average treatment effects is based on 4

56 propensity scores. foow te definition of Rosenbaum and Rubin (983), wo define it as te conditiona probabiity of receiving a treatment, given te pre-treatment caracteristics. Te metodoogy for estimating individua ypoteses is described in greater detai beow Hypotesis : Correation between importing and a better ski structure of firms is positive n order to estimate te first ypotesis, expore te differences in te ski structures of importers and non-importers. n accordance, te treatment group consists of importing firms and te contro group consists of non-importing firms. Since te aim of te first ypotesis is to anayse te correation between importing and ski structure of firms, and since tis ypotesis does not differentiate between importing starters nor does it measure te sequencing and causaity, variabes in te expressions for estimating te propensity score and te average treatment effect on te treated are not agged. t is important to discuss aso te possibe endogeneity issue, inked wit tis ypotesis. Higer sare of skied workers in importing firms migt not necessariy refect tat firms are igy-tecnoogica, as te reason for teir better ski structure migt be due to te arge voumes of importing, wic does not demand firms to ave teir own production. As a resut, arge voumes of importing migt cange te ski structure of firms in favour of te skied. Hypoteses 2 and 3 aim to overcome tis issue by focusing on importing starters, were te second ypotesis anayses weter a better ski structure is a prerequisite for te start of importing, and te tird ypotesis anayses weter ski structure canges in favour of te skied after te start of importing. Hypotesis is summarised in (43): MP t in (43) reates to importing, and Ski _ sare f ( MP ) (43) t t Ski _ saret reates to firms ski sare in te period t. Foowing existing iterature, workers are defined as skied if tey attain at east some form of coege degree (Boom et a., 20; Tokarick, 2002; and Badwin & Cain, 2000), wic is typicay 4 years of educationa attainment in Sovenia. Te foowing mode as been used for estimating te propensity score ( p X ) (probit estimation), were te propensity score equas te probabiity of being an importer, based on certain pre-treatment caracteristics of firms, X : X f ( Size, Lvae, Lke, Rims, Foreign, FD, Time, nd ) p (44) it it it it it it t t Te expanatory variabes in te mode (44) are te foowing: ogaritm of te number of empoyees in a firm ( Size it ), ogaritm of te vaue added per empoyee ( Lvae it ), ogaritm of capita per empoyee ( Lke ), regiona import sare, as a measure of regiona it 42

57 externaity ( Rims it ), dummy variabe, controing for te foreign ownersip of a firm ( Foreign it ), and a dummy variabe, controing for firms foreign direct investments abroad ( FD it ). Variabe Time t contros for year specific effects and nd t denotes industry dummy variabes (2-digit NACE rev. industries). Te average treatment effect on te treated (ATT) is ater computed in te foowing way: Ski sare px, MP ESki_ sare px ATT E, MP 0 (45) _ it t it t n equation 45, ATT reports te difference in te ski sare between importing firms ( MP t ; treatment group) and non-importing firms ( MP t 0 ; contro group), were te outcome of interest, Ski_ sareit, refers to te ski structure of firm i in year t, and X p refers to te propensity score from te equation (44) Hypotesis 2: Firms wit a better ski structure start importing Te second ypotesis expores weter firms wit a better ski structure sef-seect into importing. Testing te ypotesis foows te procedure, presented in Yang and Maick (200). Te treatment group consists of firms tat start importing in a particuar year, and te contro group consists of non-importing firms. Hypotesis 2 is summarised in (46): Prob( MPstartt ) f ( Ski _ sare t s ) (46) Te variabe MPstart t denotes importing starters, wic start importing in te period t and wic ave not been importing in te previous years. Te variabe Ski _ sare t s denotes te firm s ski sare. n order to empiricay anayse te ypotesis, te variabe of interest; i.e. te ski sare, is agged for one period (s = ) in te base resuts, wereas in te extensions of te mode, it is agged for two periods (s = 2). As a resut, wen cacuating te propensity scores, te contro variabes are agged for one or two periods, respectivey, and are parae to te ones in te mode (44). Te foowing mode as been used for estimating te propensity score ( p X 2 ) (probit estimation): p X f ( Size, Lvae, Lke, Rims, Foreign, FD, Time, nd ) (47) 2 its its its its its its t t For testing te second ypotesis, te ATT is computed in te foowing way: 2 E Ski_ sare ATT E Ski_ sare its its p X 2 p X 2, MP 0 t, MPstart t (48) n equation (48), ATT 2 reports te difference in te ski sare between importing starters ( MP start t ; treatment group) and non-importing firms ( MP t 0 ; contro group), were te outcome of interest, Ski_ sare it s, refers to te ski sare of firm i in te year s 43

58 before te start of importing, t refers to te entrance year of importing, and p X 2 refers to te propensity score from te equation (47) Hypotesis 3: Ski structure canges in favour of te skied abour after te start of importing Te next ypotesis examines weter ski structure of importing starters canges in favour of te skied after te start of importing, using differences-in-differences matcing estimator, wic enabes estimating te effect on te cange in te outcome variabe before and after te treatment. Tis step restricts te treatment variabe compared to te previous mode (equations 47 and 48) by taking into account ony te firms tat start importing in a particuar year and ave not been importing in te previous years, but continue importing at east one year after te start of importing. n te extended mode, wic concentrates on te impact in te second year after te start of importing, te treatment variabe takes into account importing starters tat continue importing at east two years after te start of importing. Te contro group again consists of non-importing firms. Wen cacuating propensity scores, contro variabes are agged for one period and parae to te ones in te mode (47). Two different metods for cacuating te ATT were used wen testing te tird ypotesis: 4 3 E Ski_ sare ATT E Ski_ sare E Ski_ sare ATT E Ski_ sare its its its its p X 2 Ski_ sare p X 2, MP Ski_ sare it it p X, MPstart t 2 0, MP p X 2 t ts, MP 0, MPstart, MP 0, MP t t ts ts, MP 0 ts (49) (50) Te basic mode concentrates on te impact one year after te start of importing (s = ), wie te extensions of te mode concentrate on te impact in te second year after te start of importing (s = 2). Tis step takes into account two different outcomes of interest and ence cacuates two different average effects of treatment on te treated; ATT 3 and ATT 4. Te outcome of interest in te former is te ski sare in year s after te start of importing ( Ski_ sare it s ), wie te outcome of interest in te atter is te cange in te ski sare in year s after te start of importing ( Ski _ sareit s Ski_ sareit ). Te treatment group in te basic mode consists of importing starters tat ave not been importing in te previous years and continue importing at east one year after te start of importing ( MP start t, MP t ), wie in te extensions of te mode, it consists of importing starters tat ave not been importing in te previous years and continue importing at east two years after te start of importing ( MP start, MP, MP ). Te contro group consists of non-importing firms t t t2 44

59 ( MP t 0, MP t s 0 ). Again, te t in equations (49) and (50) refers to te entrance year of importing Hypoteses 4 and 5: mporting capita or intermediate goods as a positive impact on te start of exporting Finay, te ast two ypoteses take into account te effect of starting to import on te start of exporting, were two different imported types of goods, intermediate and capita, were taken into account. Te procedure for estimating te propensity scores in tis step again coincides wit te equation (47), were te contro variabes are agged for one period. Te cassification of Broad Economic Categories (BEC) was used for defining intermediate and capita goods. Hypoteses 4 and 5 are summarised in (5) and (52), respectivey: Prob EXPstart ) f ( MPstart_ c ) (5) ( t s t Prob EXPstart ) f ( MPstart_ i ) (52) ( t s t EXPstartt s presents te start of exporting in period t+s, MP start_ ct presents te start of importing capita goods, and MP start_ it te start of importing intermediate goods in period t. Te two average treatment effects on te treated, ATT 5 and ATT 6, are cacuated in te foowing way: 5 ATT E EXPstart E EXPstart 6 E EXPstart its its its p X ATT E EXPstart 2 its p X p X 2 2, MPstart_ c, MP 0, MP p X 2 t, MP 0, MP ts, MPstart_ i t ts t 0 t, MP_ c 0, MP_ i ts ts (53) (54) Te outcome of interest in ATT 5 is te start of exporting in year s after te start of importing capita goods ( EXPstartit s ), were t in equation (53) refers to te year wen a firm started importing capita goods. Te treatment group in te basic mode consists of firms tat start importing capita goods and continue importing tese goods one year after te start of importing ( MP start_ ct, MP _ c t ), wie te treatment group in te extensions of te mode consists of firms tat start importing capita goods and continue importing tese goods in te first and te second year after te start of importing ( MP start_ ct, MP _ c t, MP _ c t2). Te treatment group takes into account firms tat start importing capita goods in a particuar year and ave not been importing tese goods in te previous years. Te treatment owever does not restrict imports of oter 45

60 types of goods in te years before te start of importing capita goods. Focusing ony on pure importing starters of capita goods; i.e. firms tat ave not been importing any types of goods before te start of importing capita goods, woud greaty reduce te sampe of firms in te treatment group (for 85.4 %). Te contro group consists of non-importing firms ( MP t 0, MP t s 0 ). t woud aso be interesting to contro weter firms import simiar products tat tey produce and ater se or export, and terefore contro for te pass-on trade (Damijan, Konings & Poanec, 203). However, since te data does not currenty enabe tis anaysis, it woud be noteworty to incude tis test in te future anaysis. Meanwie, te outcome of interest in ATT 6 is te start of exporting in year s after te start of importing intermediate goods ( EXPstartit s ), were t in equation (54) refers to te year wen a firm started importing intermediate goods. Te treatment group in te basic mode consists of firms tat start importing intermediate goods and continue importing tese goods one year after te start of importing ( MP start_ it, MP _ i t ), wie te treatment group in te extensions of te mode consists of firms tat start importing intermediate goods and continue importing tese goods in te first and te second year after te start of importing ( MP start_ it, MP _ i t, MP _ i t2). Te treatment group takes into account firms tat start importing intermediate goods in a particuar year and ave not been importing tese goods in te previous years. As in te case of capita goods, te treatment does not restrict imports of oter types of goods in te years prior to te start of importing intermediate goods. Focusing ony on pure importing starters of intermediate goods; i.e. firms tat ave not been importing any types of goods before te start of importing intermediate goods, woud greaty reduce te sampe of firms in te treatment group (for 65. %). Te contro group consists of non-importing firms ( MP t 0, MP t s 0 ). n order to estimate te ATT based on te particuar propensity score, severa different matcing metods were used in order to increase te significance of resuts. Tese comprise one nearest neigbour matcing wit te repacement, taking into account two different caipers, 0.05 and 0.. A tigter caiper significanty reduces bias and improves te performance of propensity score matcing. On te oter and, a narrow caiper can aso ead to te reduction of coser matces (Lunt, 204). Te caipers ave been cosen in accordance wit te anaysis made by Austin (20), wo recommends te widt of te caiper to equa 0.2 of te standard deviation of te ogit of te propensity score. Next, aso use five nearest neigbours matcing wit te repacement and appy te same two caipers. For additiona robustness cecks, radius matcing is used, were again appy caipers 0.05 and 0.. Finay, aso use kerne matcing, were te significance of resuts was furter tested by using two different bandwidts, 0.06 and 0.0. As in te severa papers tat used propensity score matcing, bootstrapped standard errors were used (see 46

61 for exampe Heckman, cimura & Todd, 997; Deejia & Waba, 999, 2002; and cino & Becker, 2002 for reference) Data and descriptive statistics Te reason for coosing Sovenia as te country of interest is due to its caracteristics of a sma and open economy. n te observation period, Sovenia increased its sare of imports of goods and services as a percentage of GDP from 47.4 % in 996 to 68.7 % in 200. Among te EU countries, Luxembourg was in te forefront wit 47. %, Spain was te ast wit 26.8 %, and te overa EU average was 37.0 % in 200. Te OECD average (25.8 %) and te word average (28.0 %) were even ower in te same year (Word Bank, 205). To estimate te reationsips between importing, exporting, and te ski structure of firms, a inked empoyer-empoyee pane dataset for Sovenian manufacturing firms tat were active during te period from 996 to 200 was used. Te dataset combines severa databases: persona income-tax data, transaction-eve data on imports and exports of goods, Statistica Registry of Empoyees, firm-eve accounting data, and data on foreign direct investments of Sovenian firms. Tese datasets were provided by te Statistica office of te Repubic of Sovenia (SORS), te Tax Autorities of Sovenia (TARS), te Bank of Sovenia, and te Agency of te Repubic of Sovenia for Pubic Lega Records and Reated Services (AJPES). Te dataset contains information on baance seet data and te income statements of Sovenian manufacturing firms, teir import and export activities, and te attributes of empoyees. Te ricness of te dataset enabes te empirica anaysis of controing for severa caracteristics of firms, i.e. number of empoyees, capita per empoyee, vaue added per empoyee, ownersip, foreign direct investments, types of imports, and takes into account te attributes of empoyees, i.e. years of scooing, educationa eve, and wages. Tabe presents some descriptive statistics of te data. For brevity, te descriptive statistics for initia years are presented wit a four-year gap, wereas te recent years ave no gaps. mporting and exporting firms sare simiar caracteristics tey are on average bigger, empoy a iger number of skied empoyees, and pay iger wages (Tabe ). Wages in importing and exporting firms are iger in tota and for skied empoyees. However, wie te sare of skied empoyees is above average in importing firms, it is usuay beow average in exporting firms. Te ony exception wen te ski sare was above te average in te observed period in exporting firms was te year Possibe reason for tis penomenon coud be te size of exporters, since exporters empoy above average number of workers. Reducing te number of workers in te recent years in exporting firms ed to te increase in te ski sare of tese firms. Tese descriptive statistics coincide wit previous empirica studies (see for exampe Atomonte et a., 203; Crino, 202; 47

62 Atomonte & Békés, 200; and Mesci et a., 2008). n te recent years, te number of empoyees as decreased in a treated firms due to te goba crisis. Tabe. Caracteristics of Sovenian manufacturing firms, broken-down by importing and exporting activities (mean vaues) Manufacturing firms - tota Year Empoyment Empoyment of skied Ski sare Gross wage 5,073 7,665 0,269,005,624,476,886 Gross wage of skied 9,96 4,37 7,567 8,45 9,406 9,07 8,985 mporting manufacturing firms Year Empoyment Empoyment of skied Ski sare Gross wage 5,587 8,503,703 2,543 3,55 3,533 4,9 Gross wage of skied 0,900 5,536 9,982 20,668 2,97 2,939 22,06 Exporting manufacturing firms Year Empoyment Empoyment of skied Ski sare Gross wage 5,724 8,548,409 2,205 2,966 2,80 3,262 Gross wage of skied 2,65 7,48 2,028 2,585 23,409 22,865 22,800 Note. Expanations of te variabes are as foows: Empoyment: te average number of empoyees; Empoyment of skied: te average number of skied empoyees; Ski sare: te average of te ski sare in firms (in per cent); Gross wage: average gross wage in ; Gross wage of skied: average gross wage of skied empoyees in. 2.4 Resuts and discussion Source: SORS, autor s cacuations n order to assure unbiased resuts, severa tests on te quaity of matcing were made. First, te Propensity Score istograms, wic present te overap between treated and untreated firms, sow tat rougy a af of te observations in te treated groups were matced wit a simiar propensity score in te untreated groups. n addition, te efficiency of te matcing for eac exogenous variabe as been cecked. Te t-test is used for testing te ypotesis tat te mean vaue of a variabe is te same in te treatment and in te contro group. Wit te exception of te t-test for te first ypotesis, a oter tests sow tat te Baancing property is satisfied. Next, te bias after te matcing procedure is measured to ceck weter te differences between te treatment and te contro group 48

63 decrease consideraby after matcing. Te test sows tat te matcing procedure on average significanty reduced te differences between te treatment and te contro groups. Atoug te t-test for te first ypotesis did not confirm tat te Baancing property is satisfied, te oter tests did, foowing to te concusion tat te matcing procedure generated an appropriate contro group to matc te observations in te treatment group. Te tests are encosed in te Appendix. Te succeeding two subsections present base resuts and some additiona robustness cecks Base resuts Te foowing tabes present te resuts on te inkages between imports, exports and te ski structure of firms. To enance te robustness of te resuts, eac treatment was evauated wit different matcing metods. Te extensions of te mode and additiona robustness cecks are presented in te next subsection. Tabe 2. Resuts of testing Hypotesis : Correation between importing and a better ski structure of firms is positive Outcome of interest: ski sare Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: *** ,90 33,289 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: *** ,90 33, *** ,90 33, *** ,90 33,289 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing firms); Contro: firms in te contro group (i.e. nonimporting firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0. Source: SORS, autor s cacuations Te resuts of te anaysis of te differences in te ski structure between importers and non-importers (Tabe 2) indicate tat importers ave a iger sare of skied empoyees tan non-importers. Depending on te cosen metod, te sare of skied empoyees in importing firms is on average iger for 25.3 to 33.3 %, compared to non-importing firms. Tese findings are confirmed by a matcing metods. 49

64 Tabe 3. Resuts of testing Hypotesis 2: Firms wit a better ski structure start importing Outcome of interest: ski sare one year before te start of importing Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: 0.05 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: *** , *** , *** , *** ,549 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters); Contro: firms in te contro group (i.e. nonimporting firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0. Source: SORS, autor s cacuations Te resuts of te anaysis of te sef-seection of firms wit a better ski structure into importing (Tabe 3) indicate tat future importers ave a iger ski sare tan nonimporters one year before starting to import. Te sare is iger for 0.4 to 20.6 %, depending on te cosen metod. Te findings are confirmed by a matcing metods. 50

65 Tabe 4. Resuts of testing Hypotesis 3: Ski structure canges in favour of te skied abour after te start of importing Outcome of interest: ski sare one year after te start of importing Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: 0.05 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: *** , *** , *** , *** ,640 Outcome of interest: cange in te ski sare one year after te start of importing Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: * ,765 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: , , ,765 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters tat import aso one year after te start of importing); Contro: firms in te contro group (i.e. non-importing firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0 Source: SORS, autor s cacuations Te resuts of anaysing te tird ypotesis (Tabe 4) indicate tat new importers ave a iger ski sare one year after te start of importing, compared to non-importing firms. Depending on te cosen metod, te sare is iger for 33.4 to 4.9 %. Tis resut supports previous resuts on te sef-seection of firms wit a better ski structure into importing. On te oter and, te majority of resuts does not sow a significant increase in te ski sare of workers one year after te start of importing and tus does not support te ypotesis tat firms improve teir ski structure one year after te start of importing. Te fina step anayses te impact of starting to import capita goods (Tabe 5) and te impact of starting to import intermediate goods (Tabe 6) on te start of exporting. 5

66 Tabe 5. Resuts of testing Hypotesis 4: mporting capita goods as a positive impact on te start of exporting Outcome of interest: start of exporting one year after te start of importing capita goods Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: 0.05 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: E , , , ,038 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters of capita goods tat import aso one year after te start of importing); Contro: firms in te contro group (i.e. non-importing firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0 Source: SORS, autor s cacuations Tabe 6. Resuts of testing Hypotesis 5: mporting intermediate goods as a positive impact on te start of exporting Outcome of interest: start of exporting one year after te start of importing intermediate goods Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: 0.05 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: *** , * , *** , ** ,038 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters of intermediate goods tat import aso one year after te start of importing); Contro: firms in te contro group (i.e. non-importing firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0 Source: SORS, autor s cacuations Te resuts suggest tat te start of importing intermediate goods as a positive impact on te start of exporting in te next period (Tabe 6), wie te start of importing capita goods does not seem to ave an immediate impact on te start of exporting in te next period (Tabe 5). One coud aso interpret tese resuts in a way tat, due to te different caracteristics and usage of intermediate and capita goods in te production process, 52

67 importing intermediate goods as a prompt effect on te start of exporting, wie importing capita goods migt ave a deayed effect on te start of exporting. To test tis caim, te foowing subsection takes aso te second period into account and makes additiona robustness cecks Extensions of te mode Foowing Yang and Maick (200), and Damijan and Kostevc (205), wo ag te outcomes of interest by one and two periods, tis subsection ags te outcome of interest for two periods wen anaysing te second ypotesis, and taking into account te second year after te start of importing wen anaysing te ast tree ypoteses. Additiona robustness cecks tat incude aterations of matcing metods (e.g. canging te caipers and te bandwidts) and additiona Baancing property tests are incuded in te Appendix. First, te extensions of te mode take into account additiona cecks of te sef-seection of firms into importing, by comparing te ski sares of non-importing firms and importing starters two years before te start of importing. Wen testing Hypotesis 2 in te basic mode and in extensions of te mode, te same treatment as been used. n addition, in order to compare te resuts of te basic and extended mode, te sampe of te treated firms as been defined in a way to focus on te same coort of firms. Te resuts suggest tat firms do not ave a significanty different ski structure two years before te start of importing. Comparing tese resuts to te resuts of te basic mode indicates tat firms decide to additionay increase teir ski sare one year before te start of importing. Tabe 7. Mode extensions of testing Hypotesis 2: Firms wit a better ski structure start importing Outcome of interest: ski sare two years before te start of importing Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: ,376 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: , , ,376 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters); Contro: firms in te contro group (i.e. nonimporting firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0. Source: SORS, autor s cacuations 53

68 Next, te ski sare and its cange two years after te start of importing are anaysed. n ine wit te base resuts and te resuts of te sef-seection of firms wit a iger sare of skied workers into importing, resuts from Tabe 8 sow tat te ski sare of importing starters is iger aso two years after te start of importing. Depending on te cosen metod, te ski sare in importing starters is iger for 4. to 52.7 % two years after te start of importing, compared to non-importing firms. n addition, resuts sow tat te increase in te ski sare in importing firms two years after te start of importing, compared to non-importing firms, is iger for 0 to 5.7 %, depending on te cosen metod. Since te coort of te treated firms was not te same in te basic and extended mode, te resuts of te two cannot be compared. More precisey, te treatment group in te extended mode was restrained, compared to te basic mode, by taking into account ony importing starters tat continue importing two years after te start of importing. Significanty positive cange in te ski sare two years after te start of importing, compared to te insignificant cange one year after te start of importing, coud be te consequence of te surviva of te fittest, were ony te most productive importing starters continue importing two years after te start of importing. On te oter and, te reason for te significanty positive cange in te ski sare in te second year after te start of importing coud be te sign of firms canging teir ski structure ony in te second year after te start of importing. t woud be interesting to anayse tese points in te future researc. Tese resuts partiay confirm te concusions of te teoretica mode, wic point to a positive impact of importing on te empoyment of skied abour. Te teoretica mode expained tat te abour productivity and consequenty aso te empoyment of skied abour increase after firms start importing. Te increased productivity is te resut of two reasons; first is troug importing intermediate goods, wic aows oter production factors to be used more productivey, and te oter is troug importing tecnoogy tat is more affordabe and/or of better quaity. 54

69 Tabe 8. Mode extensions of testing Hypotesis 3: Ski structure canges in favour of te skied abour after te start of importing Outcome of interest: ski sare two years after te start of importing Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: *** ,630 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: *** , *** , *** ,630 Outcome of interest: cange in te ski sare two years after te start of importing Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: ** ,677 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: ** , ** , ** ,677 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters tat import aso two years after te start of importing); Contro: firms in te contro group (i.e. non-importing firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0 Source: SORS, autor s cacuations Finay, anaysing te impact of starting to export in te second year after te start of importing capita or intermediate goods confirms te supposition from te previous subcapter. More precisey, te previous part presumed tat due to te caracteristics and te usage of intermediate and capita goods in te production process, importing intermediate goods as a prompt effect on te start of exporting, wie importing capita goods migt ave a deayed effect on te start of exporting. Even more, te resuts in Tabe 0 unvei tat te majority of firms start exporting in te succeeding year after te start of importing intermediate goods. Tis is evident since te resuts in te second year after te start of importing intermediate goods are insignificant or even significanty negative. On te oter and, resuts in Tabe 9 sow tat firms start exporting two years after te start of importing capita goods. As expained in te previous subcapter, tese resuts point to a different roe of intermediate and capita goods in te production process. More precisey, wie 55

70 intermediate goods usuay require additiona manufacturing processing or are used for resae, investments in capita goods take onger to sow effect. For instance, an investment in a new assemby ine requires time for instaation, testing, etc., before te start of te fina impementation of te new ine. n contrast, firms make every effort to minimise te costs of stockoding and terefore aim not to store teir intermediate inputs for onger periods but try to use tem in a manufacturing process or resae tem as soon as possibe. Tabe 9. Mode extensions of testing Hypotesis 4: mporting capita goods as a positive impact on te start of exporting Outcome of interest: start of exporting two years after te start of importing capita goods Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: 0.05 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: , ** , ** , ** ,984 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters of capita goods tat import aso two years after te start of importing); Contro: firms in te contro group (i.e. non-importing firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < 0.0 Source: SORS, autor s cacuations 56

71 Tabe 0. Mode extensions of testing Hypotesis 5: mporting intermediate goods as a positive impact on te start of exporting Outcome of interest: start of exporting two years after te start of importing intermediate goods Matcing metod ATT se Treated Contro Nearest neigbour (), caiper: 0.05 Nearest neigbour (5), caiper: 0.05 Radius, caiper: 0.05 Kerne, bandwidt: , * , * , * ,984 Note. ATT: average treatment effect on te treated; se: bootstrapped standard errors (00 repetitions); Treated: firms in te treatment group (i.e. importing starters of intermediate goods tat import aso two years after te start of importing); Contro: firms in te contro group (i.e. non-importing firms). Nearest neigbour () denotes one nearest neigbour matcing wit repacement; Nearest neigbour (5) denotes five nearest neigbours matcing wit repacement. * p < 0.; ** p < 0.05; *** p < Concusions Source: SORS, autor s cacuations Te aim of tis study was to empiricay test some of te deductions made in te teoretica mode from te previous capter. More precisey, te aim of te paper was to test te foowing: (i) weter firms wit a better ski structure sef-seect into importing; (ii) weter firms furter improve teir ski structure after te start of importing; and (iii) weter te start of importing intermediate or capita goods ater as a positive impact on te start of exporting. Tese reations between importing, exporting and te ski structure of firms in a sma and open economy were studied using te empoyer-empoyee pane dataset. Te dataset took into account Sovenian manufacturing firms in te period from 996 to 200. Wit te aim of considering evident differences between te importing and te non-importing firms, te propensity score matcing approac was appied. Te empirica anaysis confirmed tat firms wit a better ski structure sef-seect into importing. n addition, te importing firms sustain a iger ski sare tan te nonimporting firms in te first and second year after te start of importing. Wie te ski structure of firms does not seem to improve significanty one year after te start of importing, it increases in te second year after te start of importing. Te estimated resuts coincide wit some of te previous studies wic ave, to te best of my knowedge, not yet studied torougy te reationsip between importing and te ski structure of firms. Te resuts confirm te previous empirica findings on te positive impact of importing activities on te demand for skis (see for exampe Crino, 202; and Mesci et a., 2008), and add new insigts on te sef-seection of firms wit a better ski structure into importing. More precisey, tis study appies a toroug anaysis of a pases of 57

72 importing; i.e. before te start of importing, importing starters and importers in genera. Furtermore, wit te aim of capturing te eve of canges in te ski structure of firms, te differences-in-differences matcing estimator was appied. n addition, te resuts aso add new insigts regarding te impact of starting to import intermediate or capita goods on te start of exporting. Starting to import intermediate goods as a positive impact on te start of exporting aready in te first year after te start of importing tese goods. Even more, te impact on te start of exporting in te second year after te start of importing intermediate goods is insignificant or even negative, indicating tat te start of importing intermediate goods as an immediate impact on te start of exporting. n contrast, importing capita goods sows a positive impact on te start of exporting ony in te second year after te start of importing tese goods. Tese resuts indicate te different function of te capita and intermediate goods in te production process. Wie intermediate goods usuay refer to raw materias and tus require furter processing or are used for resae, capita goods usuay refer to firms fixed assets and are used to increase firms productive capacities. Consequenty, as sown by tis paper, te impact of importing capita goods may ave a deayed impact on te start of exporting, wie importing intermediate goods as an immediate effect on te start of exporting. Besides researcers in tis fied, governments, firms, workers, jobseekers and students wi aso benefit from tis study. Since firms wit a better ski structure sef-seect into importing, it is important for te governments to focus on estabising an environment tat encourages internationa cooperation and stimuation of ski upgrading in firms, in order to furter increase te productivity and competitiveness of domestic firms. Moreover, firms soud ave greater incentives for iring skied workers, wie students, workers and individuas in te job market soud in turn ave iger incentives for attaining coege degrees and acquiring additiona on-te-job training. 58

73 3 THE EFFECTS OF OUTSOURCNG AND OFFSHORNG ON SKLL STRUCTURE: EVDENCE FROM MATCHED FRM- EMPLOYEE DATA 3. ntroduction Gobaisation as canged te word dramaticay in te most recent decades. According to nternationa Monetary Fund (MF), trade iberaisation and tecnoogy improvements ed to ower trade barriers and to te drop of transportation and communication costs (MF, 203). n ine wit tese canges, transnationa companies (TNCs) cange and adjust te structure and organisation of teir vaue added activities, were offsoring and outsourcing are among teir main metods of strategic positioning. Forecasts on increasing internationaisation specify tat firms wi carry out even more of teir activities outside of teir enterprises in te future; for instance by increasing foreign direct investment (FD) fows, or by increasing foreign affiiate activity (UNCTAD, 203). By evauating te effect of offsoring and outsourcing on te ski structure of firms, empirica studies confirm an important impact of offsoring and outsourcing on te ski structure of firms in deveoped countries. On average, studies concude tat offsoring and outsourcing ave a positive impact on te empoyment of skied abour. Among tese are for exampe Mion and Zu (203), studying te effects of Cinese imports on Begian manufacturing firms, Hijzen et a. (2005), studying te UK market, Strauss-Kan (2003), studying te Frenc manufacturing industries, Egger and Egger (2003), studying te impact of trade iberaisation between Western and Eastern Europe, were te country of interest is Austria, and Feenstra and Hanson (996), studying te United States abour market. Feenstra and Hanson (996) demonstrate tat an increased import competition from ow-wage countries presents an important canne wic transfers abour demand towards more skied workers. Hijzen et a. (2005) expain tis as a consequence of reocating te unskied abour-intensive production to countries, abundant wit unskied abour, wereas ig-tecnoogy stages of productions continue to be produced in deveoped countries. Te motivation for tis paper draws upon te forecasts by UNCTAD wic assign an even greater roe to offsoring and outsourcing activities in te future, and by adding to te existing evidence on te effects of outsourcing and offsoring on te abour market. n tis study, outsourcing is defined as te ratio between firm s vaue of intermediate imports and its vaue of tota materia costs, wie offsoring is measured as te presence of te FD fows in a firm. Te first aim of te paper is to incude bot measures of strategic positioning of firms, outsourcing and offsoring, in one mode and to test weter teir positive effect on te reative empoyment of skied workers is present aso wen accounting for bot factors in one mode. To te best of my knowedge, previous empirica studies took into account ony 59

74 one of te measures in teir modes at a time. Te majority of tese modes confirmed a positive impact of offsoring and outsourcing on te abour demand, witout controing for te oter factor. Not controing for bot factors in one mode migt terefore ead to te missing variabe bias. n addition, bot factors are expected to increase in te future. Taking tese facts into account, beieve it is important to account for bot aspects in one mode. Te second aim is to test weter offsoring and outsourcing from ig-income countries ave a different impact on te ski sare of firms as compared to offsoring and outsourcing from ow-income countries. t is important to make tis distinction since aforementioned studies indicate sifts of unskied-intensive parts of production to countries, abundant wit reativey ess-skied abour. Terefore, it is essentia to account for te destination country of outsourcing and offsoring, as tis enabes a new interpretation of resuts and contros for te potentia differences of partner s performance. Regarding te abovementioned studies, one woud expect tat outsourcing from igincome countries and offsoring to ow-income countries woud ave a positive impact on te reative empoyment of skied workers. Te reasoning beind tis caim is tat outsourcing from ig-income countries enabes firms to ave access to tecnoogicay more advanced intermediate inputs, wic in turn demand te empoyment of igy skied workers. On te oter and, offsoring to ow-income countries is expected to sift some of te more manuay-intensive parts of production abroad and to keep te ig vaue added departments in te ome country (as for exampe researc, saes, marketing, finance, etc.). Te fina aim is to incude a new dimension wen defining skis, wic takes into account aso te occupationa cassification of workers and not ony te eve of teir forma education. t is important to take tis into account since workers gain teir skis not ony by forma education but aso troug various forms of vocationa trainings and during teir work career. One woud expect tat taking into consideration aso te occupationa cassification wen defining skis soud furter increase te expanatory power of te mode. Te empirica anaysis addresses te foowing ypoteses: (i) te effect of offsoring and outsourcing remains positive aso after controing for bot factors in one mode; (ii) outsourcing from ig-income countries and offsoring to ow-income countries ave a positive effect on te ski sare of firms; and (iii) te effect of offsoring and outsourcing varies between different occupationa eves of workers. To assess te effects of outsourcing and offsoring on te ski structure of firms, a matced firm-empoyee pane dataset for Sovenian firms in te period from 997 to 200 is used. Te empirica anaysis is spit into two parts. Te basic mode anayses te impact of offsoring and outsourcing on te ski structure of firms, using a conventiona definition of skied workers, wic defines skis ony by te eve of forma education. n 60

75 te mode extensions, te anaysis first differentiates between offsoring to and outsourcing from ig- and ow-income countries. An additiona extension of te mode introduces a new dimension wen defining skis, wic takes into account not ony te forma educationa eve of workers but aso teir occupationa cassification. By incuding tis, te mode does not ony contro for te educationa eve of workers, but aso for teir occupationa eve and different caracteristics of working responsibiities. By controing for te occupationa structure, te mode can aso test, weter firms differentiate between empoying workers wit different educationa eves witin te same occupationa cass. Tis concusion woud bring additiona poicy impications to previous findings. Taking into account statisticay significant resuts, find tat offsoring as a positive impact on te reative empoyment of tertiary educated workers in manufacturing firms, wie te resuts for outsourcing are uncommon. Wen controing aso for te income eve of countries, offsoring to ow- and ig-income countries sows a simiar and positive impact on te reative empoyment of skied abour in manufacturing firms. n service firms, resuts point to a weaker impact of offsoring to ig-income countries, compared to offsoring to ow-income countries. Furtermore, wen controing aso for te occupationa eve wen defining skis, resuts indicate tat offsoring to ig-income countries as a stronger impact on te reative empoyment of Professionas in manufacturing firms tan offsoring to ow-income countries. On te oter and, resuts for service firms sow tat offsoring to ig-income countries as a weaker impact on te reative empoyment of Tecnicians, compared to offsoring to ow-income countries. Wen combining bot definitions of skied; i.e. educationa and occupationa eve, te resuts for manufacturing firms do not cange significanty, wie te resuts for service firms sow tat offsoring to ow- and ig-income countries as a simiar and positive impact on te reative empoyment of tertiary educated Managers. Te resuts for outsourcing are on average not statisticay significant. Te impact of education terefore differs between occupationa groups, indicating tat firms differentiate between empoying workers wit different educationa eves witin te same occupationa group, wic is especiay true for Managers in service firms and Professionas in manufacturing firms. Tis concusion is in ine wit findings of te previous capters, confirming tat firms ave greater incentives for iring more educated workers. Te remainder of te paper is organised as foows. n te next section, a brief summary of te reevant iterature is given. Section tree introduces te metodoogy used in te empirica part, wereas section four describes te data and presents descriptive statistics. Te empirica anaysis and discussion of resuts are incuded in section five. Te ast section summarises and provides concusions. 3.2 Literature review Literature review starts wit a brief discussion on te teoretica modes wic expore te effects of outsourcing and offsoring on te abour demand. Grossman and Rossi- 6

76 Hansberg (2008) deveoped a teoretica mode tat studies te impacts of te faing offsoring costs on factor prices in te ome country. Autors differentiate between trade in goods, wic is te conventiona meaning of trade, and trade in tasks, wic reates to adding a vaue to goods in different ocations. Te mode concudes tat offsoring infuences te firms performance positivey, since it aows tem to ire some factors abroad at a ower price. On te oter and, offsoring aso brings costs as te monitoring and management of workers is indered due to ong distances. n te mode, autors take into account te effects of trade in tasks, by controing aso for skied workers. Teir mode indicates tat trade in tasks gives rise to sared gains for a domestic factors (Grossman & Rossi-Hansberg, 2008). Anoter teoretica mode was formed by Mitra and Ranjan (2009), wo study te reation between offsoring and unempoyment, were tey define offsoring as te sourcing of inputs from foreign countries. Te mode differentiates between two situations; one is wen abour is perfecty mobie and anoter wen tis is not te case. n te case of perfect abour mobiity, offsoring causes wages to increase and unempoyment to decrease, wereas in te case of imperfect abour mobiity, tere is a possibiity for unempoyment to increase in te offsoring sector, but at te same time, te unempoyment decreases in te oter sector. Finay, Egger and Egger (2003) deveoped a teoretica mode in wic tey focus on a sma country case, treated as ome country, wic produces an industriaised good and as a possibiity to outsource a ow-skied part of its production to ow-wage foreign countries. Resuts indicate tat outsourcing increases wit te decrease of trade barriers. n a competitive abour market framework, outsourcing increases reative wages of igskied abour, wie it does not affect reative empoyment. However, in a unionised framework, outsourcing increases bot, reative wages and te reative empoyment of ig-skied abour in te ome country. Compared to rater scarce teoretica anayses on te effects of outsourcing and offsoring on te abour market, empirica studies are more abundant. Feenstra and Hanson (996, 999) anayse te impact of outsourcing in te United States. Te resuts of teir earier paper point to an increase in te reative demand for skied abour due to increased outsourcing. However, te resut does not od for a time periods (Feenstra & Hanson, 996). n teir ater paper, Feenstra and Hanson (999) focus on comparing te effects of outsourcing and tecnoogy on wages. Tey concude tat bot penomena impact te reative wages of non-production workers positivey, were te impact of tecnoogy is arger, compared to outsourcing activities (Feenstra & Hanson, 999). Amiti and Wei (2005a, 2005b) expore te effects of service outsourcing and offsoring in te UK and US, respectivey. For te UK market, te autors find tat job growt and outsourcing are not negativey correated at te sectora eve (Amiti & Wei, 2005a), wereas for te US market, te autors find a positive effect of offsoring on productivity, 62

77 wie te effect on empoyment differs according to te disaggregation of industries. More precisey, wen industries are finey disaggregated, te resuts point to a negative effect. On te oter and, wen industries are defined on a broader eve, te negative effect disappears. Tis eads to a concusion tat, atoug offsoring migt affect empoyment negativey witin industries, dismissed workers renew teir empoyment in oter growing industries (Amiti & Wei, 2005b). Hijzen et a. (2005) aso examine te effects of outsourcing on te UK abour market. Te resuts indicate tat outsourcing affects te demand for unskied abour negativey and togeter wit tecnoogica cange eads to canges in te ski structure of manufacturing industries (Hijzen et a., 2005). Parae concusions on te effect of increased outsourcing were made by Strauss-Kan (2003), wo concudes tat outsourcing infuences te reative empoyment of unskied workers in Frenc manufacturing industries negativey. n addition, Egger and Egger (2003) empiricay tested te effect of outsourcing in Austria, as a consequence of trade iberaisation in te Centra and Eastern Europe. Autors find tat outsourcing increases te reative empoyment of ig-skied abour (Egger & Egger, 2003). Furtermore, wie Mice and Rycx (2009) find no major infuence of materias or business services offsoring on te empoyment in Begian firms, tey igigt te importance of distinguising between manufacturing and service industries. Traditionay, ony manufacturing industries were reated to offsoring, since teir products are easiy tradabe. However, improvements in information and communication tecnoogies ad a significanty positive impact on offsoring in service industries (Mice & Rycx, 2009). t is terefore important to compare te impact of outsourcing and offsoring in bot, manufacturing and service firms. Tis was confirmed aso by De Backer and Yamano (202), wo compare te increase of offsoring in different countries, were te anaysis was done separatey for manufacturing and service industries. Atoug offsoring increased in te observed period from 995 to 2005 in bot, manufacturing and service industries, te increase was on average bigger in te atter. mporting intermediates from abroad is owever on average sti more important in manufacturing industries (De Backer & Yamano, 202). Simiar concusions were made by Horgos (2006), using German data. Te autor concudes tat outsourcing activities are concentrated in ig-skied manufacturing industries, wie service industries sow te igest increase in outsourcing activities (Horgos, 2006). A noteworty restraint of empirica studies, presented in previous paragraps, is in te type of data used. Te studies used data, disaggregated ony at te industry eve and terefore coud not contro for firm-specific and individua-specific caracteristics tat may ave an impact on te ski structure of firms. Moreover, identifying te abour demand curve is more caenging wen using industry-eve data (Hijzen & Swaim, 200). Since firmeve data became more accessibe in te recent years, current studies estimate te effects of outsourcing and offsoring on te abour market aso in terms of firm-eve data. 63

78 Konings and Murpy (2006) evauate te substitution of workers between parents and teir affiiates in European mutinationa enterprises. Due to te ack of information on te ski composition of workers, autors were not abe to estimate te effect of outsourcing on te ski demand of workers but were abe to differentiate between regions wit different wage costs. Contrary to te common beief, teir resuts indicate empoyment reocations between parent firms and teir affiiates, bot based in te Nort EU but tey find no significant empoyment fows between te parent and affiiates, based in te Sout EU, and Centra and Eastern Europe (Konings and Murpy, 2006). n anoter study using firm-eve data, Biscourp and Kramarz (2007) differentiate between two types of imports imports of finised goods and imports of intermediate goods, wic tey define as offsoring. Tey find a strong and negative correation between imports and job destruction, were tis impact is especiay strong for imports of finised goods, imports from ow-wage countries and for arger firms. Tey aso confirm some previous findings tat tese canges on average occur witin firms. Controing for innovation does not ater teir resut (Biscourp & Kramarz, 2007). Furtermore, by estimating data on German manufacturing firms, Wagner (20) confirms tere is a sef-seection of firms into offsoring. Te anaysis concudes tat tese firms are arger, more productive and more uman capita intensive. Moreover, te autor aso confirms some previous findings, wic do not find a arge negative effect of offsoring on empoyment (Wagner, 20). Focusing on trade iberaisation in Cina after its accession to te Word Trade Organisation, Boom et a. (20) evauate ow tis event affected tecnica cange in European economies. Tey find tecnoogy improvements and productivity increases in industries, mosty affected by te increased Cinese competition, wie te effect on abour demand and surviva probabiity varies across firms. Specificay, atoug te increased Cinese competition did not affect abour demand and surviva probabiity in ig-tec firms, tey bot decreased in ow-tec firms. On te oter and, import competition from deveoped countries did not affect innovation (Boom et a., 20). Mion and Zu (203) aso studied te effects of Cinese imports, were teir main interests were Begian manufacturing firms and te Begian abour market. Te autors differentiate between imports of fina and intermediate goods, and find tat importing from Cina urts firms in te ow-tec industries. Contrary to Boom et a. (20), tey find tat import competition from Cina does not ave a negative effect on te surviva of Begian manufacturing firms. On te woe, te competition in te Begian market as increased troug te increased Cinese competition and was foowed by reduces in firm empoyment growt, and upgrades in tecnoogy and ski structure (Mion & Zu, 203). Simiar concusions were made by Lo Turco and Maggioni (202), wo focused on te effects of offsoring on te abour demand in taian manufacturing firms, were tey aso 64

79 differentiated between source countries. Te autors concude tat importing intermediates from ig-income countries does not affect empoyment, wie te effects on te empoyment are negative wen firms import intermediates from ow-income countries. Using Danis data and focusing on te effects of offsoring on wages, Hummes et a. (204) find tat offsoring as a positive impact on wages of skied abour and a negative impact on te wages of unskied abour. Te findings of presented studies in tis section sow tat iberaising trade wit deveoping countries brings opportunities for cost reductions and tecnoogy improvements, wie on te oter and it aso presents treats to abour markets in te deveoped countries. However, te majority of studies concude tis treat is not arge and is usuay concentrated on te ow-skied empoyees. Aso important is te empasis made in severa papers (see for exampe De Backer & Yamano, 202; Mice & Rycx, 2009; and Horgos, 2006) on te significance of differentiating between manufacturing and service industries, as we as te importance of using firm-eve data (Hijzen & Swaim, 200). Tis paper empoys a matced firm-empoyee pane dataset for Sovenian firms, to evauate ow significant is te impact of offsoring and outsourcing on te ski structure of firms. More precisey, te aim of te study is to make a toroug anaysis of te impact of offsoring and outsourcing on te ski structure of Sovenian firms in te period from 997 to 200. To obtain more detaied resuts and to add to te existing evidence in tis fied of knowedge, a new dimension is introduced wen defining skis, by incuding information on occupationa structure of workers. To te best of my knowedge, previous anayses differentiated between skied and unskied workers ony by ooking at teir educationa attainment or by differentiating between production and non-production workers. However, it is important to take into account aso te occupationa cassification of workers, as skis can be acquired troug empoyment and experience, and not ony troug forma education. Specificay, workers wo do not ave tertiary education aso occupy important positions in firms, wie tertiary educated workers aso occupy ess demanding positions. Te atter is especiay true for younger workers at te beginning of teir career pat, wereas te former is true for experienced workers, wo did not ave te opportunities to acieve iger forma education, but took an important position in firms owing to teir capabiities. Moreover, te anaysis is furter broadened by differentiating between occupationa groups tat define skis in order to make an even more toroug anaysis. Wit tis extension, evauate ow te empoyment effects of offsoring and outsourcing differ by task caracteristics. Neverteess, in order to fuy contro for te informa education of workers, it woud be interesting to contro aso for oter variabes, as for exampe te period of empoyment, and aso persona caracteristics of workers, as for exampe gender, engt of pursuing te studies, socia status of te famiy, etc. Tese aspects coud be incuded to te anaysis in te future extensions of te mode. 65

80 Finay, wie te buk of anayses were usuay concentrated on te effects of ony offsoring or ony outsourcing, it is important to study bot factors in one mode in order to increase te goodness of fit of te mode and avoid te missing variabe bias probem. Te reasoning beind tis is due to te findings of empirica studies, wic sow tat offsoring and outsourcing bot ave a positive impact on te abour demand, were te modes accounted for ony one of te factors in teir mode. Terefore, bot factors offsoring and outsourcing are incuded in tis anaysis. n addition, aso differentiate between outsourcing from ig- and ow-income countries and between offsoring to igand ow-income countries as tis differentiation brings new information and deepens te anaysis. Wie Lo Turco and Maggioni (202) aso differentiated between outsourcing from ig- and ow-income countries, tey did not account for offsoring, skied workers or service firms in teir anaysis. As aready presented in te introduction, te paper s main aim is to test weter: (i) incuding bot penomena offsoring and outsourcing in one mode aters te resuts of te previous empirica studies, wic contro ony for one factor in teir mode and on average find a positive impact of te particuar factor on te reative empoyment of skied workers; (ii) controing for te destination country of outsourcing and offsoring brings new insigts; especiay weter outsourcing from ig-income countries and offsoring to ow-income countries as a positive impact on te reative empoyment of skied workers; (iii) controing for te occupationa structure of workers brings additiona contributions to te resuts of te paper. 3.3 Metodoogy Tis section presents te metodoogy and sets up a framework and specification of te mode, wic wi serve for empiricay testing of preceding postuates. Since it can be deduced from te presented iterature tat definitions of outsourcing and offsoring vary significanty across different studies, tis section first presents te definitions of outsourcing and offsoring, used in te anaysis. Te framework and specification of te basic and extended modes are presented next. Definitions of outsourcing and offsoring differ widey in te iterature. Feenstra and Hanson (996) define outsourcing as te import of intermediate inputs by domestic firms, wereas in teir more recent paper (Feenstra & Hanson, 999), tey introduce two measures of outsourcing. First is te ratio between imported intermediate inputs, reative to te tota expenditure of non-energy intermediates in eac industry, and te second is defined as inputs tat are purcased from te same two-digit Standard ndustria Cassification (SC) industry as te good being produced (Feenstra & Hanson, 999). Many of te papers foow tese definitions and tis metodoogy. Simiar definition for outsourcing is aso used in te recent reports of MF (203) and UNCTAD (203), wic define outsourcing as purcasing intermediates from anoter firm, rater tan producing tem witin te firm. Taking into account tese definitions, in tis anaysis, outsourcing is 66

81 defined as te ratio between te vaue of intermediate imports and te vaue of tota materia costs of a firm i in year t: ntermedia te imports it Outsourcin git, (55) Tota materia costsit were intermediate imports are defined according to te assigned Broad Economic Categories (BEC) codes. Under BEC cassification, goods can be cassified in tree categories; capita, intermediate, and consumption goods. Besides estimating te effects of outsourcing, tis anaysis aso takes into account te effects of offsoring. For te atter, again foow te definition of MF (203) and UNCTAD (203) wic define offsoring eiter as a process of reocating a part of or a te activities to anoter firm, ocated overseas, or as foreign direct investments. For estimating te effect of offsoring, take into account te dataset from te Bank of Sovenia, wic comprises information on te FD fows for every Sovenian firm. Tis dataset gaters information on te voume of te FD and te destination country of te investment. Offsoring is denoted by introducing a dummy variabe, indicating te existence of firm s FD fows. Definitions of te outsourcing and offsoring terefore take into account ony foreign fows. Te weaknesses of te abovementioned definitions are mainy te consequence of data imitations. More precisey, since firms can buy intermediate goods aso from domestic firms, it woud be important to incude aso tis information to te anaysis, soud it be avaiabe. To contro for tis, tis study used a proxy in te form of domestic cost eve, cacuated as te difference between te tota eve of materia costs and imports. n addition, since not a FD fows affect firms ski sare, definition of offsoring soud aso incude information on te type and voume of te FD fows in order to make a more compreensive measure on its effect on te ski sare. Unfortunatey, tis data is avaiabe ony for te recent years. n future studies, it woud aso be interesting to anayse te sare of inputs from countries, were firms ave outward FD, and compare tis wit te sare of inputs from countries, were firms do not ave outward FD Framework and specification of te basic mode Tis part mainy foows te teoretica framework, introduced by Hummes et a. (204). Te production function of a firm i in year t is defined as: Y A f K, H, C ), (56) it it ( it it it were te dependent variabe, Y it, is te output, A it is productivity, K it is capita, H it is skied abour, and C it is a composite input, consisting of domestic and foreign inputs. Te atter reate to outsourcing and/or offsoring activities, and te former reate to unskied abour and domestic inputs. As presented in te iterature review, offsoring and 67

82 outsourcing activities ave distinct impacts on skied and unskied abour, were te impact on te skied abour is on average positive, wie te impact on te unskied abour is on average negative (see for exampe Hummes et a., 204; Mion & Zu, 203; Hijzen et a., 2005; Strauss-Kan, 2003; Egger & Egger, 2003; and Feenstra & Hanson, 996). Since bot factors affect te abour demand of firms, te mode of Hummes et a. (204) is extended by incuding aso domestic inputs and offsoring into te mode. Due to te abovementioned data imitations of offsoring s and outsourcing s definitions, it was not possibe to find an appropriate measure for bot factors tat woud consider a fows and is terefore not entirey consistent wit te presented mode. t woud be interesting to take tis into account in te future studies, wen data imitations are resoved. To impement te teoretica mode in te data, introduce P it as a reduced-form of te demand for firm i s products, divide te variabes in te mode (56) by te tota number of firms empoyees, separate te internationa activities of firms into offsoring (Off it ) and outsourcing (Out it ), separate domestic inputs into unskied abour (L it ) and domestic costs (DC it ), take ogaritms and rearrange te equation so tat te variabe of interest on te eft is te skied abour: )n H n P n A n K ( )n( Out Off L DC ) (57) ( it it it it it it it it Furtermore, foowing Hummes et a. (204), te ogaritm of te average wage eve (W it-s ), and te ogaritm of te vaue of exports (X it-s ) in firm i and year t are added to te mode. Te atter is introduced in order to capture time varying socks to demand for firms output. A detaied derivation of te mode is encosed in te Appendix. After rearranging, te empirica mode ence becomes: Ski_sare it = ß 0 + ß Out it + ß 2 Off it + ß 3 X it + ß 4 A it + ß 5 K it + + ß 6 W it + ß 7 DC it + Time t + nd t + ε it, (58) were te dependent variabe Ski_sare it is te ogaritm of te ratio between skied empoyees and te tota number of empoyees in firm i and year t. Simiary to Hummes et a. (204), skied workers in te first part of te anaysis are defined as tertiary educated workers, i.e. if tey attain some form of coege degree, wic is normay at east 4 years of scoo attainment in Sovenia. As aready expained, outsourcing (Out it ) is defined as te sare of intermediate imports in te tota materia costs, and offsoring (Off it ) as te dummy variabe, controing for te outward FD. Oter expanatory variabes are te foowing: X it is a ogaritm of te vaue of exports, Ait is a measure of productivity, K it is a ogaritm of capita per empoyee, W it is a ogaritm of te average annua wage eve, and DC it is a ogaritm of te domestic cost eve in firm i and year t. Domestic cost eve (DC it ) is cacuated as te difference between te tota eve of materia costs and imports. To increase te sensitivity of resuts, two different measures of productivity ( A it ) are used; vaue added per empoyee and tota factor productivity. Variabe Time t contros for year 68

83 specific effects and nd t denotes industry dummy variabes (2-digit NACE rev. industries). Foowing Hummes et a. (204), outsourcing, exports and eves of domestic costs are not scaed by firm size in order to enance te expanatory vaue of te mode. More precisey, canges in firm size migt be a consequence of te canges in tese variabes. nstead, te mode as been estimated wit and witout firm size as one of te expanatory variabes. t woud be convenient to incude aso oter contro variabes tat ave an important effect on te ski structure of firms, as for exampe information on te R&D expenditures and te number of patents. However, since tis data is not avaiabe or is imperfect, tere exists a missing variabe bias wic woud be important to take into account in te future studies Extensions of te mode Te formation of te extended mode is based on te mode, presented in te previous subsection. First, te mode is extended by differentiating between outsourcing from igand ow-income countries, and offsoring to ig- and ow-income countries. As mentioned in te iterature review, Lo Turco and Maggioni (202) aso controed for te origin of countries wen anaysing te impact of outsourcing on te abour demand in taian manufacturing firms. Te autors empasise it is important to differentiate between ig- and ow-income countries, since different origins of outsourcing can point to a different performance eve of firms. add to te anaysis of Lo Turco and Maggioni (202) by controing for skis, and incuding aso service firms and offsoring into te mode. Countries are cassified as ig- or ow-income according to te definitions, made by te Word Bank, were te ow-income, ower-midde-income and upper-middeincome economies for a particuar year are assigned as ow-income countries, and igincome economies as ig-income countries (WB, 205). Te extended mode, controing for outsourcing from ow- and ig-income countries, and offsoring to ow- and ig-income countries is te foowing: Ski_sare it = ß 0 + ß Out it + ß 2 Out_ig it +ß 3 Off it + ß 4 Off_ig it + + ß 5 Hig it + ß 6 X it + ß 7 A it + ß 8 K it + ß 9 W it + ß 0 DC it + + Time t + nd t + ε it, (59) were Out_ig it is an interaction term between outsourcing and a dummy variabe, controing for ig-income countries, Off_ig it is an interaction term between offsoring and a dummy variabe, controing for ig-income countries, and Hig it denotes a dummy variabe, controing for outsourcing from and offsoring to ig-income countries. Te rest of te mode in te expression (59) foows te basic mode (58). As aforementioned, te presented iterature suggests tat outsourcing from ig-income countries and offsoring to ow-income countries woud increase firms ski sare. Te coefficients ß 69

84 and ß 2 refect te impact of outsourcing from ig-income countries, wie te impact of offsoring to ow-income countries is refected in te coefficient ß 3. Subsequenty, te extensions of te mode aso incude a new dimension wen defining skis. n te previous anayses, workers were usuay defined as skied after acieving a particuar educationa eve or by being invoved in non-production processes. beieve tis arrangement is inadequate as forma education is not te ony factor wic defines te ski eve of workers. Stricty speaking, besides forma education and training, workers acquire skis aso troug experience and informa training. Terefore, it is important to use te occupationa eve wen defining te skis of workers, in order to take into account aso te nature of te tasks and duties of workers jobs. Four different ski eves coud be appied to ten major groups of occupations, wic are cassified by te nternationa Labour Organization (LO). Te setting of te present paper takes into account a version of te nternationa Standard Cassification of Occupations (SCO), te SCO-88 cassification, wic sares te same boundaries of te four ski eves as te SCO-08 cassification. Te top two ski eves, 3 and 4, wit te ski eve 4 being te igest, reate to tertiary education and correspond to tree major groups: Managers (ski eves 3 and 4), Professionas (ski eve 4) and Tecnicians (ski eve 3) (LO, 202; and Eias & Birc, 994). Tese tree major groups of occupations define skied workers in te extended mode. Managers incude egisators, senior officias and managers, wose main tasks consist of determining, formuating and supervising te impementation of government poicies, aws and pubic reguations, or panning, directing and coordinating te poicies and activities of enterprises, organisations, or departments. Professionas work in te fieds of pysica, ife or socia sciences, or umanities and are responsibe for increasing te existing stock of knowedge, finding soutions to te probems by appying scientific and artistic concepts and teories, and transferring teir knowedge onto oters. Finay, Tecnicians incude tecnicians and associate professionas wo ave tecnica knowedge and experience in te fieds of pysica, ife or socia sciences, or umanities. Teir main tasks incude carrying out tecnica work and teacing at particuar educationa eves, reated to te abovementioned fieds (LO, 204). 3.4 Data and descriptive statistics By combining different databases, a ric firm-eve and empoyee-eve pane dataset for Sovenian firms was obtained, covering te period from 997 to 200. Te dataset comprises information on te baance seet data and income statements of Sovenian firms, teir export and import activities (i.e. vaue of exports and imports, type of exported and imported goods, and destination of exports and imports), caracteristics of empoyees (i.e. gender, age, gross wage, educationa eve, and occupationa eve), and information on te foreign direct investments of Sovenian firms. Te atter gaters information on te FD fows for a particuar Sovenian firm. Te dataset inks te foowing databases: persona income-tax data, transaction-eve data on exports and imports of goods, Statistica Registry of Empoyees, firm-eve accounting data and FD, and was provided by te 70

85 Statistica office of te Repubic of Sovenia (SORS), te Tax Autorities of Sovenia (TARS), te Bank of Sovenia, and te Agency of te Repubic of Sovenia for Pubic Lega Records and Reated Services (AJPES). After observing vast differences between manufacturing and service firms (Tabe ), and taking into account te aforementioned empasis on te importance of separating te anaysis for manufacturing and service firms (see for exampe De Backer & Yamano, 202; Mice & Rycx, 2009; and Horgos, 2006), te empirica anaysis was carried out independenty for te two types of firms. Manufacturing firms on average empoy a iger tota number of empoyees and tertiary educated empoyees, compared to service firms. Furtermore, especiay in te more recent years, manufacturing firms on average empoy sigty oder empoyees tan service firms, were age can be considered as a proxy for te experience of empoyees (Zogi, 200). Wen comparing te average annua gross wages for te recent years, manufacturing firms on average pay teir empoyees ower wages tan service firms. However, wen comparing te average wages of tertiary educated empoyees, manufacturing firms pay iger average wages tan service firms. Manufacturing firms on average aso ave ower ski sares tan service firms. Te atter differences in te average gross wages and te ski sares coud be te outcome of a different occupationa and educationa structure of empoyees in manufacturing and service firms, wic wi be presented in one of te upcoming paragraps. For brevity, te foowing tabes present descriptive statistics for te first af of te treated period wit a four-year gap, but incude a information for te recent years. 7

86 Tabe. Caracteristics of Sovenian manufacturing and service firms Tota Year Empoyment Empoyment of tertiary educated Ski sare Age Gross wage 5,39 8,002 0,625,3,850,94 2,260 Gross wage of tertiary educated 8,696 2,804 6,32 6,993 7,993 7,752 7,703 Number of firms 25,26 27,064 30,908 32,799 35,833 36,84 37,882 Manufacturing firms Year Empoyment Empoyment of tertiary educated Ski sare Age Gross wage 5,048 7,658 0,320,066,664,547,962 Gross wage of tertiary educated 9,785 4,54 7,397 8,267 9,232 8,868 8,808 Number of firms 5,4 5,750 6,40 6,38 6,696 6,746 6,798 Service firms Year Empoyment Empoyment of tertiary educated Ski sare Age Gross wage 5,229 8,257,036,779 2,426 2,5 2,79 Gross wage of tertiary educated 8,260 2,363 5,723 6,557 7,565 7,379 7,35 Number of firms 8,037 9,047 2,527 22,729 24,773 25,647 26,495 Note. Expanations of te variabes are as foows: Empoyment: mean number of empoyees; Empoyment of tertiary educated: mean number of tertiary educated empoyees; Ski sare: te average of te sare of te tertiary educated; Age: mean age of empoyees; Gross wage: mean annua gross wage in ; Gross wage of tertiary educated: mean annua gross wage of tertiary educated empoyees in ; Number of firms: number of observations. Source: SORS, autor s cacuations n addition, aso make a comparison between offsoring and outsourcing firms (Tabe 2). Bot types of firms are bigger in size, compared to an average firm in Tabe. Te 72

87 average age of empoyees in offsoring and outsourcing firms is aso sigty iger tan in te average firm. Finay, te average gross wages and gross wages of tertiary educated empoyees are above te average, were te igest average is in te offsoring firms. A separate anaysis was done aso for firms tat offsore to and outsource from ig-income countries. Te descriptive statistics for tese firms sow tat tey are on average bigger and pay iger wages tan te average offsoring and outsourcing firms. Te tabe is encosed in te Appendix. Tabe 2. Caracteristics of Sovenian firms wic offsore and outsource Offsoring firms Year Empoyment Empoyment of tertiary educated Ski sare Age Gross wage 8,384,988 6,934 8,87 9,325 9,453 20,08 Gross wage of tertiary educated 5,25 9,729 24,759 26,282 27,707 27,380 27,487 Number of firms Outsourcing firms Year Empoyment Empoyment of tertiary educated Ski sare Age Gross wage 5,536 8,37,0,752 2,360 2,603 3,38 Gross wage of tertiary educated 0,640 5,026 8,366 9,06 9,980 20,027 20,34 Number of firms 3,520 4,7 2,84 3,089 3,53 2,75 2,495 Note. Expanations of te variabes are as foows: Outsourcing firms: firms tat import intermediate products; Offsoring firms: firms tat engage in outward FD; Empoyment: mean number of empoyees; Empoyment of tertiary educated: mean number of tertiary educated empoyees; Ski sare: te average of te sare of te tertiary educated; Age: mean age of empoyees; Gross wage: mean annua gross wage in ; Gross wage of tertiary educated: mean annua gross wage of tertiary educated empoyees in ; Number of firms: number of observations. Source: SORS, autor s cacuations Next, te occupationa structure of manufacturing and service firms is compared by using SCO-88 cassification (Figure 5). n manufacturing firms, te sare of Macinery workers as been decreasing troug te period, but it is sti te igest among a occupationa groups. On te oter and, te sare of Craft workers as been increasing troug te 73

88 period, but remained second. Te tird argest sare in manufacturing firms beongs to Tecnicians, wie te fourt and fift argest sares appertain to Eementary occupations and Cerks, respectivey. n service firms, on te oter and, Service workers occupy te argest sare and te sare remains steady trougout te observed period. Te second argest sare in service firms beongs to Tecnicians, wie te tird to Cerks. Among oter occupationa groups, Eementary occupations represent te fourt argest sare and Macinery workers te fift. Since te sares of Agricutura and Army workers represent ony a minor part of te tota sares in bot, manufacturing and service firms, tey were excuded from furter empirica anaysis (description of a major occupationa groups is incuded in te Appendix). Figure 5. Occupationa structure of manufacturing and service firms in Sovenia Source: SORS, autor s cacuations Tabe 3 beow presents te descriptive statistics of te tree major groups of occupations tat define skied workers in te extended mode; i.e. Managers, Professionas, and Tecnicians. Professionas represent te igest sare of tertiary educated among a groups, foowed by Managers and Tecnicians. Looking at te tota average in te observed period, 88.7 % of Professionas, 55. % of Managers, and 27.4 % of Tecnicians were tertiary educated. Tis aocation of sares is consistent wit te SCO-88 casses of ski eves, presented in te metodoogica part. Taking into account te average age of empoyees, Managers are on average te odest among a occupationa groups, Professionas were on average a bit oder tan te average worker in te first years of te observationa period, wie in te recent years, tey are a bit younger tan te average. n contrast, Tecnicians are te youngest of te tree groups and compared to te tota popuation of empoyees. Managers earn te igest gross wages among a occupationa groups, foowed by Professionas. Tecnicians aso ave above average wages, in parae to te tota average numbers. However, since Professionas and Managers present more tan a af of a tertiary educated workers and earn te igest wages, tertiary educated 74

89 Tecnicians earn beow te average gross wages, wen taking into account ony tertiary educated workers. Tabe 3. Caracteristics of empoyees in skied occupations Managers Year Sare in te tota empoyment Sare in te tertiary educated Age Gross wage 8,972 4,25 8,436 9,530 20,398 20,23 20,30 Gross wage of tertiary educated 2,859 8,988 24,080 25,57 26,755 26,237 26,092 Professionas Year Sare in te tota empoyment Sare in te tertiary educated Age Gross wage 9,756 4,707 7,803 8,642 9,792 9,750 9,503 Gross wage of tertiary educated 0,277 5,4 8,455 9,462 20,732 20,750 20,622 Tecnicians Year Sare in te tota empoyment Sare in te tertiary educated Age Gross wage 6,3 9,389 2,66 2,980 3,89 3,970 4,246 Gross wage of tertiary educated 8,377 2,484 5,90 6,077 7,27 6,993 7,006 Note. Expanations of te variabes are as foows: Sare in te tota empoyment: sare of a particuar occupationa group in te tota empoyment (in %); Sare in te tertiary educated: sare of a particuar occupationa group in te tota number of tertiary educated empoyees (in %); Age: mean age of a particuar occupationa group; Gross wage: mean annua gross wage of a particuar occupationa group in ; Gross wage of tertiary educated: mean annua gross wage of tertiary educated in a particuar occupationa group in. Source: SORS, autor s cacuations Te descriptive statistics of oter occupationa groups (incuded in te Appendix) revea tat oter groups present ony a minor sare in te group of tertiary educated workers. Te 75

90 igest sare in te tota empoyment is on average presented by Macinery workers, foowed by Craft workers, Eementary workers, Service workers, and Cerica workers, wie te igest earners among tese groups are on average Cerica workers, foowed by Macinery workers, Craft workers, Service workers and Eementary workers. Tis distribution of occupations is aso te reason for iger average wages in manufacturing firms and iger average ski sares in service firms. 3.5 Empirica anaysis As expained in te metodoogica part, te empirica anaysis is spit into two parts. Te basic mode measures te effect of outsourcing and offsoring on te ski structure of firms. Later, te first extension of te mode differentiates between outsourcing from igand ow-income countries and offsoring to ig- and ow-income countries, wie in te second extension an aternative definition of skied empoyees is introduced, taking into account information on te occupationa eve of empoyees Basic mode Te basic mode anayses te effect of outsourcing and offsoring on te ski structure in Sovenian firms. First, te modes are estimated wit te pooed ordinary east squares and wit metods for pane data anaysis; fixed effects and random effects. Foowing Hummes et a. (204), standard errors are custered at firm eves. Te appied procedures foow te metods of Cameron and Trivedi (2009). Due to custer-robust standard errors and an unbaanced pane dataset, a robust version of te Hausman test is needed in order to compare te modes (Cameron & Trivedi, 2009). n accordance, te metod proposed by Scaffer and Stiman (200) is appied, wie te Sargan-Hansen test is reported in te tabes. As introduced in te metodoogy part, tertiary educated workers are defined as skied in te basic mode. For brevity, ony te estimates of te variabes of interest i.e. outsourcing and offsoring are presented in te main tabes, wie te compete resuts are encosed in te Appendix. 76

91 Tabe 4. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms (observation period: ) Manufacturing firms Pooed OLS FE RE Service firms Pooed OLS FE RE Offsoring 0.32*** 0.063** 0.094*** 0.325*** ** [3.0] [2.05] [3.2] [3.55] [0.50] [2.5] Outsourcing * [.03] [0.6] [.83] [0.34] [0.27] [0.88] ndustry effects Yes Yes Yes Yes Yes Yes Year effects Yes Yes Yes Yes Yes Yes Observations 8,99 8,99 8,99 29,59 29,59 29,59 R-squared (witin) R-squared (between) R-squared (overa) Sargan-Hansen statistics *** *** Note. Econometric metods: Pooed OLS: pooed ordinary east squares; FE: fixed effects; RE: random effects. Te dependent variabe is defined as te ogaritm of te ratio between skied empoyees and te tota number of empoyees, were tertiary educated workers are defined as skied. Te expanation of variabes: Offsoring: dummy variabe, controing for te outward FD; Outsourcing: sare of intermediate imports in te tota materia costs. Contro variabes used: ogaritm of te capita per empoyee in a firm, ogaritm of te tota factor productivity per empoyee in a firm, ogaritm of te vaue of exports, ogaritm of te average annua gross wage eve, ogaritm of te domestic cost eve. *** p<0.0, ** p<0.05, * p<0., robust t-statistics in brackets, te anaysis used custer-robust standard errors. Source: SORS, autor s cacuations Taking into account ony te resuts of te most preferred mode, according to te Sargan- Hansen test, i.e. te fixed effects, offsoring sows a positive impact on te sare of skied workers in manufacturing firms, wie te effect of outsourcing and offsoring seems to ave no impact on te ski structure in service firms (Tabe 4). For robustness cecks, te modes were estimated by adding firm size as one of te contro variabes. n addition, vaue added was substituted for te tota factor productivity. n te main tabes, te tota factor productivity is cacuated using te proposed metod of Levinson and Petrin (2003). Levinson and Petrin (2003) extend te mode of Oey and Pakes (996) by substituting investments wit intermediate inputs, wen estimating te production function. Te autors argue one of te main benefits of tis procedure is data driven as te procedure can be used aso for firms wit zero investments, wie anoter advantage is te resut of intermediate inputs being more responsive to te tota productivity term tan investments (see for exampe Levinson & Petrin, 2003; and Petrin, Poi & Levinson, 2004). Bot measures te Levinson-Petrin measured tota factor productivity and te vaue added ave been for exampe used in Damijan, Konings and Poanec (204). 77

92 Robustness cecks on average confirm resuts from te basic mode on te positive effect of offsoring on te ski sare in manufacturing firms. Te robustness cecks are incuded in te Appendix Extensions of te mode n order to obtain new information, two extensions of te basic mode ave been made. Te first extension differentiates between outsourcing from ig- and ow-income countries and offsoring to ig- and ow-income countries. Te second extension incudes information on te occupationa eve of workers wen defining skis. Again, te foowing tabes incude ony te estimates of te variabes of interest, wie te estimates of te contro variabes are encosed in te Appendix Differentiation between ig- and ow-income countries Differentiating between outsourcing from ig- and ow-income countries and offsoring to ig- and ow-income countries enabes te estimation weter a particuar type of source country of offsoring and outsourcing as a more significant impact on te ski structure of domestic firms. 78

93 Tabe 5. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, differentiating between ig- and ow-income countries (observation period: ) Manufacturing firms Pooed OLS FE RE Pooed OLS Service firms Offsoring 0.83*** 0.052* 0.082*** 0.304*** ** [3.90] [.67] [2.75] [2.86] [0.9] [2.46] Offsoring_ig -0.32** * [-2.32] [.4] [.2] [0.54] [-.68] [-0.72] Outsourcing [-0.8] [-0.58] [-0.43] [-0.36] [-0.64] [-0.3] Outsourcing_ig [0.58] [0.89] [.9] [0.52] [0.73] [0.66] Hig 0.090* [.84] [0.03] [0.77] [-0.5] [0.29] [-0.5] ndustry effects Yes Yes Yes Yes Yes Yes Year effects Yes Yes Yes Yes Yes Yes Observations 8,99 8,99 8,99 29,59 29,59 29,59 R-squared (witin) R-squared (between) R-squared (overa) Sargan-Hansen statistics *** *** Note. Econometric metods: Pooed OLS: pooed ordinary east squares; FE: fixed effects; RE: random effects. Te ependent variabe is defined as te ogaritm of te ratio between skied empoyees and te tota number of empoyees, were tertiary educated workers are defined as skied. Te expanation of variabes: Offsoring: dummy variabe, controing for te outward FD; Offsoring_ig: dummy variabe, controing for outward FD to ig-income countries; Outsourcing: sare of intermediate imports in te tota materia costs; Outsourcing_ig: sare of intermediate imports from ig-income countries in te tota materia costs; Hig: dummy variabe, controing for ig-income countries. Contro variabes used: ogaritm of te capita per empoyee in a firm, ogaritm of te tota factor productivity per empoyee in a firm, ogaritm of te vaue of exports, ogaritm of te average annua gross wage eve, ogaritm of te domestic cost eve. *** p<0.0, ** p<0.05, * p<0., robust t-statistics in brackets, te anaysis used custer-robust standard errors. Source: SORS, autor s cacuations FE RE Adding te interaction terms for ig-income countries sows tat te effect of outsourcing and offsoring on te sare of ig skied empoyees is different for different source countries, especiay wen incuding aso te aternative definition of skis, wic wi be presented in te subsequent subsection. Again, according to te Sargan-Hansen statistics, te most preferred resuts are obtained wit te fixed effects metod, so te foowing concusions concentrate on te resuts of tis metod. For manufacturing firms, offsoring to ow-income countries sows a statisticay significant positive impact on te sare of skied empoyees. Te insignificant interaction term between offsoring and ig- 79

94 income countries impies tat offsoring to ig-income countries does not ave a stronger impact on te reative empoyment of skied empoyees tan offsoring to ow-income countries. Hence, offsoring to ow- and ig-income countries as a simiar and positive impact on te sare of skied empoyees in manufacturing firms. Furtermore, wie offsoring to ow-income countries in service firms does not seem to ave a statisticay significant impact on te ski sare of firms, te negative and statisticay significant interaction term indicates tat offsoring to ig-income countries as a weaker impact on te reative empoyment of skied workers tan offsoring to ow-income countries. Outsourcing does not seem to ave an impact on te ski sare of firms in neiter type of firms. n order to test if te coefficients of te variabes Offsoring and Offsoring_ig in service firms are significanty different, te Wad test was appied, confirming tat offsoring to ig-income countries as a weaker impact on te reative empoyment of skied workers (te p-vaue of te F-test was 0.089). t was presumed in te introductory part of te paper tat offsoring to ow-income countries woud sift te manuayintensive parts of production abroad, keeping te ig vaue added departments in te ome country. Contrary to tis caim, one coud deduce from te resuts for service firms tat offsoring to ig-income countries migt sift departments wit iger vaue added to ig-income countries, wic woud affect te ski structure of domestic firms negativey. Tis coud be due to a firm s takeover or an urge for cost reduction; e.g. being in te area wit iger purcasing power or coser to suppiers Aternative definition of skis To tis point, workers were defined as skied wen reacing a tertiary eve of education. However, since workers gain important skis aso by working in firms and not ony by obtaining forma education, information on te occupationa eve of workers is added to te definition of skis, as expained in te metodoogy part. n order to furter increase te contribution of te anaysis, a differentiation between ig- and ow-income countries as been made aso in tis part of te anaysis. As in te previous sections, te most preferred metod, according to te Sargan-Hansen test, are te fixed effects, so te foowing concusions reate to te resuts of tis metod. 80

95 Tabe 6. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, using occupationa cassification for defining skis (observation period: ) Manufacturing firms Pooed OLS FE RE 8 Service firms Pooed OLS FE RE Offsoring 0.45*** ** 0.6** * [3.84] [.57] [2.44] [2.6] [0.79] [.82] Offsoring_ig * [-0.39] [0.55] [0.54] [-0.53] [-.68] [-.3] Outsourcing [.5] [.09] [.34] [-0.77] [-0.24] [-0.9] Outsourcing_ig [-0.56] [-0.53] [-0.57] [0.84] [-0.20] [-0.2] Hig [.] [0.06] [0.49] [0.69] [-.5] [-.0] ndustry effects Yes Yes Yes Yes Yes Yes Year effects Yes Yes Yes Yes Yes Yes Observations 8,99 8,99 8,99 29,59 29,59 29,59 R-squared (witin) R-squared (between) R-squared (overa) Sargan-Hansen statistics, *** *** Note. Econometric metods: Pooed OLS: pooed ordinary east squares; FE: fixed effects; RE: random effects. Te dependent variabe is defined as te ogaritm of te ratio between skied empoyees and te tota number of empoyees, were Managers, Professionas, and Tecnicians in SCO-88 cassification are defined as skied. Te expanation of variabes: Offsoring: dummy variabe, controing for te outward FD; Offsoring_ig: dummy variabe, controing for te outward FD to ig-income countries; Outsourcing: sare of intermediate imports in te tota materia costs; Outsourcing_ig: sare of intermediate imports from ig-income countries in te tota materia costs; Hig: dummy variabe, controing for ig-income countries. Contro variabes used: ogaritm of te capita per empoyee in a firm, ogaritm of te tota factor productivity per empoyee in a firm, ogaritm of te vaue of exports, ogaritm of te average annua gross wage eve, ogaritm of te domestic cost eve. *** p<0.0, ** p<0.05, * p<0., robust t-statistics in brackets, te anaysis used custer-robust standard errors. Source: SORS, autor s cacuations Wen defining skied empoyees ony by teir occupationa eve; i.e. wen tey are cassified as Managers, Professionas or Tecnicians, te resuts are no more statisticay significant for manufacturing firms. On te oter and, te resuts for service firms again indicate tat offsoring to ow-income countries does not seem to ave a statisticay significant impact on te ski sare of firms, wie te negative and statisticay significant interaction term indicates tat offsoring to ig-income countries as a weaker impact on te reative empoyment of skied workers tan offsoring to ow-income countries. Once more, te Wad test was appied in order to test if te coefficients of te variabes

96 Offsoring and Offsoring_ig in service firms are significanty different. Te test confirmed tat offsoring to ig-income countries as a weaker impact on te reative empoyment of skied workers (te p-vaue of te F-test was 0.048). Next, to furter expoit te advantages of taking into account information on occupationa eve, te effect of offsoring and outsourcing on firms ski sare is estimated using disaggregated data for eac of te tree major skied occupationa groups. Te foowing tabe presents ony te resuts of te most preferred metod according to te Sargan- Hansen test (tat are te fixed effects), wie te resuts of a metods are incuded in te Appendix. Wen defining skis ony by te occupationa eve and disaggregating te data by te tree major skied occupationa groups (Tabe 7), resuts for manufacturing firms sow a statisticay significant impact of offsoring ony on te reative empoyment of Professionas, wie te resuts in service firms are statisticay significant ony for Tecnicians. Te positive and statisticay significant coefficient on te interaction term between offsoring and ig-income countries for Professionas in manufacturing firms indicates tat offsoring to ig-income countries as a stronger impact on te reative empoyment of Professionas in manufacturing firms tan offsoring to ow-income countries. Furtermore, wie offsoring to ow-income countries in service firms does not seem to ave a statisticay significant impact on te reative empoyment of te treating occupations, te negative and statisticay significant interaction term for Tecnicians indicates tat offsoring to ig-income countries as a weaker impact on te reative empoyment of Tecnicians in service firms, compared to offsoring to ow-income countries. 82

97 Tabe 7. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, for te major skied occupationa groups (observation period: ) Manufacturing firms Service firms Managers Professionas Tecnicians Managers Professionas Tecnicians Offsoring [-0.07] [.02] [-0.59] [0.59] [.37] [-0.60] Offsoring_ig ** ** [0.83] [2.20] [-0.73] [-0.37] [0.39] [-2.5] Outsourcing [0.83] [-.] [.04] [0.36] [-0.29] [0.03] Outsourcing_ig [-0.6] [.9] [-0.6] [-0.84] [0.49] [0.24] Hig [-0.38] [.2] [-.7] [0.34] [0.42] [-0.77] ndustry effects Yes Yes Yes Yes Yes Yes Year effects Yes Yes Yes Yes Yes Yes Observations 8,99 8,99 8,99 29,59 29,59 29,59 R-squared (witin) R-squared (between) R-squared (overa) Sargan-Hansen statistics *** *** *** *** *** *** Note. Econometric metod: FE: fixed effects. Te dependent variabe is defined as te ogaritm of te ratio between skied empoyees and te tota number of empoyees, were Managers, Professionas, and Tecnicians in SCO-88 cassification are defined as skied. Te expanation of variabes: Offsoring: dummy variabe, controing for te outward FD; Offsoring_ig: dummy variabe, controing for te outward FD to ig-income countries; Outsourcing: sare of intermediate imports in te tota materia costs; Outsourcing_ig: sare of intermediate imports from ig-income countries in te tota materia costs; Hig: dummy variabe, controing for ig-income countries. Contro variabes used: ogaritm of te capita per empoyee in a firm, ogaritm of te tota factor productivity per empoyee in a firm, ogaritm of te vaue of exports, ogaritm of te average annua gross wage eve, ogaritm of te domestic cost eve. *** p<0.0, ** p<0.05, * p<0., robust t-statistics in brackets, te anaysis used custer-robust standard errors. Source: SORS, autor s cacuations 83

98 Finay, bot dimensions for defining skis are combined, defining workers as skied if tey meet bot criteria; i.e. if tey attain tertiary education and are cassified as Managers, Professionas, or Tecnicians. Resuts from Tabe 8 do not sow statisticay significant effects for te most preferred metod. Wen dividing te anaysis into te tree major occupationa groups (Tabe 9), te resuts again confirm tat in manufacturing firms, offsoring to ow-income countries does not seem to ave a statisticay significant impact on te reative empoyment of tertiary educated Professionas. However, coefficient on te interaction term points to a stronger impact of offsoring to ig-income countries on te reative empoyment of tertiary educated Professionas, compared to offsoring to owincome countries. n service firms, te resut is statisticay significant ony for Managers, were offsoring to ow-income countries as a positive impact on te reative empoyment of tertiary educated Managers. Te coefficient on te interaction term between offsoring and ig-income countries is statisticay insignificant, impying tat offsoring to ig-income countries does not ave a stronger impact on te reative empoyment of tertiary educated Managers, compared to offsoring to ow-income countries. Offsoring to ow- and ig-income countries terefore as a simiar and positive impact on te reative empoyment of tertiary educated Managers in service firms. Tese resuts suggest tat te effect of educationa eve is not common, but it instead differs between different occupationa groups, were te strongest impact is on Tecnicians and Managers in service firms, and Professionas in manufacturing firms. Wie te resuts for Tecnicians are no onger statisticay significant wen te skis of empoyees are defined by bot, te educationa and te occupationa eve, te resuts for Managers confirm te concusions of te previous capters tat firms indeed ave greater incentives for iring more educated workers. n addition, wen taking into account ony te tertiary educated Professionas in manufacturing firms, te coefficient is sigty iger. Terefore, te resuts indicate tat firms differentiate between more and ess educated individuas witin te same occupationa group, were te positive effects of offsoring are concentrated on te tertiary educated Managers in service firms and tertiary educated Professionas in manufacturing firms. Among te tree occupationa groups tat define skis, te majority of Managers and Professionas were on average tertiary educated and bot groups occupied te argest sare in te group of tertiary educated. Aso, wit regard to te definitions of te SCO cassification, te two groups aso perform te most demanding tasks among a occupationa groups. 84

99 Tabe 8. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, using educationa eve and occupationa cassification for defining skis (observation period: , ony tertiary educated) Manufacturing firms Service firms Pooed OLS FE RE Pooed OLS FE RE Offsoring 0.25*** *** 0.4*** *** [4.45] [.56] [2.76] [3.76] [.52] [3.08] Offsoring_ig -0.* * [-.89] [.57] [.73] [0.78] [-0.5] [0.35] Outsourcing [-.] [-0.82] [-0.62] [-0.3] [0.5] [0.43] Outsourcing_ig 2.852* [.72] [0.92] [.5] [0.20] [-0.33] [-0.24] Hig 0.089* [.84] [0.4] [.05] [-0.03] [-0.28] [-0.53] ndustry effects Yes Yes Yes Yes Yes Yes Year effects Yes Yes Yes Yes Yes Yes Observations 8,99 8,99 8,99 29,59 29,59 29,59 R-squared (witin) R-squared (between) R-squared (overa) Sargan-Hansen statistics 44.34*** *** Note. Econometric metods: Pooed OLS: pooed ordinary east squares; FE: fixed effects; RE: random effects. Te dependent variabe is defined as te ogaritm of te ratio between skied empoyees and te tota number of empoyees, were tertiary educated Managers, Professionas, and Tecnicians in SCO-88 cassification are defined as skied. Te expanation of variabes: Offsoring: dummy variabe, controing for te outward FD; Offsoring_ig: dummy variabe, controing for te outward FD to ig-income countries; Outsourcing: sare of intermediate imports in te tota materia costs; Outsourcing_ig: sare of intermediate imports from ig-income countries in te tota materia costs; Hig: dummy variabe, controing for ig-income countries. Contro variabes used: ogaritm of te capita per empoyee in a firm, ogaritm of te tota factor productivity per empoyee in a firm, ogaritm of te vaue of exports, ogaritm of te average annua gross wage eve, ogaritm of te domestic cost eve. *** p<0.0, ** p<0.05, * p<0., robust t-statistics in brackets, te anaysis used custer-robust standard errors. Source: SORS, autor s cacuations 85

100 Tabe 9. Te effect of outsourcing and offsoring on te ski sare in Sovenian manufacturing and service firms, for te major skied occupationa groups (observation period: , ony tertiary educated) Manufacturing firms Service firms Managers Professionas Tecnicians Managers Professionas Tecnicians Offsoring ** [0.54] [0.64] [.28] [2.9] [.7] [0.53] Offsoring_ig *** [.34] [2.66] [.49] [-.06] [0.3] [.6] Outsourcing [0.05] [-.38] [-0.02] [-0.76] [0.36] [0.32] Outsourcing_ig [-3.30E-03] [.28] [0.9] [0.95] [-0.40] [0.49] Hig * E [0.8] [.74] [0.07] [0.0] [0.05] [-0.25] ndustry effects Yes Yes Yes Yes Yes Yes Year effects Yes Yes Yes Yes Yes Yes Observations 8,99 8,99 8,99 29,59 29,59 29,59 R-squared (witin) R-squared (between) R-squared (overa) Sargan-Hansen statistics *** *** *** *** *** 284.6*** Note. Econometric metod: FE: fixed effects. Te dependent variabe is defined as te ogaritm of te ratio between skied empoyees and te tota number of empoyees, were tertiary educated Managers, Professionas, and Tecnicians in SCO-88 cassification are defined as skied. Te expanation of variabes: Offsoring: dummy variabe, controing for te outward FD; Offsoring_ig: dummy variabe, controing for te outward FD to ig-income countries; Outsourcing: sare of intermediate imports in te tota materia costs; Outsourcing_ig: sare of intermediate imports from ig-income countries in te tota materia costs; Hig: dummy variabe, controing for ig-income countries. Contro variabes used: ogaritm of te capita per empoyee in a firm, ogaritm of te tota factor productivity per empoyee in a firm, ogaritm of te vaue of exports, ogaritm of te average annua gross wage eve, ogaritm of te domestic cost eve. *** p<0.0, ** p<0.05, * p<0., robust t-statistics in brackets, te anaysis used custer-robust standard errors. Source: SORS, autor s cacuations 86

101 3.6 Concusion Tis paper studies te impact of offsoring and outsourcing on te reative empoyment of skied empoyees in Sovenian manufacturing and service firms. Using a matced firmeve and empoyee-eve dataset for te period from 997 to 200, te study broadens and contributes to te previous studies in severa ways. Te anaysis incorporates bot measures; offsoring and outsourcing, into one mode, and differentiates between outsourcing from ig- and ow-income countries, and offsoring to ig- and ow-income countries. Furtermore, different dimensions wen defining skis are taken into account in order to increase te expanatory vaue of te mode. Te basic mode uses a conventiona definition of skis, defining workers as skied wen tey attain tertiary education. However, since workers deveop additiona knowedge and expertise after entering empoyment, taking into account soey te eve of forma education wen defining skis ignores te knowedge acquired during te course of empoyment. Consequenty, occupationa cassification is used to define skis in te mode extensions, were tree major occupationa groups define workers as skied; Managers, Professionas and Tecnicians. Te main findings of te anaysis are te foowing. First, offsoring as a stronger effect on te reative empoyment of skied workers tan outsourcing. Second, wen controing for ig- and ow-income countries, offsoring to ow- and ig-income countries as a simiar and positive effect on te sare of skied empoyees in manufacturing firms. n service firms, offsoring to ow-income countries does not ave a statisticay significant impact on te ski sare of firms, wie offsoring to ig-income countries sows a weaker impact on te ski sare tan offsoring to ow-income countries. Tese resuts partiay confirm te ypotesis made about te expected positive effect of offsoring to ow-income countries, wie te ypotesis for te expected positive effect of outsourcing from ig-income countries cannot be confirmed. Finay, taking into account te occupationa eve wen defining skis increases te expanatory power of te mode and serves as an additiona robustness ceck. n manufacturing firms, te resuts do not vary significanty wen defining skied workers ony by te occupationa eve or wen defining skied workers by bot, te occupationa and te educationa eve. Resuts point to a stronger impact of offsoring to ig-income countries on te reative empoyment of Professionas compared to offsoring to ow-income countries in manufacturing firms. n addition, wen taking into account ony te tertiary educated Professionas, compared to a Professionas in manufacturing firms, te coefficient on te interaction term sigty increases. On te oter and, resuts for service firms vary depending on te definition of skied workers. Wen defining skied workers ony by te occupationa eve, resuts point to a weaker impact of offsoring to ig-income countries on te reative empoyment of Tecnicians, compared to offsoring to ow-income countries. Wen defining skied workers by using bot definitions, resuts indicate a positive and simiar impact of offsoring to ig- and ow-income countries on te reative empoyment of 87

102 tertiary educated Managers. Te impact of educationa eve terefore differs between occupationa groups and is mosty concentrated on te tertiary educated Managers in service firms and tertiary educated Professionas in manufacturing firms, indicating tat firms differentiate between more and ess educated individuas witin te same occupationa group. Tis finding party confirms te deduction made in te previous capters on firms aving greater incentives for iring skied workers. 88

103 CONCLUSONS Tis doctora dissertation anaysed te inkages and causaities between importing, exporting and te ski structure of firms. Te current capter summarises key findings of te first tree capters, and presents contributions and opportunities for furter researc. Summary of te main findings Te first capter presented a teoretica mode, wic anaysed te individua s decisions for investing in ski upgrading and te firm s decisions to start tecnoogy upgrading and trading. n order to torougy anayse te individua s and te firm s beaviour, te mode was spit into two parts. Te first part anaysed te individua s decisions for ski upgrading by comparing te incentives of ig abiity and ow abiity individuas for investing in acquiring a iger eve of education. Since ow abiity individuas ave to pay disproportionay iger costs for acquiring iger skis, ony te ig abiity individuas can afford to invest in attaining a iger eve of education. However, after investing in obtaining iger skis, ig abiity and ig skied individuas demand iger wages once entering te empoyment, in order to compensate for iger educationa costs. A iger wage of a worker terefore signas te empoyment of a ig abiity and ig skied individua. Tis finding was empoyed in te second part of te mode, wic studied te beaviour of firms. Before trade iberaisation, firms decide weter to invest in iger tecnoogy or not. Te benefits of te atter are ower margina production costs, wie te drawbacks are iger fixed costs. Due to iger fixed costs, ony te most productive firms invest in iger tecnoogy. After trade iberaisation, firms decide weter to start engaging in internationa activities, wic again resuts in iger costs, but aso increases revenues and/or te empoyment of skied workers. Te most productive among ow-tecnoogy firms upgrade teir ski eves and start exporting and importing, ess productive upgrade skis and start ony importing, wie te east productive ow-tecnoogy firms continue serving te domestic market. Findings of te ow-tecnoogy firms terefore revea te importance of te start of importing for increasing te eve of productivity before te start of exporting. Tis was confirmed aso in empirica studies (see for exampe Damijan & Kostevc, 205; and Atomonte & Békés, 200). On te oter and, ony te east productive among te ig-tecnoogy firms do not start engaging in internationa activities, wie te decision of te most productive ig-tecnoogy firms to start importing and/or exporting depends on te firm s productivity eve, te ski upgrading before and after importing, and on externa factors; i.e. te eve of export and import costs. Te mode in te second capter empiricay tested some of te outcomes of te teoretica mode from te previous capter. Wen anaysing inkages and causaities between te ski structure of firms and te import status, te main ambition of te mode was to test: (i) if importers empoy a iger sare of skied empoyees, compared to non-importers; (ii) if 89

104 firms wit a reativey iger ski sare sef-seect into importing; (iii) and if firms upgrade teir ski structure after te start of importing. n addition, wen studying te impact of importing on te start of exporting, te anaysis examined te infuence of buying intermediates abroad and te infuence of aving access to tecnoogies via imports on te start of exporting. Te inked empoyer-empoyee pane dataset for Sovenian manufacturing firms was appied for te period from 996 to 200. Te dataset combined severa databases, covering te information on te baance seet data and income statements of firms (i.e. number of empoyees, capita per empoyee, vaue added per empoyee, ownersip, foreign direct investments), teir import and export activities (i.e. vaue of imports and exports, and te type of imports), and attributes of empoyees (i.e. years of scooing, educationa eve, and wages.). n order to contro for measurabe differences between importing and non-importing firms, te propensity score matcing approac was appied, were various matcing tecniques were used in order to increase te significance of te resuts. Te resuts sow tat firms wit a reativey better ski structure sef-seect into importing. n addition, atoug te ski structure of firms does not increase significanty in te first year after te start of importing, it improves in te second year. Aso, in comparison to non-importing firms, importing starters sustain a iger sare of skied empoyees in te first and second year after te start of importing. Furtermore, te start of importing intermediate goods as a positive impact on te start of exporting in te first year after te start of importing, wie te impact in te second year is insignificant or even negative. Tese resuts indicate an amost instant impact of importing intermediate goods on te start of exporting. On te oter and, te start of importing capita goods as a positive and statisticay significant impact on te start of exporting ony in te second year after te start of importing. Tese resuts point to a different roe of intermediate and capita goods in te production process. More precisey, since intermediate goods usuay present raw materias wit a sorter ifespan and terefore require furter manufacturing treatment or are used for resae, te impact of intermediate goods on te production process and consequenty aso on exports migt be immediate but sort-ived. On te oter and, te capita goods usuay present firm s fixed assets and are used over a onger period. Te tird capter anaysed te impact of offsoring and outsourcing on te reative empoyment of skied empoyees in Sovenian manufacturing and service firms. Te anaysis used a simiar matced firm-eve and empoyee-eve dataset as it was used in te previous capter, covering te period from 997 to 200, and incuding some additiona information (i.e. destination of exports and imports, destination of te FD and occupationa eve of empoyees). Te atter information was used as a suppementary measure for defining skis. Te empirica anaysis was divided into two parts, wit te first part incuding a basic mode, wic studied te impact of offsoring and outsourcing on te reative empoyment of skied workers. Te mode extensions controed aso for outsourcing from ig- and ow-income countries and offsoring to ig- and ow-income countries, and considered te eve of occupation wen defining skis. Te tree major 90

105 groups of te SCO-88 cassification defined skied workers in te extended mode Managers, Professionas and Tecnicians. Te metods for pane data anaysis were used wen estimating te effects of outsourcing and offsoring (i.e. pooed OLS, random effects, and fixed effects). Te findings of te modes point to a stronger and positive effect of offsoring on te reative empoyment of skied workers. Wen controing for te ig- and ow-income countries, te resuts vary depending on wic definition of te skied empoyees is used. Offsoring to ig- and ow-income countries indicates to ave a comparabe and positive effect on te reative sare of skied empoyees in manufacturing firms. n service firms, offsoring to ow-income countries does not ave a statisticay significant impact on te reative empoyment of skied workers. Resuts aso indicate a weaker impact of offsoring to ig-income countries on te ski sare of firms in service firms, compared to te effect of offsoring to ow-income countries. Wen introducing te occupationa eve as a measure for defining skis, te resuts for manufacturing firms do not vary significanty between different definitions of skied workers. Offsoring to ig-income countries sows a stronger impact tan offsoring to ow-income countries on te reative empoyment of Professionas and te reative empoyment of tertiary educated Professionas in manufacturing firms. n contrast, resuts for service firms vary depending on te definition of skis used. Wen defining skied workers ony by using te occupationa eve as a measure, offsoring to ig-income countries sows a weaker impact on te reative empoyment of Tecnicians, in comparison to te impact of offsoring to ow-income countries. Wen defining skied workers by te occupationa and educationa eve, a positive and simiar impact of offsoring to ig- and ow-income countries on te reative empoyment of tertiary educated Managers is confirmed. Te impact of education is terefore different among different occupationa groups. Tis indicates tat firms differentiate between empoying different workers witin te same occupationa group, wic is in ine wit concusions of te previous two capters, confirming tat firms indeed ave greater incentives for iring skied workers. Finay, severa robustness cecks were used in order to increase te significance of te resuts (for exampe controing for te firm size, and using different definitions for defining te vaue added). Robustness cecks on average confirm previous resuts. Limitations of te doctora dissertation Te potentia imitations of te teoretica mode in te first capter are severa assumptions, wic in some way constrain te vaue of te mode, as it woud be ard to test tese assumptions empiricay. Te mode for instance assumes tat te increase in te reative wage eve of skied workers after investing in ig tecnoogy and after te start of importing as to be iger tan te decrease in te eve of margina costs in tese firms. An additiona imitation of te teoretica mode is aso its more or ess static outine. More precisey, even toug te mode acknowedges tree dynamic pase sifts i.e. te individua s decision to obtain a iger educationa eve, te firm s decision to invest in 9

106 ig tecnoogy, and te firm s decision to start engaging in internationa activities it andes ony two of te firm s decisions simutaneousy, e.g. investing in ig-tecnoogy vs. not investing in ig-tecnoogy, starting to import vs. continue serving te domestic market, etc., and terefore cannot torougy refect te compexity of te current environment. Atoug te data used in tis doctora dissertation is very ric, tere are some imitations in te empirica anayses due to data usage. Since previous anayses found an important inkage between te age of te firm and te empoyment growt of te firm, it woud be interesting to incude te information on firm s age as one of te expanatory variabes. Severa studies expose tat younger firms ave iger empoyment growt tan oder firms (see for exampe Banerjee & Jesenko, 205; Adeino, Robinson & Ma, 204; and Fort et a., 203). On te oter and, younger firms are more sensitive to investment opportunities and to te business cyces wic contributed to te arge empoyment decine in younger and smaer firms in te recent downturn (Fort et a., 203). Tis information is currenty not avaiabe in te dataset. n order to furter empiricay test te concusions of te teoretica mode in te first capter, additiona information on te innovation activities of firms woud be necessary. Atoug te dataset comprises information on te Community nnovation Survey (CS), tis data proved to be inadequate for severa reasons; i.e. te survey is conducted ony every two years, and te number of observations osciates significanty between te periods. n addition, to furter increase te expanatory power of te anaysis, te extended mode coud aso contro for te ig- and ow-income countries. Wit te aim of deepening te resuts of te empirica mode in te tird capter, it woud be usefu to incude additiona information on te offsoring of firms. For instance te performance indicators of foreign partners tat are in contro of domestic firms, teir empoyment structure, te vaue added, te vaue of te investment, etc. Some of te information is aready incuded in te dataset, but ony for recent years. Terefore, by extending te observation period and adding te data for te most recent years, tis new information migt bring additiona contributions to te resuts. Neverteess, in spite of imitations, presented in tis section, tis doctora dissertation as severa contributions, wic are gatered in te foowing section. Scientific contributions and future researc suggestions Te teoretica mode in te first capter as two main contributions. First, it extends te contents of te teoretica modes of Bustos (20a, 20b) and Meitz (2003) on eterogeneous firms, by incuding aso importers in te mode. Second, te mode broadens te content of te current teoretica trade modes, wic to te best of my knowedge focused ony on firms beaviour, by anaysing aso te beaviour of individuas and teir decision for ski upgrading. Tese findings are ater used in te part 92

107 of te mode, wic studies te firm s beaviour. Te mode opens severa possibiities for future researc, were te foowing ave been tested in te empirica mode in te second capter: (i) importers ave a better ski structure prior to importing; (ii) te structure of workers canges in favour of te skied workers after te start of importing; and (iii) by aving access to tecnoogy and/or to intermediates, importing as a positive impact on te start of exporting. n addition, future studies migt aso take into account te imitations of te mode, presented in te previous section. t woud be especiay noteworty to consider adjusting te more or ess static structure of te mode, wic does not refect te compex decision-making of firms in today s environment. Te empirica anaysis in te second capter contributes to te fied of knowedge by adding new understandings of te sef-seection of importing firms and inkages between importing and te ski structure of firms. Specificay, a stages of importing are taken into account in te anayses; i.e. before te start of importing, first years of importing, and ong-term importers. Moreover, types of imports were taken into account to furter deepen te expanatory power of te anaysis, wen anaysing te impact of te start of importing on te start of exporting. Here, te anaysis distinguised between importing intermediate and capita goods. Finay, concusions of te study migt aso benefit governments, firms, jobseekers and students. Wit te aim of furter increasing te productivity and competitiveness of domestic firms, te governments soud focus on stimuating internationa cooperation and ski upgrading in domestic firms. n turn, wie firms soud ave iger incentives for iring skied workers, students, jobseekers and workers soud ave a iger stamina for increasing teir educationa eve and obtaining additiona onte-job training. As introduced in te previous section, it was not possibe to contro for innovations due to te sortcomings of te CS dataset. However, in te case of better data quaity, it woud be advantageous to contro aso for innovations in order to test additiona findings of te teoretica mode and to increase te expanatory power of te anaysis. Atoug tis was not one of te concusions of te teoretica mode, it woud aso be interesting to anayse te causaity between exporting and te ski structure of firms, and weter tere aso exists a reverse causaity between exporting and importing. Finay, te first contribution of te empirica anaysis in te tird capter incudes bot measures of strategic positioning of firms, i.e. offsoring and outsourcing, a in one mode. To te best of my knowedge, previous studies controed for ony one of te factors simutaneousy in teir modes. Since previous empirica studies confirmed tat bot of te measures infuence te abour demand and since bot are expected to increase in te future, it was important to incude bot in one mode in order to avoid te missing variabe bias. n addition, te study aso differentiated between outsourcing from and offsoring to ig- and ow-income countries, wic renders new interpretations of te resuts possibe and takes into account te differences of partner s performance. Since previous empirica studies exposed te importance of differentiating between manufacturing and service firms, te anaysis was made separatey for bot types of firms. 93

108 Furtermore, in order to increase te expanatory power of te mode and contro for te fact tat individuas acquire skis aso during a earning process wie being empoyed and not ony wen gaining te officia eve of education, te anaysis defined skis by taking into account bot measures; i.e. te educationa and te occupationa eve. As presented in te previous section, it woud be important to contro for more detaied information on offsoring, especiay since offsoring proved to ave an important impact on te ski structure of firms. 94

109 REFERENCES Adeino, M., Ma, S., & Robinson, D. T. (204). Firm Age, nvestment Opportunities, and Job Creation. NBER Working Paper Series, Paper No Atomonte, C., Aquiante, T., Békés, G., & Ottaviano, G.. P. (203). nternationaization and innovation of firms: evidence and poicy. Economic Poicy, 28(76), Atomonte, C., & Békés, G. (200). Trade Compexity and Productivity. CeFiG Working Papers, Paper No. 2. Avarez, R., Ortega, C. B. & Navarro, L. (202). Product Mix Canges and Performance in Ciean Pants. Working Papers University of Cie, Department of Economics, Paper No Amiti, M., tskoki, O., & Konings, J. (204). mporters, Exporters, and Excange Rate Disconnect. American Economic Review, 04(7), Amiti, M., & Davis, D. R. (20). Trade, Firms, and Wages: Teory and Evidence. Review of Economic Studies, 79(), 36. Amiti, M., & Konings, J. (2007). Trade Liberaization, ntermediate nputs, and Productivity: Evidence from ndonesia. American Economic Review, 97(5), Amiti, M., & Wei, S.-J. (2005a). Fear of service outsourcing: is it justified? Economic Poicy, 20(42), Amiti, M., & Wei, S.-J. (2005b). Service Offsoring, Productivity, and Empoyment: Evidence from te United State. MF Working Paper, Paper No. WP/05/238. Attanasio, O., Godberg, P., & Pavcnik, N. (2003). Trade reforms and wage inequaity in Coombia. NBER Working Papers, Paper No Austin, P. C. (20). Optima caiper widts for propensity-score matcing wen estimating differences in means and differences in proportions in observationa studies. Parmaceutica Statistics, 0(2): Aw, B. Y., Roberts, M. J., & Xu, D. Y. (20). R&D nvestment, Exporting, and Productivity Dynamics. American Economic Review, 0(4), Aw, B. Y., Roberts, M. J., & Xu, D. Y. (2008). R&D nvestments, Exporting, and te Evoution. American Economic Review: Papers & Proceedings, 98(2), Badwin, R. E., & Cain, G. G. (2000). Sifts in Reative U.S. Wages: Te Roe of Trade, Tecnoogy, and Factor Endowments. Te Review of Economics and Statistics, 82(4), Banerjee, B., & Jesenko, M. (205). Te roe of firm size and age in empoyment growt: Evidence for Sovenia, Bank of Sovenia working papers, Paper No. 2/205. Bas, M., & Strauss-Kan, V. (204). Does importing more inputs raise exports? Firm-eve evidence from France. Review of Word Economics, 50(2), Becker, S. O., & cino, A. (2002). Estimation of average treatment effects based on propensity scores. Te Stata Journa, 2(4),

110 Bernard, A. B., & Jensen, J. B. (999). Exceptiona exporter performance: cause, effect, or bot? Journa of nternationa Economics, 47(), 25. Biscourp, P., & Kramarz, F. (2007). Empoyment, ski structure and internationa trade: Firm-eve evidence for France. Journa of nternationa Economics, 72(), Boom, N., Draca, M., & Van Reenen, J. (20). Trade nduced Tecnica Cange? Te mpact of Cinese mports on nnovation, T and Productivity. NBER Working Paper Series, Paper No Brambia,., Lederman, D., & Porto, G. (200). Exports, Export Destinations, and Skis. NBER Working Papers, Paper No Burstein, A., Cravino, J., & Voge, J. (203). mporting Ski-Biased Tecnoogy. American Economic Journa: Macroeconomics, 5(2), Bustos, P. (20a). Trade Liberaization, Exports, and Tecnoogy Upgrading: Evidence on te mpact of MERCOSUR on Argentinian Firms. American Economic Review, 0(), Bustos, P. (20b). Te mpact of Trade Liberaization on Ski Upgrading: Evidence from Argentina. Barceona GSE Working Paper Series, Paper No Cameron, A. C., & Trivedi, P. K. (2009). Microeconometrics Using Stata. Coege Station, Texas: StataCorp LP. Corvers, F. (997). Te impact of uman capita on abour productivity in manufacturing sectors of te European Union. Appied Economics, 29(8), Crino, R. (202). Service Offsoring and te Ski Composition of Labour Demand. Oxford Buetin of Economics and Statistics, 74(), Damijan, J., & Kostevc, Č. (205). Learning from trade troug innovation: Causa ink between imports, exports and innovation in Spanis microdata. Oxford Buetin of Economics and Statistics, 77(3), Damijan, J., Konings, J., & Poanec, S. (204). mport Curning and Export Performance of Muti-product Firms. Te Word Economy, 37(), Damijan, J., Konings, J., & Poanec, S. (203). Pass-on Trade: Wy do Firms Simutaneousy engage in Two-Way Trade in te Same Varieties? Review of Word Economics, 49(), 85. Damijan, J., Kostevc, Č., & Poanec, S. (200). From nnovation to Exporting or Vice Versa? Te Word Economy, 33(3), De Backer, K., & Yamano, N. (202). nternationa comparative evidence on goba vaue cains. OECD Science, Tecnoogy and ndustry Working Paper, Paper No. 202/3. Deejia, R., & Waba, S. (999). Causa Effects in Nonexperimenta Studies: Reevauating te Evauation of Training Programs. Journa of te American Statistica Association, 94, Deejia, R., & Waba, S. (2002). Propensity Score-Matcing Metods for Nonexperimenta Causa Studies. Review of Economics and Statistics, 84, 5 6. Egger, H., & Egger, P. (2003). Outsourcing and ski-specific empoyment in a sma economy: Austria after te fa of te ron Curtain. Oxford Economic Papers, 55(4),

111 Eias, P., & Birc, M. (994). Estabisment of Community-Wide Occupationa Statistics. SCO 88 (COM), A Guide for Users. nstitute for Empoyment Researc, University of Warwick. Eurostat, statistica database. Retrieved June 2, 205, from: ttp://ec.europa.eu/eurostat/data/database Feenstra, R. C., & Hanson, G. H. (999). Te mpact of Outsourcing and Hig-Tecnoogy Capita on Wages: Estimates for te United States. Te Journa of Economics, 4(3), Feenstra, R. C., & Hanson, G. H. (996). Gobaization, Outsourcing, and Wage nequaity. Te American Economic Review, 86(2), Feng, L., Li, Z., & Swenson, D. L. (202). Te Connection between mported ntermediate nputs and Exports: Evidence from Cinese Firms. NBER Working Papers, Paper No Fieer, A. C., Esava, M., & Xu, D. Y. (204). Trade, Skis and Quaity Upgrading: A Teory wit Evidence from Coombia. NBER Working Papers, Paper No. w9992. Fort, T. C., Hatiwanger, J., Jarmin, R. S., & Miranda, J. (203). How Firms Respond to Business Cyces: Te Roe of Firm Age and Firm Size. NBER Working Paper Series, Paper No Foster, N., Sterer, R., & de Vries, G. (202). Offsoring and te ski structure of abour demand. wiiw Working Papers, Paper No. w9992. Fryges, H., Voge, A., & Wagner, J. (204). Te mpact of R&D Activities on Exports of German Business Services Enterprises: First Evidence from a Continuous Treatment Approac. Te Word Economy. doi: 0./twec.277. Godberg, P. K., Kandewa, A. K., Pavcnik, N., & Topaova, P. (200). mported ntermediate nputs and Domestic Product Growt: Evidence from ndia. Te Quartery Journa of Economics, 25(4), Godberg, P. K., & Pavcnik, N. (2007). Distributiona Effects of Gobaization in Deveoping Countries. Journa of Economic Literature, 45(), Greenaway, D., & Kneer, R. A. (2004). Exporting and Productivity in te United Kingdom. Oxford Review of Economic Poicy, 20(3), Greenag, C., & Rogers, M. (200). nnovation, inteectua property, and economic growt. New Jersey: Princeton University Press. Grossman, G. M., & Rossi-Hansberg, E. (2008). Trading Tasks: A Simpe Teory of Offsoring. American Economic Review, 98(5), Hapern, L., Koren, M., & Szeid, A. (20). mported nputs and Productivity. CeFiG Working Papers, Paper No. 8. Heckman, J., cimura, H., & Todd, P. (997). Matcing as an Econometric Evauation Estimator: Evidence from Evauating a Job Training Programme. Review of Economic Studies, 64(4), Hepman, E., tskoki, O., & Redding, S. (20). Trade and Labor Market Outcomes. NBER Working Papers, Paper No

112 Hijzen, A., & Swaim, P. (200). Offsoring, abour market institutions and te easticity of abour demand. European Economic Review, 54(8), Hijzen, A., Görg, H., & Hine, R. C. (2005). nternationa outsourcing and te ski structure of abour demand in te United Kingdom. Te Economic Journa, 5(506), Hi, J. L., Reiter, J. P., & Zanutto, E. L. (2004). A comparison of experimenta and observationa data anayses. n Geman, A. and X. L. Meng (Eds.), Appied Bayesian and causa inference from an incompete data perspective (49 60). New York: Wiey. Horgos, D. (2006). nternationa Outsourcing Some Measurement Probems: An Empirica Anaysis of Outsourcing Activities in Germany. AW-Report, 34(), Hummes, D., Jørgensen, R., Munc, J., & Xiang, C. (204). Te Wage Effects of Offsoring: Evidence from Danis Matced Worker-Firm Data. American Economic Review, 04(6), cino, A., & Becker, S. (2002). Estimation of Average Treatment Effects Based on Propensity Scores. Te Stata Journa, 2(4), dson, T. L., & Oi, W. Y. (999). Workers Are More Productive n Large Firms. Te American Economic Review, 89(2), LO (204). nternationa Standard Cassification of Occupations. Retrieved August 2, 204, from: ttp:// LO (202). nternationa Standard Cassification of Occupations. Structure, groups definitions and correspondence tabes. Retrieved January 8, 205, from: ttp:// pub/documents/pubication/wcms_72572.pdf MF (203). Trade interconnectedness: te word wit goba vaue cains. Retrieved August 9, 204, from: ttp:// Kasaara, H., & Rodrigu, J. (2008). Does te Use of mported ntermediates ncrease Productivity? Pant-Leve Evidence. Journa of deveopment economics, 87(), Kiriyama, N. (202). Trade and nnovation: Syntesis Report. OECD Trade Poicy Papers, Paper No. 35, OECD Pubising. Konings, J., & Murpy, A. P. (2006). Do Mutinationa Enterprise Reocate Empoyment to Low-Wage Regions? Evidence from European Mutinationas. Review of Word Economics, 42(2), Kostevc, Č. (2009). Foreign Market Competition as a Determinant of Exporter Performance: Evidence from Sovenian Manufacturing Firms. Te Word Economy, 32(6), Krugman, P. R., Obstfed, M., & Meitz, M. J. (202). nternationa Economics : Teory & Poicy (9 t ed.). Boston [etc.] : Pearson, cop. Krugman, P. R. (99). ncreasing Returns and Economic Geograpy. Journa of Poitica Economy, 99(3), Krugman, P. R. (979). ncreasing Returns, Monopoistic Competition, and nternationa Trade. Journa of nternationa Economics, 9(4),

113 Levinson, J., & Petrin, A. (2003). Estimating Production Functions Using nputs to Contro for Unobservabes. Te Review of Economic Studies, 70(2), Lo Turco, A., & Maggioni, D. (202). Offsoring to Hig and Low ncome Countries and te Labor Demand. Evidence from taian Firms. Review of nternationa Economics, 20(3), Lunt, M. (204). Seecting an Appropriate Caiper Can Be Essentia for Acieving Good Baance Wit Propensity Score Matcing. American Journa of Epidemioogy, 79(2), Meitz, M. J., & Redding, S. J. (204). Heterogeneous Firms and Trade. n Gopinat, G., Hepman, E., & Rogoff, K. (Eds.), Handbook of nternationa Economics (4 t ed., 54). Esevier. Meitz, M. J. (2003). Te mpact of Trade on ntra-ndustry Reaocations and Aggregate ndustry Productivity. Econometrica, 7(6), Mesci, E., Taymaz, E., & Vivarei, M. (2008). Trade Openness and te Demand for Skis: Evidence from Turkis Microdata. ZA Discussion Papers, Paper No Mice, B., & Rycx, F. (2009). Does Offsoring of Materias and Business Services Affect Empoyment? Evidence from a Sma Open Economy. ZA Discussion Paper Series, Paper No Mincer, J. (974). Scooing, experience and earnings. New York and London: Nationa Bureau of Economic Researc. Mion, G., & Zu, L. (203). mport competition from and offsoring to Cina: A curse or bessing for firms? Journa of nternationa Economics, 89(), Mion, G., Vandenbussce, H., & Zu, L. (200). Trade wit Cina and ski upgrading: Evidence from Begian firm eve data. NBB Working Paper, Paper No. 94. Mitra, D., & Ranjan, P. (2009). Offsoring and Unempoyment: Te Roe of Searc Frictions and Labor Mobiity. ZA Discussion Paper Series, Paper No Muender, M. A. (2004). Trade, Tecnoogy and Productivity: A Study of Braziian Manufacturers, CESifo Working Paper Series, Paper No. 48. Muus, M., & Pisu, M. (2007). mports and Exports at te Leve of te Firm: Evidence from Begium. CEP Discussion Papers, Paper No. 80. Oey, G. S., & Pakes, A. (996). Te Dynamics of Productivity in te Teecommunications Equipment ndustry. Econometrica, 64(6), Onodera, O. (2008). Trade and innovation project. OECD Trade Poicy Working Paper, Paper No. 72. Panagariya, A. (98). Variabe Returns to Scae in Production and Patterns of Speciaization. Te American Economic Review, 7(), Papke, L. E., & Woodridge, J. M. (2008). Pane data metods for fractiona response variabes wit an appication to test pass rates. Journa of Econometrics, 45(-2), Parro, F. (203). Capita-Ski Compementarity and te Ski Premium in a Quantitative Mode of Trade. American Economic Journa: Macroeconomics, 5(2),

114 Petrin, A., Poi, B. P., & Levinson, P. (2004). Production function estimation in Stata using inputs to contro for unobservabes. Te Stata Journa, 4(2), Pierce, J. R., & Scott, P. K. (202). Te Surprisingy Swift Decine of U.S. Manufacturing Empoyment. NBER Working Papers, Paper No Rave, O., & Resef, A. (203). Capita mports Composition, Compementarities, and te Ski Premium in Deveoping Countries. Retrieved November 3, 203, from: ttp://papers.ssrn.com/so3/papers.cfm?abstract_id= Rosenbaum, P. R., & Rubin, D. B. (983). Te Centra Roe of te Propensity Score in Observationa Studies for Causa Effects. Biometrika, 70(), Rubin, D. B. (997). Estimating causa effects from arge data sets using propensity scores. Annas of nterna Medicine, 27, Scaffer, M.E., & Stiman, S. (200). xtoverid: Stata modue to cacuate tests of overidentifying restrictions after xtreg, xtivreg, xtivreg2 and xttayor. Retrieved August 6, 204, from ttp://ideas.repec.org/c/boc/bocode/s tm Scank, T., Scnabe, C., & Wagner, J. (2007). Do exporters reay pay iger wages? First evidence from German inked empoyer-empoyee data. Journa of nternationa Economics, 72(), Smeets, V., & Warzynski, F. (200). Learning by Exporting, mporting or Bot? Estimating productivity wit muti-product firms, pricing eterogeneity and te roe of internationa trade. Aarus University, Department of Economics Working Papers, Paper No Smit, A. (904). An nquiry into te Nature and Causes of te Weat of Nations. Edwin Cannan, E. (Ed.), Library of Economics and Liberty. Retrieved August 8, 205, from: ttp:// Stark, O., & Wang, Y. (200). nducing Human Capita Formation: Migration as a Substitute for Subsidies. University of Bonn, Center for Deveopment Researc, Discussion Papers on Deveopment Poicy, Paper No.. Stark, O., & Cau, N. H. (998). Human Capita Formation, Asymmetric nofrmation, and te Dynamics of nternationa Migration. Reie Ökonomie / Economics Series, Paper No. 52. Stark, O., Hemenstein, C., & Prskawetz, A. (998). Human Capita Depetion, Human Capita Formation, and Migration: A Bessing in a Curse? Reie Ökonomie / Economics Series, Paper No. 55. Stark, O., Hemenstein, C., & Prskawetz, A. (997). A Brain Gain wit a Brain Drain. Reie Ökonomie / Economics Series, Paper No. 45. Strauss-Kan, V. (2003). Te roe of gobaization in te witin-industry sift away from unskied workers in France. NBER Working Paper Series, Paper No Syverson, C. (20). Wat Determines Productivity? Journa of Economic Literature, 49(2), Tokarick, S. (2002). Quantifying te mpact of Trade on Wages: Te Roe of Nontraded Goods. MF Working Paper, Paper No. 02/9. 00

115 UNCTAD, Word investment report 203. Retrieved August, 9, 204, from: ttp://unctad.org/en/pubicationslibrary/wir203_en.pdf Voge, A., & Wagner, J. (2008). Higer Productivity in mporting German Manufacturing Firms: Sef-seection, Learning from mporting, or Bot? ZA Discussion Paper Series, Paper No Wagner, J. (202a). Exports, R&D and productivity: a test of te Bustos-mode wit German enterprise data. Economics Buetin, 32(3), Wagner, J. (202b). Effects of Productivity and mport on Firm-Leve Export. Ratio Working Papers, Paper No Wagner, J. (20). Offsoring and firms performance: sef-seection, effects on performance, or bot? Review of Word Economics, 47(2), Wagner, J. (2007). Exports and Productivity: A Survey of te Evidence from Firm-eve Data. Te Word Economy, 30(), WB (205) Retrieved February 9, 205, from: ttp://data.wordbank.org/about/countryand-ending-groups. Word Bank, statistica database. Retrieved Apri 22, 205, from: ttp://data.wordbank.org/indicator/ne.mp.gnfs.zs WTO (204). Word Trade Report 204. Retrieved October 2, 204, from: ttp:// Yang, Y., & Maick, S. (200). Export Premium, Sef-seection and Learning-by- Exporting: Evidence from Cinese Matced Firms. Te Word Economy, 33(0), Zajc Kejžar, K., & Ponikvar, N. (204). Job Destruction and Productivity Gains in Heterogeneous ncumbent Firms: Comparing te Effects of mports and nward Foreign Direct nvestment. Emerging Markets Finance and Trade, 50(6), Zanutto, E. L. (2006). A Comparison of Propensity Score and Linear Regression Anaysis of Compex Survey Data. Journa of Data Science, 4, Zogi, C. (200). Measuring Labor Composition: A Comparison of Aternate Metodoogies. n Abraam, K. G., Spetzer, J. R., & Harper, M. (Eds.), Labor in te New Economy ( ). University of Cicago Press. 0

116

117 APPENDCES

118

119 LST OF APPENDCES Appendix A: Derivation of te Caim on te gross earnings of te rationa individuas wit te ow abiity being positive (Capter )... Appendix B: Propensity Score istograms and Baancing property tests... 2 Appendix C: Additiona robustness cecks of matcing resuts... 0 Appendix D: Derivation of te mode in Capter Appendix E: Description of SCO-88 major occupationa groups... 7 Appendix F: Compementary tabes to Capter Appendix G: Summary in Sovenian anguage / Dajši povzetek v sovenskem jeziku.. 33 i

120

121 Appendix A: Derivation of te Caim on te gross earnings of te rationa individuas wit te ow abiity being positive (Capter ) Caim 2: Te gross earnings of rationa individuas wit te ow abiity are positive (i.e. w U (Θ U * ) > 0). Proof: Consider rationa workers wo maximize teir gross earnings and terefore acieve te optima eve of abiity: w U (Θ U * ) = λ[n(λ / k U )] k U [(λ / k U ) ]. (A) After a simpe cacuation, one gets te foowing: w U (Θ U * ) = λ[n(λ) n(k U ) ] + k U. (A2) Knowing te foowing inequaity ods: λ > k U > k S >, it foows tat w U (Θ U * ) > 0.

122 Appendix B: Propensity Score istograms and Baancing property tests Figure B. Te Propensity Score istogram on te differences in te ski structure between importers and non-importers Note. Untreated: reates to firms in te contro group (i.e. non-importing firms), Treated: reates to firms in te treatment group (i.e. importing firms). Source: SORS, autor s cacuations Figure B2. Te Propensity Score istogram on te sef-seection of firms into importing Note. Untreated: reates to firms in te contro group (i.e. non-importing firms), Treated: reates to firms in te treatment group (i.e. importing starters). Source: SORS, autor s cacuations 2