The Impact of Trade on Technology and Skill Upgrading Evidence from Argentina

Transcription

1 The Impat o Trae on Tehnology an Skill Upgraing Eviene rom Argentina Paula Bustos * CREI an Universitat Pompeu Fabra November 2005 Abstrat In the last 20 years, wage inequality has inrease in many eveloping ountries. Most researh on this topi ouses on two possible auses: trae or skill-biase tehnial hange. Several empirial stuies oument inreases in skill intensity within all setors, avoring the tehnologial hange eplanation over trae. Instea, I present an test a moel where trae an apital aount liberalization inrease the proitability o skill-biase new tehnologies in all setors, but there is ross-irm heterogeneity in tehnology aoption ue to ierenes in proutivity, ie osts o aoption an reit onstraints. iberalization rives the aoption o new tehnologies in eveloping ountries by inreasing market size, eliminating restritions on tehnology transers an reuing interest rates. Only the most proutive irms enter eport markets, an thus make enough proits to over the ie ost o tehnology aoption. Foreignowne irms are less subjet to reit onstraints, an hene are also more likely to be able to inane investment in tehnology. The moel preits that the level o tehnology spening an skill intensity ater liberalization is isretely higher or eporters an oreign-owne irms; an the hange in tehnology spening an skill upgraing has an inverte U shape, being highest or irms in the mile range o the proutivity istribution, suh as new entrants in the eport market. I test these theoretial preitions in the ontet o the trae an apital aount liberalization in Argentina in the early 990s. I analyze a irmlevel ata set that iers rom stanar inustry surveys by inluing iret inormation on euation o workers an on spening in tehnology aoption. I in that the inrease in the relative eman o skille labor oes not ome rom labor realloation aross setors, nor aross irms but rom skill upgraing within irms. Firms that upgrae tehnology aster also upgrae skill aster. The preitions o the moel in terms o levels an hanges in tehnology are onirme by the ata, ontrolling or setor, initial proutivity an size. * I am grateul to Philippe Aghion, Pol Antras, Elhanan elpman an speially Mar Melitz or their avie an support. For helpul suggestions an omments, I also wish to thank Ivan Fernanez-Val, Manuel Amaor, Elsa V. Artai, Thomas Chaney, Pasaline Dupas, Antara Dutta, Doireann Fitzgeral, Gita Gopinath, Marius entea, Gustavo ugones, Kenneth Rogo, Karine Seraty an partiipants at the arvar International unh. All remaining errors are mine. paula.bustos@up.eu

2 . Introution In the last 20 years, wage inequality has inrease in many evelope an eveloping ountries. Most researh on this topi ouses on two possible auses, trae or skillbiase tehnial hange. The eviene suggests that the main ause o the rising skill premium has been skill-biase tehnial hange, with trae playing a minor role. An open question remains on the eets o trae on tehnology aoption an its eets on skill upgraing through that partiular hannel. The view that the rise in skill premia is ause by tehnologial hange is partly oune in empirial stuies that ontrait the preitions o the eksher-ohlin -O trae moel: when a skill-abunant ounty opens up to trae, the relative prie o skillintensive goos inreases an proution shits towars skill-intensive setors, inreasing the relative eman or skille labor an the skill premium; onversely, trae opening in skill-sare eveloping ountries woul lea to a reution in the skill premium. In the Unite States, where there has been a sharp inrease in the skill premium in the last 20 years, the relative prie o skille-labor-intensive goos has not inrease awrene an Slaughter, 993. Most o the inrease in the relative eman o skille labor has ourre within manuaturing setors, with only a minor part being eplaine by the epansion o skill-intensive setors Berman, Boun an Grilihes, 994. The ining that all setors inrease their relative eman or skille labor an that the rate o skill upgraing has been greater in omputer-intensive inustries Autor, Katz an Krueger, 998 suggests that skill-biase tehnial hange has playe a more important role than trae. In aition, several empirial stuies oument a onsierable inrease in the skill premium ater trae liberalization in eveloping ountries like Brazil, Chile, Colombia an Meio. 2 In the ase o Argentina, ater trae was liberalize in the early 990 s, the ollege wage premium inrease 0 perentage points per year, while it ha been stable in the 980 s Galeani et al In these ountries the share o skille workers has also Feenstra an anson 996, 999 argue that when trae in intermeiate inputs is introue in a -O ramework, inrease trae an ause skill upgraing within IV igit setors. 2 See Pavnik et al 2004 on Brazil, Ginling an Robbins 200 on Chile an Costa Ria; Attanasio et al 2004 on Colombia, anson an arrison 999 on Meio.

3 inrease within most inustries. These empirial inings are also onsistent with the view that skill-biase tehnial hange is the ause o wiening wage inequality, an are har to reonile with the preitions o the -O moel. 3 I will ollow a ierent theoretial perspetive that looks at the interplay o tehnologial hange an trae liberalization. Aemoglu 998, 2003 suggests that the empirial inings or the U.S. are onsistent with globalization as the root ause o wiening wage inequality i growing trae inreases the skill-bias o tehnial hange in skill-abunant ountries. Yeaple 2005 shows that inrease eport opportunities make aoption o new tehnologies proitable or more irms, thus inreasing the aggregate eman o skille labor an the skill premium. This paper presents simple moel where trae an apital aount liberalization inrease the proitability o new tehnologies an the relative eman or skille labor in eveloping ountries, an tests its preitions in the ontet o the trae an apital aount liberalization in Argentina in the early 990s. The moel buils on work by Melitz 2003 an Yeaple 2005, eparting rom the - O ramework by introuing inreasing returns to sale an monopolisti ompetition, as in Krugman 979, 980, an by ousing on within setor irm heterogeneity rather than ierenes in skill intensity aross setors. Firms are heterogeneous in an unerlying proutivity parameter whih an be interprete as managerial ability, an an hoose to aopt a lower marginal ost new tehnology, ater paying a ie ost. Within eah setor, only the most proutive irms enter the eport market an thus make 3 The inrease in the skill premium in atin Amerian ountries an be reonile with the -O ramework i unskille-labor intensive inustries were relatively more protete prior to liberalization, or i the ountries also open up to trae with more unskille-labor abunant ountries like China. These eplanations woul still work through realloations o labor towars skill intensive setors, an are not onsistent with the eviene or Colombia an Meio. In the irst ountry, Attanasio et al in no eviene o labor realloation aross setors, an in that hanges in skill premiums annot be relate to hanges in taris aross setors. Feliiano 200 ins similar results or Meio, an Verhoogen 2004 reports employment shits towars unskille labor intensive inustries, oinient with rising skill premia. Feenstra an anson 996, 997 evelop a moel that aounts or the simultaneous inrease in the skill premium in a evelope an a eveloping ountry when they open up to trae, introuing apital movements an trae in intermeiate inputs, eplaining the inrease in the relative eman or skille labor within setors. Antras et al 2005 present a moel where globalization leas to the ormation o hierarhial teams aross ountries, leaing to higher wage inequality in eveloping ountries as higher ability workers orm teams with managers in evelope ountries. 2

4 enough proits to pay the higher ie osts o aopting the new tehnology. Trae liberalization reues variable eport osts, inreasing eporting revenues an inuing more irms to enter the eport market, whih makes aoption o new tehnologies proitable or more irms. In aition, it reues the ost o aoption o new tehnologies through the elimination o taris on importe apital goos an restritions on tehnology transers, making aoption proitable or more eporters. Capital aount liberalization in a apital sare ountry reues the interest rate, urther lowering the ost o investment in new tehnologies. Aoption o skill-intensive new tehnologies inreases the relative eman or skille labor an the skill premium. But beause this aets all setors, the eet o trae annot be ientiie through variation aross setors. This an rationalize the small eets o trae oun in empirial stuies with setor-level ata. Instea, I stuy heterogeneous responses to trae an apital aount liberalization o irms within setors. I analyze a new panel o Argentinean manuaturing irms overing the perio An avantage o this ata set is that it inlues inormation on the euational level o workers, while stanar inustry surveys an ensus only lassiy workers into proution P an non proution NP oupational ategories. An aitional new eature o this ata set is that it permits to buil a omprehensive measure o tehnology upgraing, as it inlues several imensions o aoption o new tehnologies suh as spening in high teh apital goos, omputers an sotware; payments or tehnology transers an patents; an spening on equipment, materials an labor relate to innovation ativities perorme within the irm. 4 In a preliminary analysis o the ata, I in that the equilibrium relative eman o skille labor inrease 7% in the perio Galeani et al 2003 report that in the same perio the skill premium was growing at an average o 7 perentage points per 4 Suh as R&D, aaptation o new prouts or proution proesses, tehnial assistane or innovations in proution, organization, ommerialization, engineering an inustrial esign. 5 The equilibrium relative eman or skille labor is measure as the ratio o skille workers ivie by unskille workers. Skille workers are ollege grauates plus tertiary euation grauates onverte to ollege equivalents using the 992 ollege wage premium. Unskille workers are primary shool grauates an high shool grauates onverte to primary shool equivalents using the 992 high shool wage premium. As in both years workers are weighte by the 992 wage premium, measure hanges in the equilibrium relative eman o skille labor only relet hanges in quantities, an not in pries. 3

5 year in the inustrial setor, iniating that the rise in the equilibrium relative eman o skille labor must ome rom a eman shit. Out o the 7 perentage points inrease in the equilibrium relative eman o skille labor in the perio , 5 perentage points are eplaine by skill upgraing within P, NP an R&D oupational ategories, an only 2 perent by realloations o labor rom P towars NP an R&D. This eviene suggests that previous stuies that have use P an NP as a proy or unskille an skille labor might be unerestimating the egree o skill upgraing. Aitionally, I in that the inrease in relative eman o skille labor oes not ome rom labor realloation aross setors, nor aross irms, but rom skill upgraing within irms. This eviene points towars tehnology upgraing within setors an irms as the main ause o the inrease in the relative eman o skille labor. Thereore, my empirial work ouses on investigating the eets o trae an apital aount liberalization on tehnology aoption an its eets on skill upgraing through that partiular hannel. In the moel, initial heterogeneity will etermine ierential irm-level responses to liberalization. The moel has preitions both in terms o levels an hanges in tehnology spening ater liberalization or irms o ierent initial proutivity: ontinuing eporters irms that eporte both beore an ater trae liberalization, new eporters irms that starte eporting ater liberalization an never eporters irms that i not eport beore nor ater liberalization. I test the ollowing ive preitions o the moel: irst, as the only irms using the new tehnology beore liberalization are the most proutive ontinuing eporters, observe skill intensity beore liberalization is higher only or this group; seon, ontinuing eporters an the most proutive new eporters have a isretely higher level o spening in tehnology ater trae liberalization; thir an ourth, the hange in tehnology spening an skill upgraing ater trae liberalization has a an inverte U shape, being highest or irms in the mile range o the proutivity istribution new eporters an the least proutive ontinuing eporters; inally, irms that upgrae tehnology aster also upgrae skill aster. The survey also ontains inormation on the soures o inaning tehnology spening that iniates that inanial unerevelopment poses onstraints on the optimal 4

6 tehnology hoies preite by the moel, speially or small an home-owne irms. In the ontet o the moel, the presene o reit onstraints ouple with ie osts o tehnology aoption imply that oreign-owne irms will be more likely to be able to inane investment in tehnology. When taking the preitions o the moel to the ata, I test or isrete ierenes between ontinuing eporters, new eporters, oreign-owne irms an omestiallyowne non eporters, both in levels an hanges in tehnology spening an skill intensity. I in that, within eah 4 igit SIC inustry, ontinuing eporters an oreignowne irms were more skill-intensive than omestially-owne never eporters prior to liberalization. Firms whih starte eporting ater liberalization were not initially more skill intensive than never eporters, but upgrae skill aster ater trae liberalization. I also in that new eporters, ontinuing eporters an oreign-owne irms spen 53% to 69% more in tehnology than omestially-owne never eporters ater trae liberalization, ontrolling or 4 igit SIC inustry, initial proutivity an initial size. Moreover, new eporters upgrae tehnology aster than other irms. Finally, I show that irms that invest more in tehnology upgraing also upgrae skill aster, where one stanar eviation in the hange in tehnology spening eplains 30% o the average inrease in the share o skille labor. The net setion esribes the trae an apital aount liberalization in Argentina. Setion 3 esribes the ata set. Setion 4 provies preliminary empirial eviene on the inrease in the relative eman or skill. Setion 5 evelops the theoretial moel an erives the empirial preitions on the eets o trae an apital aount liberalization on within irm skill an tehnology upgraing. Setion 6 esribes the broa patterns in the ata, presents the empirial strategy an tests the preitions o the moel. Setion 7 onlues. 2. Trae an Capital Aount iberalization in Argentina At the beginning o the 990 s, Argentina unertook a broa reorm program that inlue trae an apital aount liberalization. Trae liberalization was implemente irst through unilateral poliies, an was later omplemente by regional trae 5

7 liberalization through the Merosur treaty, an the multilateral negotiations o the General Agreement on Taris an Trae GATT. Trae liberalization starte as a unilateral poliy in 988, as the result o negotiations starte in the ontet o strutural reorms supporte by the Worl Bank. The objetive o these irst steps towars reorm was to reue the sope o non tari barriers 6, whih ha been the main trae poliy instrument in the perio There was also a graual reution in import taris an surtaes, 7 that implie a lower level o protetion or the intermeiate an apital goos inustries. 8 Between Otober 988 an Otober 99, there where major revisions o tari an non tari barriers, many times relate to hanges in maroeonomi poliy aime at ontrolling hyperinlation. As a result o maroeonomi an trae poliy instability uring this perio, trae liberalization ha a an impat only ater 99, when the onvertibility plan was launhe. By Otober 99, the average nominal tari was 2%, ranging rom 0% or apital goos not proue in the ountry to 22% or onsumption goos. Almost all import lienses an quotas where eliminate, with the eeption o the automobile inustry. Ater 99, most eport taes where also eliminate, an in 992 there was an inrease in ta rebates or eports, inreasing the average rom 3.3% to 6.3% o the value o eports. The program also inlue other measures that aete trae like reorms on 6 In the 980 s, the NTB where implemente through a system o lienses an previous authorizations that regulate entry o all goos. The authorizations where organize in our lists:. Prohibite imports 0% o tari positions: sumptuary onsumption goos an inustrial intermeiate goos proue loally. 2. Imports requiring previous authorization 40% o positions : apital goos an inustrial intermeiate goos, in pratie, authorization was enie i there was loal proution. 3. Meial an pharmaeutial prouts 8% o positions. 4. Rest o imports: mainly prouts not proue loally, in this ase authorization was require but given automatially. 7 In 985 the nominal tari average was 37%. Protetion was higher or inal goos, aroun the mean or apital goos an lower or intermeiate goos. There was ierential treatment or goos proue in the ountry. 8 In 987 nominal average taris on apital goos proue in the ountry was 48% an 2% i not proue in the ountry. 6

8 ustoms aministration an port ativity, an the reintroution o the temporary amissions regime. MERCOSUR was establishe by Argentina, Brazil, Paraguay, an Uruguay in 99 with the Treaty o Asunion. The agreement inlue the progressive elimination o the tari an non tari restritions to the irulation o merhanises, the aoption o a ommon eternal tari an a ommon trae poliy with thir ountries. There was a transition phase between 99 an 994 that onsiste o progressive tari reutions aime to ahieve ree trae within the region by the en o 994. The Customs Union was establishe in 995 with the aoption o a Common Eternal Tari CET, with an average level o %. Tari varie between 0 an 20% aross inustries. Inputs an materials ha the lowest taris, ollowe by semi-inishe inustrial goos, an inal goos. There where eeptions to internal ree trae or a limite number o prouts, an speial regimes or sugar an automobiles an some prouts ae tari rates ierent rom the CET. As a result o the agreement, in 996 the import weighte average intra-mercosur tari was 0.86% or Argentina an 0.02% or Brazil, while the etra-zone average tari was 3.7% an 5.44% respetively. Trae liberalization ha a strong impat on trae lows. Between 99 an 999 imports grew at an average rate o 3.6% per year an eports at an average rate o 0.5%. While imports rom MERCOSUR grew at a similar rate than those rom the rest o the worl 4.3% eports grew twie as muh 20.8%. At the same time, ierent measures were unertaken towars apital aount liberalization. In 989 all restritions on entry an eit o oreign apital were eliminate, along with the requirement o previous authorization or Foreign Diret Investment FDI. Equal treatment o oreign an national apital was guarantee by law. In 99, the Convertibility aw establishe a ie parity between the peso an the ollar, an the ommitment o the Central Bank to sell an buy urreny at that parity. The plan also authorize eposits, ebt an ontrats to be enominate in ollars establishing a bimonetary system that eliminate all restritions on the use o oreign urreny. There was a onsierable growth o FDI uring the 990 s. The stok o FDI as a share o GDP inrease rom 7.7% in 992 to 22.2% in 999. FDI lows to the manuaturing 7

9 setor inrease rom US$ 758 million per year in the perio to US$ 2,266 million in , an US$ 3,46 million in Data The ata I analyze omes rom the Survey on Tehnologial Behavior o Inustrial Argentinean Firms [Enuesta sobre la Conuta Tenologia e las Empresas Inustriales Argentinas ETIA] onute by the National Institute o Census an Statistis in Argentina INDEC. The survey overs the perio an was onute in 997 over a representative sample o,639 inustrial irms. The sample was base on 993 ensus ata an overs 54% o total inustrial sales, 50% o employment an 6% o eports in 996. As the survey was onute in 997, it oes not ontain inormation on irms that were ative in 992 an eite aterwars. I ous my analysis on a balane panel o,57 irms present both in 992 an 996. The lak o inormation on entry an eit poses some limitations on the analysis o realloations aross irms an inustries, but as the balane panel sample still represents 49.3% o inustrial sales, the results an be interprete as highly iniative o the overall pattern o realloations. The initial year in the ata is 992, an the major trae an apital aount liberalization measures where taken in Otober 99. Still, the ata or 992 an be a goo iniation o the situation beore liberalization starte to have a onsierable impat on tehnology aoption. Between Otober 988 an Otober 99 there where major revisions on trae poliy, an a similar number o revisions in maroeonomi poliy, as the poliymakers attempte to stop hyperinlation. The etreme instability o the previous perio brought a high egree o unertainty on whether the reorms taken at the en o 99 woul be permanent. Then, even i liberalization starte having an impat in 992, many investment eisions are very likely to have been elaye until the reorm was pereive as permanent. 9 When analyzing the ata, I will use 992 as an iniator or the 9 For instane, FDI lows to the manuaturing setor inrease rom US$ 758 million per year in to US$ 2,266 million per year in Imports o apital goos in the manuaturing setor also aelerate ater 992. In 99 they where only slightly above the average or the perio , 8

10 situation beore liberalization ha a signiiative impat, an as the perio ater liberalization. Euation level o workers An important avantage o this survey over stanar inustrial surveys an ensus is that it ontains iret inormation on the euational level o workers. Table I.a reports the hange in employment by euational ategories between 992 an 996. This hange is orere by skill, with employment o engineers growing % while employment o primary shool workers ell 0%. I aggregate workers into two skill ategories to obtain a measure o the equilibrium relative eman o skille labor S/U: S U = ollege primary + + tertiaty high shool w w w ollege tertiary w higshool primary Skille workers S are ollege grauates plus tertiary euation grauates onverte to ollege equivalents. 0 Unskille workers U are primary shool grauates plus high shool grauates onverte to primary shool equivalents. This aggregation sheme orrespons to the situation where workers within skill ategories are peret substitutes, an is a goo approimation when the elastiity o substitution is higher within than aross ategories. This seems to be a reasonable assumption, as the inrease in employment o ollege grauates an tertiary euation grauates inrease by the same amount 6,5 % an employment o high shool an primary shool workers ell by similar amounts 7% an 0%. The onversion o workers to ollege an primary shool equivalents was one using the 992 inustrial setor wage premia. Then, reporte hanges in the relative eman o skille labor relet hanges in employment an not in representing.9 % o Inustrial GDP. They starte inreasing in 992 when they beame 3.2% o inustrial GDP, an ontinue growing to reah 4.8% in College grauates omplete 5 to 6 years o euation ater high shool, while tertiary grauates omplete 3 years o euation ater high shool. Estimate by minerian equations rom ousehol Survey ata in Galeani an Sanguinetti 2003 an Gasparini et al

11 wages. Overall, the equilibrium relative eman o skille labor inrease by 7% in the balane panel o 57 irms Table I.b. This survey also lassiies workers aoring to proution P, non proution NP an R&D oupational ategories, whih allows me to investigate whether the inrease in the relative eman o skille labor ame primarily rom realloations rom proution to non proution an R&D ativities. Table II.a reports the skill intensity o eah o these ativities: non proution is aroun 3 times more skill intensive than proution, an R&D aroun 5 times more skill intensive than proution. This pattern is onsistent with the inings in Berman, Boun an Grilihes 994 or the U.S. regaring the higher euational level o non proution relative to proution workers. Most empirial stuies with inustry ata use proution workers as a proy or unskille labor, an non proution workers as a proy or skille labor, given that the P/NP lassiiation is the only one available in stanar inustry surveys an ensus. These stuies apture primarily the realloations rom proution to non proution labor, but miss skill upgraing within oupational ategories, as I will show in the net setion. Spening on Tehnology The survey ontains inormation on several imensions o spening on tehnology upgraing. Firms upgrae tehnology by perorming various innovation ativities like internal R&D, paying or tehnology transers an buying apital goos that emboy new tehnologies; an with ierent purposes like hanging proution proesses, prouts, organizational orms or ommerialization. I onstrute a measure o spening on tehnology ST that inlues these ierent imensions: spening on omputers an sotware; payments or tehnology transers an patents; an spening on equipment, materials an labor relate to innovation ativities perorme within the irm. 2 The survey ontains inormation on ST or all years in the perio , while inormation on all the rest o the variables sales, eports, imports, employment by euation, investment is only available or the years 992 an ike R&D, aaptation o new prouts or proution proesses, tehnial assistane or proution, engineering an inustrial esign, organization an ommerialization 0

12 4.Preliminary Eviene: Skill Premia an the Relative Deman or Skill Skill premia starte growing in the 990 s ater having been stable uring the 980 s. Gasparini et al 2002 report wage premia estimates rom minerian regressions using ousehol Survey ata. They in that the ollege-primary shool wage premium rose 9.4% between 992 an 998, ater alling 3.7% between 986 an 992. The high shool wage premium rose muh less 4.8%, an ha been onstant uring the previous perio. Estimates or the inustrial setor in Galeani et al 2003 iniate that the ollege wage premium inrease 7 perentage points per year uring the 990 s, ater being stable in the 980 s. They o not in any signiiant tren or the high shool skill premium. The oiniene o rising skill premia an inreasing relative employment o skille workers in the perio iniates that there must have been an outwar shit in the relative eman o skille labor ater trae liberalization. As the survey oes not inlue inormation on wages, in the remaining o this setion I will analyze the inrease in the equilibrium relative eman or skille labor in the inustrial setor, measure as the relative employment o skille an unskille workers. Deompositions o the Change in the Aggregate Deman or Skille abor The inrease in the aggregate relative eman or skille labor oul be mainly riven by prout eman realloations towars skill intensive setors or ativities, holing skill intensity within ativities onstant, or by inreases in skill intensity within ativities, holing prout eman onstant. Assessing the relative importane o these two hannels is a neessary step in the investigation o the auses o the inrease in the aggregate eman or skill. Prout eman realloations an be riven iretly by trae or hanges in eman or goos, while within ativity inreases in skill intensity point towars hanges in tehnology, leaing to a ierent assessment o the role o trae through this hannel. I perorm three ierent eompositions o the inrease in the aggregate eman o skille labor: irst between an within oupational ategories P, NP an R&D; seon within an between setors; thir within an between irms.

13 To assess the importane o skill upgraing within oupational ategories relative to realloations rom proution to non proution an R&D, I perorme the ollowing eomposition o the hange in skill intensity rom 992 to 996: S = U U U S U + U U S U where = P, NP, R&D; U /U is the share o unskille workers employe in ategory ; S /U is skill intensity in ategory ; a bar over a term enotes a mean over time 992 an 996 an a beore a term enotes a hange over time rom 992 to 996. The irst term on the right reports the hange in aggregate skill intensity attributable to shits in employment shares between oupational ategories holing skill intensity within ategories onstant. The seon term reports the hange in aggregate skill intensity attributable to hanges in skill intensity within eah oupational ategory. Table II.b reports the between an within eompositions o the aggregate inrease in skill intensity in the perio O the 7 perentage points inrease, only 2 points are eplaine by realloations rom proution to non proution an R&D oupational ategories, an 5 points orrespon to skill upgraing within ategories, o whih 7.5 points orrespon to proution an 6.6 points to non proution. That most skill upgraing ours within oupational ategories suggests that stuies that use variation between these ategories as proies or skill upgraing might be missing an important part o it. In aition, it points towars hanges in the proution untion within oupational ategories, avoring the tehnologial hange over other eplanations or skill upgraing that rely on realloations o eman towars skill intensive non proution ativities ue to outsouring o proution ativities or the inreasing importane o servies over goos. The relative importane o tehnologial hange versus prout eman realloations an also be assesse by eomposing the aggregate inrease in skill intensity in hanges within an aross inustries. I the inrease in the relative eman o skille labor omes rom trae, holing tehnology onstant, there woul be no hange in skill intensity 2

14 within setors, but realloations o labor towars skill intensive setors. In this ase the eomposition is: S = U j U j U S U j + j U j U S U j where j = inustry at our-igit SIC lassiiation. Table III reports the between an within inustry eompositions o the aggregate inrease in skill intensity in the perio All the 7 perentage points inrease is eplaine by within inustry skill upgraing, the between omponent being small an negative. Moreover, all o it is eplaine by skill upgraing within irms. There is one important aveat to take into aount or interpretation o this eviene: the sample I analyze oes not ontain entry an eit, thus the realloations aross setors an irms that our through entry an eit are misse in these alulations. Still, as the balane panel represents 49% o inustrial output, this eviene points towars the relative importane o skill upgraing within setors an irms as a soure o the overall inrease in the relative eman o skille labor an the skill premium. That most skill upgraing ourre within IV igit manuaturing inustries is onsistent with the inings in Berman, Boun an Grilihes 994 or the U.S. an the iiulties o ientiying a signiiant eet o trae on the skill premium through variation aross setors in Argentina Galeani an Sanguinetti 2003 an Colombia Attanasio et al This eviene points towars tehnology upgraing within setors an irms as the main ause o the inrease in the relative eman o skille labor, an thus the net setions will ous on investigating the eets o trae on tehnology aoption, an its eets on skill upgraing through that partiular hannel. 5. Theoretial Framework The eompositions o the inrease in the relative eman o skille labor reporte in last setion point towars tehnology upgraing within setors an irms being the main ause o the inrease in the aggregate relative eman o skille labor. This setion evelops a simple moel to illustrate the links between trae, apital aount 3

15 liberalization an tehnology aoption, ousing on within setor irm heterogeneity rather than ierenes in skill intensity aross setors. The setup o the moel inorporates inreasing returns to sale an monopolisti ompetition as in Krugman 979, 980; heterogeneous irms as in Melitz 2003; an enogenous tehnology hoie as in Yeaple The main purpose o the moel is to illustrate the eets o trae an apital aount liberalization on the eporting an tehnology aoption eisions o irms, ousing in partiular on how ierenes in initial proutivity will etermine heterogeneous responses or ontinuing eporters, new eporters an never eporters, proviing thus a basis or the empirial ientiiation o the eets o trae on tehnology aoption an skill upgraing. The moel is partial equilibrium in the sense that it esribes a single inustry that is assume small enough not to aet equilibrium wages. The eposition starts by esribing the setup o the moel an analyzing the proit maimizing eporting an tehnology hoie o irms with ierent proutivity levels. I then erive omparative stati impliations or reutions in eporting osts an the in ost o aopting new tehnologies, an inally relate these preitions to the observable variables in the ata. The analysis is limite in the ollowing ways: irst, I assume that the home ountry is small enough not to aet the prie ine in the oreign ountry; seon, I abstrat rom entry an eit, as the moel only intens to esribe the within irm eets o trae an apital aount liberalization; thir I o not solve or the home inustry equilibrium prie, abstrating rom the eets o import tari reutions on the home prie, an the eets o tehnology upgraing o omesti irms. 5. Setup o the Moel Deman There is a representative onsumer with CES preerenes over a ontinuum o varieties o goo q. U N ρ ρ = q i i, 0 < ρ < 0 4

16 Consumers maimize subjet to the buget onstraint N p i q i i = E 0 Deman or a partiular variety i is: q i = E P p i P where =/-ρ > is the onstant elastiity o substitution an N P = p i 0 i Supply The supply sie is haraterize by monopolisti ompetition. Eah variety is proue by a single irm, an there is ree entry into the inustry. As in Yeaple 2005 irms an hoose to proue with two ierent tehnologies an that eature a onstant marginal ost an a ie ost. Aquisition o tehnology requires a higher ie ost in terms o payments or tehnology aoption an apital goos that emboy new tehnologies >, but guarantees a lower marginal ost <. Marginal osts are onstant an relet wage payments to two types o labor: skille S an unskille U labor, employe in ie proportions. Tehnology is more skill intensive than tehnology. As in Melitz 2003 irms are heterogeneous in their proutivity, in the sense that marginal labor ost varies aross irms utilizing the same tehnology. This iiosynrati omponent o labor proutivity will be inee by. More proutive irms nee to hire ewer workers to attain the same level o output, holing tehnology onstant. The total ost untion or tehnology T is: q TCT = T + T T =, 5

17 where S u S U S u S U a w w a a w w a + = + = U S U S a a a a > 5.2 Firm Proit Maimization: Tehnology Aoption an Eport Deisions Proits in the Domesti Market With CES preerenes, the proit maimizing prie is a onstant mark-up over marginal ost, then a irm with proutivity using tehnology T will harge the ollowing prie in the omesti market: T p T T, = = ρ Quantity sol, revenues an proits woul be: T T T T T T T T T r P E q p r EP q = = = = ρ ρ Proits in the Eport Market As in Melitz 2003 there are two types o trae ritions:. A per-unit ieberg ost, so that τ units nee to be shippe per unit sol abroa. 2. An initial ie ost to start eporting. 6

18 Eporting proits will be: T P E = * * ρ τ * where E * an P * are spening in goo q an the prie ine in the oreign ountry, an serve only the omesti market ly the omesti market The assoiate proit levels woul be: porting an tehnology hoie or eah p. is the amortize per-perio portion o the initial eporting ost. Tehnology Aoption an Eporting eisions Eah irm has our options:. Use tehnology an 2. Use tehnology an eport 3. Use tehnology an serve on 4. Use tehnology an eport. [ ] [ ] P E P E P E P E P E + = = + = * * * * P = E ρ ρ τ ρ ρ ρ τ ρ To solve or the proit maimizing e routivity level, I will eompose the proit untions into our omponents:. Proits rom serving the omesti market using the low tehnology: 2. The inrease in revenues rom eporting using the low tehnology: * * = ρ τ P E r 7

19 3. The inrease in revenues rom omesti sales when swithing to the high tehnology: = ρ P 4. The inrease in revenues rom eporting sales when swithing to the high E r tehnology: * * = r ρ τ P E The proit untions an then be written as: r r r r r = + = + = Proposition : I a irm ins eporting proitable uner tehnology, then that irm w This is true beause high teh irms will sell at a lower prie, an thus have higher re > > + = r r r From omparison o one an eine the uto proutivity level = ill also in eporting proitable uner tehnology :. > > venues rom eporting than low teh irms: 0 > > r 0 an : * * = > > ρ τ P E 8

20 Proposition implies that all irms with will be eporters, regarless o tehnology hoie. > Proposition 2: I a irm oes not in tehnology proitable when eporting, that irm will not in tehnology proitable when only serving the omesti market: > > This is true beause when a irm is eporting, the reution in marginal osts erive rom aoption o tehnology inreases both revenues rom omesti sales an rom eporting while it only inreases omesti revenues i the irm is only serving the omesti market: 0 0 < = < + > r r r From omparison o one an eine the uto proutivity level : an [ ] * * + = > > ρ ρ τ P E P E Proposition 2 implies that all irms with will use tehnology, regarless o eporting status. < Finally the least proutive irms o not in it proitable to aopt any o the two tehnologies an I assume they eit. As long as is small enough relative to an, the minimum proutivity observe will be eine by 0 = > > ρ P E Eporting an Tehnology Aoption Threshols There are two possible onigurations or the tehnology an eporting status eisions:. : < 9

21 -irms with < < only serve the omesti market an use the low tehnology. -irms with < < eport an use the low tehnology. -irms with < eport an use the high tehnology. In this ase there will not be any irms using the high tehnology an serving only the omesti market. 2. < : In this ase all eporters will use the high tehnology. The tehnology an eporting hoies in this ase are analyze in Appeni. I will not ous on this ase here as I observe eporters that use the low tehnology both in 992 an 996 in the ata. The onition or < is: E P < + τ E P * * < The ie ost o tehnology aoption must be big enough relative to the ie eporting ost or there to be eporters using the low tehnology. As in the net subsetion I will only analyze the ase where those proutivity threshols to simpliy notation: <, I will rename - The proutivity threshol or eporting to be proitable or irms using the low tehnology will now be. - The proutivity threshol or aoption o tehnology to be proitable or eporters will now be. 5.3 Trae an Capital Aount iberalization This setion esribes the eets o trae an apital aount liberalization on the eporting an tehnology aoption eisions. In partiular, I analyze the eets o a reution in eporting an tehnology aoption osts on the proutivity threshols an. 20

22 Reution in eporting osts A reution in eporting osts an our beause the variable eporting ost τ or the ie eporting ost has allen. I onentrate on the ase where variable eporting osts all, as there is iret eviene that trae liberalization reue variable eport osts, not ie eporting osts. The preitions are very similar in the ase where the ie osts are alling. A reution in τ inreases eporting revenues, thus more irms in it proitable to pay the ie osts o entering the eport market, an more eporters in it proitable to pay the ie ost o aoption o tehnology. Proposition 3: A reution in variable eport osts τ inues more irms to enter the eport market. This an be seen in a reution o the uto : * = P ρ > 0 * τ E Proposition 4: A reution in variable eport osts τ inues more eporters to aopt tehnology. This results rom a reution in the uto : τ = * * * * [ ] τ E P ρ E Pρ τ E P ρ 0 > + Reution in the ost o aopting new tehnologies The ost o aopting tehnology, i this tehnology is esigne in evelope ountries an thus must be importe, will be aete by import taris on apital goos that emboy new tehnologies an taes on payments or international tehnology transers τ. In aition, i investment in new tehnologies must be mae one perio beore olleting revenues, part o the ost o aoption will be given by the interest rate R. Thus, i F is the ost o tehnology in terms o apital goos an payments or 2

23 tehnology transers, the ost o aoption woul be = + R τ F. Then a reution o taris on importe apital goos an taes on tehnology transers τ will reue the ost o aoption. In aition, apital aount liberalization in a apital sare ountry will bring a reution in the interest rate, urther reuing the ost o aoption. Proposition 5: A reution in will inue more eporters to aopt tehnology. This results rom a reution in the uto : = [ ] * * {[ τ E P ρ + E Pρ ] } > 0 The eets o liberalization on within irm eporting an tehnology aoption eisions The preise preitions o the moel in terms o the eets o trae an apital aount liberalization on the tehnology aoption eisions o never eporters, new eporters an ontinuing eporters epen on the orering o threshols beore an ater liberalization. Net I will analyze all the possible orering o threshols to ientiy whih one is onsistent with the broa patterns in the ata, an then erive the preitions o the moel or that ase. There are three possible orering o threshols beore t=0 an ater t= trae an apital aount liberalization: 0 0. < < < < In this ase irms with: - < < : remain serving only the omesti market an using tehnology. - < < : start eporting an remain using tehnology. 0 - < < : start eporting an swith to tehnology < < : ontinue eporting an swith to tehnology. 22

24 0 - < : ontinue eporting an using tehnology < < < < In this ase all new eporters remain using tehnology. In the net setion o the paper I will show that this ase is not onsistent with what I observe in the ata, as some new eporters aopt tehnology < < < < In this ase there woul be no new eporters using tehnology ater liberalization, but I o observe new eporters using the low tehnology, thus this ase is not onsistent with the ata either. The onition to obtain ase is that the reution in the ost o aopting tehnology an in the variable eporting ost are big: < + < < 0 0 E P E P 0 < < + 0 τ τ < E * P * E * P * 5.4 Moel Preitions on Tehnology Spening an Skill Upgraing:. The level o spening on tehnology ater liberalization is or irms in the range < : never eporters an the least proutive new eporters; or irms in the range < : the most proutive new eporters an all ontinuing eporters. 2. The hange in spening on tehnology ater liberalization is: 0 or irms in the range < : never eporters an the least proutive new eporters; - or irms in the range < < 0 : the most proutive new eporters an the least proutive ontinuing eporters; 0 or irms in the range proutive ontinuing eporters. 0 < : the most 3. The level o skill intensity beore liberalization is: a a S U or irms in the range 0 < : non eporters, new eporters an the least proutive ontinuing 23

25 eporters; eporters. a S or irms in the range 0 < : the most proutive ontinuing a U 4. The hange in skill intensity is: 0 or irms in the range < : non eporters an the least proutive new eporters; a a 0 S S or irms in the range : U the most proutive new eporters an the least proutive ontinuing eporters; 0 0 or irms in the range < : the most proutive ontinuing eporters. a a U < < 5. Skill upgraing is ause by tehnology upgraing, then irms that upgrae tehnology also inrease their skill intensity. 6. Empiris In this setion I try to ientiy the eets o trae an apital aount liberalization on tehnology aoption an skill upgraing within irms. As liberalization aets all irms an setors in the eonomy, the ientiiation strategy will be base on the heterogeneous responses o irms o ierent initial proutivity levels, as preite by the moel presente in the previous setion. First I esribe broa patterns in the ata an relate them to the laim in the theoretial setion that there was only one orering o utos onsistent with these patterns. Seon, I report eviene on inaning onstraints that implies that the optimal tehnology hoies preite in the previous setion are unattainable or small an omestially owne irms, suggesting a role or FDI both as a ontrol an as an aitional measure o the eets o apital aount liberalization on skill an tehnology upgraing. Finally, I isuss the empirial strategy an onut empirial tests o the 5 preitions erive orm the moel. 6. Broa Patterns in the Data: Eporter Premia Aoring to the moel in the previous setion, there shoul be systemati ierenes in proutivity, size, spening in tehnology an skill intensity, between ontinuing eporters, new eporters an non eporters. 24

26 In the moel heterogeneity is given by labor proutivity holing tehnology onstant, whih is not observable in the ata. As a proy or proutivity I use initial labor proutivity eine as sales ivie by employment in primary shool equivalents in 992 Ptiv. 3 This proy also inorporates initial ierenes in hoie variables like apital stok per worker an tehnology, whih I o not observe in the ata, but these are epete to be positively orrelate with iiosynrati proutivity ierenes, so that the orering o irms is preserve by the proy. As measures or irm size I will use employment, employment in primary shool equivalents an sales. Skill intensity is measure as the share o skille labor in employment in primary shool equivalents: s t S = S t t ws / wu 992 ws / wu + Ut 992 where t=992, 996. As skille labor is weighte by the skill premium in 992, hanges in this share will only relet hanges in quantities o skille an unskille labor, an not hanges in the skill premium. Table IV reports the ierenes between irms that eporte both in 992 an 996 ontinuing eporters, irms that eporte in 996 but not in 992 new eporters, an irms that only serve the omesti market non eporters. The ontinuing eporter an new eporter premia are estimate rom a regression o the orm ln Y = α + α NE + α EE + α EN + I + ε NE EE EN j where i inees irms, j inees inustries our igit SIC lassiiation; NE are new eporters, EE are ontinuing eporters, EN are irms that eporte in 992 but in t in 996, 4 an the reerene ategory relative to whih ierenes are estimate is non 3 Value ae woul be a better measure than sales, but it is not available in the ata. As ierenes in proutivity will always be ompute relative to the our igit SIC inustry average, i irms within eah IV igit inustry have a similar value ae over sales ratio, the orer o proutivity woul be similar using sales or value ae in the proutivity measure. Employment in primary shool equivalents is ompute as: where t=992, 996. abor proutivity is ompute as sales t = St ws / wu + U 992 t ivie by employment in primary shool equivalents rather than employment per worker as it intens to be a proy the or iiosynrati omponent o labor proutivity an thus shoul not inlue ierenes in proutivity ue to ierenes in skill. 4 Only 28 out o 59 irms are in this ategory, thus it is har to interpret the oeiients on this group, speially beause some o the zeros or 996 oul be impute. I only inlue them as a ontrol group. 25

27 eporters; I j are inustry ummies, an Y is the irm harateristi or whih the premia are estimate. Firm harateristis inlue labor proutivity, size, the share o skille labor; an spening on tehnology per worker. Eporter premia in size an proutivity are positive an signiiant at % both in 992 an 996, an bigger or ontinuing eporters than or new eporters. This pattern is onsistent with the moel, as ontinuing eporters are more proutive thus bigger than new eporters, whih in turn are more proutive than non eporters, both initially an ater liberalization. In the theoretial setion I mentione that there was only one orer o utos prior to liberalization that was onsistent with the ata <. In that orering, only ontinuing eporters were using the high tehnology beore liberalization. This is the only ase onsistent with the ollowing patterns in eporter premia: irst, spening in tehnology per worker is 38% higher or ontinuing eporters in 992 an their share o skille labor is 6.5 perentage points higher than that o non eporters 4% higher than the overall average share o skille labor; seon, irms that woul start eporting ater 992 o not invest more in tehnology that year, an their skill intensity is only 2. perentage points higher than that o non eporters, the ierene being signiiative only at 0% level. Aitionally, I mentione that there was only one orering o utos beore an ater 0 0 liberalization that was onsistent with the ata: < < < <. In this ase, only the most proutive new eporters woul upate tehnology, whih is onsistent with average skill intensity an spening in tehnology or new eporters being higher than or non eporters, but lower than or ontinuing eporters in 996. Also, the least proutive ontinuing eporters woul upate tehnology, whih is onsistent with the share o skille labor an spening in tehnology inreasing in 996 or always eporters. 5 5 There where two other possible orerings o utos. In the seon one, all new eporters woul remain using tehnology ater liberalization, whih is not onsistent with the share o skille labor an spening in tehnology per worker being higher or new eporters than omesti irms in 996; In the thir one, all new eporters woul use tehnology ater liberalization, whih is not onsistent with spening in 26

28 6.2 Broa Patterns in the Data: Eviene on Finanial Unerevelopment The survey also ontains inormation on the soures o inaning tehnology spening that an be use to qualiy an eten the preitions o the moel regaring optimal tehnology hoie. Aess to reit will be partiularly important in the presene o ie osts o tehnology aoption, as the inane nees o irms woul be high relative to their ash low in this ase. I irms are reit onstraine in the sense that there is a ollateral requirement to reeive a loan, bigger irms woul beneit more rom the reution in the ost o aopting new tehnologies, as bigger irms are more likely to be above the ollateral threshol require to inane the ie osts o tehnology aoption, or an inane it with urrent proits. Foreign owne irms woul also be at an avantage as they an obtain uns rom their parent irms in evelope apital markets. Table V.a reports the soures o inaning tehnology spening. On average, irms inane 60% o their spening with own uns, whih suggests that inanial markets are unerevelope. Table V.b reports the oeiients o a regression o the orm Y = α + α NE + α EE + α EN + α FO + β log + γ log Ptiv + I + ε NE EE EN FO j where i inees irms, j inees our igit SIC inustries; FO is an iniator variable or oreign ownership, is a measure o irm size given by employment in primary shool equivalents; Ptiv is labor proutivity; I j are inustry ummies an Y is the share o tehnology spening inane by eah partiular soure. Size has a positive an signiiant eet on the share inane by private banks, as preite by stanar reit onstraint moels. Aitionally, oreign owne irms inane 3% more o their spening with uns rom the parent irm, an 2% less with own uns, whih is onsistent with them being less aete by the lak o evelopment o loal inanial markets than nationally owne irms. This eviene on inanial unerevelopment suggests there might be a ierential eet o trae an apital aount liberalization on tehnology an skill upgraing or tehnology per worker the average share o skille labor being lower or new eporters than or ontinuing eporters. 27