Trade liberalization and labour markets:

EMPLOYMENT PAPER 2002/41 Trade lberalzaton and labour markets: Perspectve from OECD economes Mchael Landesmann Robert Stehrer Sandra Letner Employment Sector INTERNATIONAL LABOUR OFFICE GENEVA

EMPLOYMENT PAPER 2002/41 Trade lberalzaton and labour markets: Perspectve from OECD economes Mchael Landesmann Robert Stehrer Sandra Letner The Venna Insttute for Internatonal Economc Studes (WIIW) Employment Sector Internatonal Labour Offce Geneva

ISBN 92-2-113210-2

Foreword Ths s one of the case studes prepared wthn the framework of the research project on global economc ntegraton and employment polcy carred out by the Employment Polcy Department. The case studes are desgned to provde detaled emprcal assessment of the effects of growth of manufactured trade, nduced by trade lberalzaton, on manufacturng employment and wages n a carefully selected set of countres. Whle there s wdespread concern about these effects, they reman nadequately understood. There exsts a large lterature on the experence of ndustralzed countres, but controverses abound. On developng countres, the lterature s extremely lmted. Gven ths backdrop, the case studes are expected to make a substantal contrbuton to our understandng of the changes beng engendered by globalzaton, a subject of much current nterest to nternatonal organzatons, natonal polcy makers, the academc communty and cvl socety organzatons. Ths study analyses the effects of growng North-South trade n manufactures on manufacturng employment and wages n Northern economes. It s based on a thorough and careful scrutny of the emprcal evdence from seven advanced ndustral economes France, Germany, Japan, the Netherlands, Sweden, the Unted Kngdom and the Unted States for the perod 1980-96. The man fndngs can be brefly stated as follows. Frst, growth of mports from South has been mportant not just n labour-ntensve ndustres but also n skll-ntensve ndustres. These developments reflect a loss of comparatve advantage n labour-ntensve ndustres and a growth of outsourcng n skll-ntensve ndustres. Second, growth of mports from South had an adverse effect on employment n the 1980s but not n the 1990s. Thrd, growth of mports from South was accompaned by growth of exports to South. And the growth of exports to South had a postve mpact on employment n both the 1980s and the 1990s. Fourth, these fndngs mply that the net employment effect of the growth of North-South trade was nsgnfcant n the 1980s and postve n the 1990s. Fnally, North-South trade had no effect on the growth of real wages, whch was drven solely by the growth of labour productvty. Ths mples that wage nequalty n Northern economes grew manly because technologcal change was faster n skllntensve ndustres. Rashd Amjad Drector a.. Employment Strategy Department

Acknowledgement The authors would lke to thank Ajt Ghose for useful comments on earler drafts of ths paper. Further we receved helpful comments and suggestons from partcpants at the Semnar on Trade Lberalzaton and Employment organzed by the ILO n Geneva n September 2001 where the paper was presented.

Contents Page Foreword Acknowledgement 1. Introducton.. 1 2. Theoretcal approaches 2 2.1 The Heckscher-Ohln framework. 2 2.2 The Rcardan approach.. 3 2.3 Labour economcs approach 4 3. The data and descrptve analyss 4 3.1 The data sets used n ths study 4 3.2 Descrptve features of OECD trade ntegraton wth the South 6 3.2.1 Aggregate mport penetraton and export ratos, 1980-96 6 3.2.2 Import penetraton and export ratos to groups of Southern economes 9 3.2.3 Import penetraton and export orentaton by ndustry 11 3.2.4 Import penetraton, export actvty, employment and wage growth 15 4. Econometrc model and analyss.. 18 4.1 Theoretcal model. 18 4.2 Econometrc analyss: Employment and wage effects of nternatonal trade. 20 4.2.1 Results for the total sample.. 20 4.2.2 Results for hgh- and low-skll-ntensve ndustres 23 5. Conclusons 29 Appendx Tables 33 References 71

1. Introducton Ths study attempts to contrbute to the already extensve lterature on the lnk between ncreasng nternatonal trade ntegraton and labour market outcomes n OECD economes. It s a well-known fact that post-second World War trade growth has outstrpped producton growth. However, at the begnnng of the 1990s, concern about the mpact of ncreasng trade lberalzaton (as a result of the North Amercan Free Trade Agreement (NAFTA), the Uruguay Round, and Common European trade polcy) on trade and producton ncreased, n partcular as the share of developng, newly ndustralzng countres n world trade ncreased sgnfcantly. In ths paper we see trade lberalzaton as an ongong process, although we test for evdence of potentally ncreased effects n the 1990s by dstngushng ths perod from an earler perod. Prevous research nto ths area has been surveyed n a number of contrbutons (see, for example, Wood, 1995; Slaughter, 1998; and Lawrence, 1996 and for studes on European countres Dewatrpont et al., 1999). The justfcaton for another quanttatve study n ths feld s the followng: 1. So far there has been much less research on the European economes then on the Unted States. Furthermore, the avalable studes on Europe are already somewhat outdated (for example, n Dewatrpont, Sapr and Sekkat, 1999 the tme seres underlyng the work mostly end at 1992). We shall see that, unlke the Unted States, trade ntegraton wth the "South" pcked up n Europe, partcularly n the 1990s; thus an up-to-date database coverng much of the 1990s s needed to explore the ssue for the European economes. 2. In the 1990s ntegraton n Europe was extended to nclude the formerly planned economes of Central and Eastern Europe. Prevously members of ther own, relatvely autarkc trade bloc, snce 1990 these countres underwent a dramatc process of trade lberalzaton and trade reorentaton, partcularly towards the European Unon (EU). Evdence suggests that the Central and Eastern European countres (CEECs) are dfferent n terms of ther technologcal productvty levels from ther EU tradng partners, and mght also be dfferent n terms of factor endowments (.e. skll composton of ther labour force, ease of access to captal). Hence both Rcardan and Heckscher-Ohln reasonng would apply regardng labour market mplcatons of ncreasng trade ntegraton between the CEECs and the EU. 3. In Europe another group of countres that mght requre attenton n ths context are the Southern European countres (Greece, Portugal, Span and Turkey) as they represent the "South" from the "Northern" European perspectve. The 1980s and 1990s also saw ncreasng trade ntegraton between these countres and the rest of the EU. 4. Lastly, we analyse on the decomposton of the "non-european South". We dstngush a group of countres whch have undergone dramatc "upgradng" of ther ndustral and export structures over the perod under study. The resultng changes n ther postons n the nternatonal dvson of labour have, n turn, mplcatons for the labour markets n the Northern economes. Much of the analyss of nternatonal ntegraton and labour markets has worked wthn a bascally statc analytcal framework. We have shown n a number of contrbutons (see Landesmann and Stehrer, 2000, 2001; Stehrer, 2003) that adoptng a dynamc framework of trade and catchng-up changes the perspectve on labour market effects as well.

2. Theoretcal approaches Whle ths study s an appled economc exercse, a short revew of exstng theoretcal approaches on the ssue of trade lberalzaton and labour markets mght be approprate. Broadly, we can dstngush three approaches: 1. The Heckscher-Ohln (HO) framework, from whch a number of mplcatons for the mpact of nternatonal trade lberalzaton on labour markets can be deduced. It has been the most wdely used theoretcal framework n the lterature so far. 2. A Rcardan approach, whch emphaszes dfferences n technologcal (productvty) levels amongst tradng partners. Ths approach has also been dynamcally extended to nclude changes n relatve productvty postons (for example through a process of catchng up whch can be dfferentated across dfferent types of ndustres). 3. A "Labour economcs" approach, often adopted by labour market economsts who are deterred by the other two approaches' assumptons of perfectly compettve markets and who emphasze the role whch labour market nsttutons (unonzaton and barganng nsttutons, mnmum wage legslaton, etc.) and labour market segmentaton phenomena (on the demand and supply sde) play n determnng labour market outcomes. In the followng subsectons, we brefly address some ssues n each of these three approaches: 2.1 The Heckscher-Ohln framework In the context of trade lberalzaton, ths approach consders world market product prces as determnng factor ncomes. Gven the tradtonal assumptons concernng producton functons and moblty of factors across ndustres, full trade lberalzaton mples, n a stuaton wthout complete specalzaton, that all producers across the globe wll produce up to a pont where ther margnal products are equalzed; ths leads to factor prce equalzaton. A "Northern" (N) economy whch s relatvely well endowed wth ether captal or sklled labour and whch opens up trade wth a "Southern" (S) economy whch, n turn, s relatvely well endowed wth unsklled labour, wll fnd that factor rewards for captal and sklled labour wll rse and those for unsklled labour wll fall, and vce versa for the Southern economy. Ths occurs through the mechansm of specalzaton n skll- and captal-ntensve products/ndustres n the North and n unsklled labour-ntensve products/ndustres n the South. Emprcal studes whch attempted to apply ths approach to North-South trade ntegraton, emphaszed that the mpact on factor prces occurs va changes n relatve product prces (Lawrence and Slaughter, 1993; Sachs and Shatz, 1994). Hence evdence has to be found whch shows that, wth greater openness, the relatve prces of captal- and skll-ntensve goods rse n Northern economes and fall n Southern economes. The emprcal evdence for ths s rather weak; a major complcaton s that over tme, technology cannot be assumed to be constant, and hence uneven productvty growth rates wll co-determne relatve product prce movements. On the employment sde, the H-O framework would predct shfts n employment (and output) structures n the N economes towards captal- and skll-ntensve producton and the opposte for the S economes. It s clear that a study based on the H-O framework requres data dsaggregated by ndustry. Full employment of the stocks of sklled and unsklled labour and of captal s assured through a shft wthn ndustry n the choce of technques.

2.2 The Rcardan approach The core of ths approach consders productvty level dfferences across economes and, n partcular productvty dfferentals across ndustres. Formulated usually n a one-factor constant returns modellng framework, t predcts complete specalzaton across ndustres n those areas where a partcular economy has the largest productvty dfferental. Tradtonally, the model assumes a unque wage rate across all sectors and therefore relatve productvty dfferentals determne relatve unt costs and thus comparatve advantage. Just as n the Heckscher-Ohln framework, there s an assumpton of full employment; hence labour market clearng determnes the wage rate (Dornbusch, Fscher and Samuelson, 1977; Krugman, 1986). The analyss of the real wage rate (.e. nomnal wage rate dvded by the prce ndex of the basket-of-wage goods consumed) becomes dffcult as productvty levels of both domestc producers and mporters affect the real wage rate. Hence, workers real ncomes are a functon of productvty levels n all economes. Nonetheless, snce the productvty levels of the ndustres n whch ther own economes specalze also determne relatve wage rates across economes, no factor prce equalzaton s predcted. In dynamc formulatons of the Rcardan model, productvty catchng up (or fallng behnd) of the S economy relatve to the N economy can be consdered. Catchng up mples that the cut-off pont of specalzaton shfts n the drecton of hgher-productvty ndustres (from the S economy's pont of vew) whch are beng added to the S economy's specalzaton and are beng lost to the Northern economy (where these ndustres are lowproductvty ndustres). There are two mportant changes whch could be made to the Rcardan model of nternatonal trade and catchng up: one s that one mght not want to assume the same wage rate n a country across sectors. In the labour market approach dscussed below, numerous reasons for wage dfferentals across sectors are consdered even for homogeneous labour (dependng on varous entry and ext barrers nto/out of partcular jobs and ndustres). In such a stuaton, comparatve advantage s not smply determned by relatve productvty dfferentals across sectors but also by wage structures. The other change s that productvty catchng up mght not proceed at the same rates across sectors (as assumed, for example, n Krugman, 1979). For example, the fastest catchng-up rates mght take place n ndustres n whch there are the largest gaps n technology levels. Ths amounts to an applcaton of the Gerschenkron thess ("advantage of backwardness") at the ndustry level. Both these two changes affect the dynamcs of comparatve advantage and the mpact on labour markets (employment and wages) n both the advanced and the catchng-up economes. What about the emprcal predctons of a dynamc Rcardan model wth regard to trade lberalzaton and labour markets? Two ssues can be dstngushed: the move from autarky to trade (.e. the effects of trade lberalzaton), and the effects of catchng up (.e. closng the productvty dfferental between the N and S economes). The move from autarky to trade mples that the ndustry mx shfts n the N economy towards the hgh productvty ndustres (where productvty dfferentals are the greatest) and the opposte occurs n the S economes. If we thnk of hgher productvty ndustres as requrng more sklled labour (thus extendng the model to a two factor framework, but restrctng ourselves to lmtatonal producton functons) then trade lberalzaton would mean n the N economy an ncrease n the relatve demand for sklled labour as aganst unsklled labour and vce versa n the S economy. Ths conforms wth the H-O model predcton. In the dynamc context of catchng up, however, both types of economes' output mx would shft towards hgher productvty growth ndustres: n the case of the S economy because t acqures some hgher productvty growth ndustres, as the cut-off pont of nternatonal specalzaton moves n ts favour, and n the N economy because t loses these ndustres, whch, from ts pont of vew are relatvely low productvty ndustres. Ths dynamc verson of the model mples, contrary to the H-O predctons, that the demand for labour would n both economes shft towards sklled labour.

2.3 Labour economcs approach The approach adopted by labour market economsts vews the effects of nternatonal ntegraton as shftng labour demand schedules for dfferent types of workers. It takes, as a startng pont, a model smlar to the Heckscher-Ohln framework, n that trade lberalzaton would, through specalzaton, shft labour demand towards the factor wth whch an economy s relatvely better endowed. However, there s a dfference n the approach to how product and labour markets functon. In partcular, they do not envsage wage rates n an economy to be entrely determned by world market prces as n the H-O model (through an adjustment of ndustry mx and a choce of technques respondng to specalzaton advantages and dsadvantages) rather, they allow for mperfectons both n product and labour markets. In an economy that has lberalzed trade, there reman mperfectons n product markets due to "home market bas", nformaton and other advantages of domestc producers and the nature of market structures (olgopolstc or monopolstc competton), among others. The exstence of mark-ups n an mperfectly compettve product market opens up scope for barganng n labour markets. The postons of dfferent types of workers (e.g. skll groups) at ndustry (and frm) level barganng depend n turn on entry and ext barrers to the labour market of the ndustry, barrers to skll acquston, hdden cost advantages n human captal and nsder relatonshps, as well as on the organzatonal and nsttutonal features of the barganng process (such as ndustry- or economy-wde barganng, organzatonal strength of dfferent skll groups, the mpact of mnmum wages and outsde ncome.) These mperfectons undermne (put a wedge between) the tght lnk between world market prces and factor prces postulated n the H-O model. Nonetheless, nternatonal competton on export and home markets does affect product and labour market outcomes. From the emprcal angle, labour market economsts focus on shfts n the demand for dfferent types of workers (n terms of skll, gender and age) brought about by nternatonal competton. The H-O predctons wth regard to nternatonal specalzaton stll take nto account such shfts, but so does technology whch mght shft demand n favour of partcular types of workers (n terms of sklls, gender and age). Wage determnaton n the (non-tradable) sector has an mpact on wage determnaton n the tradable sector and vce versa. Furthermore, the factors mentoned above regardng mperfectons on product and labour markets leave consderable scope for dfferent responses by dfferent economes to the process of nternatonal ntegraton. In partcular, the varyng mpact of nternatonal ntegraton on labour market outcomes n the Unted States, the Unted Kngdom and the countres of contnental Europe has tradtonally been emphaszed. In ths paper, we shall adopt an eclectc approach. Elements of each of the above approaches wll be brought nto the dscusson, both n specfyng the model for our econometrc analyss and n nterpretng the results. 3. The data and descrptve analyss 3.1 The data sets used n ths study The analyss n ths paper has sngled out the followng countres for whch trade, producton and employment statstcs have been compled: France Germany Japan

the Netherlands Sweden the Unted Kngdom the Unted States sets: We call ths group the "reportng countres". We have reled manly on the followng data OECD STAN database, whch comples producton, employment and nvestment statstcs at the (ISIC) ndustry level; and OECD Blateral Trade Database (BTD), whch contans blateral trade data for each of the reportng OECD economes by 4-dgt ISIC ndustres. The above two data sets have been matched resultng n a consstent producton, employment and trade data set for 23 manufacturng ndustres (for a lstng of these ndustres see table A.1 n the appendx). In general, the STAN database contans data from 1970 to 1997, although data n the last year for some countres s ncomplete. The OECD Blateral Trade Database contans data only from 1980-1998. Thus, n the analyss we were constraned to consder two dfferent tme perods wthn the perod 1980 to 1996. Apart from the above two data sets we used also data compled by Papaconstantnou from the OECD on employment by four skll categores (blue collar unsklled, blue collar sklled, whte collar unsklled and whte collar sklled) also avalable at the ISIC ndustry level (OECD, 1998) (see table A.2 n the appendx for a lst of ndustres ranked by the share of the whte collar hghsklled labour Force n total employment). As regards trade relatons of the above reportng OECD countres, we compled trade fgures for the followng sx groups of tradng partners (see also table A.3 n the appendx for a lst of countres ncluded): 1. Tradng partners from the North (OECD countres): Group A: Australa, Austra, Belgum/Luxembourg, Canada, Denmark, Fnland, France, Germany, Italy, the Netherlands, New Zealand, Norway, Sweden, Swtzerland, the Unted Kngdom, the Unted States. Group C: Japan. 2. Tradng partners from the South: Group B: Southern European countres (Greece, Portugal, Span and Turkey). Group D: Four Asan tgers (Hong Kong (Chna), Sngapore, Tawan, Provnce of Chna and the Republc of Korea). Group E: Selected developng countres comprsng Argentna, Brazl, Chna, Inda, Indonesa, Malaysa, Mexco, Phlppnes, Thaland, and rest of the world (RoW). (d) Group F: Eastern European transton countres comprsng Czechoslovaka (before 1993), the Czech Republc (after 1993), Poland and Hungary. At tmes we also look at aggregates of Southern economes comprsng groups D+E or B+D+E or B+D+E+F. The growth rates of the respectve varables used n the descrptve analyss below and the econometrc research were calculated as lnear tme trends over the two sub-perods 1980 to 1988 and 1989 to 1996.

4. Descrptve features of OECD trade ntegraton wth the South 3.2.1 Aggregate mport penetraton and export ratos, 1980-96 Fgure 3:1 Aggregate mport penetraton and export ratos Graph a: Import penetraton from South (B+D+E) Graph b: Import penetraton from North Graph c: Export ratos to South (B+D+E) Graph d: Export ratos to North

Fgure 3.1 shows the shares of mports from the North (group A) and South (groups B+D+E defned above) nto the seven reportng countres lsted earler n relaton to producton levels. It s clear that, wth the excepton of the Unted States and Japan, the shares of mports n producton are much hgher for mports from the North than from the South. For the European countres, manufactured mports from the South amounted to between 5 and 10 per cent of manufacturng producton, whle mports from the North (whch nclude mports from other EU economes) amount to between 15 and 50 per cent of manufacturng producton. Thus the order of magntude s qute dfferent as regards mports nto EU markets from other Northern or Southern economes. Wth regard to exports, the pcture looks rather smlar. Excludng ntra-eu trade would, of course, dramatcally change the pcture for EU economes, brngng the mportance of trade relatons wth the rest of the North down to the level of the Unted States where, by the md-1990s, the relatve mportance of mports/exports from/to the South and the North s qute smlar. The latter s also the case for Japan. The pattern of mport penetraton descrbed above can be further explored by lookng at the trend growth rates of mport penetraton. Fgure 3.2 gves the growth rates of mport penetraton and of export ratos (exports/producton) over the perod 1982-96. Fgure 3.2: Average growth rate of export ratos and mport penetraton, 1982-96 It shows that over the entre perod 1982-96, the growth rates of mport penetraton from the South were almost double of those from the North. The dscrepances were smaller n the case of the smaller EU economes the Netherlands and Sweden and Japan than n the other selected OECD economes. If we subdvde the overall perod nto two sub-perods, 1982-89 and 1989-96 (fgure 3.3), we can see that the sharp ncreases n Europe (and Japan) n mport penetraton rates occurred n the second perod, whle n the Unted States they had already occurred n the earler sub-perod. Thus mport penetraton from the South nto Europe and Japan s a more recent phenomenon than t s nto the Unted States. It s no wonder then that the debate on North-South ssues took off earler n the Unted States than n Europe n spte of Wood s poneerng study!

Fgure 3.3: Average growth rates of export ratos and mport penetraton Graph a: Perod 1982-89 Graph b: Perod 1989-96

5. Import penetraton and export ratos to groups of Southern economes Let us now consder trade relatons wth dfferent groups of countres from the South (groups B, D, E and F); we shall also refer at tmes to trade wth Japan, desgnated as group C. We examne whether there are dfferental patterns of mport penetraton and export ratos from/to these dfferent groups of Southern economes vs-à-vs the Northern economes. The followng pattern emerges (see tables 3.1 and 3.2, respectvely): Table 3:1 Import penetraton by country groups France France France Country groups Year A B C D E F B+D+E+F 1980 0.162 0.010 0.006 0.005 0.024 0.001 0.040 1989 0.227 0.020 0.014 0.011 0.027 0.002 0.059 1996 0.215 0.028 0.011 0.013 0.035 0.003 0.078 Sweden Sweden Sweden 1980 1989 1996 0.268 0.287 0.305 0.006 0.010 0.010 0.011 0.025 0.010 0.008 0.015 0.016 0.024 0.018 0.017 0.005 0.004 0.005 0.043 0.046 0.049 Germany Germany Germany 1980 1989 1996 0.135 0.166 0.161 0.006 0.010 0.016 0.008 0.016 0.016 0.007 0.011 0.013 0.020 0.021 0.033 0.004 0.004 0.014 0.037 0.047 0.076 The Netherlands The Netherlands The Netherlands 1980 1989 1996 0.408 0.505 0.505 0.010 0.015 0.023 0.013 0.022 0.031 0.014 0.024 0.045 0.053 0.044 0.072 0.003 0.004 0.009 0.079 0.087 0.149 The Unted Kngdom The Unted Kngdom The Unted Kngdom 1980 1989 1996 0.150 0.218 0.291 0.006 0.012 0.020 0.010 0.021 0.024 0.016 0.026 0.038 0.022 0.027 0.058 0.001 0.002 0.004 0.046 0.067 0.120 The Unted States The Unted States The Unted States 1980 1989 1996 0.040 0.059 0.073 0.001 0.002 0.002 0.018 0.034 0.031 0.010 0.023 0.024 0.019 0.030 0.060 0.000 0.000 0.000 0.031 0.055 0.086 Japan Japan Japan 1980 1989 1996 0.027 0.033 0.043 0.000 0.000 0.001 0.006 0.011 0.013 0.017 0.015 0.029 0.000 0.000 0.000 0.023 0.027 0.043

Table 3:2 Export ratos by country groups France France France Country groups Year A B C D E F B+D+E+F 1980 0.141 0.012 0.002 0.003 0.065 0.002 0.081 1989 0.191 0.024 0.006 0.008 0.051 0.001 0.084 1996 0.204 0.035 0.006 0.013 0.053 0.005 0.106 Sweden Sweden Sweden 1980 1989 1996 0.264 0.308 0.356 0.010 0.016 0.019 0.004 0.007 0.016 0.007 0.010 0.019 0.058 0.033 0.062 0.005 0.003 0.010 0.080 0.062 0.110 Germany Germany Germany 1980 1989 1996 0.174 0.239 0.212 0.009 0.020 0.022 0.003 0.008 0.009 0.004 0.009 0.014 0.049 0.038 0.051 0.005 0.005 0.017 0.067 0.072 0.104 The Netherlands The Netherlands The Netherlands 1980 1989 1996 0.417 0.515 0.569 0.015 0.029 0.038 0.002 0.006 0.008 0.006 0.011 0.023 0.074 0.051 0.067 0.003 0.003 0.011 0.099 0.095 0.138 The Unted Kngdom The Unted Kngdom The Unted Kngdom 1980 1989 1996 0.122 0.144 0.243 0.008 0.012 0.027 0.003 0.005 0.011 0.019 0.020 0.040 0.057 0.035 0.057 0.003 0.001 0.005 0.086 0.068 0.130 The Unted States The Unted States The Unted States Japan Japan Japan 1980 1989 1996 1980 1989 1996 0.043 0.047 0.061 0.053 0.069 0.061 0.002 0.002 0.002 0.002 0.002 0.002 0.007 0.012 0.014 0.007 0.012 0.017 0.018 0.023 0.033 0.033 0.026 0.044 0.046 0.024 0.037 0.000 0.000 0.000 0.000 0.000 0.000 0.042 0.041 0.064 0.066 0.048 0.072 Import penetraton by the South as a whole nto the European economes and Japan ncreased manly durng the second perod, 1989-1996, and nto the Unted States t ncreased over both sub-perods. On the export sde, the second perod was the man perod of exports/producton growth to the South from all the Northern countres; ths mght be due to the openng up of these Southern economes to mports. There s an nterestng pattern wth respect to the man perods n whch mport penetraton nto Group A ncreased from the dfferent country groups, B, C, D, E, F: As one would expect, the man growth of mports from the Eastern European countres occurred durng the perod after 1989. Ths mport penetraton s only relevant for the European reportng countres n our sample. From the Southern European countres (group B) we observe a contnuous ncrease n mport penetraton (and also n export ratos) over both perods. For Japan (group C) rsng mport penetraton nto the markets of the other Northern economes occurred manly durng the frst perod, and ths was also the case for the group of Asan tgers (group D). In contrast, for the group of Southern developng countres (group E, whch ncludes Chna) penetraton nto Northern countres markets took place manly durng the second perod.

On the export sde, the man dfference from the mport penetraton pcture s that the expanson of Northern exports nto the frst group of Asan tgers (D) and Japan I contnued to reman mportant even over the second perod, due possbly to the above- mentoned process of mport lberalzaton. 6. Import penetraton and export orentaton by ndustry Appendx tables A.4 and A.5 gve an overvew of the eght ndustres wth the hghest mport penetraton rates and export ratos from/to the North and South. Industres wth the hghest mport penetraton rates from the South were: non-ferrous metals, textles, apparel and leather products, other manufactured products (ncludng toys), petroleum products, wood products and furnture, professonal goods, food, beverages and tobacco. Lookng at mport penetraton over tme (see tables A.6 and A.7, respectvely, n the Appendx), we fnd that the ndustres wth the hghest growth rates n mport penetraton from the South over the perod 1980-96 were: other transport equpment, offce and computng equpment, machnery and equpment nec, electrcal machnery, textles, clothng and leather products, metal products, rado, TV and communcatons equpment, and professonal goods. Hence there was a sgnfcant dfference between ndustres wth the hghest levels of mport penetraton and ndustres n whch the hghest rates of change n mport penetraton occurred! The hghest growth rates n mport penetraton from other Northern economes were n: other transport equpment, rado, TV and communcatons equpment, ron and steel, drugs and medcnes, offce and computng equpment, arcraft and professonal goods. Lookng at the second perod, 1989-96, we found that qute a few of these ndustres also appear as ndustres wth the hghest rates of growth of mport penetraton from the South. Thus, whle there are some dfferences n the ndustres n whch mport penetraton proceeds fastest wth Southern and Northern economes, there s also consderable overlap. Industry composton of mport penetraton and export ratos over the perods 1982-89 and 1989-96: Appendx tables A.8 and A.9 gve a detaled pcture of the ndustres wth the hghest (average) rates of mport penetraton and exports/producton ratos over the two perods 1982-89 and 1989-96 n relaton to each of the groups of tradng partners (A-F). Table 3.3 selects a few ndustres n whch most of the mport penetraton occurred wth these groups of tradng partners.

Table 3.3: Industres wth hghest levels of mport penetraton from groups A-F for 1982-89 and 1989-96 Perod 1 (1982-1989) Perod 2 (1989-1996) Professonal goods Offce machnes and computng Arcraft Motor vehcles Chemcals Group A Professonal goods Offce machnes and computng Arcraft Chemcals Motor vehcles Professonal goods Other transport equ. Offce machnes and computng Rado, TV, communcatons equpment Motor vehcles Group B Offce machnes and computng Other transport equpment Professonal goods Motor vehcles Rado, TV, communcatons equpment Textles, apparel and leather Non-ferrous metals Motor vehcles Offce machnes and computng Machnery and equpment Group C Textles, apparel and leather Motor Vehcles Machnery and equpment Mneral products Non-ferrous metals Textles, apparel and leather Other manufactures Offce mach. And computng Professonal goods Rado, TV and communcatons Group D Offce machnes and computng Other manufactures Textles, apparel and leather Professonal goods Rado, TV and communcatons Textles, apparel and leather Non-ferrous metals Petroleum products Other Manufact. Wood pds. Group E Textles, Apparel and leather Other Manufactures Non-ferrous metals Petroleum products Offce machnes and computng Irrelevant f or ths group Group F Textles, apparel and leather Non-ferrous metals Machnery and equpment Wood products

A few comments on ths ndustry pattern: North-North ntegraton (.e. wth groups A and C respectvely) shows a stable and smlar pattern, wth Japan (group C) showng a relatvely stronger presence n offce machnery and computng and n rado, TV and communcatons equpment than does the rest of the north (group A). Amongst the Southern groups of tradng partners there are consderable dfferences: whle amongst groups B, E and F, textles, apparel and clothng tops the lst n both perods (we checked group F, the CEECs, only for the second perod), n group D (the Asan tgers ), offce machnery and computng tops the lst n the second perod. Ths ndustry group also starts to feature n group E s lst of top ndustres n the second perod. In group B, on the other hand, motor vehcles head the lst, partcularly n the second perod. In groups B, E and F a number of natural-resource-based ndustres also feature (non-ferrous metals, petroleum products, wood products and mneral products). Revealed comparatve advantage n dfferent skll groups of ndustres: Lookng at revealed comparatve advantage (RCA) ndcators, 1 we use an ndustry rankng constructed on the bass of nformaton avalable on the skll composton of the labour force n the varous ndustres. We use OECD statstcs on the share of hghly-sklled employees n total employment (see secton 3.1 and appendx A.2) and look at RCAs and patterns of mport penetraton n those ndustres whch employ the most/least sklled workers. Agan our focus s on patterns between tradng partners and changes over tme. 1 The RCA measure was calculated as RCA X M = + ( X M ) where = 1,...,n stands for ndustres.

Fgure 3.4: Revealed comparatve advantages wth country groups Graph a: Industres wth hghest skll ntenstes Graph b: Industres wth lowest skll ntenstes Graph c: Industres wth hghest skll ntenstes Graph d: Industres wth lowest skll ntenstes

Our frst set of fgures 3.4a-d shows the RCAs n the ndustres (always an aggregate of 8 out of 23) wth hghest and lowest skll content where the RCAs are always calculated from the pont of vew of the reportng countres. The graphs show the three-year averages for the perods 1984-86 and 1994-96. We can see that n relaton to the North, Japan and, to a lesser extent, the Unted States, mantaned relatvely hgh postve RCAs n the most skll-ntensve ndustres. In relaton to the South (groups B and the combned group D+E) the reportng countres all had hgh postve RCAs n hgh-skll ndustres n the frst perod and, whle these remaned postve n the second perod, they dmnshed sgnfcantly n relaton to groups D+E. And, n some nstances, also n relaton to group B. To complete the pcture, we also show the graph for the ndustres wth lowest skll content (3.4b) wthout further dscussng t. In graphs 3.4c and 3.4d we concentrate on the South (groups B, D, E). We can see here that the reportng countres, except Japan were buldng up farly hgh negatve RCAs n the ndustres wth the hghest skll content vs-à-vs country group D (wth the excepton of Sweden due to communcatons equpment). Wth group E there were sgnfcantly shrnkng postve RCAs n ths group of ndustres. Ths s also the case, n some nstances, n relaton to group B. To avod overload, we dd not nclude group F (the CEECs) n the graphs, but here the reportng countres showed postve RCAs n the group of hgh-skll ndustres. Graph 3.4d shows RCAs for ndustres wth the lowest skll content. In summary, at the begnnng of the perod there exsted a classcal pattern of specalzaton whch has changed over tme n the sense that an mportant group of catchng-up countres started to do qute well n medum- to hgh-tech (medum- to hgh-skll) ndustres as mport penetraton ratos were rsng, especally n these ndustres. 7. Import penetraton, export actvty, employment and wage growth Before proceedng wth our econometrc analyss on the relatonshp between trade ntegraton and labour markets, we present a few descrptve features of ncreasng trade ntegraton on the one hand, and labour market developments on the other. We do ths by lookng at varous groups of ndustres occupyng top or bottom postons n a number of rankngs and check at the same tme the other characterstcs of these ndustres. Thus, for example, we look at the ndustres whch experenced the hghest growth rates n mport penetraton over the perod 1980-96 from the South and fnd out what happened to employment or wage growth n these ndustres relatve to the total set of ndustres.

Fgure 3.5: Dfferences n growth rates of total manufacturng Graph a: Industres wth hghest skll ntenstes 1980-96 Graph b: Industres wth hghest skll ntenstes 1989-96 Graph c: Industres wth lowest skll ntenstes 1980-96 Graph d: Industres wth hghest mport/producton growth from South (D+E)

Startng agan wth the ndustres whch have the hghest/lowest skll content. Graphs 3.5a-d show the rates of change of mport penetraton from the dfferent groups of tradng partners and the patterns of employment and wage growth n these ndustres, all relatve to the average n total manufacturng. Graphs 3.5a and 3.5b (coverng respectvely the full perod 1980-96 and the more recent perod 1989-96) reveal that the ndustres wth the hghest skll ntenstes experenced hgh growth of mport penetraton partcularly from the South (group of countres D+E and B), a pattern even more marked durng the second perod. Wth few exceptons (Sweden over the whole perod, Germany over the more recent perod), these were also ndustres n whch employment growth was above average, n whch (labour) productvty growth was above average (the Unted Kngdom beng an excepton over the whole perod, and the Netherlands and Sweden over the more recent perod) and wage growth was above average (agan wth the excepton of the Unted Kngdom over the longer perod, and a few other exceptons over the more recent perod). Graph 3.5c presents the nformaton for the perod 1980-96 for the group of low-skll ndustres: these showed below average growth rates of mport penetraton, especally from the South (groups D+E and B); they were (wth he excepton of the Unted Kngdom) ndustres wth below average (labour) productvty growth, below average wage growth (Sweden and the Netherlands beng exceptons) and, nterestngly, lke the hgh-skll ndustres, they showed above average employment growth. The latter result mples that the ndustres n between (.e. nether the hgh- nor low-skll ndustres) experenced below average employment growth. Fnally, we look at one more groupng of ndustres for a prelmnary analyss of the relatonshp between trade ntegraton wth the South (group D+E) and labour market varables. Graph 3.5d presents the ndustres wth the hghest growth rates of mport penetraton wth the South and shows whether these ndustres experenced above/below average wage, employment and productvty growth. We can see that the ndustres wth hghest growth n mport penetraton from the South had above average productvty growth rates. Interpretng ths causally, ths could be seen as evdence for defensve restructurng. However, we cauton aganst ths nterpretaton as we also found that these ndustres were also more skll-ntensve than average manufacturng and ths alone could be the cause of uneven productvty growth. However, we could not fnd clear evdence from ths group concernng relatve wage or employment growth for these ndustres. Generally, these results may suggest that there are two mportant factors whch could determne the ongong dynamc processes. Frst, the Southern countres had a comparatve advantage n low-skll-ntensve goods as seen n the hgh level of mport penetraton at the begnnng of the perod, reflectng the classcal pattern of specalzaton as suggested by the Heckscher-Ohln nterpretaton. However, n terms of growth, the most dynamc sectors, wth the hghest growth rates of mport penetraton, were the hgh-skll-ntensve sectors. These began wth low levels of mport penetraton but experenced above average growth rates. Here, outsourcng actvtes of partcular producton stages may have played a sgnfcant role, although ths study, for data reasons, cannot embark upon an explct analyss of these processes (on ths, see Arndt and Kerzkowsk, 2001); or t could reflect an ndustry level catchng-up process such as the one dscussed n secton 2. The relatve strength of these two effects would then result n dfferent effects on labour markets n advanced economes. Thus, whle ths prelmnary analyss of the relatonshp between ncreasng trade ntegraton and labour market varables gves some nsghts, t s tme to turn to an econometrc estmaton whch can explot the varaton n the full data set (across ndustres and countres), and allows for jont estmatons of the mpact of a varety of relevant varables.

8. Econometrc model and analyss In the followng, we frst derve a smple model that can be used to analyse the mpact of global trade ntegraton upon labour market varables (employment and wage rates) then present the estmates of ths model wthn the constrants of our data set. 9. Theoretcal model We start wth a smple producton functon for a frm n ndustry I as q = α 1 α ( A l ) K wth 0 < α < 1, where we assume that techncal progress A s labour augmentng. Ths producton functon s the same for each frm n ndustry I, but may dffer across ndustres. Further, we assume that prces p are determned exogenously n world markets. The producton factor K refers, nter ala, to nfrastructure and can thus be seen as a fxed factor for each frm. Standard proft maxmzaton (wth respect to effcency unts of labour) yelds labour demand for each frm n ndustry I as determned by l = 1 1 α α 1 1 ( α p ) w A K 1 Gven p, the ndustry wage rate w, K and A, we can determne the employment levels l and output levels q for the ndvdual frm. Total ndustry demand Q, ths determnes the number Q of frms n the ndustry as n =.Total employment demand n ndustry I s then L = nl. q Assumng that the producton functons are lnearly homogeneous, total demand for labour n ndustry I can be wrtten as L = 1 α α 1 1 ( α p ) w A K n = ( α p ) 1 1 1 1 α w 1 α 1 A 1 K Q q In ths formulaton rsng ndustry demand (other parameters remanng constant) ncreases the number of frms whle the output level of each frm remans constant. The demand components are domestc demand D and foregn demand (exports) X. Further one has to subtract mports M whch are assumed to be perfect substtutes for domestc producton. To calculate the effect on home labour demand one can calculate the labour equvalent of home producton as L m = 1 α α 1 1 m ( α p ) w A K n = ( α p ) 1 1 1 1 α w 1 α 1 A 1 K M q

Ths s just the amount of labour whch would be demanded f these goods were to be produced n the home country. Producton at home s Q = D + X + M, and total employment demand can be calculated as L = 1 1 1 q 1 α α 1 1 ( α p ) w A K ( D + X M ) Takng logarthms and the dervatve wth respect to tme yelds L & L 1 p& 1 = 1 α p 1 α w& w A& A K& + K D + Q D& D + X Q X& X M Q M& M Ths equaton shows that employment demand s negatvely affected by rsng wage rates and labour productvty growth and postvely affected by rsng prces and an ncrease n the fxed factor (nfrastructure). Further, the growth rates of domestc demand and exports have a postve mpact on the growth rate of labour demand, and rsng mports have a negatve mpact. These growth rates are weghted by the shares of the respectve components n domestc producton Q. In a more advanced settng, one would have to take nto account the fact that these shares change over tme f domestc demand, exports and mports experence dfferent growth rates. We shall skrt around ths theoretcal problem by usng fxed md-perod ratos of these shares n the econometrc model we estmate below. The underlyng model to estmate the effects of export and mport growth on wage rates (by ndustry) s straghtforward: hgher demand for goods n ndustry I shfts the demand curve for labour outwards whch leads to hgher wages, assumng a normally postvely sloped supply curve. Symmetrcally, an ncrease n mport penetraton shfts the demand curve for labour nwards whch results n lower wages. An ncrease n labour productvty rases the margnal product of labour, whch results n hgher (real) wages. Ths assumes that the pressure on wages by lad-off workers s small.

10. Econometrc analyss: Employment and wage effects of nternatonal trade In ths secton we use the same data as n secton 3 and we do not nclude the CEECs as the relevant data perod after trade lberalzaton s too short. We dvde the perod nto two sub-perods 1980 to 1988 and 1989 to 1996. The growth rates of the respectve varables were calculated as above. Further we dropped ndustry 384d (dscrepancy scrap metals) for reasons of poor data qualty. Labour productvty s measured as output per employee. Unfortunately, data for prce movements and trends n domestc demand are not avalable. The effects of these varables (such as dfferent shfts n domestc demand across countres and ndustres) together wth the not specfed stock of the complementary factor (nfrastructure) K s taken nto account by ndustry and country dummes. As mentoned above, the changes n the shares of exports and mports n domestc producton over the estmaton perod should be taken nto account. However, n the cross-sectonal framework used below (whch works only wth trend growth rates over the relevant perod), ths cannot be modelled dynamcally over tme. In the econometrc analyss, we proxed the shares by takng averages over the respectve tme perods. The respectve growth rates were then multpled by these shares (see last equaton n secton 4.1 above). The mplcaton s that mports and exports have a larger mpact on employment or wages f ether the share of mports or exports n domestc producton s hgh or the growth rates of mports or exports are hgh, or both. The smple theoretcal framework ntroduced above results n the followng econometrc model. EMPN LCPE + d + d c cg cg c = β 1EXP + β2 IMP + β3opr + β 4 c c Thus, the varables ncluded n the analyss were the growth rates of employment n sector I ( EMPN ) as a dependent varable, the growth rates of labour productvty ( OPR ), and fnally the c growth rates of labour costs per employee ( LCPE ); d and d c denote ndustry and country cg cg dummes, respectvely. The varables EXP and mports IMP denote the growth rates of exports and mports by ndustry I from country group g (g=a,b,c,d,e) multpled by the export shares and mport penetraton ratos for each ndustry (tme averages over the two sub-perods) to whch the reportng country was exportng and from whch t was mportng. Smlarly the equaton for wage rates s specfed as: c c LCPE OPR + d + d c cg cg = β 1EXP + β2imp + β3 c c The regressons are estmated for the two perods separately. The ntroducton of dummes for each ndustry and country means that we estmate a two-way panel data set usng the LSDVestmator. These two sets of dummy varables account for the effects of ndustry and country specfc characterstcs (such as prce changes and shfts n domestc demand). 11. Results for the total sample The tables below report the results for both tme perods and both dependent varables. Further the results are dfferentated by the fve country groups: OECD-North (A), Southern Europe (B), Japan (C), Asan tgers (D), and developng countres (E). The tables n the text only report the estmated parameters for exports, mports, productvty and labour costs per employee. The complete set of estmates, ncludng the ndustry and country dummes are presented n the appendx (tables A.10 to A.14). In the frst analyss we ncluded all ndustres n the sample and estmated the equatons above for the fve country groups and for both perods separately. Tables

4.1 and 4.2 report the results for employment growth and the growth of labour costs per employee, respectvely. Below the coeffcents, the p-values are reported: ***, **, and * denote sgnfcance at the 1, 5, and 10 per cent levels respectvely. Tables A.10 to A.14 n the appendx show the results for all varables, ncludng ndustry and country dummes and further statstcs. Table 4.1: Regresson results for employment growth ncludng all ndustres Perod 1 A B C D E Exports 0.093-0.940-1.293 1.977 ** -0.138 0.476 0.162 0.466 0.024 0.671 Imports -0.194* -1.363** -0.072-0.757*** -1.364 *** 0.060 0.041 0.847 0.000 0.002 Productvty -0.259*** -0.254*** -0.284*** -0.267*** -0.253 *** 0.000 0.000 0.000 0.000 0.000 Labour costs -0.294** -0.288** -0.179-0.245** -0.212 * 0.023 0.026 0.206 0.047 0.098 R square 0.784 0.777 0.792 0.794 0.779 R square adjusted 0.727 0.719 0.728 0.739 0.720 Prob > F 0.000 0.000 0.000 0.000 0.000 N 154 154 132 154 154 Perod 2 A B C D E Exports 0.179 *** 1.863 *** 6.008*** 1.299 0.834 * 0.001 0.001 0.000 0.266 0.097 Imports -0.310** 1.021 0.388* 0.108 0.125 0.014 0.316 0.089 0.484 0.649 Productvty -0.044-0.029-0.119-0.022-0.075 0.595 0.725 0.191 0.792 0.394 Labour costs -0.468*** -0.489*** -0.551*** -0.453** -0.404** 0.007 0.005 0.004 0.013 0.026 R square 0.517 0.515 0.592 0.480 0.487 R square adjusted 0.391 0.388 0.467 0.344 0.353 Prob > F 0.000 0.000 0.000 0.000 0.000 N 154 154 132 154 154

Table 4.2: Regresson results for growth of labour costs per employee ncludng all ndustres Perod 1 A B C D E Exports 0.207 ** -0.346 0.860-0.184 0.337 0.023 0.462 0.490 0.774 0.140 Imports -0.181 ** -0.516-0.147-0.071-0.043 0.011 0.000 0.575 0.588 0.890 Productvty 0.138 *** 0.139 *** 0.126 *** 0.136 *** 0.144 *** 0.000 0.917 0.001 0.000 0.000 R square 0.967 0.966 0.970 0.965 0.966 R square adjusted 0.959 0.958 0.961 0.957 0.957 Prob > F 0.000 0.000 0.000 0.000 0.000 N 154 154 132 154 154 Perod 2 A B C D E Exports 0.007 0.135 0.135 0.135 0.135 0.796 0.649 0.612 0.861 0.319 Imports 0.006 0.466 0.466 0.466 0.466 0.925 0.378 0.930 0.621 0.281 Productvty 0.197 *** 0.194 *** 0.194 *** 0.194 *** 0.194 *** 0.000 0.000 0.000 0.000 0.000 R square 0.868 0.869 0.879 0.869 0.869 R square adjusted 0.835 0.836 0.843 0.836 0.836 Prob > F 0.000 0.000 0.000 0.000 0.000 N 154 154 132 154 154 The growth rates of productvty and labour costs per employee had the expected negatve effect on employment growth n the frst perod but not so n the second, when only the estmate for labour costs per employee was sgnfcant. Productvty had a sgnfcant and, as expected, postve effect on labour costs per employee (wages) n both perods. For the trade varables, t turned out that the dvson nto the two perods proved to be qute mportant. Import penetraton had a sgnfcant negatve effect on employment growth n the frst perod (1982-1988), wth the excepton of mports from Japan I, but ths effect vanshed n the second perod for all country groups, wth the excepton of mports from OECD Northern countres (A) where t remaned sgnfcantly negatve and was even hgher than n the frst perod. For the other countres, the coeffcents became postve, but only sgnfcant at the 10 per cent level for mports from Japan. Exports had a sgnfcant and postve effect on employment growth only for exports to country group D ( Asan tgers ) n the frst perod; the estmates for ths varable were postve and sgnfcant n the second perod wth the excepton of ncreased exports/producton growth to country group D. We found no sgnfcant mpacts of the trade varables on the growth rates of labour costs per employee, wth the excepton of a postve and sgnfcant effect of ncreased export ratos to OECD Northern countres. Summarzng, t seems frst, that trade ntegraton (.e. growth n mport penetraton and n exports/producton) had an mpact, but dfferently for the two perods: whereas n the frst perod the negatve effects on employment growth were sgnfcant, n