Sectoral employment effects of trade and productivity in Europe

This article was downloaded by: [York University Libraries]On: 17 November 2014, At: 18:41Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Applied EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/raec20

Sectoral employment effects of trade and productivityin EuropeFilip Abraham a & Ellen Brock aa Faculty of Economics and Applied Economics, Katholieke Universiteit Leuven,Naamsestraat 69, B-3000 Leuven, BelgiumPublished online: 05 Oct 2010.

To cite this article: Filip Abraham & Ellen Brock (2003) Sectoral employment effects of trade and productivity in Europe,Applied Economics, 35:2, 223-238, DOI: 10.1080/00036840210161828

To link to this article: http://dx.doi.org/10.1080/00036840210161828

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose ofthe Content. Any opinions and views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources of information. Taylor and Francis shall not beliable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilitieswhatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out ofthe use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/loi/raec20

http://www.tandfonline.com/action/showCitFormats?doi=10.1080/00036840210161828

http://dx.doi.org/10.1080/00036840210161828

http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/page/terms-and-conditions

Sectoral employment effects of trade and

productivity in Europe

FILIP ABRAHAM and ELLEN BROCK*

Faculty of Economics and Applied Economics, Katholieke Universiteit Leuven,Naamsestraat 69, B-3000 Leuven, Belgium

The impact of trade and technology in the European case is assessed. A framework isdeveloped which incorporates employment effects of (i) export expansion (ii) importcompetition and (iii) labour-saving productivity improvements. In this context,evidence is found for the hypothesis that international trade induces adjustmentsin technology.

I . INTRODUCTION

In recent years, a growing number of articles investigate the

impact of international trade on labour markets. Most of

the papers focus on the US experience and view interna-

tional trade specialization as one possible explanation why

relative wages and employment prospects of unskilled

workers have deteriorated. Although dissenting opinions

exist (Wood, 1994), the consensus view is that international

trade only plays a minor role and that the primary driving

force of US labour market trends is skill-biased technolo-

gical change (Berman et al., 1994; Krugman and Lawrence,

1996).

This paper studies the impact of international trade on

sectoral employment in Europe. There are three valid

reasons for doing so. First, European countries are much

more open than the USA. In 2000, the export/GDP ratio

was 32% for the EU-15 whereas it was 10% for the USA.1

Especially the Netherlands and Belgium, which have

respectively export/GDP ratios of 53% and 68%, are typi-

cal examples of small open economies with a high degree of

openness. Krugman (1995) among others argues that the

evolution of exports and imports cannot explain US labour

market developments because the US economy is just not

open enough for trade to matter a lot. Turning this argu-

ment around, we expect significant employment effects

from European trade.

The second motivation refers to the rigid labour marketsin Europe. For the EU-15, the unemployment rate cur-

rently stands at 7.7%.2 Minimum wages, strong unions

and other institutional features prevent wages from going

down in order to bring down unemployment. Moreover,

reallocation between sectors in the European economies isslow and quite limited due to insufficient labourmobility and

generous unemployment benefits (see Decressin and Fatas,

1995, for a comparison of the US and the European

experience). This justifies a detailed look at trade-related

employment changes at the sectoral level and would imply

that the assumption of full labour mobility in conventionaltrade models serves at best as a long-run approximation. In

the European case, it is also important to consider hiring

and firing costs, which make changes in employment more

difficult.

A third and final reason concerns the dichotomy betweentechnical change and international trade that underlies the

debate on US labour market developments. International

trade and technological progress are usually seen as

mutually exclusive explanations of employment changes.

In Europe however, business leaders and company surveysusually emphasize the link between international trade and

the introduction of new technologies and production meth-

ods (see Abraham and Konings, 1999). Export expansion

gives rise to productivity gains and in the case of import

competition companies may end up restructuring.

Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online # 2003 Taylor & Francis Ltdhttp://www.tandf.co.uk/journalsDOI: 10.1080/00036840210161828

Applied Economics, 2003, 35, 223–238

223

*Corresponding author. e-mail: [email protected] These figures are calculated with data obtained form the OECD Main Economic Indicators (see http://www.oecd.org).2 This figure refers to the EU-15 standardized unemployment rate in January 2002 and can be found on the OECD webpage (seehttp://www.oecd.org).

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

Theoretically speaking, this means that trade variablesinfluence productivity and therefore indirectly affectemployment. Hence, a distinction should be made betweendirect and technology-induced indirect effects of trade onemployment. An innovation of this paper is to develop aframework that allows us to empirically distinguishbetween those direct and indirect effects.

The remainder of the paper is structured as follows. Inthe next section, the relation between trade, technology andlabour markets is analysed based on the existing theoreticaland empirical literature. In Section III a European modelfor employment adjustments is derived that serves as thebasis for the empirical work. The fourth section presentsthe estimation results. The paper ends with a summary ofthe main insights.

II . REVIEW OF THE LITERATURE

The Heckscher–Ohlin–Samuelson (HOS) theory is themost common framework to address the employmenteffects of trade. This theory is based on assumptions ofperfect intersectoral mobility of all production factorsand full employment.3 International trade leads – througha world-wide equalization of product prices – to specializa-tion according to relative abundance of production factors.

In a stylized version of the model that is typically appliedto industrialized countries like Europe, two types of labour(skilled and unskilled) are employed in a skill-intensiveexport sector and an import-competing sector with rela-tively more unskilled labour. From a sectoral perspective,an expansion of trade leads to an increase in demand forgoods of the export sector which creates new jobs in thissector. The positive relation between an expansion ofexport demand and total sectoral employment is definedas the export demand effect.

On the other hand, the import-competing sector experi-ences increased competition from countries with a relativeabundance of cheap unskilled labour. Due to this importcompetition, jobs are lost. The negative impact of importcompetition on total sectoral employment is called theimport competition effect in this paper.

The import competition and export demand effects arisenaturally in a theoretical framework that analyses the roleof unions in the transmission of trade shocks (Brander andSpencer, 1988; Driffil and van der Ploeg, 1995; Gaston andTrefler, 1995; Huizinga, 1993; Mezzetti and Dinopoulos,1991; Naylor, 1998; Vandenbussche and Konings, 1998).While the theoretical set-up of the model varies, thesepapers are cast in terms of interfirm rivalry in one industry.Import competition takes the form of a decline in output ormarket share of the domestic firm with respect to its foreign

competitor(s). In a straightforward way, this leads to thereduction in the domestic sectoral employment level cap-tured by the import competition effect. The export demandeffect follows from the fact that trade integration offersdomestic firms the opportunity to penetrate in the foreignmarket, raising exports and sectoral employment in thehome country.Among the extensive empirical work in this area, focus is

here on studies that concentrate on trade-related totalemployment changes at the sectoral level. Revenga (1992)for the US and Neven and Wyplosz (1999) for German,French, Italian and UK manufacturing use a product pricemethodology and focus on the import competition effect.Revenga finds a statistically significant but small impact ofimport prices on sectoral employment. Neven and Wyploszfind no clear pattern for the HOS effect of import competi-tion on employment but they do observe a drastic restruc-turing in unskilled labour intensive industries. The work byFreeman and Revenga (1999) and by Larre (1995) draws adirect link between trade flows and employment in OECDcountries. This has the advantage of not having to useinternational price data, which are often of lower qualitythan the reported trade volumes (see Slaughter and Swagel,1997, pp 17–18). Like most of the empirical research, thosestudies find only small labour market adjustments to inter-national trade.A growing literature looks at the relation between inter-

national trade and productivity. As one of the first, Wood(1994) asserted that international competition leads firmsin the advanced economies to raise productivity by pursu-ing labour-saving innovations. This implies that there is anindirect negative employment effect which we call in thispaper the productivity effect of international trade onemployment.Looking at this reasoning more carefully, the trade-

related productivity effect on employment requires that:(i) exports and/or import competition affect technology(measured by productivity) and (ii) this increase in produc-tivity affects employment. Note that, in principle, the firstcondition can go in both directions. On the one hand,domestic companies may not be able to cope with foreigncompetition. In this case, internal restructuring in the formof lay-offs does not keep up with the decline in sales so thatdomestic firms are confronted with falling productivity.This situation is more plausible with large hiring and firingcosts which are present in the European economies (seeBertola, 1990; Grubb and Wells, 1993; Bentolila andBertola, 1990; Garibaldi, 1998; Booth, 1997). Such hiringand firing costs unambiguously decrease employment var-iation in response to various economic shocks. On theother hand, trade may induce firms to successfully intro-duce productivity-enhancing technologies. For the USA,

224 F. Abraham and E. Brock

3 Recent work by Davis (1998) drops the hypothesis of full employment by considering minimum wages.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

Bernard and Jensen (1999, 2001) find no evidence for apositive impact of exports on productivity. Causality goesin the other direction: more productive firms become betterexporters. The results of Lawrence (2000), who uses bothprice and quantity measures, indicate that import competi-tion has a positive impact on US total factor productivity.This effect is mainly present in skill-intensive sectors andindustries competing with developing countries. With stu-dies of Europe lacking, the link between international tradeand productivity is further explored. Additionally, thisstudy explicitly computes the employment effects oftrade-related productivity changes.

III . A EUROPEAN MODEL FOREMPLOYMENT ADJUSTMENTS

The set-up of the model

In the following pages, a European model for employmentadjustments is proposed which explicitly relates changes inlabour and total factor productivity to exports and importsin addition to capturing the export demand and importcompetition effects on sectoral employment. This modelserves as a background for the empirical analysis which isthe primary focus of this paper. One representative sectorfor which a model of monopolistic competition is con-structed is considered. Product differentiation is consistentwith the observation that the EU-countries’ trade witheach other and the rest of the world is mainly of theintra-industry type. In the derivations, the subscript i refersto a specific country. Assume that there are m countries. Ineach country, there are ni identical firms in the representa-tive industry. Therefore, firms within the same countrycharge the same price.

The worldwide real consumption Xð Þ of the products ofa representative industry is expressed in a Dixit–Stiglitzframework. In the following expression, Xi refers to thesectoral production of country i. Because of the assump-tion of ni identical firms in the sector of this country i, Xiequals nixi where xi denotes the production of an indivi-dual firm. � (with � > 1) is the elasticity of substitution.

X ¼Xmi¼1

Xð��1Þ=�i

!�=ð��1Þ

¼Xmi¼1

nixið Þ ��1ð Þ=� !�= ��1ð Þ

ð1ÞStandard utility maximization (see Dixit and Stiglitz,

1977) yields the following demand function for the outputof country i:

Xi ¼ nixi ¼�piP

��E

Pð2Þ

with pi as the price charged by all firms in country i,

P ¼Pm

i¼1 p1��i

� �1=1��as the price index of the representative

sector and E ¼Pm

i¼1 piXi as the worldwide expenditureson the products of the representative sector. The inversedemand function is then equal to:

pi ¼E

X

�XiX

��1=�

¼ E

X

�nixiX

��1=�

ð3Þ

Next, the supply side of the model is considered. Totalcosts of an individual firm in country i are the sum of fixedcosts Fið Þ and variable costs Ci: The variable costs aredetermined by the cost of labour wið Þ and capital rið Þ.For the variable cost function, a Constant Returns toScale Cobb–Douglas function is used (see Varian, 1984,p. 29). Therefore, total costs Ctot

i are:

Ctoti ¼ Fi þ Ci ¼ Fi þ KiA

�1i w

�ii r

1��ii xi ð4Þ

where Ki ¼ ��ii 1� �ið Þ �i�1ð Þ: In this expression, Ai refers to

technological progress and Ki refers to a constant. Whenusing expression (4), declining average costs and economiesof scale are introduced. The profits of an individual firmare given by: �i ¼ pixi � Ctot

i . From expression (3) andassuming that firms are sufficiently small so that they arenot able to influence aggregate production when their indi-vidual production rises, the perceived elasticity of demandequals �: When ci denotes marginal costs, the first ordercondition reduces to:

pi 1� 1

�

� �¼ ci ð5Þ

with ci ¼ KiA�1i w

�ii r

1��ii : Combining expressions (3) and (5)

gives the equilibrium sectoral demand/output of a repre-sentative sector in country i:

Xi ¼�

�

�� 1

��

c��i E�X1�� ð6Þ

For deriving the conditional labour demand lið Þ of anindividual firm, Shepard’s lemma is applied to Equation 4:

li ¼ KiA�1i �iw

�i�1i r

1��ii xi ð7Þ

Because firms within a sector are identical, total sectoralemployment equals Li ¼ nili:

Li ¼ KiA�1i �iw

�i�1i r

1��ii Xi ð8Þ

Substituting Equation 6 into Equation 8 and using theexpression for ci, labour demand of a representative sectorreduces to:

ln Lið Þ ¼ Gi þ � ln Eð Þ � �� 1ð Þ ln Xð Þ

� �i �� 1ð Þ þ 1ð Þ ln wið Þ � 1� �ið Þ �� 1ð Þ ln rið Þ

þ �� 1ð Þ ln Aið Þ ð9Þ

with Gi ¼ ð1� �Þ ln Kið Þ þ ln �ið Þ � � ln �ð Þ þ � ln �� 1ð Þ:

Sectoral employment effects of trade and productivity in Europe 225

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

The export demand effect

In Equation 9, the variable E captures the effect of anexpansion in world-wide expenditures on sectoral employ-ment. According to the theoretical model where � is largerthan 1, an increase in this variable is expected to positivelyinfluence sectoral labour demand. For the purpose of thispaper, this variable E is related to the export demand effectdiscussed earlier. For this reason, E is measured by totalsectoral real exports (EXPÞ.

The import competition effect

In the model, enhanced foreign competition is captured byan increase in the sectoral output of a foreign country j.With the aid of Equations 1 and 9, the impact of increasedforeign output on sectoral employment of country i is:

d ln Lið Þd ln Xj

� � ¼ d ln Lið Þd ln Xð Þ

d ln Xð Þd ln Xj

� � ¼ � �� 1ð ÞPjXjE

� �< 0

ð10Þ

As seen in the above equation, this effect is negative.Moreover, the higher the foreign market share (asmeasured by PjXj=E), the stronger the negative impacton domestic sectoral employment. In empirical work,foreign import competition is measured by the importpenetration ratio which is defined as imports divided bythe difference between production and net exports.

The productivity-related and total effects of internationaltrade on employment

In this theoretical model, Ai measures technology. In theempirical work, it is required to distinguish between thecases of labour-saving and labour-augmenting technologi-cal progress. In addition, an indicator that captures the roleof technology is needed. Following a similar methodologyas Card et al. (1999), the Ai-variable in expression (9) ismeasured by two productivity variables in the regressionequation for sectoral employment. The first variable used isvalue added per worker ðVAÞ, which reflects gains in aver-age labour productivity. The other variable, total factorproductivity ðTFPÞ;4 measures gains that raise productivityof all production factors. Let PRODi represent the variableused (VAi or TFPi) to proxy the Ai-variable of Equation 9:

�� 1ð Þ ln Aið Þ ¼ ln PRODið Þ þ "i ð11Þ

Based on Equations 9 and 11, the sectoral employmentequation is specified as:

ln ðEMPLitÞ ¼ �i1 þ �1 ln EXPitð Þ þ 1 ln IMPitð Þ

þ�1 ln WAGEitð Þ þ 1 ln PRODitð Þ þ u1it

ð12Þ

In this expression, i and t now denote respectively industryand time, ai1 refers to a dummy which captures omittedindustry specific effects and uit is the error term whichrepresents a combination of the error term of expression(11) and other error terms due to estimation of Equation12. This regression equation provides an estimate of theimpact of productivity on employment which is one aspectof the productivity effect of international trade on employ-ment. When is positive, increases in productivity arelabour-augmenting. If is negative, increases in pro-ductivity are labour-saving.5

The other aspect of the productivity effect of interna-tional trade on employment concerns the impact of tradeintegration on productivity. For this purpose, a secondequation is introduced where productivity is regressedupon trade and other variables:

ln ðPRODitÞ ¼ �i2 þ �2 ln ðEXPitÞ þ 2 ln ðIMPitÞ

þ �2 ln ðRDitÞ þ�2PATit þ’2 ln ðCAPitÞþ u2it

ð13Þ

The main focus in Equation 13 is on the regression co-efficients for the export and import variable. Thesecoefficients can be positive or negative depending onwhether companies successfully improve productivitywhen faced with international competition or are insteadstruggling with internal restructuring. The CAP variablerefers to the capital stock per employee and is includedbecause labour-saving technologies are usually accompa-nied by investment in new machinery.6 RD are Researchand Development (R&D) expenditures per employee whichact as an input indicator of innovation and PAT are therelative granted patents which are a measure for innovativeoutput.7

Combining the two aspects just mentioned yields theproductivity effects of international trade on employment.For this purpose, it is useful to substitute Equation 13 intoEquation 12:


4 The construction of this variable is discussed in Appendix I.5 According to expression (9), it is also clear that the wage elasticity should be negative. Also note that capital costs are omitted from theregression equation. Capital costs can however be captured by time dummies. Introducing these time dummies did not significantlychange the results. Therefore, time dummies were not included in the regression equations.6 Note that when total factor productivity is used as the productivity variable, capital per employee will not be included in the regressionequation because this last variable is included in the total factor productivity variable.7 By relative granted patents, it means the granted patents in a certain year relative to the total granted patents in that year. Note that thisvariable is not expressed in logarithms because for a certain year the data show a value of 0.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

ln ðEMPLÞit ¼ �i þ � ln ðEXPitÞ þ ln ðIMPitÞ

þ � ln ðWAGEitÞ þ � ln ðRDitÞ þ �PATit

þ ’ ln ðCAPitÞ þ uit ð14Þ

with �i=�i1 þ 1�i2, � ¼ �1 þ 1�2, ¼ 1 þ 1 2;� ¼ �1; � ¼ 1�2, � ¼ 1�2, ’ ¼ 1’2 and uit ¼ u1it þ 1u2it:

In this equation, the effect of an increase in exportdemand on employment that occurs via an increase in pro-ductivity equals 1�2. Analogously, 1 2 refers to the pro-ductivity effect of increased import competition onemployment. Equation 14 also yields the total impact ofexport demand on employment as measured by the � co-efficient. This total effect is the sum of the export demandeffect, �1, and the productivity effect of exports on employ-ment, 1�2. Similarly, captures the total impact of importcompetition on trade and consists of the direct ( 1) and theproductivity induced effects (1 2) of import competitionon sectoral employment.

IV. AN EMPIRICAL INVESTIGATION

Data description and econometric methodology

Regression Equations 12 and 13 are estimated for ten EU-countries: Belgium, Denmark, Finland, France, Germany,Italy, the Netherlands, Spain, Sweden and the UnitedKingdom. For each country, the data set covers nine sectorsof manufacturing classified according to the ‘InternationalStandard Industrial Classification’ (ISIC) revision 2. Exceptfor Spain, the data set starts at 1978 and ends at 1994 formost countries. The sources and the construction of the dataare discussed in Appendix I. Except for import penetrationand relative granted patents, all variables are deflated.

Like other trade-related empirical work (e.g. Coe andHelpman, 1995), first it is necessary to confirm whetherthe data are nonstationary or not. When the data are non-stationary, it is necessary to use first differences for theestimations. However, using first differences when variablesturn out to be cointegrated would be counterproductive,since the long-term relationship between these variableswould become obscured (Greene, 1997). Therefore severalunit root and cointegration tests are performed. First,Dickey–Fuller (DF) and Augmented Dickey–Fuller(ADF) tests are computed for the data of the individualindustries of the different countries. In general, these DFand ADF-tests show that the hypothesis of the presence ofa unit root cannot be rejected. However, the power of thesetests is very low since only 17 annual observations are dealtwith in the panel. The work of Im et al. (1996) shows DFand ADF tests which are quite easy to compute whenworking with panels with a small time and cross-sectiondimension. Their tests are based on the mean of the unitroot statistics of the individual industries. Using the unitroot and cointegration tests based on the work of these

authors, it turns out that in general the hypothesis of thepresence of a unit root of the different variables could notbe rejected. Moreover, the cointegration tests showed thatthe error terms of the regressions are nonstationary most ofthe time. It was decided however to give the regressionresults in levels as well as in first differences (seeAppendix II and III) since the econometrics on unit rootand cointegration tests for panel data is still in progressand the emphasis is on the theoretical model and the inter-pretation of the estimated coefficients.The employment and productivity regression equations

(see Equations 12 and 13) constitute a recursive modelbecause the productivity equation does not contain anyendogenous variables from the labour demand equation,while this latter equation contains endogenous variablescoming from the former equation. More specifically, theemployment equation depends on productivity which inturn is explained by a set of exogenous explanatory vari-ables. To capture in the employment equation only theproductivity changes that are explained by import penetra-tion, export demand, capital intensity and technology vari-ables, a two stage least squares approach is used bysubstituting the fitted values of the productivity measureobtained by the regression results from Equation 13 intoEquation 12. Furthermore a fixed effects approach is used.An exhaustive sample is dealt with and Hausman testsshow that this fixed effects approach in comparison withrandom effects is appropriate (Matyas and Sevestre, 1996).Early regression results pointed to autocorrelation. In

order to combine two stage least squares and a correctionfor autocorrelation, the methodology of Fair (1970) isused. The productivity equation is estimated with the aidof a Cochrane–Orcutt iterative technique. In order to esti-mate the labour demand equation, Equation 13 is esti-mated without a correction for autocorrelation. Thefitted values of this equation are then substituted inEquation 12. Then, this equation is also estimated with aCochrane–Orcutt iterative method. When using thisapproach, the standard errors of the employment equationare corrected for the use of the generated regressors fromthe first stage.To check the robustness of the results, several consis-

tency checks were performed which are not reported butcan be obtained from the authors. Regressions were esti-mated taking lags of several explanatory variables. It wasconsidered that this was most of all necessary for the R&Dvariable because investments in R&D take a long time tomature. In general, the results did not significantly changewhen lagged variables were introduced, the estimationresults without any lags are presented.

The export demand effect

One important theme of this paper is the contribution ofinternational trade to total sectoral employment in the


Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

European economy. Tables A1–A4 (see Appendix II) showthe results for regression Equation 12 expressed in levelsand first differences when using value added or total factorproductivity as the productivity variable. The estimatedresults in these tables are quite satisfactory. Except forItaly and Spain, estimated coefficients of the exportdemand, import penetration, productivity and labour costvariables usually carry the expected signs and are in mostcases significant at conventional confidence levels.

Table 1 focuses on the export demand effect in greaterdetail by presenting the regression coefficients for �1 inEquation 12, using various estimation techniques. Thistable shows evidence of a robust export demand effect inmost European countries. When using the estimationresults in either levels or in first differences, statisticallysignificant regression coefficients for exports with VA arein general higher than when TFP is used as productivityvariable. Comparing elasticities across countries, statisti-cally significant regression coefficients range from 0.06 inFrance to 0.49 in Denmark. Strong export demand effectsare also found in Sweden and the UK. With the exceptionof this last country, export growth appears to create morejobs in the smaller than in the larger European countriesalthough differences between estimation techniques blurthis picture somewhat.

To get an idea of the importance of the export demandeffect, the accumulated export-induced employment crea-tion were computed for Danish and French manufacturing,the countries with the highest and lowest statistically sig-nificant elasticities. Tentatively applying an elasticity to theobserved growth in real exports in the period 1978–1994 of0.49 in the case of Denmark and 0.06 in the case of France,an accumulated employment creation of 143 850 jobs inDenmark and 186 357 jobs in France were obtained. Thisamounts to 28% of the average workforce of 500 547 peo-ple during this period in Denmark and to 3% of the aver-age workforce of 4 792 176 people during the same period

in France. Those numbers are substantial and it is thereforeunfortunate that the sectoral employment effects of exportexpansion receive scant attention in the trade-related litera-ture.

The import competition effect

Turning to the import competition effect in Table 2,increased import penetration destroys sectoral employmentin a number of countries. This effect is most present in theScandinavian countries but also observed in the UK, theNetherlands and Belgium. Focusing again on the magni-tude of the employment adjustments involved, Sweden withthe strongest import competition effect of �0.97 andBelgium with the lowest statistically significant elasticityof �0.04 were selected. It is noted that import penetrationin manufacturing grew by 46% during 1978–1994 inSweden and by 49% in Belgium during this same period.During 1978–1994, import competition costs the jobs of411 630 workers or 44% of the average manufacturinglabour force in Sweden. For Belgium, 16 034 jobs – whichrepresents 1% of the average labour force during 1978–1994 – were lost. These findings suggest that for someEuropean countries import competition matters a lot,while for others import competition is not the main drivingforce for employment adjustments. This clearly is in con-trast with the basic thrust of the literature where there islittle evidence for the import competition effect. As arguedearlier, most studies focus on the USA which is less openthan Europe. Therefore, employment effects caused byincreased import competition may be higher in Europethan in the USA.

Trade and productivity

In the analysis of the productivity effects of internationaltrade on employment, one important condition was related


Table 1. The export demand effect

VA TFP

Country Levels First differences Levels First differences

Belgium 0.16** 0.31** 0.09** 0.20**Denmark 0.16** 0.40** 0.11* 0.49*Finland 0.13** 0.41** 0.05 0.14*France 0.06** 0.27** 0.07** 0.05Germany 0.13** 0.12** 0.09* 0.22*Italy �0.02 0.05 �0.03 0.03Netherlands 0.13** 0.29** 0.11** 0.11Spain 0.03 0.07* 0.02 0.01Sweden 0.19** 0.32** 0.12** 0.22**UK 0.17** 0.34** 0.21** 0.26**

Note: ** Significance at the 5 % level (two�tailed test).* Significance at the 10 % level (two�tailed test).

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

to the impact of exports and imports on productivity. The

regression results of the productivity Equation 13 are given

in Tables A5–A8 and summarized in Tables 3 and 4. From

Table 3, exports emerge as a statistically significant source

of productivity gains in virtually all cases. Belgium tops the

list with the strongest impact of exports on productivity.

The ranking of the other countries varies depending on

which estimation method and productivity variable was

chosen. Interesting and plausible are the consistently higher

coefficients when the total factor productivity variable is

used. Export growth appears to raise the efficiency of

both capital and labour such that the adjustment in labour

productivity captures only one quarter to half of the gains

in total factor productivity.

Turning to the productivity effect of import penetration

(see Table 4), negative regression coefficients are obtained

that are statistically significant for all (several) countries

when total factor productivity (value added per worker)

is used as the dependent variable. Large negative effects

are found in Sweden, the Netherlands and Denmark (in

the regressions with TFP) and to a lesser extent in

Belgium and Germany. Opposite to Wood’s hypothesis,

it is concluded that increased import competition causes

a loss in productivity. This supports the view that compa-

nies are unable to scale down their factor use at the same

rate as rising foreign competition reduces their sales.

Restructuring is a difficult process in Europe.

As Lawrence (2000) and Bernard and Jensen (1999 and

2001) point out, the link between international trade and

productivity can go in both directions. Import and export

competition can trigger higher productivity, while at the

same time sectors confronted with falling (growing) pro-

ductivity may tend to have high levels of import competi-

tion (exports). One way to deal with this problem is to run

Granger causality tests which suggest that in the present

dataset the link goes most often from trade to productivity.

Another way to confront this reverse causation is the use of

an instrumental variables (IV) approach. As suggested by


Table 2. The import competition effect

VA TFP


Belgium �0.04** �0.17** �0.01 �0.05Denmark �0.45** �0.66** �0.49** �0.82**Finland �0.18** �0.18 �0.21** �0.18**France 0.007 �0.16* 0.003 �0.003Germany 0.05 �0.03 0.06 �0.06Italy �0.01 �0.05* �0.01 �0.06 �

Netherlands �0.10* �0.34** �0.08 �0.08Spain 0.05 �0.04 0.05 �0.01Sweden �0.45** �0.97** �0.41** �0.38**UK �0.12** �0.27** �0.16** �0.20**


Table 3. The productivity effect of export

VA TFP


Belgium 0.33** 0.38** 0.82** 0.83**Denmark 0.20** 0.25** 0.81** 0.79**Finland 0.16** 0.15** 0.33** 0.35**France 0.28** 0.30** 0.64** 0.65**Germany 0.20** 0.19** 0.58** 0.59**Italy 0.24** 0.23** 0.52** 0.48**Netherlands 0.20** 0.22** 0.77** 0.81**Spain 0.10** 0.16** 0.26** �0.15Sweden 0.26** 0.23** 0.61** 0.59**UK 0.18** 0.14** 0.44** 0.39**

Note: ** Significance at the 5% level (two�tailed test).* Significance at the 10% level (two�tailed test).

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

theHOS theory, skill- and capital intensity form appropriateinstruments (see Lawrence, 2000). Because of data limita-tions, the lagged trade variables were taken as instruments.This clearly leads to lower significance of the trade variablesin the productivity equation, especially when estimating infirst differences. Unfortunately, the standard errors of theIV-estimates are quite high, so that the lack of significancemight be caused by the use of weak instruments.8

Productivity and employment

The second condition for a meaningful productivity effectof trade on employment concerns the employment adjust-ment to productivity changes. To capture this relationship,Equation 12 is estimated where the productivity variable isinstrumented by using the fitted values for VA and TFPfrom the productivity regression (13). Based on the results

in Appendix II, Table 5 presents the relevant information.Moderate to strong negative regression coefficients areobtained for the productivity effect on employment whenvalue added per worker is used as the productivity variable,in particular when the employment equation is estimated infirst differences. With a few exceptions, a statisticallysignificant relation is not found between employment andproductivity when TFP is considered. Apparently, shocksthat raise labour productivity are labour-saving, most of allin the Scandinavian countries and in the UK. But factorneutral productivity changes do not destroy jobs.

Productivity-related and total effects of international tradeon employment

Direct and productivity-induced employment adjustmentsto export demand are brought together. First earlier esti-


Table 4. The productivity effect of import competition

VA TFP


Belgium �0.15** �0.16** �0.35** �0.34**Denmark �0.09 �0.17 �0.96** �0.98**Finland �0.04 �0.04 �0.11** �0.13**France �0.08 �0.14** �0.21** �0.22**Germany �0.14** �0.20** �0.33** �0.35**Italy �0.04 �0.03 �0.23** �0.20**Netherlands �0.30** �0.30** �0.85** �0.87**Spain 0.004 0.0009 �0.24** 0.39**Sweden �0.52** �0.57** �0.35** �0.46**UK �0.13 �0.11* �0.33** �0.31**


Table 5. Productivity and employment

VA TFP


Belgium �0.10** �0.71** 0.04 �0.15Denmark �0.17** �1.45** �0.02 �0.49**Finland �0.39** �2.80** �0.04 �0.42**France �0.09** �0.71** �0.01 �0.04Germany �0.13** �0.34** 0.004 �0.29 �

Italy �0.04 �0.46** �0.01 �0.17Netherlands �0.03 �0.85** 0.03 �0.03Spain �0.09 �0.26* �0.04* �0.07Sweden �0.15** �1.61** 0.03 �0.19UK �0.13** �1.34** �0.03 0.005


8 Especially for the data in first differences, the correlation between the current and lagged trade variables is extremely weak. This is apossible indicator of weak instrumental variables.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

mates for the export demand are reproduced where VA isused as the productivity variable. Subsequently, the elasti-

cities that measure the productivity effect of exports andimport competition on employment are computed from the

parameter values reported in earlier tables.9 Since onlychanges in labour productivity affect employment, onlythe estimates for the value added per worker productivity

variable are reported. Finally, the total elasticities ofrespectively export demand and import competition on

employment are calculated by summing up the figures inthe previous two columns.

Table 6 makes an interesting point. Export growth raises

labour productivity which offsets part of the employmentcreated by rising export demand. While level estimates vary

from country to country, export-induced gains in labourproductivity typically destroy one out of every five to sevenjobs created by export expansion. This ratio rises to

one out of three workers in France and even half of theemployment creation in Finland. These export-related

productivity effects are much stronger when using first dif-ferences in the estimations. This reduces considerably the

positive employment effects of an expansion in exportdemand.

In Table 7 similar information is provided for the importcompetition variable. The positive productivity effect ofimport competition on employment is not as intuitive as

the jobs lost when exporting firms adopt more efficienttechnologies. Most plausibly, the lay-offs prevented by

the inability of companies are captured – when facedwith stronger import competition – to smoothly readjusttheir labour force to falling output levels. In the case of

estimations in levels, the productivity effect is – except forBelgium and France – relatively small in comparison with

the import competition effect. However, when using estima-tions in first differences, the opposite holds. For Belgium,Finland, Germany, the Netherlands, Sweden and the UK,

half of the jobs lost by increasing import competition arecompensated by the positive productivity effect.


9 This effect is not estimated directly so that no statistical significance tests can be reported.

Table 6. The productivity and total effects of exports on employment with VA

Levels First differences

Export demand Productivity- Total Export demand Productivity- TotalCounty effect related effect effect effect related effect effect

Belgium 0.16 �0.03 0.13 0.31 �0.26 0.05Denmark 0.16 �0.03 0.13 0.40 �0.36 0.04Finland 0.13 �0.06 0.07 0.41 �0.42 �0.01France 0.06 �0.02 0.04 0.27 �0.21 0.06Germany 0.13 �0.02 0.11 0.12 �0.06 0.06Italy �0.02 �0.009 �0.029 0.05 �0.10 �0.05Netherlands 0.13 �0.006 0.124 0.29 �0.18 0.09Spain 0.03 �0.009 0.021 0.07 �0.04 0.03Sweden 0.19 �0.03 0.16 0.32 �0.37 �0.05UK 0.17 �0.02 0.15 0.34 �0.18 0.16

Table 7. The productivity and total effects of imports competition on employment with VA

Levels First differences

Import Productivity- Total Import Productivity- TotalCountry competition effect related effect effect competition effect related effect effect

Belgium �0.04 0.01 �0.03 �0.17 0.11 �0.06Denmark �0.45 0.01 �0.44 �0.66 0.24 �0.40Finland �0.18 0.01 �0.17 �0.18 0.11 �0.07France 0.007 0.007 0 �0.16 0.09 �0.07Germany 0.05 0.01 0.06 �0.03 0.06 0.03Italy �0.01 0.001 �0.009 �0.05 0.01 �0.04Netherlands �0.10 0.009 �0.09 �0.34 0.25 �0.09Spain 0.05 �0.0003 0.04 �0.04 0.0002 �0.04Sweden �0.45 0.07 �0.38 �0.97 0.91 �0.06UK �0.12 0.01 �0.11 �0.27 0.14 �0.13

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

V. CONCLUSION

This paper deals with the sectoral employment effects ofinternational trade in Europe. It offers several contribu-tions to the rapidly expanding literature on trade, technol-ogy and employment.

First, evidence that trade matters for employment inEurope is provided. With this finding, the findings goagainst the thrust of the US-focused literature which attri-butes only a secondary role to trade-related labour marketadjustment. The fact that most European countries aremore open economies undoubtedly explains part of thisresult. But it is also believed that, in a European contextwith rigid labour markets and limited intersectoral labourreallocation, the modelling of sectoral employment adjust-ments explains why significant trade-related employmenteffects are observed. Most importantly, this empiricalwork highlights the importance of export growth as a keyengine of job creation. This point is often lost in the litera-ture which focuses heavily on the detrimental impact ofincreased import competition on employment and wages.

As a second contribution, this paper puts in doubt thedistinction between trade and technology as independentsources of employment changes. In most European coun-tries of the sample, international trade affects productivity.More specifically, rising export demand increases labourand total productivity, a point frequently made byEuropean business leaders. By contrast, sectors confrontedwith increased import competition experience a decline inproductivity indicating that the reduction in factor use doesnot keep up with the loss in market share. This scenario ofinflexible sectoral restructuring is consistent with the con-ventional view that European labour markets are rigid andconstrained by stringent hiring and firing conditions.

The consequences for employment of the relationshipbetween trade and productivity leads to a third majortheme of this paper. It is shown that employment respondsto trade-induced changes in average labour productivity.The possibility of such an indirect technology-related linkfrom international trade to employment was alreadyemphasized by authors such as Wood. This contributionis to develop and to apply a framework that quantifies thisproductivity effect of international trade on employment.Perhaps the most striking outcome of this empirical exer-cise is that part of the employment created by an expansionof export demand is neutralized by the negative employ-ment effect of an export-induced increase in labour produc-tivity. Interestingly, this suggests that export growth ratherthan import competition encourages the introduction oflabour-saving production methods that are so common inthe European industry.

The results of this paper leave open several paths forfuture research. One suggestion is a closer look at the dif-ferences in employment responses across countries. Quiteoften in the regression results, the Scandinavian countries

but – to a lesser extent – also the UK and the Beneluxcountries, are affected more profoundly by trade and pro-ductivity than the other European countries of the sample.Another promising route would be to exploit the sectoraldimension of the panel and to compare the employmentadjustments in sectors with different structural character-istics. Finally, it is possible to geographically disaggregatetrade flows to analyse the contribution of various tradingpartners to sectoral employment changes. These issues willbe addressed in future work.

ACKNOWLEDGEMENTS

This paper benefited from comments at conferences inPurdue University, Orebro and from seminar presentationsat ZEI in Bonn and the Centre Universitaire duLuxembourg.

REFERENCES

Abraham, F. and Konings, J. (1999) Does the opening of Centraland Eastern Europe kill jobs in the West?, The WorldEconomy1, 221 (4), 585–601.

Bentolila, S. and Bertola, G. (1990) Firing costs and labourdemand: how bad is Eurosclerosis?, Review of EconomicStudies, 57 (3), 381–402.

Berman, E., Bound, J. and Griliches, Z. (1994) Changes in thedemand for skilled labor within US manufacturing evidencefrom the annual survey of manufacturers, Quarterly Journalof Economics, 109 (2), 367–97.

Bernard, A. B. and Jensen, J. B. (1999) Exceptional exporterperformance: cause, effect or both?, Journal of InternationalEconomics, 47 (1), 1–25.

Bernard, A. B. and Jensen, J. B. (2001) Exporting andProductivity, NBER Working Paper Series, No. 7135,National Bureau of Economic Research, Cambridge,Massachusetts.

Bertola, G. (1990) Job security, employment and wages, EuropeanEconomic Review, 34 (4), 851–86.

Booth, A. (1997) An analysis of firing costs and their implicationsfor unemployment policy, in Unemployment Policy:Government Options for the Labour Market (Eds) G. de laDehasa and D. J. Snower, Cambridge University Press,Cambridge, pp. 359–88.

Brander, J. A. and Spencer, B. (1988) Unionized oligopoly andinternational trade policy, Journal of InternationalEconomics, 24 (3–4), 217–34.

Card, D., Kramarz, F. and Lemieux, T. (1999) Changes in therelative structure of wages and employment: a comparison ofthe United States, Canada and France, Canadian Journal ofEconomics, 32 (4), 843–77.

Coe, D and Helpman, E. (1995) International R&D spillovers,European Economic Review, 39 (5), 859–87.

Davis, D.R. (1998) Does European unemployment prop upAmerican wages? National labor markets and global trade,American Economic Review, 88 (3), 478–94.

Decressin, J. and Fatas, A. (1995) Regional labour marketdynamics in Europe, European Economic Review, 39 (9),1627–55.


Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

Dixit, A. and Stiglitz, J. E. (1977) Monopolistic competition andoptimum product diversity, American Economic Review, 67(3), 297–308.

Driffil, J. and van der Ploeg, F. (1995) Trade liberalisation withimperfect competition in goods and labour markets,Scandinavian Journal of Economics, 97 (2), 223–43.

Fair, R.C. (1970) The estimation of simultaneous equation modelswith lagged endogenous variables and first order serially cor-related errors, Econometrica, 38 (3), 507–16.

Freeman, R. and Revenga, A. (1999), How much has tradeaffected Western job markets? in Trade and Jobs in Europe.Much Ado about Nothing? (Eds) M. Dewatripont, A. Sapir,and K. Sekkat, Oxford University Press, Oxford, pp. 8–32.

Garibaldi, P. (1998) Job flow dynamics and firing restrictions,European Economic Review, 42 (2), 245–75.

Gaston, N. and Trefler, D. (1995) Union wage sensitivity to tradeand protection: theory and evidence, Journal of InternationalEconomics, 39 (1/2), 1–26.

Greene, W. H. (1997), Econometric Analysis, Prentice HallInternational, New York.

Griliches, Z. (1979) Issues in assessing the contribution ofresearch and development to productivity growth, BellJournal of Economics, 10 (1), 92–116.

Griliches, Z. and Ringstad, V. (1971) Economies of Scale and theForm of the Production Function: An Econometric Study ofNorwegian Manufacturing Establishment Data, NorthHolland, Amsterdam and London.

Grubb, D. and Wells, W. (1993) Employment regulation andpatterns of work in EC countries, OECD Economic Studies,21, OECD., Paris, 7–56.

Huizinga, H. (1993) International market integration and unionbargaining, Scandinavian Journal of Economics, 95 (2), 249–55.

Im, K. S., Pesarin, M. H. and Shin, Y. (1996) Testing unit roots inheterogeneous panels, University of Cambridge, mimeo.

Krugman, P. (1995) Growing world trade: causes and conse-quences, Brookings Papers on Economic Activity, 1, 327–77.

Krugman, P. and Lawrence, R. (1996) Trade, jobs and wages, inPop Internationalism (Ed.) P. Krugman, MIT Press,Cambridge, pp. 35–48.

Larre, B. (1995) The Impact of Trade on Labour Markets: anAnalysis by Industry, OECD Jobs Study Working PaperSeries, No. 6, OECD, Paris.

Lawrence, R.Z. (2000) Does a kick in the pants get you going ordoes it just hurt? The impact of international competition ontechnological change in US manufacturing, in The Impact ofInternational Trade on Wages (Ed.) R. C. Feenstra,University of Chicago Press, Chicago, pp. 197–224.

Maskus, K. (1991) Comparing international trade data and pro-duct and national characteristics data for the analysis oftrade models, in International Economic Transactions (Eds),P. Hooper and J. D. Richardson, University of Chicago,Chicago, pp. 17–60.

Matyas, L. and Sevestre, P. (1996) The Econometrics of PanelData: A Handbook of Theory and Applications, KluwerAcademic, Dordrecht.

Mezzetti, C. and Dinopoulos, E. (1991) Domestic unionizationand import competition, Journal of InternationalEconomics, 31 (1–2), 79–100.

Naylor, R. (1998) International trade and economic integrationwhen labour markets are generally unionised, EuropeanEconomic Review, 42 (7), 1251–68.

Neven, D. and Wyplosz, C. (1999), Relative prices, trade andrestructuring in European industry, in Trade and Jobs inEurope. Much Ado About Nothing? (Eds) M. Dewatripont,

A. Sapir, and K. Sekkat, Oxford University Press, Oxford,pp. 33–59.

OECD (1994) The Measurement of Scientific and TechnologicalActivities: Proposed Standard Practice for Surveys ofResearch and Experimental Development: Frascati Manual1993, OECD, Paris.

OECD (1995) Research and Development Expenditure in Industry,OECD, Paris.

OECD (1997) The OECD Stan Database for Industrial Analysis,OECD, Paris.

Revenga, A. (1992) Exporting jobs? The impact of import com-petition on employment and wages in US manufacturing,Quarterly Journal of Economics, 107 (1), 255–84.

Slaughter, M. and Swagel, P. (1997) The Effects of Globalisationon Wages in the Advanced Economies, IMF Working Paper(97/43), IMF, Washington DC.

Vandenbussche, H. and Konings, J. (1998) Globalization and theeffects of national versus international competition on thelabour market, World Economy, 21 (8), p. 1151–77.

Varian, H. R. (1984), Microeconomic Analysis, Norton, NewYork.

Verspagen, B., van Moergastel, T. and Slabbers, M. (1994)MERIT Concordance Table: IPC-ISIC (rev.2), mimeo.Merit Research Memorandum 2194 (004).

Wood, A. (1994), North-South Trade, Employment and Inequality:Changing Fortunes in a Skill-Driven World, OxfordUniversity Press, Oxford.

APPENDIX I

The following countries are covered in the data set:Belgium, Denmark, Finland, France, Germany, Italy,Netherlands, Spain, Sweden and the United Kingdom.Nine sectors of the manufacturing sector which are classi-fied according to the ‘International Standard IndustrialClassification’ (ISIC) revision 2 are used: food, beveragesand tobacco (ISIC 31), textiles, apparel and leather (ISIC32), wood products and furniture (ISIC 33), paper, paperproducts and printing (ISIC 34), chemical products (ISIC35), non-metallic mineral products (ISIC 36), basic metalindustries (ISIC 37), fabricated metal products (ISIC 38)and other manufacturing (ISIC 39). The starting year ofthe data set is 1978 for all countries except for Spain whereit is 1980. Data are used until 1994 except for Denmark,Spain and Germany. The end points for the first two coun-tries are 1992, whereas the data for Germany go until 1993.The data for employment, labour costs, import, export,

production and value added are obtained from the OECDStan Database for Industrial Analysis (1997). The data foremployment cover the number of employees as well as self-employed, working proprietors and unpaid family workers.The gross wage data per employee cover wages and varioussupplements such as employer’s compulsory pension ormedical payments. Except for import penetration and rela-tive granted patents, all variables are expressed in constantprices. The deflators are calculated with the aid of valueadded in current and constant prices per industry.


Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

For productivity, there are two variables: value added

per worker and total factor productivity which is trans-formed into indices where 1990 is the base year. The per-centage change of the total factor productivity can be

expressed as follows:

AA ¼ QQ� LL� �

� � KK � LL� �

ðA1Þ

In this expression, the first term refers to the percentagechange in the output–labour ratio. In the second term, �refers to the capital share in production. Therefore, 1� �ð Þrefers to the labour share in production which is calculatedas the share of labour costs in value added. For

some sectors, labour costs exceed value added due tothe existence of e.g. losses in these sectors or because theindustry receives significant net subsidies. Therefore, an

average is calculated.10 KK � LL� �

refers to the percentagechange in the capital–labour ratio. For the capital data,

two series are distinguished. For Belgium, Denmark,Finland, France, Germany and Sweden gross capitalstock data are expressed in 1990 prices and are obtained

from the International Sectoral Data Base from the OECDStatistical Compendium 1998/2. For Italy, the

Netherlands, Spain and the UK, capital data are con-structed with the aid of deflated investment data Itð Þfrom the OECD Stan Database for Industrial Analysis

(1997).11 Following Griliches (1979), an initial capitalstock for 1978 (and 1980 for Spain) is computed first. Ifit is assumed that both the depreciation rate �ð Þ and the

annual growth rate �ð Þ of the past investments are con-stant, the initial capital stock K1978ð Þ equals:

K1978 ¼ I1978 þ 1� �ð ÞI1978 þ 1� �ð Þ22I1978

þ 1� �ð Þ33I1978 þ � � �

¼ I19781

1� 1� �ð Þ

� �ðA2Þ

with ¼ 1=ð1þ �Þ where � is the mean annual rate ofgrowth of investments over the period 1970-1978. LikeMaskus (1991), we use a depreciation rate of 13.33%.After having obtained the initial capital stock, deflatedinvestments series are accumulated and depreciated from1978 onwards.The patent variable measures the relative granted

patents. Under relative granted patents, the grantedpatents in one industry relative to all granted patents in acertain year are understood. The patent data cover patentswithin the EPO (European Patent Office). The classifica-tion is the International Patent Classification (IPC). Theconversion to the ISIC-classification is computed with theaid of the correspondance table of Verspagen et al. (1994).Except for Belgium, the data for expenditures in R&Dcome from the Analytical Business Enterprise Researchand Development (ANBERD) series from ‘Research andDevelopment Expenditure in Industry’ (OECD, 1995). ForBelgium, data for 1978–1991 are obtained from the OfficialBusiness Enterprise Research and Development(OFFBERD) series and for the period 1994–1995 data ofthe DWTC (Dienst voor Wetenschappelijke, Technische enCulturele Aangelegenheden) are used. All data are con-verted to the ISIC revision 2 according to the correspon-dance table in OECD (1994). With the aid of a cubic splineinterpolation technique, missing observations are filled in.


10 Another way to calculate the cost share of capital is to construct a figure for capital services using the concept of user cost of capital(see Griliches and Ringstad, 1971).11 A more complete description of how the second capital series is constructed is available from the authors upon request.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

APPENDIX II


Table A1. The regression results of the employment equation with VA, estimated in levels

Country ln (IMP) ln (EXP) ln (WAGE) ln (VA) R2

Belgium �0.04** 0.16** �0.15** �0.10** 0.99(�2.09) (5.20) (�4.55) (�3.70)

Denmark �0.45** 0.16** �0.15** �0.17** 0.99(�4.49) (2.82) (�2.25) (�3.56)

Finland �0.18** 0.13** �0.14** �0.39** 0.99(�3.26) (3.76) (�2.30) (�7.42)

France 0.007 0.06** �0.09** �0.09** 0.99(0.17) (2.45) (�2.22) (�3.50)

Germany 0.05 0.13** �0.22** �0.13** 0.99(1.11) (3.18) (�2.42) (�2.67)

Italy �0.01 �0.02 �0.04 �0.04 0.99(�0.45) (�1.09) (�0.84) (�1.57)

Netherlands �0.10* 0.13** �0.23** �0.03 0.99(�1.85) (3.86) (�5.16) (�0.80)

Spain 0.05 0.03 �0.05 �0.09 0.99(1.53) (0.90) (�0.58) (�1.18)

Sweden �0.45** 0.19** �0.22** �0.15** 0.99(�6.36) (4.96) (�4.22) (�4.71)

UK �0.12** 0.17** �0.31** �0.13** 0.99(�2.26) (4.83) (�4.95) (�3.88)

Note: ** Significance at the 5% level (two-tailed test).* Significance at the 10% level (two-tailed test).

Table A2. The regression results of the employment equation with VA, estimated in first differences

Country ln (IMP) ln (EXP) ln (WAGE) ln (VA) R2

Belgium �0.17** 0.31** �0.04 �0.71** 0.24(3.33) (3.85) (�0.77) (�3.72)

Denmark �0.66** 0.40** �0.01 �1.45** 0.07(�2.53) (2.19) (�0.07) (�2.64)

Finland �0.18 0.41** �0.06 �2.80** 0.12(�1.21) (2.22) (�0.36) (�2.62)

France �0.16* 0.27** -0.05 -0.71** 0.17(�1.95) (3.58) (�0.71) (�3.54)

Germany �0.03 0.12** �0.17* �0.34** 0.31(�0.66) (2.43) (�1.81) (�2.30)

Italy �0.05* 0.05 �0.01 �0.46** 0.17(�1.69) (1.03) (�0.22) (�2.05)

Netherlands �0.34** 0.29** �0.21** �0.85** 0.45(�3.15) (4.05) (�3.83) (�3.04)

Spain �0.04 0.07* �0.03 �0.26* 0.43(�1.44) (1.82) (�0.31) (�1.92)

Sweden �0.97** 0.32** 0.03 �1.61** 0.14(�3.20) (2.34) (0.19) (�3.30)

UK �0.27** 0.34** �0.18 �1.34** 0.34(�3.15) (4.91) (�1.37) (�3.45)


Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14


Table A3. The regression results of the employment equation with TFP, estimated in levels

Country ln (IMP) ln (EXP) ln (WAGE) ln (TFP) R2

Belgium �0.01 0.09** �0.15** 0.04 0.99(�0.74) (2.48) (�4.25) (1.61)

Denmark �0.49** 0.11* �0.12* �0.02 0.99(�4.83) (1.88) (�1.85) (�0.47)

Finland �0.21** 0.05 �0.12* �0.04 0.99(�3.66) (1.46) (�1.83) (�0.72)

France 0.003 0.07** �0.12** �0.01 0.99(0.08) (2.37) (�2.79) (�0.49)

Germany 0.06 0.09* �0.12 0.004 0.99(1.35) (1.94) (�1.45) (0.11)

Italy �0.01 �0.03 �0.04 �0.01 0.99(�0.55) (�1.32) (�0.93) (�0.87)

Netherlands �0.08 0.11** �0.23** 0.03 0.99(�1.39) (2.70) (�4.76) (1.22)

Spain 0.05 0.02 �0.09 �0.04* 0.99(1.55) (0.57) (�1.11) (�1.67)

Sweden �0.41** 0.12** �0.24** 0.03 0.99(�5.46) (2.84) (�4.32) (1.02)

UK �0.16** 0.21** �0.34** �0.03 0.99(�3.06) (5.66) (�5.32) (�0.87)


Table A4. The regression results of the employment equation with TFP, estimated in first differences

Country ln (IMP) ln (EXP) ln (WAGE) ln (TFP) R2

Belgium �0.05 0.20** �0.12** �0.15 0.55(�1.54) (2.68) (�4.02) (�1.61)

Denmark �0.82** 0.49* �0.07 �0.49** 0.23(�3.79) (2.83) (�1.01) (�2.35)

Finland �0.18** 0.14* �0.04 �0.42** 0.27(�3.36) (1.87) (�0.68) (�2.08)

France �0.003 0.05 �0.06* �0.04 0.58(�0.10) (1.05) (�1.94) (�0.58)

Germany �0.06 0.22 � �0.13 �0.29* 0.27(�0.85) (2.23) (�1.47) (�1.87)

Italy �0.06* 0.03 �0.02 �0.17 0.17(�1.81) (0.54) (�0.53) (�1.18)

Netherlands �0.08 0.11 �0.16** �0.03 0.39(�0.60) (0.97) (�3.73) (�0.22)

Spain �0.01 0.01 �0.08 �0.07 0.30(�0.27) (0.34) (�0.88) (�1.06)

Sweden �0.38** 0.22** �0.14** �0.19 0.49(�3.87) (2.55) (�2.85) (�1.45)

UK �0.20** 0.26** �0.44** 0.005 0.50(�2.60) (3.03) (�6.65) (0.02)


Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14


Table A5. The regression results of the productivity equation with VA, estimated in levels

Country ln(IMP) ln(EXP) ln(R&D) PAT ln(CAP) R2

Belgium �0.15** 0.33** 0.01 0.001 0.50** 0.98(�4.07) (7.01) (0.63) (0.001) (6.92)

Denmark �0.09 0.20** 0.03 �0.19 0.12 0.93(�0.65) (3.65) (1.26) (�1.27) (1.18)

Finland �0.04 0.16** �0.008 0.002 0.44** 0.98(�0.77) (4.72) (�0.35) (0.03) (5.08)

France �0.08 0.28** �0.03* �0.06 0.58** 0.98(�1.02) (5.78) (�1.87) (�0.24) (6.47)

Germany �0.14** 0.20** �0.05 �0.73 0.39** 0.98(�2.23) (3.55) (�1.32) (�1.39) (3.29)

Italy �0.04 0.24** 0.005 �0.16* 0.03 0.98(�1.12) (7.49) (0.96) (�1.74) (0.49)

Netherlands �0.30** 0.20** 0.02 � 0.23 0.04 0.98(�3.19) (3.83) (1.88) (0.74) (0.52)

Spain 0.004 0.10** 0.03** �0.02 0.05** 0.99(0.14) (2.56) (2.54) (�0.19) (2.64)

Sweden �0.52** 0.26** 0.01 0.08 0.87** 0.98(�3.37) (3.80) (0.52) (0.28) (5.97)

UK �0.13 0.18** 0.05** �0.02 0.13 � 0.97(�2.05) (3.97) (2.03) (�0.11) (1.93)

Note: ** Significance at the 5% level (two=tailed test).* Significance at the 10% level (two-tailed test).

Table A6. The regression results of the productivity equation with VA, estimated in first differences

Country ln (IMP) ln (EXP) ln (R&D) PAT ln (CAP) R2

Belgium �0.16** 0.38** 0.008 0.01 0.54** 0.51(�3.94) (8.51) (0.33) (0.17) (4.22)

Denmark �0.17 0.25** 0.006 �0.18 0.25 0.29(�1.04) (4.96) (0.23) (�1.21) (1.61)

Finland �0.04 0.15** �0.01 �0.01 0.15 0.16(�0.75) (4.64) (�0.56) (�0.16) (1.53)

France �0.14* 0.30** �0.02 �0.06 0.78** 0.35(�1.69) (6.09) (�1.35) (�0.27) (3.84)

Germany �0.20** 0.19** �0.03 �0.21 0.008 0.16(�3.40) (3.57) (�1.01) (�0.73) (0.04)

Italy �0.03 0.23** 0.005 �0.14 0.12** 0.38(�0.96) (7.00) (1.02) (�1.60) (1.98)

Netherlands �0.30** 0.22** 0.01 0.20 0.15** 0.17(�3.33) (4.37) (1.48) (0.75) (1.96)

Spain 0.009 0.16** 0.05** 0.02 0.07** 0.46(0.02) (4.27) (3.62) (0.16) (3.51)

Sweden �0.57** 0.23** 0.005 0.05 0.64** 0.19(�3.70) (3.48) (0.16) (0.22) (3.65)

UK �0.11* 0.14** 0.05* �0.01 0.17** 0.22(�1.87) (3.28) (1.91) (�0.06) (2.61)


Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14


Table A7. The regression results of the productivity equation with TFP, estimated in levels

Country ln (IMP) ln (EXP) ln (R&D) PAT R2

Belgium �0.35** 0.82** 0.009 0.01 0.96(�11.50) (21.15) (0.46) (0.27)

Denmark �0.96** 0.81** 0.04** �0.18 0.96(�7.88) (19.97) (2.13) (�1.59)

Finland �0.11** 0.33** 0.05** 0.01 0.87(2.10) (9.65) (2.42) (0.23)

France �0.21** 0.64** �0.02 �0.28 0.89(�2.92) (15.33) (�1.53) (�1.23)

Germany �0.33** 0.58** �0.01 �0.27 0.89(�5.03) (10.31) (�0.45) (�0.59)

Italy �0.23** 0.52** 0.008 �0.27* 0.90(�3.62) (10.34) (1.05) (�1.81)

Netherlands �0.85** 0.77** 0.01 �0.30 0.92(�9.85) (15.56) (1.25) (�1.03)

Spain �0.24** 0.26** 0.01 0.35 0.73(�3.33) (2.41) (0.35) (0.86)

Sweden �0.35** 0.61** �0.001 �0.09 0.92(�3.42) (11.99) (�0.04) (�0.43)

UK �0.33** 0.44** 0.10** �0.24 0.91(�4.42) (8.46) (3.11) (�0.80)


Table A8. The regression results of the productivity equation with TFP as productivity variable,estimated in first differences

Country ln (IMP) ln (EXP) ln (R&D) PAT R2

Belgium �0.34** 0.83** 0.009 0.004 0.83(�11.75) (23.15) (0.47) (0.07)

Denmark �0.98** 0.79** 0.05** �0.17 0.83(�8.27) (20.89) (2.50) (�1.52)

Finland �0.13** 0.35** 0.04** � 0.02 0.48(�2.39) (10.40) (1.97) (�0.29)

France �0.22** 0.65** �0.01 �0.19 0.70(�3.15) (15.95) (�0.84) (�0.91)

Germany �0.35** 0.59** �0.01 �0.29 0.54(�5.57) (11.08) (�0.53) (�0.94)

Italy �0.20** 0.48** 0.003 �0.23* 0.47(3.34) (9.70) (0.50) (�1.64)

Netherlands �0.87** 0.81** 0.008 �0.09 0.71(�10.80) (17.08) (0.85) (�0.37)

Spain 0.39** �0.15 0.002 0.333 0.09(2.33) (�1.09) (0.04) (0.75)

Sweden �0.46** 0.59** �0.02 �0.009 0.55(�4.06) (11.51) (�0.95) (�0.04)

UK �0.31** 0.39** 0.11** �0.26 0.41(�4.21) (7.72) (3.40) (�0.91)

Note: Significance at the 5% level (two-tailed test).Significance at the 10% level (two-tailed test).

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 1

8:41

17

Nov

embe

r 20

14

Documents

Sectoral employment effects of trade and productivity in Europe