migration the controversies and the evidence

MigrationThe Controversies and the Evidence

This volume takes a critical look at the current divide over immigrationpolicies. It hopes to shed new light on the debate by bringing togetherpapers that investigate the link between trade and factor mobility, partic-ularly labour migration, from theoretical and empirical perspectives. Itexamines the substitutability between trade and migration, the impact ofregional integration on the location of economic activity, the role of publicgoods provisions, and the political economy of migration. Several papersquantify the link between trade, trade policies, migration and income dis-tribution in sending and receiving nations using econometric methods andgeneral equilibrium simulations. Case studies of past and present migra-tion episodes are also presented: the impact of NAFTA on migratorypressure and wage gaps; the trade–migration links between Eastern andWestern Europe; and the historical experience with migration flows in thenineteenth century.

is Research Director at the Centro Studi Lucad’Angliano and Professor of Economics at the Università di Brescia. Heis a Research Fellow in the International Trade Programme of CEPR.

is Professor of Economics at the University of Genevaand Professeur Invité at CERDU (Université d’Auvergne). He has workedat the research department of the World Bank (1980–93) and has taughtat Georgetown University. He is a Research Fellow in the InternationalTrade Programme of CEPR.

. is Professor of Economics and Director of theInstitute for the Study of Labor (IZA). He is Co-Director of CEPR’sLabour Economics research programme.

Centre for Economic Policy Research

The Centre for Economic Policy Research is a network of over 400 ResearchFellows, based primarily in European universities. The Centre coordinates itsFellows’ research activities and communicates their results to the public and privatesectors. CEPR is an entrepreneur, developing research initiatives with the produc-ers, consumers and sponsors of research. Established in 1983, CEPR is a Europeaneconomics research organisation with uniquely wide-ranging scope and activities.

CEPR is a registered educational charity. Institutional (core) finance for theCentre is provided by major grants from the Economic and Social ResearchCouncil, under which an ESRC Resource Centre operates within CEPR; the EsméeFairbairn Charitable Trust; the Bank of England; the European Central Bank andthe Bank for International Settlements; 22 national central banks and 43 compa-nies. None of these organisations give prior review to the Centre’s publications, nordo they necessarily endorse the views expressed therein.

The Centre is pluralist and non-partisan, bringing economic research to bear onthe analysis of medium- and long-run policy questions. CEPR research may includeviews on policy, but the Executive Committee of the Centre does not give priorreview to its publications, and the Centre takes no institutional policy positions. Theopinions expressed in this report are those of the authors and not those of theCentre for Economic Policy Research

Executive Committee

Chairman Anthony LoehnisCo-Chairman Guillermo de la Dehesa

Jan Krysztof Bielecki Peter MiddletonDiane Coyle Bridget RosewellQuentin Davies Mario SarcinelliDenis Gromb Kermit SchoenholtzPhilippe Lagayette Bernard Dewe Matthews

Officers

President Richard PortesChief Executive Officer Stephen YeoResearch Director Mathias Dewatripont

90–98 Groswell Road 7 March 1999London EC1V 7RRTel: (44 171) 878 2900Fax: (44 171) 878 2999Email: [email protected]

The Centro Studi Luca d’Agliano

Luca d’Agliano was born in Turin, Italy, in 1961. After obtaining his Bachelor ofArts degree at Cambridge University he was admitted to St Antony’s College,Oxford University. Under the guidance of Amartya Sen, he studied questions relat-ing to developing countries and welfare economics. He died in a road accident in1984.

The Centro Studi Luca d’Agliano was founded in Turin in 1986 in Luca’smemory. Its main activities, in keeping with an approach similar to his, compriseresearch in development economics, with particular attention to the conflictbetween efficiency and equity and the fight against poverty.

MigrationThe Controversies and the Evidence

Edited by

RICCARDO FAINI

JAIME DE MELO

and

KLAUS F. ZIMMERMANN

The Pitt Building, Trumpington Street, Cambridge, United Kingdom

The Edinburgh Building, Cambridge CB2 2RU, UK http://www.cup.cam.ac.uk40 West 20th Street, New York, NY 10011-4211, USA http://www.cup.org10 Stamford Road, Oakleigh, Melbourne 3166, Australia

© Centre for Economic Policy Research 1999

This book is in copyright. Subject to statutory exception and to the provisions ofrelevant collective licensing agreements, no reproduction of any part may takeplace without the written permission of Cambridge University Press.

First published 1999

Printed in the United Kingdom at the University Press, Cambridge

Typeset in Monotype Times 10/12 []

A catalogue record for this book is available from the British Library

Library of Congress cataloguing in publication data

Migration : the controversies and the evidence / edited byRiccardo Faini, Jaime De Melo, and Klaus F. Zimmermann.

p. cm.‘Papers presented at a conference on “Trade and factor mobility”,

held in Venice on 24–25 January 1997’–Frwd.Includes index.ISBN 0-521-66233-8 (hc.)1. Free trade–Econometric models–Congresses. 2. Emigration and

immigration–Economic aspects–Econometric models–Congresses.3. Labor mobility–Econometric models–Congresses. 4. Free trade–Europe–Econometric models–Congresses. 5. Europe–Emigration andimmigration–Econometric models–Congresses. 6. Labor mobility–Europe–Econometric models–Congresses. 7. International trade–Econometric models–Congresses. I. Faini, Riccardo. II. De Melo,Jaime. III. Zimmermann, Klaus F.HF1703.T7 1999382–dc21 98-55715 CIP

ISBN 0 521 66233 8 hardback

Contents

List of figures page xiList of tables xiiiForeword xvAcknowledgements xviList of conference participants xvii

21 Trade and migration: an introduction 1Riccardo Faini, Jaime de Melo and Klaus F. Zimmermann1 Trade and migration: the main trends 22 The controversies 53 Insights from theory 84 Quantifying the link between trade and migration 125 Historical and contemporary evidence 156 Conclusions and policy implications 17

PART ONE: INSIGHTS FROM THEORY

22 Trade liberalisation and factor mobility: an overview 23Anthony J. Venables1 Introduction 232 Competitive models 253 Increasing returns and cumulative causation 344 Concluding comments 45Appendix 45Discussion 48André Sapir

23 Regional integration, trade and migration: are demand linkages relevant in Europe? 51Rodney D. Ludema and Ian Wooton1 Introduction 51

vii

2 The model 533 Labour supply 574 Equilibrium 595 Economic integration 626 Conclusions 67Discussion 68Giorgio Basevi1 Introduction 682 The first issue: stable and unstable equilibria 693 The second issue: different initial countries’ sizes 744 Duality between trade and factor movements 745 Workers’ preferences dependent on the outcome 75

24 Beyond international factor movements: cultural preferences,endogenous policies and the migration of people: an overview 76Arye L. Hillman and Avi Weiss1 Policy asymmetries 772 Cultural preferences and immigration policy 803 Models of the migration of people 834 Endogenous immigration policy 855 Conclusions 88Discussion 91Francesco Daveri

25 Trade liberalisation and public-good provision: migration-promoting or migration-deterring? 94Konstantine Gatsios, Panos Hatzipanayotou and Michael S. Michael1 Introduction 942 The model 963 Import restriction and international migration 994 Import restrictions and public-good provision 1035 Conclusions 106Appendix 1 107Appendix 2 108Discussion 112Ignazio Musu

PART TWO: QUANTIFYING THE LINKS BETWEEN TRADE AND MIGRATION

26 Trade and migration: a production-theory approach 117Ulrich Kohli

viii Contents

1 Introduction 1172 The production-theory approach to modelling the demand

for imports and foreign-labour services 1213 Comparative statics 1244 Empirical implementation 1305 Empirical results 1336 Total factor productivity 1407 Conclusions 143Discussion 147Marzio Galeotti

27 Migration, dual labour markets and social welfare in a small open economy 151Tobias Müller1 Introduction 1512 Social welfare effects of migration and protection 1533 A model of dual labour markets and discrimination 1554 Labour market effects of immigration and protection 1585 Immigration, protection and social welfare: a simulation

analysis 1626 Conclusions 174Appendix 1751 The simulation model 1752 Data and calibration 176Discussion 185Rudolf Winter-Ebmer

28 Globalisation and migratory pressures from developing countries:a simulation analysis 190Riccardo Faini, Jean-Marie Grether and Jaime de Melo1 Introduction 1902 A Ricardo–Viner model of migration 1933 Decomposing the effects of globalisation on the supply of

migrants for two archetype economies 1974 Simulation results 2025 Sensitivity analysis 2086 Conclusions 211Appendix 2121 The simulation model 2122 The data 217Discussion 221Alessandra Venturini

Contents ix

PART THREE: HISTORICAL AND CONTEMPORARY EVIDENCE

29 Were trade and factor mobility substitutes in history? 227William J. Collins, Kevin O’Rourke and Jeffrey G. Williamson1 Introduction 2272 Theory 2283 Time series: macro-instability and the long swing,

1870–1940 2334 Panel data from history: the Atlantic community,

1870–1940 2445 Political-economy connections: tariffs and immigration

restrictions 2486 History’s bottom line 252Appendix 253Discussion 260Gianni Toniolo

10 Liberalisation and incentives for labour migration: theory with applications to NAFTA 263James R. Markusen and Steven Zahniser1 Introduction 2632 NAFTA 2663 Investment liberalisation and income distribution: I – the

Feenstra–Hanson model 2734 Investment liberalisation and income distribution: II – the

Markusen–Venables model 2805 Technology and the maize sector 2846 Summary and conclusions 290Discussion 294Pasquale M. Sgro

11 East–West trade and migration: the Austro-German case 296Rudolf Winter-Ebmer and Klaus F. Zimmermann1 Introduction 2962 Trade, migration and the labour market consequences 2973 East–West flows of goods and people 3044 Effects on the labour markets 3145 Conclusions 324Discussion 327Marina Schenkel

Index 329

x Contents

Figures

2.1 Incentives for factor mobility page 262.2 Mobility dynamics for the sectorally mobile factor and one

specific factor 302.3 Parameters of the model 312.4 Two sector-specific factors internationally mobile and factor

A immobile 322.5 Lower price of intermediates 382.6 x-sector B-intensive 412.7 Steep FPE set 432.8 Trade barriers lower 433.1 KK, labour demand schedule 563.2 MM, distribution of preferences 583.3 LL, labour supply schedule 593.4 Equilibrium allocations 603.5 Changing trade costs 643.6 Trade liberalisation and location of industry 65D3.1 Supply of labour function horizontal 70D3.2 Supply of labour function positively-sloped 726.1 Input quantities 1196.2 Gross output and aggregate input 1206.3 Total factor productivity 1417.1 Impact of immigration on native employment: the

‘melting-pot’ case 1597.2 Impact of immigration on native employment: the

‘guest-worker’ case 1617.3 Impact of different immigration levels on native social

welfare: the ‘melting-pot’ and ‘guest-worker’ cases 1697.4 Impact of different tariff levels on social welfare 1717A.1 Structure of nested production functions in the simulation

model 175

xi

8.1 Labour-allocation decision 1959.1 Long swings in trade and factor flows: USA, 1872–1913 2359.2 Long swings in trade and factor flows: USA, 1920–39 2359.3 Long swings in trade and factor flows: UK, 1870–1913 2369.4 Long swings in trade and factor flows: UK, 1920–37 23610.1 Investment liberalisation and wage gaps: the Feenstra–

Hanson model 28010.2 Investment liberalisation and wage gaps: the Markusen–

Venables model (activity shifts) 28210.3 Investment liberalisation and wage gaps: the Markusen–

Venables model (factor-price effects) 28310.4 Competitive model with multiple techniques of agricultural

production 28510.5 Competitive model with multiple techniques of agricultural

production: trade liberalisation 28610.6 Specific-factors model with an outside opportunity 28810.7 Model with a public intermediate good: public infrastructure 28911.1 Immigration and trade, 1986–96 30711.2 East–West migration and trade: Germany, 1986–94; Austria,

1990–4 308

xii List of figures

Tables

1.1 Foreign population, early 1980s and early 1990s page 31.2 Intercontinental migration, 1871–1913 41.3 European integration and migration, 1960 and 1970 56.1 Input data – summary statistics 1336.2 Parameter estimates 1346.3 Allen–Uzawa elasticities of substitution, selected years 1356.4 Hicksian elasticities of complementarity, selected years 1356.5 Price elasticities of input demand (cost-function setting) 1366.6 Price elasticities of input demand (production-function setting) 1366.7 Price and quantity elasticities (GNP-function setting) 1376.8 Price and quantity elasticities (immigration-quotas setting) 1386.9 Price and quantity elasticities (variable-resident-employment

setting) 1397.1 Policy experiments and labour market specifications 1647.2 Impact of immigration and protection on native social

welfare 1667.3 Impact of immigration and protection on output and the

labour market 1677.4 Social welfare effects of immigration: sensitivity analysis 1727.5 Education level of the work force in Switzerland, by

nationality 1747A.1 Equations of the simulation model 1777A.2 Variables of the simulation model 1797A.3 Employment, Switzerland, 1985 1817A.4 Parameters of the simulation model 1828.1 Structure of production, demand and factor allocation 1998.2 Factor–household mapping 2008.3 Policy experiments 2018.4 Macro results: labour supply and home–foreign-labour

allocation 203

xiii

8.5 Household incomes and labour-allocation decisions 2058.6 Micro results: net price and sectoral output shifts 2078.7 Sensitivity analysis: labour supply and home–foreign-labour

allocation 2098A.1 Model equations 2148A.2 Variables and parameters 2168A.3 Output and final demand structure 2178A.4 Factor shares in value-added 2189.1 Factor mobility correlates with trade, 1870–1913 and 1919–36 2379.2 Trade and factor mobility, time-series complements or

substitutes? 2429.3 Trade and factor mobility correlations: decade averaged

panel data for the Atlantic economy, 1870–1940 2479.4 The determinants of immigration policy, c.1860–1930 25110.1 Legal immigration to the USA by persons born in Mexico

and deportations of Mexicans, FY 1981–95 26410.2 Canadian, Mexican and US civilian employment, by

economic activity, 1994 26710.3 Comparison of US and Mexican employment structures,

1988 26810.4 Canadian, Mexican and US maize production, 1992–6 26910.5 Population, median age and population growth, NAFTA

countries, estimates and projections, 1970–2025 27110.6 Selected economic statistics for Mexico, 1980–95 27210.7 Relative wage and wage shares, by Mexican region, 1975–88 27410.8 Average annual compensation, by sector, 1980–93 27510.9 Monthly compensation per worker in Mexican

manufacturing, 1987–95 27610.10 Foreign investment in Mexico, 1980–95 27810.11 Simulation results for the alternative-technologies model 28711.1 Empirical studies on the effects of trade and migration on the

labour market in Austria and Germany, 1990–7 29911.2 Economic development in Central and Eastern Europe,

1991–7 30511.3 Labour, migration and trade, Austria and Germany, 1986–95 30911.4 Foreign direct investment, 1989–95 31211.5 Employment growth, Austria, 1987–94 31511.6 Native employment growth, Austria, 1987–94 31611.7 Wage growth, Austria, 1987–94 31711.8 Employment growth, Germany, 1986–94 31911.9 Native employment growth, Germany, 1986–94 32011.10 Wage growth, Germany, 1986–94 321

xiv List of tables

Foreword

Trade and migration have often been considered as substitutes. Economictheory treats international trade as the result of differences in factor endow-ments. Goods thus embody capital and labour, and trade in goods is thus indi-rectly trade in factors such as labour. Trade in goods embodying labour is thusa substitute for the movement of labour itself. A similar perspective informspolicy discussions concerning migration. Increased trade, by fostering thegrowth of income and wages in poorer countries, can reduce the incentive formigration from the poorer to the richer countries. The historical evidenceseems more mixed, however. The rapid growth of trade which took place theturn of the century, and again in the post-war period, was accompanied bymore, not less, migration. In the 1980s and 1990s, on the other hand, tradegrew rapidly without any corresponding growth of migration.

The CEPR has played an active role in the economic analysis of Europeanmigrations, beginning with an exploratory workshop in 1985, organised byLouka Katseli, followed by research networks in the late 1990s, funded bythe European Commission’s SPES and HCM Programmes.

This volume draws on the work of these networks, and in particular onthe papers presented at a conference on ‘Trade and Factor Mobility’, heldin Venice on 24–25 January 1997. The conference was organised as part ofthe Centre’s research programme on ‘European Migration: FromEconomic Analysis to Policy Response’, led by Klaus F. Zimmermann andsupported by the European Commission’s Human Capital and MobilityProgramme. The Centro Studi Luca d’Agliano provided generous financialsupport, as did the the Dipartmento di Scienze Economiche at theUniversità di Venezia.

We are particularly grateful to Riccardo Faini, Jaime de Melo and KlausF. Zimmermann for their efforts in organising the conference and preparingthe papers presented at the conference for publication.Stephen YeoChief Executive Officer 7 March 1999

xv

Acknowledgements

The editors and publishers acknowledge with thanks permission fromNBER to reproduce Table 10.7, from R.C. Feenstra and G.H. Hanson,‘Foreign Direct Investment and Relative Wages: Evidence from Mexico’sMaquiladoras’ (1995).

xvi

List of Conference Participants

Giorgio Barba Navaretti, Centro Studi Luca d’AglianoGiorgio Basevi, Università di BolognaCarlo Carraro, Università degli Studi di Venezia and CEPRWilliam Collins, Harvard UniversityFrancesco Daveri, Università di BresciaJaime de Melo, Université de Genève and CEPRGil Epstein, Bar-Ilan University, IsraelRiccardo Faini, Università degli Studi di Brescia and CEPRMarzio Galeotti, Università degli Studi di BergamoJean-Marie Grether, Université de GenèvePanos Hatzipanayotou, Aristotelian University of ThessalonikiArye Hillman, Bar-Ilan University, IsraelUlrich Kohli, Université de GenèveJames Markusen, Universitat Pompeu Fabra and CEPRIgnazio Musu, Università Ca’Foscari di VeneziaSimona Orlano, Centro Studi Luca d’AglianoAndré Sapir, ECARE, Université Libre de Bruxelles and CEPRMarina Schenkel, Università degli Studi di VeneziaPasquale Sgro, Deakin UniversityGianni Toniolo, II Università di Roma, ‘Tor Vergata’ and CEPRHeinrich Ursprung, Universität KonstanzAnthony Venables, London School of Economics and CEPRAlessandra Venturini, Università degli Studi di BergamoJeffrey Williamson, Harvard UniversityRudolf Winter-Ebmer, University of Linz and CEPRIan Wooton, University of Glasgow and CEPRKlaus F. Zimmermann, Institute for the Study of Labor (IZA)

xvii

1 Trade and migration: anintroduction1

RICCARDO FAINI, JAIME DE MELO AND

KLAUS F. ZIMMERMANN

International migration is the absentee in the current wave of globalisation,particularly in Europe. Helped by falling communication and transportationcosts and by the reduction in policy barriers to commodity and capital flows,trade flows and foreign direct investment (FDI) have increased in the last 20years at a faster rate than world production. Migration flows, on the otherhand, have shown little change during the same period when one excludes thetemporary surge following the collapse of the socialist regimes in EasternEurope. This contrasts sharply with previous integration episodes: in thenineteenth and early parts of this century, but also in the 1960s, internationallabour mobility played a central role in fostering economic integration.

The changing stance towards migration policies goes a long way inexplaining these trends. At the turn of the century, the attitude towardimmigration used to be quite liberal. Similarly, in the 1960s governments inreceiving countries often took an active role in encouraging migration.Nowadays, the policy imperative has become to limit or even to stop anyfurther immigration. In part, this new attitude reflects the fears thatimmigration may worsen the domestic income distribution by widening theskilled–unskilled wage gap and aggravate unemployment.2 There is littleevidence, however, that these concerns are well founded. Nonetheless,recent popular thinking in receiving countries has it that migration is exces-sive3 and therefore detrimental to the welfare of natives, and that thissomehow provides a reason for highly restrictive policies. Clearly, the neg-ative stance toward immigration reflects more than simple economic con-cerns. Those opposing immigration fear that it may exacerbate socialtensions and blur national identities in host countries as well as aggravatedomestic economic problems.

Pressures to tighten immigration laws have been quite strong, particu-larly in Europe. However, immigration policies remain a highly divisiveissue in many receiving countries. Attempts to tighten such policies havetypically led to bitter conflicts among domestic constituencies. They also

1

irritate relationships with sending countries that rely on emigration toalleviate structural imbalances in their labour markets and earn valuableforeign exchange through workers’ remittances. Finally, and perhaps morecrucially, immigration controls have so far proved to be quite ineffective instemming undesired population inflows.

Are there more palatable alternatives for migration policies ? If trying tocontrol the symptoms does not work, treating the problem directly by pro-moting equitable and sustained growth in origin countries might be moreeffective. Trade integration seems a particularly commendable strategy toalleviate migration pressure, for at least two reasons. First, trading goodsrepresents a way to exchange the services of the factors embodied in thosegoods. To the extent that barriers to trade are eliminated and commoditytrade increases, the exchange of factor services will also increase and there-fore the incentive for factors to move should diminish, in which case, asshown by Mundell (1957), trade in goods and the international mobility offactors are ‘substitutes’. Second, deeper trade integration is often advocatedas a means to achieve faster convergence between countries with differentincome levels. The EU experience, where poorer regions have been rapidlycatching up with relatively better-off regions, is often cited as evidence (Ben-David, 1993). US and, to a lesser extent, EU policy-makers seem to be con-vinced by these arguments and have negotiated integration agreements withtheir relatively poorer neighbours. Yet, both theoretical and empirical evi-dence on the effectiveness of trade integration is far from being conclusive.

In sum, the debate on immigration is still occupying centre stage, and isthe subject of much research. This volume hopes to shed new light on thisdebate by bringing together studies that investigate the link between tradeand factor mobility, particularly labour migration, both from theoreticaland empirical points of view. In this Introduction, we highlight the mainissues and controversies, describing the new evidence brought by the chap-ters in the volume. Section 1 gives evidence on the importance of interna-tional migration. Section 2 briefly introduces the debate and surroundingcontroversies indicating in passing the dimensions that go beyond thetrade–migration link. The remaining sections introduce the contributions inthe volume. Section 3 discusses theoretical contributions on the linksbetween trade and migration, while section 4 indicates attempts to modelthese links and section 5 discusses the evidence from historical and contem-porary episodes. Section 6 draws some conclusions for policy implications.

1 Trade and migration : the main trends

Trade flows have increased at a systematically faster rate than productionduring most of the post-war era. For example, during the 1980s, world trade

2 Riccardo Faini et al.

increased at 4.1 per cent per annum while output grew at 3.1 per cent. Thispattern has been even stronger in the 1990s with trade growing at 9.2 percent in 1994–5 and output at 2.9 per cent. Growth in FDI was even stronger.According to UNCTAD, FDI growth between 1981 and 1993 was almostdouble that of exports. Multinational corporations now account for two-thirds of world trade.

This expansion in trade and foreign investment largely reflects an increas-ingly liberal trade regime brought about by the successful negotiations ofthe Uruguay Round that achieved significant tariff reductions, particularlyin those countries where tariffs were initially high (Finger et al., 1996).More importantly it extended the rules of GATT and of the WTO to hith-erto exempted or excluded sectors such as textiles, agriculture and services.Similarly, the policy climate for foreign investment has become steadilymore liberal. An UNCTAD survey of some 60 countries shows that out of212 legislative changes during 1993–4, 209 went in the direction of a moreliberal FDI framework (UNCTAD, 1995).

International labour flows, however, have not followed suit. Considertable 1.1 that provides data for the 1980s. The share of foreigners in thepopulation barely changed during the decade in most European countriesand Canada, though it registered a significant increase in the USA. This isnot because the propensity to migrate has declined. Quite the opposite:income and wage differentials between rich and poor countries have, if any-thing, increased during the period. Moreover, falling communication andtransportation costs mean that residents in sending countries are increas-ingly well informed about economic conditions in receiving countries andcan afford more easily than in the past the cost of migrating.

Trade and migration 3

Table 1.1. Foreign population, early1980s and early 1990s (per cent of totalpopulation)

Early 1980s Early 1990s

Belgium 29.0 29.1France 26.8 26.3Germany 27.4 28.5Italy 20.7 21.7UK 22.8 23.5USA 26.2 27.9Canada 16.1 16.1

Source: SOPEMI (1995).

The secular fall in mobility costs highlights the contrast between con-temporary migration trends and the events at the turn of the century. Weknow that both trade and capital flows were instrumental in boostingintegration among national economies in the late nineteenth century(Krugman, 1995; Sachs and Warner, 1995). Between 1870 and 1913, exportflows increased at an average annual rate of more than 4 per cent, whereasthe rate of growth of GDP stood ‘only’ at 2.6 per cent. Unsurprisingly, in1913 exports accounted for a substantially larger share of GDP than in 1870(Krugman, 1995). The pattern is not very different from what we observetoday. However, the growth in trade was then accompanied by a surge inmigration flows. Table 1.2a shows that between 1871 and 1880 and 1901 and1910 the population share of emigration rose in many sending countries.The increase was particularly pronounced in the ‘new’ emigration countries,such as Italy, where the population share of emigrants soared from 0.1 percent to 1.6 per cent, and in Spain. The pervasive role of migration is evenmore apparent if we focus on host countries. Between 1871 and 1880,average annual immigration was equal to 0.5 per cent of the domesticpopulation in both the USA and Canada and to slightly more than 1 percent in Argentina. 30 years later, these figures had doubled or even trebled.At its peak in 1913, immigration was equal to 3.8 per cent of Argentina’sand Canada’s population, and 1.2 per cent of the USA’s (table 1.2b).


Table 1.2. Intercontinental migration, 1871–1913

a European emigration (per cent of home-country population)

1871–80 1901–10 1913

Austria-Hungary 0.03 0.48 0.61British Isles 0.50 0.65 1.04Germany 0.15 0.05 0.04Ireland 0.67 0.70 0.68Italy 0.11 1.08 1.63Spain 0.36 0.57 1.05

b Immigration from Europe (per cent of the host country population)

1871–80 1901–10 1913

USA 0.55 1.02 1.22Canada 0.55 1.68 3.84Argentina 1.17 2.92 3.83

Source: Ferenczi and Willcox (1934).

We find a similar picture when we focus on the more recent integrationepisode in Europe during the 1960s, during which the share of intra-ECtrade for the EC-6 rose rapidly (table 1.3a). This growth in intra-EC tradewas accompanied by rising migration (table 1.3b). Between 1956 and 1973more than 11 million people migrated internationally, mostly fromSouthern to Northern Europe. The share of foreign workers in the labourforce increased substantially from 2 per cent in 1960 to 6 per cent in 1970in Germany, from 6 per cent to 8 per cent in France and from 5 per cent to7 per cent in Belgium.

In sum, the contrast between table 1.1 on the one hand, and tables 1.2and 1.3 on the other is clear: whereas increased trade integration at the turnof the century and in the 1960s was accompanied by increased migration,this was not so during the increased trade integration of the 1980s.

2 The controversies

It is widely recognised that the limited role of international labour mobil-ity largely reflects the policy stance of governments in receiving countries.Left to market forces, thanks to cheaper transportation and improved


Table 1.3. European integration and migration, 1960 and 1970

a Intra-European trade (per cent of imports originating in EC-6)

1960 1970

Belgium 51 59France 32 49Germany 31 44Italy 29 41Netherlands 49 56

b Foreign workers (per cent of labour force)

1960 1970

Belgium 25 27France 26 28Germany 22 26Italy — —Netherlands 21 23

Source: Molle (1990).

information on living conditions abroad, international migration wouldhave most probably registered a massive increase. As mentioned above, thepolicy stance toward migration in receiving countries has undergone aradical change beginning in the mid-1970s. Whereas in the 1950s and 1960s,governments in Northern Europe sought to encourage immigration andwere actively recruiting workers from Southern Europe and NorthernAfrica, nowadays in most receiving countries the policy imperative hasbecome to halt any further immigration.

Policy choices are, of course, not made in a vacuum. The changing stanceof policy-makers in receiving countries largely reflects the attitudes ofvoters there, in particular the belief that immigration may aggravateunemployment and wage conditions of unskilled workers in Europe andthe USA. This is a hotly debated issue in the press on both sides of theAtlantic. However, evidence for the USA (Friedberg and Hunt, 1995) andfor Europe (Zimmermann, 1995) concludes that immigration has playedvirtually no role in explaining the worsening labour market conditions ofunskilled workers. Yet, voters appear to be undaunted by this evidence andhold an altogether different opinion on these matters. Other elements needto be brought in to understand voters’ attitudes and policy choices.

This contrast between voters’ attitudes and evidence of immigration is allthe more pronounced once one recognises that a more complete evaluationof immigration in receiving countries should extend beyond the (short-run)impact on the labour market. In particular, immigration will have long-runeffects on the demographic structure, and hence on the social securitysystem in the host country. Under the ‘pay-as-you-go’ social securitysystems with contributions pegged to wage income prevailing in mostreceiving countries, immigration may indeed yield a further benefit byslowing population ageing, alleviating the dependency burden and allow-ing a reduction in social security contributions, with a favourable impacton net wages of natives (see Hillman and Weiss, chapter 4 in this volume).However, demographic projections indicate that immigration needs toincrease to unprecedented levels to make a dent in the social security prob-lems of receiving countries (Stalker, 1994).

The contrast between the predictions of standard economic theorisingand the evidence on the one hand and citizen attitudes on the other is con-troversial and puzzling. At a theoretical level, utilitarianism cannot justifyrestrictive immigration quotas. Other elements must be taken into account.An obvious extension to the standard utility-maximisation framework is torecognise that voters’ welfare includes other arguments than income (e.g.social capital or ‘cultural cohesiveness’ provided by a relatively homogene-ous population, as recognised by Schiff, 1997, for example). Then immigra-tion in receiving countries can be welfare-reducing owing to the disutility


from ‘excessive’ immigration, and likewise for sending countries. It is alsopossible that immigration affects the consumption of public goods. It isconceivable that natives perceive that immigrants have relatively intensiveconsumption patterns of public goods like education, and lower theirquality of life.

How to reduce immigration in receiving countries is also controversial.In the post-1974 period, governments tended to rely mostly on border andpopulation controls (Collinson, 1993). Yet, as shown by the sequence ofimmigration regularisation programmes, invariably implemented togetherwith the promise to apply tight immigration limits, controls on populationflows are largely ineffective. Porous borders, the reluctance to impose sti-fling domestic controls in a democratic society and the pull effect of labourdemand from labour-intensive sectors in receiving countries have combinedto undermine the effectiveness of direct controls and have resulted in asteady flow of immigration.

Alternative indirect approaches to limiting migration have been sought.These have revolved around the alternative ways policies can affect the tradelinks between sending and receiving countries. One approach has focused onencouraging capital flows and directing aid toward emigration countries(Lucas, 1995, ILO–UNCHR, 1992). However, (private) capital flowstowards sending countries cannot be easily influenced by policy. Moreover,there is little evidence that even unrestricted capital mobility may result infalling migration. Lucas’ (1990) computations suggest that substantial wagedifferentials across countries are compatible with free capital mobility.Similarly, aid flows are probably too small to affect the economies in mostemigration countries (Albania may be an exception). Moreover, targetingaid toward sending countries, most of which are in the middle-income range,may conflict with the ultimate motivations of aid policy.

Under these circumstances, trade policy may represent a more effectivestrategy to deal with migratory pressures. It holds some clear advantages.First, and most obviously, trade policy is to a large extent under the controlof policy-makers. It is largely recognised that both the USA and the EUhave concluded regional integration agreements (RIAs) with their relativelypoorer neighbours to stem migratory pressures. The NAFTA treaty in par-ticular was explicitly targeted, at least by Mexican officials, at reducingmigratory pressure. In former President Salinas’ words: ‘Mexico wants toexport goods, not people.’ Similarly, the EU’s activism in concluding tradeagreements both with Northern African and Eastern European countriesis typically attributed to the desire of limiting migration. Second, but morecontroversial, are the effects of trade liberalisation: the increases inefficiency should decrease migratory pressures, both in sending and inreceiving countries. In sending countries, higher incomes should lower


outward migratory pressures, especially if unskilled-labour-intensive indus-tries are the main beneficiaries. In receiving countries, trade liberalisationshould reduce the demand for unskilled labour, thereby reducing thedemand-pull effect on migrants.

The effectiveness of trade liberalisation as a way to alleviate migrationpressure is not, however, unquestioned. First, theory does not provide anunambiguous answer on the link between trade and migration. Second, asmentioned earlier, historical evidence seems to suggest that periods ofgreater trade integration were also characterised by large population flows.From these episodes, however, we cannot infer what would have happenedto migration in the absence of the drive toward greater trade integration. Inthe nineteenth century, in particular, trade and mobility costs fell sharply.Similarly, in the 1960s, governments in host countries were actively pro-moting both trade and labour immigration. It is not surprising then to findthat in both episodes trade and migration flows moved in the same direc-tion and showed a concomitant increase. But this does not mean that tradeliberalisation by itself provided a boost to migration. This example illus-trates the difficulties of inferring the link between trade and migration fromsheer observations of the data. Empirical evidence must be accompaniedby theory to help identify the main factors at work in shaping the relation-ship between trade and factor mobility.

3 Insights from theory

In the Heckscher–Ohlin (HO) model, commodity trade arises becausenations have different factor endowments. If factor mobility leads to areduction in such differences it will undermine the basis for trade. In this set-up, therefore, trade and factor movements are substitutes. In the standardmodel with two factors, two goods and two diversified countries with perfectcompetition and identical technologies shared by both countries, if labourmigrates from the labour-abundant country, the production of the export-able good will fall and the output of the importable sector will increase alongtraditional Rybczynski lines. On both counts, trade will decline. Similarly,in the labour-scarce country, the migration of labour will boost productionof the labour-intensive importable good and lead to a decline in the outputof the exportable good, and hence to a reduction in trade. It is therefore clearthat trade liberalisation should reduce incentives to migrate.

Adding internationally mobile capital need not change the basic predic-tions of the HO model. Consider indeed a simple set-up with three factors –capital, skilled and unskilled labour. Suppose that capital is fully mobile, butskilled and unskilled labour migrations are restricted. Then, trade liberalisa-tion in skill-abundant countries will reduce the incentive for unskilled-labour


immigration. Once again, therefore, commodity trade and labour mobilityare substitutes.

Yet, this HO prediction is not shared by all trade models, even those wheretrade is based on differences in factor endowments. As the contributions byVenables and by Markusen and Zahniser (chapters 2 and 10 in this volume)show, the link between trade and migration can be complex. Rather thansummarising these two contributions, consider the following approachesthat yield results opposite to the simple HO model outlined above.

(1) Modified HO models – With more than two factors, the impact of tradeliberalisation on the incentive for factor mobility is generally ambigu-ous. Even with two factors, though, trade liberalisation may not lead toa convergence in factor prices, unless we rule out the possibility offactor-intensity reversal. Otherwise, with the countries having oppositefactor-intensity ranking of the two goods, convergence in commodityprices brought by trade liberalisation will lead relative factor prices tomove in the same direction in both countries. Lower barriers to goodstrade could as a result cause factor prices to diverge and would there-fore enhance the incentive for factors to move.

(2) Ricardian models – Suppose that countries have the same endowmentbut different technologies. Suppose furthermore that the ‘rich’ countryhas a more productive technology in the labour-intensive sector. It willthen export the labour-intensive good and, in the absence of factormobility, will enjoy a higher wage. If barriers to factor mobility areeliminated, labour will migrate from the ‘poor’ to the ‘rich’ country.The increase in the labour–capital ratio will, again through standardRybczynski effects, strengthen the specialisation of the rich country inthe labour-intensive sector and lead to more trade, opposite to the HOprediction.

(3) Specific-factors models – In this set-up, some factors (say, labour) arefully mobile across sectors, while other factors (say, land or capital) aretied to a given sector. By lowering the price of the importable good,trade liberalisation will depress the demand for labour (i.e. the mobilefactor) in the import sector and lower both the nominal wage and thereal wage in terms of the exportable good (whose price is unchanged).Real wages will, however, increase if expressed in terms of the import-able good. Whether real wages as a whole increase or not will thendepend on the weight of importable goods in the consumption basket.The effect of trade liberalisation on the incentive to migrate cannottherefore be determined a priori.

(4) Financial-constraint models – Consider a simple case where labourmigration entails a monetary cost. With imperfect capital markets,


some would-be migrants may be financially constrained and thereforeunable to migrate to the desired destination. Suppose now that tradeliberalisation in the labour-abundant sending country raises the wagethere, along traditional HO lines. This will relax the financial constrainton would-be migrants and may, somewhat paradoxically, increasemigration (Lopez and Schiff, 1995). In a related paper, Schiff (1997)shows that migration costs are often substantial and may thereforeimpose a severe constraint on would-be migrants. The strength of thisargument is boosted by the observation that, for some (low-)incomeranges, migration is an increasing function of the wage in the origincountry (Banerjee and Kanbur, 1981; Faini and Venturini, 1993).

(5) ‘New’ trade theory – With increasing returns and monopolistic competi-tion, the reduction in trade costs will not generally lead to factor priceconvergence. Consider, for instance, a standard model in the ‘new’ tradetheory with one factor (say, labour), one constant-returns-to-scalesector producing a homogeneous good and one increasing-returns-to-scale sector producing a set of differentiated goods. In the pre-tradeequilibrium, the wage in each country is pinned down by the assump-tion of constant returns to scale in the homogeneous good sector.Wages in the two countries are therefore equal. When trade opens,however, it is entirely possible that one country will specialise in theincreasing-returns-to-scale sector. Wages may therefore diverge andlabour will have an incentive to move.

Which type of model is relevant will, of course, depend on the character-istics of the economies that are integrating. For example, based on thecontributions in this volume, one could argue that HO-based models mightbe the most relevant approximation to study the factor-movement incen-tives of NAFTA, and monopolistic competition models to study regionalintegration in Europe. It is hard to isolate the driving forces in each modelbecause they are typically drawn along several dimensions. It is thecontribution of Venables’ lucid survey in chapter 2 to show how trade liber-alisation affects the incentives for factor mobility in a unified framework forall the type of models described above (except Ricardian models and factor-endowment models that emphasise financial constraints to migration).

Venables’ technique is to consider a two-country world in which coun-tries differ only in their endowments of each of two factors with (exoge-nous) impediments to factor and goods trade. He then studies how,starting from an initial equilibrium with no incentives for further factormobility, goods-trade liberalisation affects the incentives for factor mobil-ity. Rather than generating new results, his contribution is to show the richset of outcomes for substitutability/complementarity relations among


factors of production in the most widely used trade models: HO, specific-factors and monopolistic competition.

A somewhat different, but related, issue is whether trade liberalisationwill foster convergence among regions or instead will increase theattractiveness of the relatively rich areas. With positive transport costs and(localised) increasing returns to scale, two opposite forces determine thelocation of economic activity. First, there is an incentive to concentrate pro-duction in only one region to exploit scale economies. However, theconcentration of production in one region would push up wages there andwould also imply higher transportation costs to meet the demand in thebackward region. A reduction in trade costs could make it easier for firmsto serve consumers located in the relatively poor region while producing inthe rich region. Moreover, this would not be the end of the story, given thatthe greater attractiveness of the richer region would prompt workers tomigrate there. Migration, in turn, will increase the market size of the moreprosperous region and induce more firms to move there. This cumulativecausation process was analysed by Krugman (1991) and is clearly relevantto assess the relationship between trade liberalisation and migration.

The issue of the agglomerative forces of integration is raised by Venablesand is also picked up by Ludema and Wooton in chapter 3. They developa more realistic set-up where potential migrants are only partially mobile(not, as in Krugman, fully mobile) because of natural barriers such as thepreference for the home country or differences in language and culture. Inthis set-up, a reduction in trade costs resulting from integration is less likelyto lead to a polarised outcome. Ludema and Wooton show the intriguingpossibility that, for sufficiently mobile labour, progressive integration mayinitially lead to agglomeration, then again to diversification as trade costsare lowered further. This leads them to suggest a sequence of integrationpolicies to avoid temporary dislocations of labour: a restriction on labourmovements during the initial phases of trade liberalisation, with barriers tolabour mobility removed once free trade in manufactures has been estab-lished.

If the link between trade and migration can be complex, it still remainsthat there is a fundamental asymmetry in the policies adopted in receivingcountries: liberal policies in trade and capital exports, both of which dis-advantage low human-capital residents and strict policies against immigra-tion. This puzzling asymmetry is the point of departure of thepolitical-economy perspective taken by Hillman and Weiss in chapter 4. Inthis setting, labour is not apersonal and cultural preferences (a public good)matter in policy choices. Can then jurisdictions with voluntariness of loca-tional choice replicate the efficiency of the market in the provision of thispublic good? Unfortunately not, because typically immigration policies are


not based on discriminatory pricing and people who are willing to pay tobelong to a jurisdiction are excluded.

Hillman and Weiss also review political-economy models of immigrationpolicy that seek to explain why inefficient immigration policies are adopted.For example, in a median-voter framework, it can be shown that the medianvoter could simultaneously choose to ban immigration but that she may atthe same time support tacitly selective illegal migration in selected indus-tries because she would benefit from immigrants’ presence. In their words:‘the “illegality” of illegal immigrants has the effect of transforming immi-grants from mobile factors of production who would compete in domesticlabour markets with the median voter, to sector-specific factors of produc-tion whose presence is beneficial.’ They also show how a public-policy per-spective can help explain differences in immigration policies acrosscountries.

Trade liberalisation may have other non-conventional effects, once oneincludes the effects of migration on the allocation of benefits coming froma public good. Gatsios, Hatzipanayotou and Michael in chapter 5 considera simple trade-theoretic model of a small economy where factor incomesare taxed at a fixed rate and the tax revenue is used to purchase a public-consumption good. They argue that the fall in trade barriers may influencethe provision of public goods through both income and induced migrationeffects. First, any change in trade policy should affect income, leadingtherefore to a change in tax revenues and in the provision of public goods.Second, trade policy will also have an impact on factor prices and therebyon migration and aggregate income. Finally, the unit cost of the publicgood may also be affected, depending on the relative labour intensity of theimportable good and on the impact of (relative) prices and labour endow-ment changes on the cost of producing the public good. There is also afurther channel of interest (that the authors, however, do not consider)where migration itself may depend on the availability of public goods. Thisis clearly an interesting avenue for further research.

4 Quantifying the links between trade and migration

The ambiguities of theoretical predictions highlight the need for both mod-elling (which models conform to the data and how sensitive model predic-tions are to model choice or model closure), and for evidence (from currentand historical episodes). Such a two-pronged approach is needed becausewe are far from being able to give direct evidence that the data support onemodel over another. And even if we were able to choose confidently onemodel over alternatives, not all relevant models could be tested.Furthermore, one would still be left with the possibility that the evidence is


circumstantial, and therefore not relevant for other episodes. The contribu-tions in Part Two represent efforts to fit structural models that correspondclosely to the theoretical models discussed in Part One whereas thecontributions in Part Three rely on historical episodes (the nineteenthcentury, the recent wave of regional integration agreements – NAFTA andEastern Europe) and use reduced-form methods that are often distant fromthe models they purport to represent. The two approaches are clearlycomplementary4

Kohli in chapter 6 studies the impact of immigration on the receivingcountry. Rather than relying on a reduced-form approach, Kohli takes theGNP function approach, where the value of national output is maximisedfor given output prices and factor endowments. His model has four inputs:imports, capital, non-resident labour and resident labour – where residentlabour and capital are the two factors with endogenously determined prices.His approach is well rooted in trade theory, since it corresponds to the stan-dard specification of the price-taking economy. Although he does notspecify a full general equilibrium model, the supply side is complete, whichallows him to specify precisely the sense in which inputs and output are sub-stitutes or complements. And by specifying flexible functional forms, Kohliis able to let the data decide on the nature of the trade–migration linkwithout imposing too much structure on his estimates (beyond the assump-tion that perfect competition prevails in factor and product markets).

Kohli’s estimates are for Switzerland over the period 1950–86, wherenon-resident labour made up to a quarter of the labour supply. He findsthat trade and labour mobility are complements both in the Allen–Uzawasense (i.e. for given input prices) and in the Hicksian sense (i.e. for givenquantities). He estimates that an increase in the supply of non-residentworkers depresses only slightly the wages of resident workers (a 1 per centincrease in non-resident workers reduces the wage of resident workers by0.1 per cent), confirming the findings of small effects of migration on wagesestimated in other settings.

Chapters 7 and 8 use computable general equilibrium models to simulaterespectively the effects of exogenous immigration in receiving countries andthe effects of alternative policy packages to stem the supply of migrants insending countries. Both are calibrated models though they differ in theirtreatment of the labour market and the household-maximisation decision.

Müller in chapter 7 studies the impact of immigration in an efficiency-wage model of a dual labour market in which immigrants work only in thesecondary sector while natives are employed in both sectors. Hence there isdiscrimination against immigrants of the type ‘equal pay for equal work,but unequal work’. His estimates are also for Switzerland. Besides pro-posing a labour market specification applicable in some other European


countries, Müller’s approach allows him to illustrate the trade-off fornatives of an increase in immigration. Whereas in the standard labourmarket specification, an increase in immigration has a very small impact onthe welfare of natives, this is not so with a dual labour market when dis-crimination confines immigrants to the secondary labour market (this casecorresponds to the ‘guest-worker’ system often used in Europe). The rea-soning is simple. Initially, primary-sector wages are kept high to induceworkers there not to shirk. Indeed, if caught shirking, primary-sectorworkers are fired and are forced to take employment in the (low-wage) sec-ondary sector. Consider now the impact of immigration. By assumption,immigrants are confined to the secondary sector, leading to a fall of wagesthere. The decline in secondary-sector wages in turn increases the punish-ment for shirking in the primary sector. Firms can therefore pay lowerwages in the primary sector and employment expands, with a first-ordereffect on welfare.

Müller’s simulations show that, under this dual labour market specifica-tion, the welfare benefits for natives of a given immigration is four timeslarger than in the standard perfectly competitive labour market model. Ofcourse, the improvement in natives’ welfare comes at the cost of increasedinequality among natives. He also shows that protection (defined to yieldthe same pattern of sectoral output levels as under immigration) would bepreferred by natives in terms of social welfare (here defined by Atkinson’ssocial welfare function) to immigration in the absence of labour market seg-mentation in spite of small efficiency losses. On the other hand, with a duallabour market, protection is detrimental for efficiency as secondary-sectorjobs are concentrated in the protected industries. Indeed, protection wouldraise secondary-sector wages and crowd out primary-sector employment.

Müller then explores the degree of inequality-aversion necessary for awelfare-maximising government to choose a no-immigration policy underdifferent closure rules for the labour market. He also examines the fear of‘over-foreignisation’ coming from the perceived failure of the policy of rota-tion aiming at promoting return migration. In that case, immigrantsbecome permanent residents and the government cannot be indifferent totheir welfare. Then one should no longer discriminate against immigrants(as the Swiss government did in 1970 when it improved the legal status ofimmigrants), and Müller computes a welfare estimate of the welfare gainof removing the operation of the guest-worker system for the population ofimmigrants in Switzerland in 1985. Though this was not his focus, hisresults can help explain the change in attitude towards immigration.

Faini, Grether and de Melo in chapter 8 use a standard Ricardo–Viner(RV) model (similar in structure to the model used by Müller except for thelabour market specification) to study the migration decision in sending


countries. In their model, the focus is on household decisions. They assumefour household categories, each facing a labour supply and migration deci-sion. (Because of the assumption of household heterogeneity, migration isonly partial.) This model is then applied to two archetype economies cor-responding to the case of a middle-income and of a low-income economy,respectively. The two archetypes differ because of the different weights ofagricultural production, the skill composition of their labour force andtheir pattern of comparative advantage. The authors confront two strate-gies aimed at reducing migratory pressure: unilateral trade liberalisation (inthe North and the South) versus direct measures (increased transfers to theSouth or increased migration costs) under different model closures (perma-nent versus non-permanent migration, presence or absence of financialconstraints to migration for certain groups, etc.).

In this setting, trade liberalisation is bound to elicit a different migratoryresponse in the two archetype economies and it will be sensitive to modelclosure. For example, in the middle-income economy, the reduction inimport barriers is accompanied by a strong export response and, as a result,leads to an appreciation of the real exchange rate which in turn raises thedomestic real wage and lowers the incentive to migrate. Conversely, in thelow-income country trade liberalisation elicits a relatively weaker exportresponse and brings therefore a depreciation of the real exchange rate. Theincentive for migration increases. Overall, the simulation results suggestthat if policy-makers want to alleviate migration pressure they should liber-alise trade with middle-income rather than with low-income countries.Interestingly enough, this is exactly what the existing North–South regionalintegration agreements seem to be doing.

5 Historical and contemporary evidence

If structural econometric estimates and simulation studies provide theneeded parameter estimates and a feel for orders of magnitude, they are stillconditioned on model choice, and perhaps do not sufficiently confront thedata, at least in the eyes of the sceptics. Carefully chosen historical episodesover a long time span and contemporary case studies of clearly definedpolicy episodes provide a useful complement to the modelling approach. Inother words, the contributions in Part Three ‘let the data speak’.

Collins, O’ Rourke and Williamson in chapter 9 take a close look at thehistorical evidence on the link between trade and migration. They arguethat the direction of this relationship may well depend on the historicalperiod under analysis. Consider, for instance, the New World situation inthe nineteenth century, where an ‘open frontier’ meant that new land couldbe easily brought under cultivation. In this context, a fall in trade costs will


increase the demand for the land-intensive goods in the New World andlead to a frontier extension whose exploitation would most probablyrequire more capital. In this set-up, therefore, trade and capital mobilitymay be complements. As reasonable as this argument sounds, the authorsfail to find much empirical support for it: when they regress total trade onthe absolute values of capital and migration flows, they find that the formervariable has a positive coefficient (which they interpret as denoting comple-mentarity) mainly in the post-frontier period in 1919–36. Overall, theauthors find little indication that capital and migration may have acted assubstitutes for trade. The authors, however, are careful to point that they‘are only exploring correlations’. Even a positive association between, say,trade and migration may mean only that both benefited from the fall intransatlantic transportation costs.

Markusen and Zahniser in chapter 10 look for reasons why the trade andinvestment liberalisation under NAFTA might not lead to the convergencein unskilled wages in the two countries that would be predicted by the HOmodel. Though it is too early to detect the effects of NAFTA, much of theliberalisation that had to be carried out by Mexico had already occurredduring the 1984–94 period during which the wage gap between skilled andunskilled increased in all regions, particularly in the North where liberal-isation had the largest impact. A widening gap also occurred in manufac-turing.

Three different types of models (two of which are rooted in familiarfactor-proportions theory) consistent with the stylised facts of the technol-ogy in manufacturing in the USA and Mexico are found to be consistentwith a widening gap following trade and investment liberalisation. In onemodel, goods are ranked by skill intensity. Investment liberalisation leadsto a shift of investment towards Mexico. However, those goods whose pro-duction is shifted from the USA to Mexico are the least skill-intensive fromthe US point of view, but the most skill-intensive from Mexico’s standpoint.The demand for skills therefore increases in both Mexico and the USA.Similar HO reasoning also works towards explaining why the reduction ofprotection in the production of labour-intensive maize in Mexico (but verycapital-intensive in the USA) can lead to a rise in the wage–rental ratio inboth countries, because from each country’s point of view the price of thecapital-intensive good has risen. It is clear from these two examples thatonly few modifications would need to be brought to the standard HOmodel to produce different results.

Finally, Winter-Ebmer and Zimmermann in chapter 11 focus onEast–West migration and trade. One major concern in some EU countriesstems from the potential labour market effects of integration and enlarge-ment to the East. Opening markets will encourage factor flows and trade,


and hence very probably cause adjustments in wages and employmentopportunities. Given the ever-rising unemployment rates, and the relativedecline of unskilled wages in Western Europe, the Eastern enlargement isseen as a threat to native labour markets. Because of geography and his-torical ties, Austria and Germany have already received disproportionallymore immigrants and stronger increases of trade flows than other coun-tries. It has to be expected that this trend will continue in any process of eco-nomic integration in the East. Hence, the Austro-German situation is aninteresting case study.

The authors first provide an extensive survey of the previous empiricalliterature on the labour market in Austria and Germany. This is important,since a large part of the European literature deals with the situation in thosetwo countries. Here the findings at large are that neither immigration nortrade had a relevant harming effect on employment and wages. They thenprovide new econometric evidence on the issue. Using industry-panel datafor both countries, they employ a reduced-form approach to examine theeffects of trade and immigration (from the East and elsewhere) on total andnative employment, and wages. The Austrian findings suggest thatimmigration has a negative impact on native employment and wages, andno effect on total employment. Imports affect employment negatively andexports have a positive effect on wages. The German results indicate thatimmigration and trade is not harming employment and wages. Nativesseem to be complements to migrants, at least to those from East Europe.Trade does not affect wages at all, and hardly affects employment. Theseresults are in line with other recent studies for both Austria and Germany.

6 Conclusions and policy implications

Theory tells us that factor movements and commodity trade are jointlydetermined by technology and tastes, pointing out where to look for thefactors that determine the link between trade and factor mobility. We alsoknow that new trade models often predict that trade and migration arecomplements, and economic-geography models point out that integrationmay lead to an agglomeration of economic activity. Yet, as Sapir notices inhis discussion of Venables’ chapter 2, the applicability of increasing returnsmodels to North–South trade may be limited. This does not imply,however, as shown by Markusen and Zahniser, that we should rely exclu-sively on standard HO theory. Moreover, financial and institutional con-straints may reverse the direction of the trade–migration link. In sum,both the theoretical and empirical contributions in this volume suggestthat trade liberalisation will not always alleviate the incentives for factormobility.


Overall, the chapters in this volume suggest several observations. First,initial conditions matter. Trade liberalisation of high-income countrieswith middle-income countries is more likely to foster convergence and dis-courage migration, though liberalisation in investment flows could alterthis outcome. By the same token, integration of goods markets betweeneconomies with very different initial conditions could lead to opposite out-comes. This is because the fall in trade costs would lead to more polarisa-tion and more migration, in part because of the relaxation of financialconstraints on migration costs. There are therefore good reasons to be rel-atively optimistic about the migration outlook from Eastern Europe toWestern Europe following the Europe Agreements. These are middle-income countries, demographic conditions are stable and, provided that thetransition to a market economy is successful, massive migrations shouldnot constitute a significant threat.

Second, short-run effects may be important, even when integration isbetween countries that are not at the extremes, as in the case of Mexico andthe USA. For instance, it would appear that NAFTA encouraged migra-tion, at least in agriculture in the short run. Disruption of Mexico’s maizeproduction to the benefit of its counterpart in the USA has put downwardpressure on unskilled wages on both sides of the border as this highly pro-tected sector in Mexico is a very labour-intensive activity in Mexico and avery capital-intensive one in the USA. This may be a special case, but itserves to point out that even if technology is available to all off the shelf, thesame technologies are not always profitable everywhere at the same time. Inthe longer run, though, improved conditions in the Mexican economywould still stem migratory pressures.

The policy message is therefore clear. Trade liberalisation and migrationcontrols are not alternative policy strategies as suggested by a straightfor-ward application of trade theory. They work with differing effectiveness overdifferent time horizons. Migration controls are likely to be somewhat moreeffective in the short run and, in any case, remain the main tool to avoidmassive, and largely undesired, immigration in receiving countries. But, iftheir objective is to stem migratory pressures, policy-makers’ reliance onmigration controls in the short run should not dispense them from search-ing for more forward-looking strategies to alleviate migration pressure in themedium run. Despite theoretical ambiguities and policy disputes, the evi-dence continues to point towards benefits from trade liberalisation.

NOTES

1 We thank Jean-Marie Grether and Sanoussi Bilal for comments. Financialsupport to Jaime de Melo by FNRS no. 12-42011.94, to Riccardo Faini byConsiglio Nazionale delle Ricerche and to Klaus F. Zimmermann by the EU


HCM Programme through the research network ‘European Migration: FromEconomic Analysis to Policy Response (no. ERBCHRXCT940515) is gratefullyacknowledged.

2 Clearly, there are strong analogies between these concerns about the impact ofimmigration and the controversy surrounding the effects of globalisation andtrade integration on the labour market outcome in OECD countries.

3 A poll conducted on the behalf of the European Commission shows that a sig-nificant majority of those interviewed believe that ‘immigrants are too many’.This is particularly true in the largest four European countries, with 54.7 per centof Frenchmen, 56.7 per cent of Germans, 51.3 per cent of Britons and – some-what surprisingly, given the very low population share of foreigners – 64 per centof Italians believing the number of migrants to be excessive.

4 Econometric evidence from reduced-form models produces correlations whilesimulation results from structural models usually correspond closely to thegeneral equilibrium models of trade theory, but they can capture only the effectsincorporated in the model, and when calibrated to real-world data, the fit is onlyfor an arbitrarily chosen base period.

REFERENCES

Banerjee, B. and S.M. Kanbur (1981). ‘On the Specification and Estimation ofMacro Rural–Urban Migration Functions: With an Application to IndianData’, Oxford Bulletin of Economics and Statistics, 43, 7–29

Ben-David, D. (1993). ‘Equalizing Exchange: Trade Liberalization and IncomeConvergence’, Quarterly Journal of Economics, 108, 653–79

Borjas, G. (1995). ‘The Economic Benefits of Immigration’, Journal of EconomicPerspectives, 9, 3–22

Collinson, S. (1993). Europe and International Migration (London: Pinter)Faini, R. and A. Venturini (1993). ‘Trade, Aid and Migration: Some Basic Policy

Issues’, European Economic Review, 37, 435–42Ferenczi, I. and W. Willcox (1934). International Migrations (Geneva: United

Nations)Finger, M., M. Ingco and U. Reincke (1996). The Uruguay Round. Statistics on

Tariff Concessions Given and Received (Washington, DC: World Bank)Friedberg, R. and J. Hunt (1995). ‘The Impact of Immigration on Host Country

Wages, Employment and Growth’, Journal of Economic Perspectives, 9, 23–44ILO–UNCHR (1992). ‘Informal Summary Record: Joint ILO–UNCHR Meeting

on International Aid as a Mean to Reduce the Need for Emigration’ (Geneva:ILO)

Krugman, P. (1991). Geography and Trade (Cambridge, MA: MIT Press)(1995). ‘Growing World Trade: Causes and Consequences’, Brookings Papers on

Economic Activity, 1, 327–62Lopez, R. and M. Schiff (1995). ‘Migration and Skill Composition of the Labor

Force: The Impact of Trade Liberalization in LDCs’, PRWP, 1493(Washington, DC: World Bank)

Lucas, R. E. Jr. (1990). ‘Why Doesn’t Capital Flow from Rich to Poor Countries ?’,American Economic Review Papers and Proceedings, 80, 92–6


Lucas, R. E. B. (1995). ‘International Trade, Capital Flows and Migration:Economic Policies Toward Countries of Origin as a Means of StemmingMigration’, Boston University, mimeo

Molle W. (1990).The Economics of European Integration (Aldershot: Darmouth)Mundell, R. (1957). ‘International Trade and Factor Mobility’, American Economic

Review, 47, 321–35Ruffin, R. (1984). ‘International Factor Movements’, in P. Kenen and R. Jones

(eds.), Handbook of International Economics (Amsterdam: North-Holland)Sachs, J. and A. Warner (1995). ‘Economic Reform and the Process of Global

Integration’, Brookings Papers on Economic Activity, 1, 1–95Schiff, M. (1997). ‘Migrations et échanges: Aspects normatifs et positifs’, in J. de

Melo and P. Guillaumont (eds.), Commerce Nord–Sud, Migration etDélocalisation (Paris: Economica)

Siebert, H. (ed.) (1994). Migration: A Challenge for Europe (Tübingen: Mohr)SOPEMI (1995), Tendances des Migrations Internationales (Paris: OECD)Stalker, P. (1994). The Work of Strangers: A Survey of International Labour

Migration (Geneva: International Labour Office)UNCTAD (1995). World Investment Report (Geneva: United Nations)Zimmermann K. (1995). ‘Tackling the European Migration Problem’, Journal of

Economic Perspectives, 9, 45–62; reprinted in M. N. Jovanovich (ed.),International Economic Integration. Critical Perspectives of the World Economy(London: Routledge, 1998)


Part OneInsights from theory

2 Trade liberalisation and factormobility: an overview

ANTHONY J. VENABLES

1 Introduction

A reduction in the cost of one type of international transaction – lowertransport costs for goods trade or the liberalisation of a factor flow –generally changes the incentives to make other international transactions.There are many examples of the way in which changing trade barriers maychange the incentives for factor mobility, although some of these examplespull in opposite directions. For example, in discussion of EU trade policytowards Eastern Europe it is suggested that making trade in goods easiermight reduce the incentives for labour to emigrate (Begg et al., 1992). Inthe study of migration from Europe to the USA in the late nineteenthcentury, it is argued that reductions in transport costs for agriculturalproducts increased the incentives for labour migration, both inter-region-ally and internationally (Harley, 1978, 1980). In a regional context, it isoften suggested that falling transport costs were a trigger for migration tocities, as the costs of feeding the urban population were reduced (Bairoch,1988).

The objective of this chapter is to provide an overview of theory on therelationship between trade liberalisation and factor mobility. What doestheory suggest about the way in which reducing the costs of making sometransactions changes the incentives to make others? Under what circum-stances will goods-trade liberalisation reduce factor flows, and under whatcircumstances will it increase them? What are the consequences of thesefactor flows for trade?

This chapter addresses these issues in a number of commonly used trademodels, each of which illustrates different forces that may be important. Westart with perfectly competitive models, in which there are no distortions ofany sort. In such models we can be sure that narrowing a goods-pricedifference between two countries also narrows any difference in the price ofthe factor used intensively in production of that good.1 But what does it doto differences between the prices of other factors? The answer depends on

23

the relationship between factors of production in the economies’ technolo-gies, and we illustrate two possibilities. The first is based on theHeckscher–Ohlin (HO) model, in which trade liberalisation does indeedreduce the incentives for mobility of both factors. But moving to the richerset of substitution–complementarity relationships between factors, asoccurs in the specific-factors model, we show how trade liberalisation maycause factor mobility, which in turn makes the two economies more dis-similar in their relative endowents, increasing trade flows.

We then move on from a perfectly competitive environment to situationswith increasing returns to scale, and study two models based on theHelpman–Krugman (1985) model of international monopolistic competi-tion. With increasing returns to scale absolute as well as relative factorendowments become important, and a number of additional forces comeinto play. First, cost levels in the economy may become a function ofcountry size, and we capture this in a variant of the model in which one ofthe sectors is used as an input to production. Second, market size becomesan important determinant of the location of production. And, third, thecost of living index also becomes a function of country size. Each of theseconsiderations creates the possibility that factor flows are self-reinforcing,so cumulative causation comes into effect: factor inflows enlarge theeconomy, increasing the real return to factors and creating further inflows.As we shall see, trade liberalisation in models of this type may create factormobility, possibly causing agglomeration of activity in one location. Indeveloping these ideas, we provide a generalisation of Krugman’s (1991)‘core–periphery’ model of economic geography, placing his model in amore standard two-country (or region), two-sector and two-factor trademodel.

The method of analysis employed in the chapter is as follows. Supposethat economies are identical, except in their shares of the world factorendowment. For a given set of goods trade barriers we map out the factorprices associated with each possible division of the world endowment offactors between the two countries – i.e. we construct a contour map offactor prices over the box diagram representing the division of the worldendowment between countries. From this, we can derive factor-pricedifferences between economies. If there are barriers to factor mobility thensmall factor-price differences will not generate factor flows, but large oneswill. We can therefore map out the regions of the endowment box in whichthere are incentives for movement of different factors, turning the endow-ment box into a phase diagram. If we start the world economy at an initiallong-run equilibrium (that is, given the barriers to goods trade and factortrade, at an endowment point at which the gains from factor movementshave been exhausted) and then consider a goods-trade liberalisation, is the

24 Anthony J. Venables

initial equilibrium allocation of factors between economies still an equilib-rium, or are there now incentives for further factor mobility? If so, in whatdirection, and with what long-run outcome? We answer these questions byseeing how the trade liberalisation alters the phase diagram for factormobility, and using the diagram to trace the direction of factor movements.

It should be emphasised that the models presented in this chapter are notnew, and neither are the results. However, the chapter poses a somewhatdifferent question from most of the preceding literature. Existing literatureon the relationship between goods and factor trade asks whether the twoare substitutes or complements. Probably the best known paper in this lit-erature is Markusen (1983), who assumes that economies have identicalendowments but – for a variety of reasons – have different factor prices.Starting from a situation where factor mobility is ruled out, Markusen theninvestigates the consequences for trade of allowing factor mobility and, likethe present chapter, concludes that the HO substitution story is not general.The present chapter continues this line of research, but differs from it in thefamily of models analysed, in the assumption that economies are identicalexcept for factor endowments and, most importantly, in its emphasis onstudying the effects on the long-run equilibrium of changes in the cost ofinternational transactions.

2 Competitive models

2.1 Heckscher–Ohlin model

Our point of departure is the HO model, and Mundell’s (1957) observationthat if there is free trade and endowments are in the diversification cone,then there is no incentive for factor mobility. What happens if trade is lessthan perfectly free and as trade barriers change? We can answer thisdrawing on analysis undertaken by Norman and Venables (1995).

Figure 2.1 illustrates the incentives for factor mobility for different divi-sions of the world endowment of the two factors between the twoeconomies. The factors are labelled A and B, with prices a1, a2, and b1, b2 inthe two countries. (The interpretation of the factors will vary from exampleto example, so it best not to label them as labour or capital a priori.) Theheight of the vertical axis is the world endowment of factor A, and thelength of the horizontal axis is endowment of B. A point in the box repre-sents a division of the world endowment of the two factors between the twocountries (labelled 1 and 2), with country 1’s endowment given by thevector from O1 to the point, and country 2’s endowment measured from O2.At any such point we can compute the equilibrium finding, amongst otherthings, the equilibrium factor and goods prices.

Trade liberalisation and factor mobility 25

The figure illustrates a situation in which trade in the B-intensive indus-try (call it the x-sector) has real ad valorem trade costs of T21 per unit,while trade in the A-intensive sector (the z-sector) is perfectly free. Theprices of x-sector output in each country are denoted p1 and p2, and theoutput of the z- sector will be used as the numeraire. The curves in the boxdefine regions within which international differences in goods and factorprices lie within pre-set bounds.

Consider first the lens-shaped region between the two bold curveslabelled T. This is the area in which there is no goods trade – autarky pricesdiffer by less than the trade barrier, i.e. p1/p2 ,T,p2/p1, so trade is notprofitable. Clearly, the region includes the main diagonal (where relativeendowments and goods prices are the same in both countries) and extendsto regions on either side of the diagonal, where differences in relativeendowments are small enough to give differences in goods prices of lessthan T. In the area outside this lens relative prices are fixed, at p15Tp2 inthe upper left of the figure where country 1 is importing x, and at p25Tp1in the lower right, where country 2 is importing.

What about factor prices? In studying the location decisions of factorswe have to distinguish between factor prices in terms of the numeraire andfactor prices deflated by cost of living indices in each location. If a factoralways consumes in the same place then mobility decisions will be guidedby the former, but if consumption occurs in the country of residence, then


Figure 2.1 Incentives for factor mobility

the latter are appropriate. We shall assume throughout this chapter thatfactors consume in the country of residence, so that to look at mobilitydecisions it is always appropriate to deflate factor prices by a cost of livingindex, which we shall denote e(p1), e(p2) in countries 1 and 2, respectively.We shall call these deflated factor prices ‘real factor prices’.

Looking now at figure 2.1, the presence of trade costs means that factor-price equalisation (FPE) occurs only on the main diagonal of the box.Above and to the left country 1 is poorly endowed with B relative to A andas we have seen p1.p2 so a1 ,a2 and b1 .b2. In the HO model goods-pricedifferences map into proportionately larger factor-price differences, soinequalities between factor prices in terms of the numeraire also holdbetween real factor prices. If there is free mobility of factors the story istherefore straightforward. Factor movement takes place until the economyreaches the main diagonal, and mobility of just one of the factors issufficient to reach this diagonal and achieve FPE. The point on the diago-nal to which the system converges depends on the initial endowment andthe dynamic specification of factor mobility. Since factor mobility gives theeconomies identical relative endowments, it substitutes for trade.

Assuming perfectly frictionless mobility of factors is of course unre-alistic, and is an assumption that sits uncomfortably with our assumed fric-tion in goods trade. What if there are costs associated with the mobility offactors as well as of goods? Let us call the ad valorem costs of movingfactors A and B, TA21 and TB21, respectively. The lines TA and TB onfigure 2.1 give the loci of endowments at which factors are indifferentbetween migrating and not migrating, and are drawn for a relatively highvalue of TA and lower value of TB. Thus, the curves TA give the set of factorendowments at which a1/e(p1)5TAa2/e(p2) and a2/e(p2)5TAa1/e(p1).Between these curves differences in ai are small enough that it is not worth-while for factor A to move. Above and to the left of the upper TA-curvea1/e(p1) is small enough relative to a2/e(p2) for factor A to move fromcountry 1 to country 2, as illustrated by the vertical arrows on the figure.Similarly, the curves TB are the sets of endowments at which b1/e(p1)5TBb2/e(p2) and b2/e(p2)5TB b1/e(p1). Outside these curves differences in bi /e(pi)are large enough for B-mobility to follow the horizontal arrows, and insidethe difference is small enough for B not to move.

What do we learn from figure 2.1? First, we can use it to investigate whatcombination of goods and factors is traded at equilibrium. Standard HOtheory begs this question by partitioning commodities into goods, usuallyfreely tradable, and factors, usually totally immobile. Here we can see thatthe answer depends on the relative positions of the T-, TA- and TB-curves.Look only at the upper left part of figure 2.1 (arguments for the lower rightpart are symmetric) and now suppose that we start out in the far upper left.


At points above the T-, TA- and TB- curves there is, respectively, trade ingoods, factor A and factor B. Factor mobility induces movement down andto the right, but as we cross the TA line so factor A mobility ceases – real aidifferences are not large enough for further movements to occur. We nowmove to the right with factor B coming into country 1. This narrows goodsprice differences, and as the T-contour is crossed goods trade ceases.2

However, real bi differences are still large enough for factor B to move fromcountry 1 to country 2, and the system will end up at a point on the TB-curve, at which point there is no trade of any sort.

Armed with this, how does a trade liberalisation affect factor mobility?A reduction in T has two effects. First, it shifts the T-curves inwards; asexpected the no-trade region contracts. Second, if the TA- and TB-curves lieoutside the T-curve then they are shifted further out (given the levels ofmobility barriers TA and TB). The reason is that goods- and factor-pricedifferences outside the TT lens are reduced, so the areas of endowmentspace in which there is an incentive for factors to move are reduced. Theconclusion is then rather obvious. Trade liberalisation reduces the incen-tives for factor mobility, and will not create factor mobility where there wasnone before. For example, if prior to the trade liberalisation the economywas on the TB frontier, then the outwards movement of the frontier pushesthe system into the interior of the region where there is no incentive forfactor B to move.

How does this relate to the examples mentioned in the Introduction?Think of country 1 as Eastern Europe and 2 as the EU, and suppose thatthe initial endowment is in the lower right corner of the box, where country1 has much labour (factor B) and little capital (factor A). Eastern Europe(country 1) exports good x, which is the labour (B)-intensive good, butfaces trade barriers to these exports. These barriers reduce the price oflabour and raise the price of capital in Eastern Europe, creating the direc-tions of factor mobility illustrated – labour (B) outflow and capital (A)inflow. A reduction in the barriers pushes the TA- and TB- loci outwards,reducing the international differentials in factor prices, and possibly beinglarge enough to staunch the flow of emigrants.

Can this model illuminate nineteenth-century European–US migrantflows? Suppose that country 1 is now the USA and country 2 Europe.Country 1 has little labour (factor B) and much land (rechristened factorA). This places us in the top left corner of the box with the USA exportingthe land-intensive good and receiving an inflow of labour. However, tradeliberalisation or falling transport costs reduce the incentive for labourinflow. The reason is that the liberalisation raises land rents and – since landand labour are the only factors used in the two sectors – this reduces USwages.


The failure of the HO model to capture US–European migration flowsin a satisfactory way is of course due to the assumption that the same twofactors are used in both sectors. We need to separate out specific factors foragriculture and industry, and it is to this that we now turn.

2.2 Specific factors

In the HO model an increase in an economy’s endowment of one factorcannot reduce the return to the other, but in a model with sector-specificfactors this complementarity becomes possible and, as we shall see, itgreatly increases the richness of the model.3

We continue to assume that factor A can be used in either sector, but nowassume that factor B is used only in the x-sector, and introduce factor C,with price ci, used only in the freely traded z-sector. As is well known, withfree trade the FPE set in this model is a two-dimensional sub-space of thethree- dimensional factor space. What happens if there is not free trade but,as in sub-section 2.1, x-sector output faces trade costs at rate T ?

In the HO model the presence of trade costs reduces the dimensionalityof the FPE set from a two-dimensional set to a line, and so in the specific-factors model trade costs reduce the FPE set from a line to a point. Toanalyse mobility dynamics in the three-dimensional endowment space weproceed in two stages, looking first at the mobility of factors A and C, andthen at B and C.

2.2.1 Mobility of the non-sector-specific factorFigure 2.2 gives the mobility dynamics for the sectorally mobile factor (A)and one of the specific factors (C). It has factors A and C on the axes, andis drawn under the assumptions that the two economies are equally wellendowed with the other sector-specific factor, B, that half of costs in eachsector go to factor A and half to the specific factor, and that consumers’expenditure is equally divided between the two goods.

The bold dashed line is the FPE set with free trade. It has a positive slopeless than unity, because increasing country 1’s endowment of the specificfactor C must be associated with a less than proportional increase in thenon-specific factor, A, if factor intensities in each sector are to be held con-stant and FPE maintained. Above and below the line factor-pricedifferences and consequent mobility are as suggested by the arrows in thetop left and bottom right corners of the box.

What differences are made by trade costs on good x? On a vertical line inthe middle of the box, where C1 5C2, there is no trade and p1 5p2, while tothe left of this country 1’s abundance of B relative to C means that p1 ,p2.Trade, if it occurs, takes the form of country 1 exporting x (the B-intensive


good) and importing z. The goods-price difference unambiguously reducesthe real return to factor B in country 1 relative to country 2, and creates aregion in which real returns to both factors A and C are higher in country1 than in country 2. The edges of this region are the lines aa and cc, givingthe sets of endowments at which a1/e(p1)5a2/e(p2) and c1/e(p1)5c2/e(p2).While the precise positions of these lines depend on consumption shares,the important (and general) point is that the curves intersect only at thecentral point of the box, dividing endowment space up into four regions,with real factor-price differences as suggested by the arrows.

Figure 2.2 is perhaps easiest to interpret if we think of it in terms of ourUS–Europe migration example. The two economies have equal endow-ments of land (factor B), but economy 1 has little labour (factor A) andcapital (factor C), so we are in the lower-right-hand area of the box. Thiseconomy exports agriculture (x-sector, intensive in factor B) which has hightransport costs. We see that such an economy can very well find itself withinflows of both capital (C) and labour (A). Even if the economy starts in aregion where it is losing one of the factors (e.g. points d or e), it may,depending on adjustment speeds, move into the region in which it is receiv-ing inflows of both factors.4 These factor movements are trade-reducing asthey bring convergence of the countries’ relative endowments.5

What is the effect of trade liberalisation in this setting? The answer tothis turns on parameters of the model. Figure 2.3 is constructed with the


Figure 2.2 Mobility dynamics for the sectorally mobile factor and one specificfactor

x-sector relatively more factor A-intensive than the z-sector, and maps outiso-relative factor-price loci for factor A. The three dashed lines are drawnfor the case of free goods trade, and are the loci of endowments which giveconstant relative real ai in the two economies; the central line is a2/e(p2)5a1/e(p1), the lowest line a1/e(p1)5TAa2/e(p2) and the uppermost a2/e(p2)5TAa1/e(p1), where TA.1 is the barrier to mobility of factor A. The three solidlines are the analogous curves when there are positive barriers to goodstrade.

Comparing the dashed lines with the solid, there are two points ofdifference. First, the dashed lines are further apart than the solid. Exactlyas we saw with the HO model, trade liberalisation has the effect of shiftingthe iso-relative factor-price loci outwards – further away from each other.This is because a variation in endowments has a smaller effect on factorprices, the more open is the economy. But second, and more importantly,trade liberalisation rotates the iso-relative factor-price lines clockwise. Thereason is that, at a point near O1 country 1 is exporting x, so trade liberal-isation increases p1. This raises a1 and e(p1) so, in general, has an ambigu-ous effect on the real return, a1/e(p1). However, since figure 2.3 is drawn withthe assumption that the x-sector is more factor A-intensive than the z-sector, the increase in a1 is relatively large, giving an increase in a1/e(p1). Thisenlarges the region in which factor A flows into country 1, giving the clock-wise rotation indicated.


Figure 2.3 Parameters of the model

In this case, then, trade liberalisation can promote factor inflow. IgnoringC-mobility and assuming a cost to A-mobility of TA, the point f is an initiallong-run equilibrium. Trade liberalisation means that the dashed, not thesolid, lines become applicable, so factor A flows into country 1 and thesystem moves to point g. If we were to also allow capital mobility, then therewould be an associated capital inflow, moving the economies towards thecentre of the endowment box.

The result that liberalisation causes inflow of the sectorally mobile factordepends on trade liberalisation increasing the real return to the factor. Thisarises if the export sector is the more intensive user of the sectorally mobilefactor (as assumed here) or if the factor consumes relatively little of the exportgood (so that the increase in e(p1) is small). Whether this case captures theissues raised by transport-cost reductions and immigration to the USA istherefore an empirical matter, turning on details of technology and preferences.

2.2.2 Mobility of the specific factorsFigure 2.4 maps out the case in which the two sector-specific factors areinternationally mobile, and factor A is held immobile. As in figures 2.2 and2.3, the heavy dashed line is the FPE set with perfectly free trade. The linehas a negative slope, indicating complementarity of the two factors. Forexample, if country 1 were to receive an inflow of factor B then this wouldreduce b1 and also reduce c1 (because factor A moves between sectors,


Figure 2.4 Two sector-specific factors internationally mobile and factor Aimmobile

raising the intensity of the specific factor in both sectors). Above and belowthe line factor-price differences are as suggested by the arrows in the upper-right- and lower-left-hand corners of the box.

What differences are made by trade costs on good x? First, there is a no-trade region which forms a lens-shaped area similar to that in figure 2.1. Itis not illustrated on figure 2.4, but the kinks in the cc - and bb -lines occurat the edge of this region. On the main O1O2 diagonal p1 5p2, and aboveand to the left of this diagonal country 1 is relatively poorly endowed withfactor B (used in the x-sector), so if trade occurs country 1 imports x andp1.p2. Whereas point e had FPE under free trade, the goods-price gap p1.p2 now means that b1.b2 and c1 ,c2. Point e is therefore below the linealong which there is FPE for factor B and above the line with FPE for factorC, and both factor prices are equal only at the central point of the boxwhere the two economies are identical.

The lines bb and cc give the real FPE sets for each factor, i.e. they are thesets of endowments at which b1/e(p1)5b2/e(p2) and c1/e(p1)5c2/e(p2). Theirprecise positions depend on parameters but the qualitative configuration ofthe figure and associated directions of factor mobility are general.

We are now in a position to see the effects of trade liberalisation. Forsimplicity, we look only at cases in which mobility of factors B and C iseither impossible or completely free. Suppose first that the system is at pointf, and that only factor B is mobile. With the initial goods-trade barriers theeconomy would move from point f to point g. Trade liberalisation nowshifts the bb- and cc-loci inwards, towards their free trade positions (theheavy dashed line), so point g becomes inside the region in which country1 loses factor B: factor movement occurs in the direction of the arrow tosome point such as h. The interpretation of this movement is straightfor-ward. At point g country 1 is importing x-sector goods, but the presence oftransport costs creates protection for country 1 x-industry. Trade liberal-isation reduces this protection, so reduces the real return to the x-sector-specific factor (B), causing its outflow. There are several noteworthy pointsabout the move from g to h. First, the movement makes the relative factorendowments of the economies more different. At point g country 1 has rel-atively much C, half of world A and approximately half of world B, but thefactor mobility created by trade liberalisation moves it to a point with rel-atively little B. And, second, the volume of trade increases along the pathfrom g to h. Of course, the trade liberalisation alone increases trade, butthere is an additional effect coming from outflow of the factor used in theimport-competing sector. What we see in this case is trade liberalisation cre-ating factor movements which make the economies more different in rela-tive factor endowments and which increase trade flows.

A similar analysis holds if we assume that only C is mobile, and B is fixed


at some level. For example, if the initial long-run equilibrium is at point ion figure 2.4, the trade liberalisation creates the downward movementalong the dashed arrow to j. At i country 1 is relatively well endowed withB, so exports x-sector output. Trade costs act to protect country 2’s x-sectorand reduce country 1’s exports. Reducing this protection causes country 1’sx-sector to expand, drawing factor A into this sector, which in turn reducesthe country 1 price of factor C. There is therefore factor C outflow from 1,making the countries’ factor endowments more different and increasingtrade volumes.

To what real-world situations might this analysis be applicable? Supposethat the initial equilibrium is at point i where country 2 is a developedcountry (endowed with much skilled labour (C) and little unskilled labour(B)) and country 1 is a developing country which is exporting the unskilled-labour- intensive product, x. The developed country (country 2) now cutsits barriers to imports of x-sector products. As a consequence the develop-ing country (country 1) reallocates its sectorally mobile factor to exportproduction of x-sector output, thus reducing the real wage of skilled labourin the z-sector, and creating a ‘brain drain’ – the movement from i to j.

Summarising results from the two competitive models – HO and specific-factors – we see quite a complex set of possible responses. Although the HOmodel predicts that trade liberalisation reduces the incentives for factormobility, this need not hold once we allow for the richer set of interactionsbetween factors given in the specific-factors model. Trade liberalisation cancause inflow of the inter-sectorally-mobile factor (if the liberalising sectoris a relatively intensive user of the factor) and can cause factor flows thatmake economies more different (as it may reduce the return to aneconomy’s scarce, and import-competing, factor). Empirical work isneeded to apply these models to particular circumstances.

3 Increasing returns and cumulative causation

A recurrent question in the study of factor mobility is whether mobilitytends to reinforce the advantages of existing concentrations of economicactivity, or lead to a spatial smoothing out of activity. For example, in theEU, does factor mobility tend to draw factors of production out of periph-eral regions and into the centre, or does it lead to convergence of regions?Do reductions in the costs of goods trade between countries or regions tendto reinforce centripetal or centrifugal factor flows?

So far, we have worked with models that cannot begin to address thesequestions, because of the assumption of non-increasing returns to scale inall activities. To address them, we must add some element of increasingreturns to scale to the theory, and this is what we now do. Our tools for this


will be the much-used Dixit–Stiglitz (1977) model of monopolistic competi-tion, and its multi-country extensions (Helpman and Krugman, 1985). Weknow from Krugman (1991) that this model can support destabilisingmigration, and we seek in the remaining sections of the chapter to analysethis within the framework we have developed. This will involve some gener-alisations of Krugman’s model, and we shall show how factor-supply andmarket-access considerations interact to determine the incentives for factormobility.

3.1 An increasing-returns model

In this section of the chapter we develop in full a model with trade, factormobility, and increasing returns to scale. The model has the factor marketstructure of the HO model of sub-section 2.1, but superimposes monopo-listic competition on the x-sector. As in sub-section 2.1 the factors are Aiand Bi with prices ai and bi in countries i51, 2. The z-sector is perfectlycompetitive, is relatively A- intensive and produces output that is freelytraded. The x-sector has trade frictions and is now also assumed to beimperfectly competitive, with Dixit–Stiglitz product differentiation.

Using this approach, consumers have a constant elasticity of substitutionsub-utility function defined over x-industry-differentiated products. Thissub-utility function is given in the appendix (p. 45), but it is more econom-ical to work with the dual to the utility function, and so write the sub-expenditure function (or price index) for x-sector products in each country,qi, as

qi5 [ni pi12e1nj(Tpj)

12e]1/(12e), e.1, i,j51,2, i? j, (1)

e is the elasticity of substitution between varieties, T is the ad valorem tradecost, and ni and pi are, respectively, the number of varieties of x-sectorgoods produced in country i, and their price. Given expenditure on x-sectorproducts in aggregate in each country, Ei, the demand for the output of asingle country i firm in its home and export market are, from the demandsystem implied by the price index (2.1)

xii5pi2eqi

e21Ei, xij5pi2eT12eqj

e21Ej. (2)

Each firm sets price at a fixed mark-up over marginal cost, the size ofwhich is given by the elasticity of substitution between varieties, so

pi(121/e)5g(ai,bi) (3)

where g(ai, bi) is x-sector unit cost in country i. At a zero profit equilibriumeach firm’s total sales, xii1xij, must reach a level determined by the size ofthe fixed costs. We can choose units such that this level equals unity, so


xii1xij5pi2eqi

e21Ei1pi2eT12eqj

e21Ej 51. (4)

(The choice of units involves setting each firm’s fixed input requirement Fequal to 1/(e21).) This equation holds provided ni.0; if an economy hasno x-sector production then the total sales of a potential entrant must beless than or equal to unity. Assuming n1, n2 .0 it is possible to use (2.1) and(2.4) to obtain an explicit expression for the number of firms operating ineach country and this takes the form

ni pi 5 2 i,j51,2, i? j. (5)

Turning to the z-sector, we assume that this sector uses the two primaryfactors as input, and may also use x-sector products as intermediate goods.These enter x-sector production through a cost function which we assumetakes the same form as the consumer’s expenditure function (2.1), so thatthe country i price index for intermediate goods is qi.

6 The z-sector unit costfunction is then h(ai, bi, qi). We take z as the numeraire good, so z-sectoractivity satisfies

h(ai, bi, qi)#1, zi.0, complementary slack, i51,2. (6)

Characterisation of equilibrium can be completed as follows.Expenditure on x-sector output, Ei, comes from two sources, final andintermediate demand. Country i income is aiAi 1biBi of which proportionm is spent on the x-sector. The input of intermediates per unit output of zis given by the partial derivative hq(ai, bi, qi), so adding consumption andintermediate demands, Ei is

Ei 5m[aiAi1biBi)1qihq(ai,bi,qi)zi. (7)

Turning to factor markets, equality of supply and demand for factors A andB takes the form

Ai5zi ha(ai,bi,qi)1ni[xii1xij1F]ga(ai,bi) (8)

Bi 5zi hb(ai,bi,qi)1ni[xii1xij1F]gb(ai,bi) (9)

where the two terms on the right-hand side are factor demand from the z-and x- sectors, respectively.

We are now in a position to pose the question: given endowments andtrade barriers, what are factor prices, and what are the incentives for factormobility? There are three new forces at work, and they mean that absoluteas well as relative, endowments are now important in determining factorprices.

The first is that production costs may now on depend on absolute, as wellas relative endowments. This arises because a large economy will, other

EjT12e

(pj/pi)e 2 T12e

Ei

1 2 (pj/pi)eT12e


things being equal, produce more x-sector products and this reduces itsprice index, qi. If the x-sector products are used as intermediates, then thisis a cost reduction for z-sector firms which, in equilibrium, will be trans-lated into a higher price for one or both of the factors of production.

The second force is that the combination of intra-industry trade, increas-ing returns to scale and trade barriers, implies that firms tend to locate inthe market with larger expenditure on x-sector products. This means thatthe large economy will tend to have disproportionately more x-sector pro-duction, raising demand for factors of production, and again tending togive higher factor prices in the large economy. This can be seen by inspec-tion of (5). Set T.1 and suppose for simplicity that p1 5p2. Production inthe two economies is the same if E1 5E2, but increasing E1 causes dispro-portionately larger increases in n1. Notice also that this effect becomes morepowerful as T is reduced.

The third force is a cost of living effect. As we have already noted, thelarger economy will tend to have a lower price index for differentiated prod-ucts, and hence a lower cost of living index. Given factor prices in terms ofthe numeraire, this effect will increase real factor returns in the largereconomy.

To analyse the implications of these forces for factor prices and factormobility we separate them out, first looking at a model in which just thecost effect is operating, and then moving to a model where both the expen-diture effect and the cost of living effect are important.

3.2 Country size and costs

Suppose first that x-sector production is used only as an intermediate good,so m, the share of consumption expenditure going to the x-sector, is zero.7

What, then, does the FPE set look like, and what are the incentives forfactor mobility outside this set?

It is easy to see that we have both a15a2 and b1 5b2 only if the twoeconomies have identical endowments. It results from the following chainof reasoning. If factor prices are equalised, then so too are x-sector prices(from (3)) and the price indices (from (6)). This means that the numbers offirms in the two economies are equal (from (1), provided T.1) in whichcase so are levels of x-sector expenditure and z-sector production (from (5)and (7)). This implies the same factor demand in the two countries, so thatcountries must have identical endowments. Notice that this argumenthinges on trade barriers. If T51 then increasing returns are entirely global,not national, so the FPE set is a two- dimensional sub-set of endowmentspace, as in HO and Helpman and Krugman (1985).

What happens if we give country 1 more of both factors, holding the A/B


endowment ratios in the two countries equal? Supposing that techniques ofproduction are unchanged this raises z1 and n1 equiproportionately (fromfactor market clearing, (8) and (9)). But as this happens so q1 falls, from theprice index, and this generates a cost saving on country 1 z-sector produc-tion. This lower price of intermediates is captured by factors of production,so raising one or both of the factor prices in the larger economy.

The picture is given in figure 2.5. Axes are as before, and the solid linesaa and bb are the loci of endowments at which factor prices are equalised.8

These divide endowment space into four regions, with directions of factormobility as illustrated.

The first thing to note about figure 2.5 is that the point where theeconomies is identical has FPE but, if both factors are mobile, is unstable.Unsurprisingly, having built increasing returns into the production sectorof the economy, there are gains from cumulatively moving factors into thelarger economy, and long-run equilibrium involves complete agglomera-tion of activity in one economy. The more interesting case is when only oneof the factors is mobile. The immobile factor prevents complete agglomera-tion from occurring, and gives a long-run equilibrium in the interior of thebox, lying on either curve aa or bb.

Second, how do the loci depend on the size of the barriers to trade in thex-sector? Whereas the solid lines in figure 2.5 are drawn for T51.5, thedashed lines are for T51.1, and we see that the regions in which both factorsmove from one economy to the other are increased by trade liberalisation.


Figure 2.5 Lower price of intermediates

Suppose that only one of the factors – say, factor B – is mobile. Then thepoint labelled e is an initial long-run equilibrium. Trade liberalisation nowcreates the incentive for movement along the dashed arrow, horizontally topoint f, this making the two economies more different in size. The reason isthat the need to locate near final consumers is reduced, enabling the benefitsof agglomeration to be more fully realised, so the larger economy becomes amore attractive location for the increasing-returns industry. This raises bi inthe large economy relative to the small, giving the movement indicated. Theeffect of liberalisation is then to create factor mobility, and this makes theeconomies less similar. It also increases the volume of trade – the twoeconomies are more dissimilar in both relative and absolute endowments.

A note of caution needs to be sounded about these results. The analysisis based on perfect factor mobility – what if there are frictions TA and TB infactor movements? We could add a15TA a2, a25TA a1 and b1 5TB b2 andb25TB b1 loci to the figure, and they would take the form of bands aroundthe a15a2 and b15b2 lines that we have illustrated. The effects of the tradeliberalisation is both to move the centres of the bands, as illustrated infigure 2.5, and also to widen the bands, because lower goods trade barriersmake the returns to scale less national in scope. This reduces the likelihoodthat trade liberalisation will cause the factor mobility we saw in figure 2.5.For example, following trade liberalisation factor- price differences at pointe might be insufficient to cause factor movement – e might lie in the bandbetween b1 5TB b2 and b25TB b1 for the new goods-trade barriers.

The message from this section is then that if country size feeds into lowercosts via increasing returns of some sort, then factor mobility can bedestabilising, and lead to agglomeration of activity in one economy.However, full agglomeration occurs only with perfect mobility of bothfactors. If a single factor is mobile then there are interior equilibria, such aspoint e of figure 2.5, but trade liberalisation might cause the smallereconomy to become still smaller (the move from e to f ). Frictions in factormobility reduce the likelihood of this occurring.

3.3 Country size, expenditure and the core–periphery model

We now turn to the market size and cost of living effects of factor mobilityin an increasing-returns model. We do this by switching off the intermedi-ate goods linkage – so hq 50 from now on – and instead having x-sectoroutput go to final consumption, som .0. If there are no transport coststhen this is exactly the textbook Helpman–Krugman model, in which stan-dard HO properties hold for net trade and for factor prices. However, as weshall see, the story changes radically when we put some trade frictions inthe picture.


Our first task is to characterise the FPE set. To do this, it is important tonote that the location of the x-industry is determined by two different setsof forces – product market demand, as given in (5), and factor supply, asgiven by (8) and (9). If trade costs were zero, then (5) would cease to hold(it is undefined when T51). Intuitively, free trade relaxes the productmarket conditions as supply has to equal demand only at the world level,not country by country. In this case the FPE set is the full diversificationset, as in HO and in Helpman–Krugman. But if T.1 then both the productand the factor market conditions have to hold, and this requirementreduces the dimension of the FPE set.

To find the FPE set we look first at the product market conditions. Iffactor prices, and hence also goods prices, are equal in the two countries,then (5), giving the value of production in each country, become

n1p 5 , n2 p 5 (10)

where a bar over a variable denotes FPE. The equations say that locationof the industry depends on the location of x-sector expenditures, Ei, and ontransport costs. If there are positive transport costs, (1.T12e), then pro-duction will be skewed towards the location with the larger expenditure,and this effect is more powerful the smaller is T (and closer is 12T 12e tozero), since production is less anchored by the need to supply local con-sumers. Expenditure, Ei, depends on factor endowments – we have, from (7)

Ei 5m(aiAi 1 biBi).

Using this in (10) gives x-sector output in each country as a linear functionof factor endowments.

Equation (10) captures the effect of product market access on industriallocation. We also have the usual factor market considerations. Assumingthat both the z- and x-sectors are operating in both countries and denotingtechnical coefficients at FPE factor prices by ga, etc. we can use factormarket-clearing conditions (8) and (9), together with zero profits (implyingxii1xij1F5 p/g), to derive,

n1p 5 g, n2 p 5 g. (11)

These say that the value of x-sector output in each country is (given FPEtechnical coefficients) a linear function of factor endowments.

Equations (10) and (11) give the product and factor market determinantsof x-sector production. Requiring them both to be satisfied restricts endow-ments, and this restriction defines the FPE set. Eliminating n1 (or n2 – thetwo pairs of equations are not independent) from these equations, noting

1haB22hbA2

hagb2gahb21haB12hbA1

hagb2gahb2

E2 2 T 12eE1

1 2 T 12e

E1 2 T 12eE2

1 2 T 12e


that both expressions are linear in Ai and Bi and using world endowmentsA5A11A2 and B5B1 1B2 we can characterise the FPE set as a linear rela-tionship between A1 and B1. The FPE set is no longer the full diversifica-tion set of standard HO or Helpman–Krugman analysis, but instead aone-dimensional sub-space of this set.

The expression for the FPE set is straightforward, and we need note onlythat it includes the point at which the two economies are identical, is linear,and has gradient

5 5

(12)

where D is the determinant of the matrix of technical coefficients

D; hagb2 gahb.0,

positive since the x-sector is B-intensive. (The two different forms given forthe gradient are useful in different contexts, and can be derived from eachother using homogeneity of the cost functions.)

This is illustrated in figure 2.6, in which the FPE set constructed above is

hag(1 2 m) 1 gahm 2 T 12e(hag1 mbD)hbg(1 2 m) 1 gbhm 2 T 12e(hbg1 maD)

hag(1 2 T 12e) 2 mbD(1 1 T12e)hbg(1 2 T 12e) 2 maD(1 1 T 12e)

dA1

dB1


Figure 2.6 x-sector B-intensive

the line segment de, and the continuation of this set into regions in whichone country is specialised is given by the segments, O1d and eO2.

9 The FPEset goes through the mid-point of the endowment box and has gradient(given by (12)) less than the O1O2 diagonal. The intuition for this can beseen by considering a point below the O1O2 diagonal. At such a pointcountry 1 is relatively well endowed with B, and hence has relatively muchx-sector production. If there are transport costs, this is consistent with FPEonly if country 1 has relatively high demand for x-products – i.e. the pointis closer to O2 than to O1. de must therefore cut the O1O2 diagonal fromabove, as illustrated. On either side of the FPE set, nominal factor pricesare as illustrated on figure 2.6. Above the FPE set country 1 is relatively wellendowed with A, so we have a1 ,a2, b1.b2, and p1 .p2.

It is important to note for our purposes that the slope of the FPE setdepends on trade costs. Increasing T, we see that as T→` so de goes to theO1O2 diagonal of the box, because under autarky there is FPE only if theendowment ratios of the two economies are equal. (This can be seen byletting T in the second part of (12) which then gives the ratio of worlddemand for A to world demand for B). Reducing T reduces the gradient ofde, rotating the FPE set clockwise around the central point, and at somevalue of T the FPE set is horizontal (see the first part of (12)). At still lowervalues of T the FPE set has negative gradient, and as we go to free trade(T→1) the gradient tends to the ratio of factor prices, b/a, so FPE occursonly if the economies have the same incomes and expenditures. The reasonis that at low T the tendency for industry to be drawn into the economy withthe larger market becomes stronger (as noted in our discussion of (10)), soFPE is possible only if the two markets are of very similar size. In otherwords the expenditure effect becomes crucial in determining the form of theFPE frontier.

To look at the incentives for factor mobility we need to take the furtherstep of deflating factor prices by the cost of living index. Consumers benefitfrom the full range of products supplied, so the argument in the cost ofliving index is the price index, qi, which as we have already seen, is decreas-ing with country size.

The implications of this are illustrated in figures 2.7 and 2.8. Figure 2.7is drawn for a relatively high value of T, and hence relatively steep FPE set– the dashed line O1deO2 . Above this line country 1 is poorly endowed withB relative to A, so b1.b2 and a1,a2, and conversely below the line.Adjusting for differences in price indices, qi, the lines aa and bb plot the setsof endowments on whicha1/e(q1)5a2/e(q2)andb1/e(q1)5b2/e(q2). The intui-tion behind the positions of these lines can be seen by considering point f.At f factor prices are equalised in terms of the numeraire, but economy 1has more x-sector production than economy 2 (it has a larger market and



Figure 2.7 Steep FPE set

Figure 2.8 Trade barriers lower

a more B-abundant endowment). It consequently has q1,q2, so country 1has a higher real return to both factors. This means that f must lie belowthe iso-real-a line, and above the iso-real-b line.

Given these real factor-price differences, the directions of factor mobil-ity are illustrated by the arrows on the figure. Qualitatively the picture issimilar to figure 2.5 (p. 38). If there is free mobility of both factors the equi-librium at C is unstable, but mobility of a single factor gives interior long-run equilibria.

Figure 2.8 gives the case when trade barriers are lower. All the lines in thefigure have been rotated clockwise around the central point, as we saw inour earlier discussion of the gradient of the FPE set. The importance of thisis that it brings a qualitative change in the stability of the system withrespect to factor B movements, as the segment DE of the bb line now hasnegative gradient. This segment is therefore unstable with respect to factor-B movements, so B- movements alone will have the effect of making theeconomies less similar, and causing agglomeration of x-sector productionin one or other of the economies – i.e. movement to the edge of thediversification set, along O1D or EO2.

To see this result most starkly, suppose that A is immobile, and the worldendowment is equally divided between economies. B is mobile, so givenhigh enough goods-trade barriers there is a stable equilibrium at C, withtwo economies identical, as illustrated in figure 2.7. Trade liberalisationnow has the effect of destabilising this equilibrium as it rotates the FPE setaround point C to a position such as that illustrated in figure 2.8. There arethree equilibria in the model. C remains an equilibrium although it is nowunstable, and points horizontally across on O1D or EO2 are stable equilib-ria. Nothing in the model predicts which of the two stable equilibria is likelyto be reached.

If this result seems familiar, it is of course, because it is no more than ageneralisation of the ‘core–periphery’ model of Krugman (1991). We haveplaced his analysis in a 232 technology more in line with standard tradetheory, and the generalisation of his result is that a ‘core–periphery’ struc-ture will emerge as a consequence of trade liberalisation if the factor usedintensively in the increasing returns sector is internationally mobile.

Pulling together results from the models with imperfect competition andincreasing returns to scale gives the following conclusions. If both factorsare mobile then the models predict complete concentration of economicactivity in one location. This is unsurprising – the models are constructedwith increasing returns and, if both factors are mobile, with no anchor tohold any activity in a declining country. The more interesting cases arethose in which one factor is mobile, and the other immobile. In the inter-mediate-goods model of sub-section 3.2 we see that trade liberalisation will


accentuate differences between the economies, as it raises the returns toagglomeration of industry. The ‘core–periphery’ model of section 3.3 givesa similar result, although in more dramatic form. In earlier cases tradeliberalisation affects factor mobility in a continuous way, but in this case wesee the possibility of discontinuous change. As trade barriers fall below acritical point so they destabilise the symmetric equilibrium, causing a‘core–periphery’ structure to emerge.

4 Concluding comments

The very simplest models of international trade offer the comforting resultthat trade liberalisation brings about a convergence of real incomes thatreduces migration pressure. We have seen in this chapter that there are manymore forces at work, both in perfectly and imperfectly competitive environ-ments, creating the possibility that trade liberalisation may promote factormobility, rather than substitute for it. Of course, we know that it is alwayseasy to add distortions of one sort or another to a model to overturn resultsand establish perverse cases. This chapter has tried to strike a balancebetween moving beyond simple HO theory, yet not falling into the trap ofadding arbitrarily many distortions of various sorts. It has done so by stayingwithin the confines of what are probably the three most widely used modelsof international trade – HO, specific-factors and Helpman–Krugman –except for the addition of some costs to trade. Even within these confines avery rich set of outcomes is possible. Trade liberalisation can promote migra-tion into a labour-scarce economy (as with our US economy example) orskilled labour outflow (the ‘brain-drain’ case). Adding imperfect competitionand increasing returns to scale, trade liberalisation can destabilise a symmet-ric equilibrium, and cause a core–periphery structure to emerge.

It is rather unsatisfactory to conclude with the message that differentmodels produce different outcomes. However, it would indeed be surpris-ing if a single theory were appropriate to study trade and migration in suchwidely different contexts as migration between countries of the EU or statesof the USA, and between countries with widely different factor endow-ments, economic structures and per capita income levels. The challenge forresearch is now to identify circumstances where each model is appropriate.

APPENDIX

Figures in the chapter are generated by numerical simulation using the fol-lowing functions:


Consumer preferences are Cobb–Douglas with shares 1⁄2, 1⁄2 except insection 3.1 where m50.

(i) In the HO model cost functions take the form,

g5ai1/3bi

2/3, h5ai2/3bi

1/3

for the x- and z-sectors, respectively.(ii) In the specific-factors model cost functions take the form

g5ai1/2bi

1/2, h5ai1/2ci

1/2.

(iii) In the increasing-returns model the sub-utility function dual to theprice index is

Ui5 [nixii(e21)/e1njxji

(e21)/e]e/(e21).

In construction of figures, e55.Cost functions in section 3.2 take the form

g5ai1/3bi

2/3, h5ai1/2qi

1/2,

and in section 3.3 take the form

g5ai1/3bi

2/3, h5ai2/3bi

1/3.

NOTES

This research is supported by by the UK ESRC-funded Centre for EconomicPerformance at the London School of Economics and by the British TaiwanCultural Institute. Thanks to participants in the Venice conference and to K.Harley, K. Middelfart-Knarvik, D. Puga and A. Sapir for helpful comments.1 Dixit and Norman (1980) resolve the difficulty of defining relative factor inten-

sity in multi-factor models precisely by defining it in terms of the effects of goodsprices on factor prices.

2 Or to be precise, x-trade ceases. If factors consume in the country of residencethis implies that all trade ceases, but if factors were to repatriate income then z-trade would be required to balance international payments.

3 See also Neary (1995) for analysis of factor mobility in the specific-factors model.

4 Although the cc-curve is constructed with returns in the host country equalised,the picture is qualitatively similar if capital moves in response to nominalreturns.

5 As country 1’s endowments of A and C expand so does its production of man-ufactures (z-sector), so if intra-industry trade is superimposed on the modeltotal trade volumes might increase even though inter-industry trade volumesdecline.

6 In the applications of the model developed below we shall look only at cases inwhich x-sector output is used either as a consumer good or as an intermediate, sothis assumption is not restrictive.


7 See Venables (1996) for analysis of a variant of this model, and Faini (1984) andMarkusen (1989) for studies of the role of non-traded inputs and producerservices in determining trade and growth. In this model there may be multipleequilibria even with immobile factors. We rule this out by setting T sufficientlyhigh to ensure uniqueness.

8 Since the numeraire is the only consumption good (m50) we do not need to dis-tinguish according to where the consumption of mobile factors takes place.

9 Expressions (10) and (11) were derived under the assumption that both sectorsare active in both countries.

REFERENCES

Bairoch, P. (1988). Cities and Economic Development: From the Dawn of History tothe Present (Chicago: Chicago University Press)

Begg, D. et al. (1992). Is Bigger Better? The Economics of EC Enlargement (London:CEPR)

Dixit, A.K. and V.D. Norman (1980). The Theory of International Trade(Cambridge: Cambridge University Press)

Dixit, A.K. and J.E. Stiglitz (1977). ‘Monopolistic Competition and OptimumProduct Diversity’, American Economic Review, 67, 297–308

Ethier, W.J. (1982). ‘Decreasing Costs in International Trade and Frank Graham’sArgument for Protection’, Econometrica, 50, 1243–68

Faini, R. (1984). ‘Increasing Returns, Non-traded Inputs and RegionalDevelopment’, Economic Journal, 94, 308–23

Harley, K. (1978). ‘Western Settlement and the Price of Wheat, 1872–1913’, Journalof Economic History, 38, 865–78

(1980). ‘Transportation, The World Wheat Trade and the Kuznets Cycle’,Explorations in Economic History, 17, 218–50

Helpman, E. and P. Krugman (1985). Market Structure and Foreign Trade(Cambridge, MA: MIT Press)

Krugman, P.(1980). ‘Scale Economies, Product Differentiation and the Pattern ofTrade’, American Economic Review, 70, 950–9

(1991). ‘Increasing Returns and Economic Geography’, Journal of PoliticalEconomy, 99, 483–99

Markusen, J.R. (1983) ‘Factor Movements and Commodity Trade asComplements’, Journal of International Economics, 13, 341–56

(1988). ‘Production, Trade and Migration with Differentiated Skilled Workers’,Canadian Journal of Economics, 21, 492–506

(1989). ‘Trade in Producer Services and in Other Specialized IntermediateInputs’, American Economic Review, 74, 85–95

Markusen, J.R. and J. Melvin (1981). ‘Trade, Factor Prices and the Gains fromTrade with Increasing Returns to Scale’, Canadian Journal of Economics, 14,450–69

Markusen, J.R. and A.J. Venables (1996). ‘The Theory of Endowment, Intra-indus-try and Multinational trade’, CEPR, Discussion Paper, 1341

Mundell, R.A. (1957). ‘International Trade and Factor Mobility’, AmericanEconomic Review, 47, 321–35


Neary, J.P. (1995). ‘Factor Mobility and International Trade’, Canadian Journal ofEconomics, special issue, 28, S4–S23

Norman, V.D. and A.J. Venables (1995). ‘International Trade, Factor Mobility andTrade Costs’, Economic Journal, 105, 1488–1505

Sapir, A. (1983). ‘Foreign Competition, Immigration and Structural Adjustment’,Journal of International Economics, 13, 381–94

Venables, A.J. (1996). ‘Equilibrium Locations of Vertically Linked Industries’,International Economic Review, 37, 341–59

Discussion

ANDRE SAPIR

Ever since Mundell’s (1957) path-breaking contribution, economists areaccustomed to thinking that trade liberalisation reduces the incentive forfactor mobility. The notion that trade and factor flows are substitutes haswidely percolated outside the economics profession. For instance, politi-cians in the European Union (EU) and in the United States (USA) haveembraced regional trade liberalisation largely on the assumption thatincreased trade would reduce the incentive for immigration of labour fromtheir poorer regional partners.

Mundell’s intuitive result rests crucially on two neo-classical assump-tions. First, trade conforms to the Heckscher–Ohlin (HO) model, whichimplies, inter alia, that trade liberalisation tends to equalise factor rewardsacross countries. Second, international factor mobility is solely a functionof differences in international factor prices. Therefore, provided that itachieves full factor-price equalisation, free trade completely eliminates theincentive for factor mobility.

Over the past 15 years, a number of contributions have investigated theimpact of alternative trade models on the relationship between trade liberal-isation and factor mobility. Chapter 2 by Tony Venables, who is himself oneof the contributors in this field, provides a splendid overview of this recentliterature. The chapter considers two alternatives to the HO workhorse: theRicardo–Viner (RV) specific-factors model, and the Helpman–Krugman(HK) imperfect-competition model. Using a unified approach encompassingall three models, Venables shows that the Mundellian result does not neces-sarily carry through to the other two models.

48 Discussion by André Sapir

While I have nothing to add to the finding that, in theory, there is a‘possibility that trade liberalisation may promote factor mobility, ratherthan substitute for it’ (p. 45), I would like to dispute the conclusion of thechapter. There, Tony states that, since ‘different models produce differentoutcomes . . ., [t]he challenge for research is now to identify circumstanceswhere each model is appropriate’ (p. 45). My feeling is that in the cases listedabove, namely regional trade liberalisation between the EU and EasternEurope or the Mediterranean countries, or between the USA and Mexico,we already know fairly well which model is most appropriate. Because eachof these regional trade schemes involves countries with significantlydifferent factor endowments, I feel confident that the HO model is moreappropriate than any of the other two models considered in the chapter.Does this mean that I conclude that EU and US politicians are necessarilywell founded in accepting that trade and factor flows are substitutes?

The answer is ‘yes’, but with an important proviso. The literaturereviewed by Tony Venables takes for granted the assumption that factormobility is purely a response to international factor-price differences.Clearly, this is a major weakness which needs to be addressed by the liter-ature. We know that foreign direct investment (FDI) responds to manyother economic forces besides international profit differences. In particular,FDI often occurs between countries with similar factor rewards.

Equally, international labour migration is affected by other factorsbesides international wage differentials. In particular, I find it sensible toassume that migration depends not only on wage differentials, but also onthe income level in the sending (i.e. low-income) country. Hence, the rela-tionship between trade liberalisation and migration could be quitecomplex. One possible such relationship might be as follows. Below acertain lower-income threshold in the sending country, trade liberalisationand migration might be complements since the former tends to fosterincome, thereby allowing potential migrants to meet the cost of movingabroad. Above this lower-income threshold, trade liberalisation and migra-tion are substitutes, at least until an upper-income threshold. Trade liber-alisation and migration are likely to foster income in the sending country.Once a certain upper threshold is reached, migration is likely to slow downor perhaps stop completely, even though the absolute income differencewith richer countries might remain substantial. This is due to the fact thatindividuals are likely to prefer living in their own rather than in a foreigncountry. Therefore, trade liberalisation and migration might be neither sub-stitutes nor complements above this upper-income threshold.

Assuming that migration indeed responds not only to wage differentialsbut also to wage levels, the important proviso in assessing the relationshipbetween regional trade liberalisation and migration therefore relates to the


income thresholds discussed above. In particular, the practical issue iswhether wage (or income) levels in Eastern Europe, the Mediterraneancountries or Mexico are above the lower, but below the upper, incomethreshold. My guess is that the answer is clearly ‘yes’ for Eastern Europeand Mexico, but only marginally so for the Mediterranean countries. Theimplication is that the hunch of EU and US politicians is probably correct:regional trade liberalisation is likely to reduce migration flows towards thenations they govern.

REFERENCES

Mundell, R. (1957). ‘International Trade and Factor Mobility’, American EconomicReview, 47, 321–35

50 Discussion by André Sapir

3 Regional integration, trade andmigration: are demand linkagesrelevant in Europe?

RODNEY D. LUDEMA AND IAN WOOTON

1 Introduction

In this chapter we examine the consequences of increased economicintegration between nations within a region. We assume that the countriesalready have made significant strides towards a common market by lower-ing trade costs and impediments to factor mobility. However, some barri-ers remain, and we shall consider economic integration as a combinationof dropping barriers to factor mobility and reducing transport costs.1 Weare interested in determining the circumstances under which economicintegration will lead to agglomeration of economic activity, whereby onepart of the region will accumulate virtually all of the manufacturing activ-ity, while the remainder will become largely de-industrialised. This clearlyis an important issue for a regional grouping such as the EU, particularlygiven its recent efforts to deepen the level of integration between memberstates.

We adopt Krugman’s general equilibrium, economic-geography model(1991a) to characterise what we refer to as the ‘labour-demand’ side of themodel.2 This produces the agglomerative forces that can lead to acore–periphery pattern through backward and forward ‘demand linkages’.The backward linkage captures the notion that manufacturing productionwill tend to concentrate where there is a large market, but that the marketwill be large where manufacturing production is concentrated (because thisis where the manufacturing workers live and consume). This is reinforcedby the forward linkage in which, other things being equal, the cost of livingwill be lower in the country with the larger manufacturing sector, becauseconsumers in that location can rely less on imports (which are subject totransport costs). Both of these linkages will tend to draw manufacturingworkers into the core, leaving the remainder of the region a rural hinter-land. In addition to the barriers to trade in manufactures, we allow for theimperfect mobility of manufacturing labour, which results in an upward-sloping international labour-supply schedule. Thus workers are assumed to

51

have preferences for living in a particular country and, in order to inducethem to migrate, a relatively higher real wage will have to be offered in theother country. We believe this to be a useful innovation for economic-geog-raphy models applied to regions in which language and cultural barriers aresignificant, despite the lack of formal, legal impediments to mobility.

The benefits of introducing imperfect labour mobility into this frame-work are more than merely adding realism. First, it substantially expandsthe possible outcomes arising from economic integration. Depending onthe relative importance of agglomerative forces and labour mobility,integration may have no impact on the location of manufacturing activityor it may drive it to complete agglomeration, as in Krugman (1991a). Anintermediate and more interesting possibility is that integration, in the formof the progressive reduction of trade barriers, may be accompanied by athree-stage pattern of industrial production: international diversification,followed by the agglomeration of activity into a core–periphery pattern,followed by the re-industrialisation of the periphery.

Second, introducing imperfect labour mobility allows us to examine theeffects of a dual approach to integration, that of reducing trade barriers andreducing barriers to mobility. We show that it may be possible to eliminatetemporary agglomeration, and its concomitant disruptions, by appropri-ately sequencing the two forms of integration.

Krugman (1991a) originally set out the demand-linkages model, in part,to explain the core-periphery patterns of North America and Europe.3

However, Krugman (1991b) cites data suggesting that Europe’score–periphery pattern is more of an income than an employment pattern.That is, as compared to the USA, Europe exhibits more pronounceddifferences in per capita income across its constituent states, but less pro-nounced differences in the concentration of manufacturing labour.Krugman’s conclusion is that the demand linkages model does not explainthis. This is echoed by Venables (1994), who writes, ‘migration in Europe isperhaps insufficiently responsive to economic factors for [the demandlinkage] mechanism to be of much relevance to European integration’.4

A second point mentioned in Krugman (1991b), as well as in Krugmanand Venables (1995), is that there are signs of a gradual re-industrialisationof the periphery, in North America, Europe and East Asia. Krugman andVenables (1995) use an alternative model, in which labour is internationallyimmobile (but intersectorally mobile), to generate a three-stage patternsimilar to the one described above. They contrast this alternative model tothe demand-linkage model (p.860) as follows:

Simple geography models like Krugman (1991a) respond in a monotoneway to declining transport costs: when these costs fall below a critical level,industry concentrates in one region. Here, because labor is immobile (and

52 Rodney D. Ludema and Ian Wooton

thus wage differentials between regions emerge), continuing reductions intransport costs eventually lead to a reindustrialization of the low-wageregion. We believe this is not just an artifact of the model: it represents areal distinction between interregional and international economicsbecause labor is in fact much less mobile between than within nations.

All of this appears to cast doubt on the demand-linkage model, particu-larly in its application to Europe. However, it is our view that the results ofthe demand-linkage model that appear at odds with the facts are essentiallyartefacts of the assumption of perfect labour mobility. With the assump-tion of imperfect labour mobility (not complete immobility), we show thatthe experience of Europe is completely consistent with the demand-linkagemodel. It is entirely possible for a region like the USA, because of its greaterlabour mobility (flatter labour-supply schedule), to exhibit more pro-nounced differences in employment and less pronounced differences in percapita income than Europe. It is also entirely possible to generate a three-stage integration pattern.

The layout of the chapter is as follows. In section 2 we give the details ofthe structure of Krugman’s model, which provides the labour-demandschedule. We follow this, in section 3, by an examination of the influence oflocational preference on labour supply. The resulting equilibria are dis-cussed in section 4, while section 5 consists of some comparative staticsexercises of more liberal trade and factor mobility. Section 6 draws someconclusions.

2 The model

We model labour demand using Krugman’s (1991a) model of economicgeography. This has two sectors: agriculture and manufacturing.Agriculture produces a homogeneous good according to constant returnsto scale with farm labourers as the sole, sector-specific factor. A unit of theagricultural product requires the input of one labourer. The agriculturalworkers are internationally immobile. In contrast, the manufacturingindustry produces differentiated products with increasing-returns-to-scaletechnology. It employs internationally mobile labour. Each variety of themanufactured good is produced with increasing returns, these scaleeconomies leading to the concentration of production of each variety in asingle plant.

The homogeneous agricultural commodity can be traded costlessly,while trade in the differentiated good involves iceberg-type trade costs. Asa result, there may be a tendency for plants to locate in the country in whichdemand for the variety is greater, so as to minimise these trade costs. Thusthe nominal wage may be higher in the larger country. In addition, the cost

Are demand linkages relevant in Europe? 53

of living differs internationally as imported varieties attract trade costs.Workers may therefore have an incentive to move to the industrialised ‘core’where the real wage is higher than in the agricultural ‘periphery’, and thisconcentration of workers increases the size of the market in the core andreinforces the decisions of firms to locate there.

All individuals share the same Cobb–Douglas utility function:

U5CMm CA

12m, (1)

where CA is consumption of the agricultural good and CM is consumptionof the aggregate manufactured good. Therefore agents will always spend theproportion m on manufactures. The manufactures aggregate is defined by:

CM5 , (2)

where N is the (large) number of potential products and s.1 is the elastic-ity of substitution among the varieties.

There are two countries in the economic region and two sector-specificfactors of production in each country. Agricultural workers are interna-tionally immobile, the supply in each country being (12m)/2.Manufacturing workers may move between countries. Let L1 and L2 be thenumbers of workers in country 1 and country 2, respectively, where:

L1 1L25m. (3)

Production of a representative variety of the manufactured good pro-duced in country i involves a fixed cost and a constant marginal cost:

LMi5a1bxi. (4)

where xi is output of that variety and LMi is the labour used in its produc-tion. Agricultural output is freely traded and hence agricultural workers’earnings are the same in both countries and will be our numeraire. Formanufactures, a proportion t,1 that is shipped from one country arrivesin the other.

There is a large number of manufacturing firms, each producing a singleproduct and facing an elasticity of demand equal to s. Profit-maximisingbehaviour will lead firms to price their variety at a mark-up over marginalcosts:

pi5 bwi, (5)

where wi is the wage rate of workers in country i. Relative prices must then be:

. (6)p1

p2

5w1

w2

1 s

s 2 12

s

(s 2 1)4s 2 1

s3oN

i51

ci


Free entry implies zero profits, hence:

(pi 2bwi)xi 5awi, (7)

which implies:

x15x25 . (8)

The numbers of varieties of the manufactured good produced in country iis ni. Combining (4) and (8) yields:

. (9)

The quantities of each variety consumed will depend on their prices. Pricesof imports must take into account the transport cost. Hence, for country 1,

5 5 . (10)

where cij is the quantity consumed in country i of a variety manufacturedin country j.

The share of the manufacturing labour force in country 1 is denoted f;L1/m. Define z11 to be the ratio of country 1’s expenditure on local productsto its expenditure on varieties imported from country 2, while z12 is the ratioof country 2’s expenditure on manufactured imports to its expenditure onlocal goods. Therefore:

z11 5 ;

z12 5 . (11)

The total income of workers equals the total spending on the products thatthey produce. If Yi is the national income of region i (including agriculturalincome), then the nominal wages of an individual worker in each region are:

w15 Y11 Y2 ;

w25 Y11 Y2 (12)

where:

Y15 1w1 fm

Y25 1w2(12 f )m. (13)1 2 m

2

1 2 m

2

41 11 1 z12

231 11 1 z11

211 2 f

41 z12

1 1 z1221

f 31z11

1 1 z112

f1 2 f

12s

1w1

w2t2

f1 2 f

12s

1w1tw2

2

2s

1w1tw2

22s

1p1tp22c11

c12

n1

n2

5L1

L2

a(a 2 1)b


For a given allocation of manufacturing labour f, (11)–(13) are a systemthat determines nominal wages, incomes and spending patterns.

Because of transport costs, the prices of manufactures differ in the twocountries and therefore the cost of living will be a determinant of a worker’sdecision as to location. The price indices of manufactured goods are:

P1 5 fw112s1(12 f) ;

P2 5 f 1 (12f)w212s . (14)

Real wages of workers in each country are:

v1 5w1P12m;

v2 5w2P12m. (15)

Figure 3.1 illustrates the labour-demand schedule KK, plotting f, the pro-portion of the (manufacturing) labour force in country 1, against v5v1/v2,the real wage in country 1 relative to that in country 2. Using an elasticityof substitution between manufactures (s54, as in Krugman, 1991a), thefigure shows dramatic differences in the schedules for different trade costs.With high trade costs, the forces of agglomeration are low – when there areconsumers in both locations, firms will produce in both countries. This isdrawn as KK(t50.5). As trade costs fall, the stronger the agglomerativeforces become, as illustrated by KK(t50.6) and KK(t50.75).

11 2 s41w1

t 212s

3

11 2 s1w2

t 212s

43


Figure 3.1 KK, labour demand schedule

It is worth noting that having even more liberal trade makes the KK curveflatter. Indeed, as t approaches one, the KK curve approaches horizontal.That the slope of the KK curve rises and then falls as the transport cost isreduced is largely irrelevant in the presence of perfect labour mobility (thatis, as in Krugman, 1991a), but it turns out to be very important in the pres-ence of imperfect labour mobility, as we shall see in section 5.

3 Labour supply

Mobile workers are assumed to have different preferences over the locationin which they would rather live and work. Because of these, the relativewage that will have to be offered to induce each worker to migrate willdepend on the strength of her locational preference. We shall show that thiscan have a significant effect in mitigating the forces of industrial agglomera-tion that arise in the presence of increasing-returns-to-scale technology.

An industrial worker will make her decision as to where to live and workbased on a comparison of real wages, in the light of her strength of prefer-ences for living in each country. Why a worker prefers one country or theother is not explicitly modelled, merely being represented as a discount ratethat would have to be applied to the relative real wage in order to inducethe worker to move. However, real-world experience certainly provides agreat deal of casual evidence of non-pecuniary benefits that individualsenjoy from living in one country or the other.5 These preferences will berepresented by a density function and we shall examine how changes in thedistribution may alter the structure of industrial production within theregion.6

Let (gi be worker i’s discounting of the real wage offered in country 1,while (12gi) is her discounting of the real wage in country 2.7 Worker i willbe indifferent between the two locations when the discounted real wages areequalised, that is:

giv15(12gi)v2. (16)

We assume that workers’ preferences are distributed across the interval [0,1] and we will investigate the consequences of different distributions ofnational sentiment. Let the (gi be distributed according to a truncatednormal distribution:8

g(gi;s)5 , (17)

0,

f(gi;s) represents the cumulative normal distribution with mean of 0.5 andstandard deviation s.0. f(gi;s) is the probability density function of thesame distribution. By adopting this distribution, we are assuming that

0,gi,1;elsewhere.

5 f(gi;s)F(1;s) 2 F(0;s)


preferences are symmetric about the mean of 0.5 – that is, indifferencebetween the two locations. The greater is s, the stronger the location prefer-ences of some of the workers.

Figure 3.2 illustrates the distribution of workers for various values of s.When s50.05, the values of g are spread fairly widely from below 0.4 toabove 0.6.

A worker at one or other of these values would require a 50 per centpremium to induce her to move from her preferred location. This distribu-tion of preferences is drawn as MM(s50.05). Reducing the standard devia-tion concentrates preferences much more tightly around the mean,requiring smaller migration premia. As an example we have drawn MM(s50.005). Were s50, workers would have identical preferences and beindifferent between the two countries. Thus MM(s50) is a spike at themean of 0.5.

Let the marginal worker have preference weight gi5g, such that a workerwhose gi .g will choose to live and work in country 1. Then f, the propor-tion of manufacturing workers in country 1, will be determined by:

f512G(g;s), (18)

where G is the cumulative distribution function, the integral of (17). Define u;(12g)/g. This defines the marginal worker and hence, from (16), v5u.

Thus we can calculate the labour-supply schedule, showing the relativesupply of workers to country 1 given the relative real wage and the degreeof disparity in preferences over location. Figure 3.3 plots various labour-supply schedules. LL(s50.05) shows a steep, relatively inelastic schedule,


Figure 3.2 MM, distribution of preferences

corresponding to a fairly broad distribution of tastes. LL(s50.005) plots amore elastic labour-supply curve where workers need less to induce them tomigrate. LL(s50) is a perfectly elastic supply schedule corresponding to aworker being indifferent between locations, which is the case implicitlyassumed by Krugman (1991a).

4 Equilibrium

We have now established a labour-demand schedule, showing the propor-tion of the manufacturing labour force that will be employed in country 1as a function of the relative real wage offered by that country, given the levelof international transport costs. We have also found the labour-supplyschedule, showing the willingness of workers to take employment incountry 1 as a function of the real wage and for particular distributions ofpreferences over locations in the two-country region. Our next task is todetermine the equilibrium employment pattern, resulting from thesepreferences and transport costs.

An equilibrium arises either at a point of intersection between thedemand and supply schedules (an interior solution) or at either end-point,where all of the manufacturing labour is employed in a single country (acorner solution). Not all equilibria are stable solutions: a stable interiorequilibrium requires that the labour demand schedule (KK) have a slopeless than that of the labour supply schedule (LL) at the point of intersec-tion. A stable corner solution occurs when the LL schedule lies above the


Figure 3.3 LL, labour supply schedule

KK schedule at f50 or when the LL schedule lies below the KK scheduleat f51 (all manufacturing workers employed in country 2 or country 1,respectively).

If labour is in perfectly elastic supply, as in Krugman (1991a), thelabour-supply schedule LL(s50), illustrated in figure 3.4a, is flat at rela-tive wage v51. Three possible configurations of equilibria are possible,depending crucially on the level of the trade costs. The forces of


Figure 3.4 Equilibrium allocations

(a)

(b)

agglomeration are weak when transport costs are high. Thus, when tradecosts are high as in KK(t50.5), the labour demand curve is downward-sloping, crossing LL(s50) from above at the single equilibrium of diversi-fied production, f5 1⁄2. As trade costs are reduced the labour-demandschedule changes slope as the forces of agglomeration become stronger. Atsome intermediate levels of trade costs, such as KK(t50.6), the labour-demand schedule cuts v51 at three points, yielding five equilibria. Theinterior equilibrium of diversified production (f5 1⁄2) remains stable and,in addition, the two corner equilibria (corresponding to completeagglomeration in one or other of the countries) are stable, while the othertwo equilibria are unstable. If transport costs are reduced sufficiently, thelabour-demand curve becomes everywhere upward-sloping, as with KK(t50.75), and diversification becomes unstable and only agglomerated equi-libria are stable.

Now consider the implications of labour being in less-than-perfectly-elastic supply – that is, when the labour-supply schedule slopes upwards.One thing is immediately clear: when the labour-demand schedule is down-ward-sloping, as is the case for KK(t50.5) in figure 3.4a, a diversified equi-librium will continue to be the only possibility, irrespective of the degree ofinternational labour mobility. However, an inelastic labour supply will mit-igate agglomerative forces and diversification may remain a stable equilib-rium even when demand linkages are relatively strong. Thus in figure 3.4a,the curve LL(s50.05) is steeper than KK(t50.75) and consequently man-ufacturing will be spread across the region, despite the agglomerativeforces. Agglomeration of manufacturing activity is not an equilibrium inthe case of LL(s50.05), as this corresponds to a distribution of workers’preferences over locations such that substantial premia would have to bepaid to induce many workers to migrate.

What would be the result were workers, on the whole, less concerned withlocation? With LL(s50.005) workers are closely concentrated around themean (indifference) and the tails of the distribution are virtually empty. Yeteven this degree of locational preference may be sufficient to offset theforces of industrial agglomeration and result in diversification across theregion. In figure 3.4b, the labour demand schedule KK(t50.63) is posi-tively-sloped which, if labour were freely mobile, would result in anagglomerated equilibrium. But even with a distribution close toindifference the equilibrium can be diversified production.

However, agglomerative forces may dominate, resulting in agglomerationoutcomes becoming stable equilibria in addition to or in place of the diver-sified equilibrium. Thus KK(t50.65) intersects LL(s50.005) five times: astable, diversified equilibrium bounded by two unstable equilibria whichare themselves flanked by stable, agglomerated equilibria. But there may


now be some workers for whom the economic inducements will never besufficient for migration. Consequently, the agglomeration is not completeand a small share of the manufacturing labour force remains in the periph-ery. This is also the case for outcomes where only the agglomerated equi-libria are stable, as when the labour-demand schedule KK(t50.67)intersects LL(s50.005).

This situation is interesting, both in being consistent with casual empir-icism (no country is left entirely without a manufacturing sector) and inproviding a genuine role for government policy to encourage industriallocation (beyond the ‘bang-bang’ nature of policies when agglomeration iscomplete and migration incentives lead to a total move of all manufactur-ing from one country to the other).9

How consistent is our model with reality? Let us compare some stylisedfacts for the USA and the EU. The American economy can be character-ised as a fully integrated market, with few or no barriers to inter-state factormovements and low internal-trade costs. Consequently, it will have anupward-sloping (but not very steep) labour-demand curve and a virtuallyflat labour-supply schedule. The resulting equilibrium yields a highlyagglomerated economy, with a relatively low income differential. InEurope, trade barriers are low, but higher than those in the USA, whileimpediments to factor movements are also greater (for both administrativeand cultural reasons). Agglomeration forces may still be strong (KKupward-sloping), but both the labour-demand and labour-supply curveswill be steeper than those of their American counterparts. Manufacturingwill have a partially agglomerated equilibrium fairly close to diversificationand there will be a higher wage differential than that in the USA. Thus theexperience of Europe relative to the USA is quite consistent with thedemand-linkage model.

We have shown that is possible to construct cases where the equilibriumis diversification or partial agglomeration depending on the strength of theagglomerative forces on the labour-demand side and the immobility orunwillingness of labour to move on the supply side. What we now want toshow is that, even when labour is initially relatively immobile (as, say, inEurope), the demand linkages that generate agglomerative forces will beimportant in determining the response of the region to market integration.

5 Economic integration

We model economic integration in two ways: the lowering of trade barriersand an increase in the willingness of labour to migrate. Trade is liberalisedas t is increased, free trade being attained when it reaches unity. Improvedlabour migration will be reflected in weaker locational preferences of all


workers (that is, reductions in s), such that a lower premium is necessary toinduce each worker to move to the other country.

While liberalisation of trade and factor movements will lead to moreefficient production of manufactures within the region, it may have seriousdistributional consequences. Specifically, whenever agglomeration occurs,those factors left in the periphery will suffer lower real incomes as fewermanufactures are locally produced, resulting in a higher cost of living. Thelocal governments will then have an incentive to try to ensure that the coredevelops in their country, an issue examined in another paper (Ludema andWooton, 1998).

5.1 Trade liberalisation

With freely internationally mobile labour, trade liberalisation will eventuallyreach the point at which agglomerative forces are sufficiently strong that theregion will jump from having industry dispersed between the countries to acore–periphery pattern of production. Once this jump has occurred, thepattern will be maintained for any further trade liberalisation. This will not,in general, be the outcome of trade liberalisation when labour is less freelymobile between the countries in the region. Instead, two possibilities arise.

First, lowering trade costs will affect the allocation of labour betweencountries, leading to agglomeration only in situations in which the slope ofthe labour-demand schedule is positive (so that the forces of agglomerationare strong) and steeper than the labour-supply schedule. Therefore, if thelabour-supply line remains steeper than the labour-demand schedule at anylevel of trade costs, 0# t#1, then agglomeration will never occur, despitethe trade liberalisation. Thus, unless the region has sufficiently mobileworkers, agglomerative forces can be held at bay and the region will alwayshave widely dispersed industry. This situation seems to be what Venables(1994) has in mind when he suggests that European migration is ‘perhapsinsufficiently responsive to economic factors’ for agglomerative forces ‘tobe of much relevance to European integration’.

The second, and more intriguing, possibility arises when labour is morefreely (but not completely) mobile. This is illustrated in figure 3.5 in whichLL(s50.005) represents a fairly elastic labour supply and the KK schedulesshow labour demand under increasingly liberal trade (from fairly hightrade barriers with KK(t50.6) to free trade for KK(t51)). In these cir-cumstances demand linkages can certainly have a role to play. The initialstages of trade liberalisation will shift the labour-demand schedule from anegatively-sloped function (at the diversified equilibrium of f5 1⁄2) to onethat is positively-sloped. As the slope increases, there may be a range oftrade costs for which the labour-demand curve becomes steeper than the


labour-supply curve at f5 1⁄2. Consequently, manufacturing will graduallymove to a partially agglomerated equilibrium with all but the least mobileworkers from one country having moved into the industrial core.

However, this agglomeration will be only a temporary phenomenon astrade liberalisation proceeds. If trade costs continue to fall, the labour-demand schedule gets flatter again, as trade liberalisation weakens theagglomerative forces. (At free trade, labour demand is perfectly elasticbecause, even with demand linkages, there are no benefits to agglomerationif goods can be costlessly moved around the region.) Therefore workers willbegin to desert the core and return to their preferred locations. Eventually,trade liberalisation will have progressed sufficiently for the region to returnto a stable, diversified equilibrium.

Thus weak locational preferences (less-than-perfectly mobile labour)may result in three phases of trade liberalisation. It may initially drive pro-duction from diversification into partial agglomeration and then back intodiversification, as the curves cross then re-cross. These effects are qualita-tively different from those derived under the assumption of freely mobilelabour, both in the temporary nature of the agglomerated equilibrium andthe fact that the degree of industrial concentration (when agglomerationoccurs) varies with the transport cost.

We sketch out the shift from diversification to agglomeration (and backagain) as trade is liberalised in figure 3.6. Trade liberalisation is indicatedby declining (12 t). Stable equilibria are drawn as solid lines, while thedotted lines indicate unstable equilibria. Were labour freely mobile, we


Figure 3.5 Changing trade costs

would see the now familiar ‘pitchfork’ diagram where trade liberalisationinitially adds stable agglomerated equilibria to the initial unique equilib-rium of diversified production.10 Further reduction in trade barrierschanges the slope of the KK curve and only the diversified equilibria remainstable until all trade impediments are eliminated. Instead of this, we findthat the agglomerated equilibria do not involve such an extreme concentra-tion of manufacturing. As trade liberalisation progresses, initially theagglomerative forces increase and the equilibria shift further apart.Eventually though, as was pointed out above, continued demolition oftrade barriers lessens the incentive for agglomeration. Consequently, theagglomeration will become less pronounced, ultimately disappearing as theregion returns to diversified production. Thus, in place of the pitchfork wehave a shape somewhat more reminiscent of a spearhead.

5.2 Increased labour mobility

If economic integration can also take the form of reduced barriers to labourmigration (say, in the form of improvements in foreign-language instruc-tion), then the labour-supply schedule will rotate, becoming flatter. Whenagglomerative forces are weak (the labour-demand schedule being nega-tively- sloped) increased labour mobility, for any given level of trade costs,will have no effect and diversification remains the sole equilibrium.

For stronger demand linkages, the labour-demand schedules will beupward-sloping. If initially the labour supply is relatively inelastic, increasedlabour mobility may cause the equilibrium allocation to change fromdiversification to increasing agglomeration as the slope of the labour-supply


Figure 3.6 Trade liberalisation and location of industry

schedule falls. Thus improving labour mobility will enable agglomerativeforces to have an increasing effect.

5.3 Comprehensive regional integration

We now consider moves towards establishing a common market in theregion with free trade and unrestricted international labour movements.Trade liberalisation is clearly a worthwhile objective in this setting. Indeed,we have shown that completely free trade in goods eliminates the need ofmovements of factors of production and, consequently, this may be theonly policy that the governments of the region need pursue. Though labourmobility would not be necessary to achieve economic efficiency, it may bedesirable in meeting the social or political aspirations of the region. Theresults of sub-sections 5.1 and 5.2 indicate that the timing of the two ele-ments of regional integration – trade liberalisation and increased labourmobility – may be crucial in avoiding temporary dislocations in productionand local employment.

If labour is initially relatively immobile, trade liberalisation will takeplace without inducing any international migration. Once a (fairly) liberaltrade regime is established, a diversified equilibrium will arise. In that situa-tion factor movements could be allowed, or even encouraged, though nonet international movements will be necessary to maintain the diversifiedequilibrium. If, however, trade liberalisation takes place in the presence ofreasonably footloose labour, it might lead to swings towards agglomerationand then back to diversification, with the associated emigrations and returnmigrations of manufacturing workers. These temporary dislocations in thelabour market could cause a great deal of undesirable social upheavalthroughout the region.

These problems might be avoided by sequencing the integration policies.Specifically, a restriction on international movements of labour while thetrade barriers are being dismantled would suppress the labour migrationsthat would be induced by the agglomerative forces. These forces decline astrade barriers fall. Hence, once free trade in manufactures has been estab-lished, barriers to factor mobility could be reduced.

This policy prescription is already foowed by the EU in agreements withnew members. The EU and acceding countries initially eliminate their bilat-eral tariffs, while restrictions on factor mobility are maintained for severalyears before being lowered. This ensures that industrial structures canadjust without inducing problems of large, temporary international labourmigrations. In particular, it ensures that the acceding country has a betterchance of retaining its manufacturing sector and does not experience de-industrialisation through its union with a larger economic entity.


6 Conclusions

This chapter uses elements of the new economic geography to examineregional integration. A common market is attained by the elimination ofboth barriers to trade in goods and impediments to factor mobility. Whilea lot of attention has been given to the effects of trade liberalisation inmodels of economic geography, our explicit consideration of issues offactor mobility is both novel and yields outcomes that differ significantlyfrom those where a single regional market for homogenous labour isassumed.

If workers differ in their willingness to move between regions, the equi-librium regional distribution of industry may be affected. First, the reluc-tance of some workers to move may constrain the forces of industrialagglomeration, leading to countries retaining shares of industrial produc-tion which would have been drawn into a core area of the region if labourwere more freely mobile. Second, even when agglomeration occurs, it willnot be complete, as some workers will remain unwilling to migrate in equi-librium.

Trade liberalisation will induce a smoother transition in regional manu-facturing activity than occurs with freely mobile labour. However, as adiversified industrial structure will be the equilibrium for both high tradecosts and low trade costs, it may be advisable to restrict labour movementsuntil the trade liberalisation phase is complete.11

We conclude that economic-geography models are indeed relevant tocommon markets, such as Europe. When demand linkages are sufficientto generate strong agglomerative forces, the national governments willhave to take some care in choosing both the depth of integration and thetiming of its achievement.

NOTES

Paper presented at the Venice conference (24–25 January 1997). We wish to thankGiorgio Basevi, Julia Darby, Riccardo Faini and Jim Malley for their help, com-ments, and suggestions.11 Thus the single-market initiative of the EU sought to advance the ‘four free-

doms’, two of which were free trade in goods and free mobility of labour.12 Krugman considers regions within a country. However, given our interest in the

EU, we re-label the model to have international migration between countries thatare part of a common factor market.

13 Europe’s hub is ‘somewhere in or near Belgium’ (Krugman, 1991a, 484).14 If agglomerative forces are strongest at the level of particular industries, and not

manufacturing as a whole, Krugman and Venables (1995) suggest that the eco-nomic geography of Europe will be affected, despite a relatively low degree ofinternational labour mobility.


15 Indeed, a small differential in real wage rates between Glasgow and Washington,DC would not be sufficient to induce either author to change location.

16 This approach to modelling labour supply is similar to that of Faini (1996).17 For analytical convenience we have assumed 0#gi#1. The level of gi is not

important, the relevant number is the ratio gi /(12gi), which can take any non-negative value.

18 We adopt a normal distribution for convenience, and are grateful to GiorgioBasevi for encouraging us to use the normal. We wish to have a distribution oftastes such that some workers are more willing to migrate than others and thatallows us to make parametric changes to the degree of labour mobility.

19 We investigate the roles of national governments in influencing location else-where (Ludema and Wooton, 1998).

10 See, for example, Puga (1996).11 Indeed, countries acceding to membership of the EU typically lower trade bar-

riers before intra-regional factor mobility is permitted.

REFERENCES

Faini, R.(1996). ‘Increasing Returns, Migrations and Convergence’, Journal ofDevelopment Economics, 49, 121–36

Krugman, P. (1991a). ‘Increasing Returns and Economic Geography’, Journal ofPolitical Economy, 99, 483–99

(1991b).Geography and Trade (Cambridge, MA: MIT Press)Krugman, P. and A. J. Venables (1995). ‘Globalization and the Inequality of

Nations’, Quarterly Journal of Economics, 110, 857–80Ludema, R. D. and I. Wooton (1998). ‘Economic Geography and the Fiscal Effects

of Regional Integration’, CEPR, Discussion Paper, 1822Puga, D. (1996). ‘The Rise and Fall of Regional Inequalities’, CEPR, Discussion

Paper, 314Venables, A. J. (1994). ‘Economic Integration and Industrial Agglomeration’,

Economic and Social Review, 26, 1–17

Discussion

GIORGIO BASEVI

1 Introduction

In their chapter 3, Ludema and Wooton (hereafter, LW) integrateKrugman’s (1991) model by considering an imperfectly elastic supply oflabour with respect to international differences in real wages. Thus, they get

68 Discussion by Giorgio Basevi

a larger set of possible outcomes, some of which throw useful insights onthe question of the preferred sequence between liberalisation of trade ingoods, or of labour movements, as the process of international economicintegration proceeds. I shall concentrate my comments on two main issues,and briefly mention a few other points.

2 The first issue: stable and unstable equilibria

The number of possible outcomes identified by LW is incomplete. In orderto see it, notice that the demand for labour in Krugman’s model, and thesupply of labour in LW’s model, are both nonlinear in the (f, v) space;while their analytical expression is not easily derived, numerical simulationssuggest that they may present at least one point of inflection. Moreover, thedemand function is not monotonic in t. It is negatively-sloped for hightransport costs (i.e. for a low value of t), it gets positively-sloped as trans-port costs are reduced (i.e. for a high value of t), but it flattens as they tendto zero (i.e. as t tends to unity). As for the supply function, it is positively-sloped but, in LW’s example, it gets flatter as labour becomes internation-ally more mobile (i.e. as a5b tend to infinity).

Consider first the case analysed by Krugman (i.e. when the supply oflabour function is horizontal – when a5b tend to infinity in LW’s model).We have three possible sub-cases:

(1a) The demand for labour function is negatively-sloped (high transportcosts, or a low value of t). In this case (figure D3.1a) there are threeequilibria, but the two corner ones are unstable, while the uniqueinterior equilibrium is stable. The two countries get the same (half)share of manufactures.

(1b) The demand for labour function is positively-sloped (low transportcosts, or a high value of t). Also in this case (figure D3.1b) there arethree equilibria, but the two corner ones are now stable, while theunique interior equilibrium is unstable. One or the other country getsall of the manufacturing sector.

(1c) The slope of the demand for labour function alternates in sign (inter-mediate transport costs, or an intermediate value of t). In this case(figure D3.1c) there are five equilibria, the two corner ones are stable,like the middle one, while the two intermediate ones are unstable. Thusthere are two possibilities: complete agglomeration of industry in oneor the other country, or the two countries get half of the industrialsector each.

Consider now the case analysed by LW – i.e. when the supply of labour func-tion is positively-sloped (when 0#a5b #1`). We have four possible sub-cases:



Figure D3.1 Supply of labour function horizontal

(a)

(b)

(c)

(2a) The demand for labour function is negatively-sloped (high transportcosts, or a low value of t). In this case (figure D3.2a) there are threeequilibria, but the two corner ones are unstable, while the uniqueinterior equilibrium is stable. The two countries get the same (half)share of manufactures. Thus the implications of this case are the sameas those of case (1a).

(2b) The demand for labour function is positively-sloped, it cuts the supplyof labour function only once, and with a larger slope in the neighbour-hood of the crossing point. In this case (figure D3.2b) there are threeequilibria, the two corner ones are stable, while the unique interiorequilibrium is unstable. One or the other country gets all of the man-ufacturing sector. Thus the implications of this case are the same asthose of case (1b).

(2c) The demand for labour function is positively-sloped, it cuts the supplyof labour function only once, and with a smaller slope in theneighbourhood of the crossing point. In this case (figure D3.2c) thereare three equilibria, the two corner ones are unstable, while the uniqueinterior equilibrium is stable. The two countries get the same (half)share of manufactures. Thus the implications of this case are the sameas those of cases (1a) and (2a).

(2d) The demand for labour function is positively-sloped, it cuts the supplyof labour function three times, with a smaller slope in the neighbour-hood of the first crossing point, with a larger slope in the neighbour-hood of the second crossing point, and again with a smaller slope inthe neighbourhood of the third crossing point. In this case (figureD3.2d) there are five equilibria, the two corner ones are unstable, likethe middle one, while the two intermediate ones are stable. Thus thereare two alternative solutions, both with incomplete agglomeration ofindustry in one or the other country. This is the case emphasised byLW.

(2e) The demand for labour function is positively-sloped, it cuts the supplyof labour function three times, with a larger slope in the neighbour-hood of the first crossing point, with a smaller slope in the neighbour-hood of the second crossing point, and again with a larger slope in theneighbourhood of the third crossing point. In this case (figure D3.2e)there are five equilibria, the two corner ones are stable, like the middleone, while the two intermediate ones are unstable. In this case thereare two possibilities: complete agglomeration of industry in one or theother country, or the two countries get half of the industrial sectoreach. This is the same result as in case (1c).

It follows from this last case, which LW have not considered, that theirconclusions (‘if workers have some, limited location preferences, . . .



Figure D3.2 Supply of labour function positively-sloped

(a)

(b)

(c)

complete agglomeration is not an equilibrium outcome, while an equalallocation of workers ... is an unstable equilibrium’), and (‘If workers differin their willingness to move between regions, ... even when agglomerationoccurs, it will not be complete’), are not correct generally. Moreover, theirresult is based on the particular inflection of the supply of labour function,which initially is concave downward and then turns concave upward. This,however, depends on the assumption that a5b in the frequency distribu-tion of workers’ preferences. With a large enough relative to b, it is possi-ble to get the opposite sequence of concavity, as in case (2e) representedabove in figure D3.2e. In fact it would have been natural, it seems to me,


Figure D3.2 (cont.)

(d)

(e)

that, having selected a b-distribution, which allows us to control for boththe concentration of workers around the mode and the skewness of theirdistribution, LW should not have abandoned this second possibility byimposing a5b: if so, they might as well have used a simpler symmetricaldistribution, like the normal one.

3 The second issue: different initial countries’ sizes

The above analysis implies that in cases (1c) and (2e) an important elementis the initial difference in the relative size of a country’s industrial sector. Ifthe countries that decide to integrate are close in their initial shares of theindustrial sector, then they will convergence to the even solution, each ofthem getting half of the sector. On the other hand, if the countries havewidely different initial shares of the industrial sector, then equilibrium willmove to the complete agglomeration solution. This suggests that, in choos-ing the countries to integrate, it is important to consider how wide apartthey are in their relative economic structure. This is a conclusion reminis-cent of a similar one, obtained with a growth model by Dallas and De Vries(1995).1 Consideration of the relative size of the industrial sectors in thecountries that decide to integrate their economies also provides additionalsuggestions about the sequence of liberalisation of trade and of labourmobility, on which LW concentrate most of their conclusions. In fact, sincewe have seen that cases (1c) and (2e) may induce a country to lose all of itsindustrial sector if it has a relatively small share of it to start with, it followsthat countries that plan to integrate with others endowed with a relativelylarger share of the industrial sector should avoid falling in the traps of thesetwo cases. As figures D3.1c and D3.2e indicate, this can be obtained byrotating counterclockwise the supply of labour (or clockwise the demandfor labour), so as to eliminate the crossing points with unstable equilibria.This can indeed be done by increasing the obstacles to labour mobilitywhen liberalising trade in goods.

4 Duality between trade and factor movements

The preceding section followed the suggestions of LW’s chapter by consid-ering as most appropriate the sequence of liberalising trade first, and thenallowing also international labour movement. As pointed out by LW, thisis indeed the approach taken in the EU. However, other areas of integra-tion have experimented a different sequence. Consider NAFTA, but alsothe EU vis-à-vis its Eastern and Southern neighbours, where liberalisationof trade has followed the (possibly illegal) mobility of labour, and in fact ithas been implemented in order to avoid or limit migration.


This leads me to consider that the neatness of the Heckscher–Ohlin–Samuelson (HOS) framework, where mobility of products and offactors are dual to each other, is lost in the geo-economic model à laKrugman. It would be interesting to explore under what conditions such aduality could be recovered, without losing the richer characterisation of thegeo-economic structure that Krugman’s models allow.

5 Workers’ preferences dependent on the outcome

In the model used by LW to characterise labour preferences about location,these are predetermined and do not change with the result. It might beinteresting to consider an alternative model, where preferences about loca-tion are partly exogenous and partly dependent on the results of the moreor less diffused density of the industrial population in the two countries;this, however, would require making the model explicitly dynamic.2

NOTES

1 Their model was also used by Basevi (1991) to analyse the prospects of Germanre-unification.

2 Attempts along these lines are those by Bertola (1993) and Faini (1995).

REFERENCES

Basevi, G. (1991). ‘Eastern Europe, Germany and the EC: A Framework forAnalysing their Economic Integration’, in W. Heisenberg (ed.), GermanUnification in European Perspective (London: CEPS, Brassey’s)

Bertola, G. (1993). ‘Models of Economic Integration and Localized Growth’, in F.Torres and F. Giavazzi, Adjustment and Growth in the European MonetaryUnion (Cambridge: Cambridge University Press)

Faini, R. (1996). ‘Increasing Returns, Migrations and Convergence’, Journal ofDevelopment Economics, 49, 121–36

Dallas, H. and C.G. De Vries (1995). ‘Piecemeal versus Precipitous Factor MarketIntegration’, International Economic Review, 36, 569–82

Krugman, P. (1991). ‘Increasing Returns and Economic Geography’, Journal ofPolitical Economy, 99, 483–99


4 Beyond international factormovements: cultural preferences,endogenous policies and themigration of people: an overview

ARYE L. HILLMAN AND AVI WEISS

While international migration can be studied as a case of internationalfactor mobility in response to factor-reward differences, we can alsoproceed beyond theories which view immigration in terms of internationalfactor movements.

(1) Rather than considering migration only in terms of location responsesof apersonal factors of production, we might wish to consider thegreater complexities that are introduced by recognition of the personalrelationships and characteristics of people.

(2) We might wish to consider whether and how national and culturalpreferences influence immigration policies.

(3) We might wish to view countries’ immigration policies as endogenouslydetermined by the self-interest of pre-existing residents.

Such directions of investigation suggest that the economic considerationsthat are stressed by the theory of international factor movements only par-tially explain the propensity of national states to control and restrictimmigration. For immigration policies are also seen as reflecting culturalpreferences and affinities, and perhaps likes and dislikes that are containedin the collective memories of different peoples.

Although these latter considerations add relevant dimensionalities to thestudy of international migration, incentives derived from comparativeincomes retain a primary role in providing a market foundation for interna-tional migration. In section 1 we provide a brief overview of internationalmigration, as presented by the theory of international trade and factormobility (for more comprehensive accounts than we present here, see Wong,1995, and chapter 2 in this volume), but we observe how differences in poli-cies adopted by countries toward trade and immigration are inconsistentwith policy symmetries implied by the theory. To investigate the possiblesource of the policy asymmetry, in section 2 we place cultural preferencesinside a Tiebout locational-choice model as a means of describing collective

76

attributes of culturally homogeneous populations. We observe, however, theexplanatory limitations of the Tiebout model that are due to inconsistenciesbetween the locational analogue of voluntary market choice and exclusionfrom jurisdictions effected by restrictive immigration policies. In section 3we review approaches to international migration which go beyond thetheory of international factor movements by endowing people with personallocational preferences and attributes and by acknowledging that people mayhave families. Section 4 reviews models where the migration policies chosenby countries are endogenous – that is, policies are explained by economicand political self-interest and the institutional mechanisms of collectivechoice. Section 5 draws conclusions.

While this chapter complements the picture of international migration asdescribed in the traditional models of international factor movements, thecomplete range of complementary issues is broad, and their investigationat the time of writing remains incomplete and awaits further research intohow preferences based on cultural affinity affect economic behaviour andthe endogeneity of economic policies.

1 Policy asymmetries

In the classic models of international trade, factors of production are inter-nationally immobile. This is indeed the means of defining the country. A‘country’ is a collection of factors of production located in the samenational jurisdiction. Free trade in goods among such collections of factorsof production is efficient (absent various second-best arguments that aredue to assumptions of market imperfections) and, likewise, market incen-tives lead factors of production to move internationally to maximiseincome. Depending on the assumptions regarding technology, interna-tional trade in goods and international movements of factors of productionmay be substitutes or complements (that is, trade in goods may enhance ordiminish the incentives for factors of production to move internationally).

The efficiency of free trade has not pre-empted protectionist policies,instituted ostensibly and primarily for income-distribution reasons (seeHillman, 1989). The barriers against free movement in goods have histori-cally been less intense than the restrictions countries have imposed on thefree international movement of labour, and policies with regard to the inter-national movement of real capital have in general also been more liberalthan labour immigration policies (see Schultze, 1998). The standard inter-national-economics models do not address these asymmetries in policy. Letus adopt a Heckscher–Ohlin (HO) view of the world, where internationalmovements of goods and factors of production are substitutes. We can re-phrase the above observations as follows: countries have been more open

Cultural preferences, endogenous policies and the migration of people 77

to indirect inflow of factor services embodied in goods and to direct capitalflows than to direct inflows of labour.

In HO and in the other international-economics models, factors of pro-duction are apersonal entities. Indeed, in principle in HO it matters not forthe basic results whether goods move or factors move – and if factors moveinternationally, it does not matter which factor moves; the results are thesame (see Mundell, 1957). The Rybczynski theorem tells us that residentsof a country diversified in production should not care at all about interna-tional movements of labour or capital, since international-factor move-ments do not affect their real incomes (changes occur only in domesticproduction patterns and the content of international trade). The factor-price equalisation (FPE) theorem tells us that it is a matter of indifferencefor relative incomes of factor owners in production-diversified countrieswhether goods or factors move internationally. Depending on the specifica-tion of the model, more complex and varied outcomes are also possible.Davies and Wooton (1992) distinguish between the incomes of pre-existingresidents and immigrants in a model with two traded goods and threefactors of production, and formulate a taxonomy which includes resultswhich deviate from the usual presumptions about the income-distributionconsequences of immigration of skilled and unskilled labour.

The relation between immigration and domestic income distribution istied to a debate regarding the extent to which the widening disparitybetween incomes of skilled and unskilled workers in the USA, and theunemployment of unskilled workers in the EU, is due to international trade,or to technology which has advantaged skilled and disadvantaged unskilledworkers (see the overviews by Richardson, 1995; Wood, 1995; Slaughterand Swagel, 1997). The evidence reveals the presence of both effects. To theextent that the predictions of the international FPE theorem have been real-ised, we are confronted with the question: why have the income disparitiesbeen permitted to develop via trade, when policies have nonethelessrestricted the direct immigration of foreign unskilled labour which wouldhave had the same adverse consequences for unskilled labour in the devel-oped countries? That is, if policies reflect voters’ preferences, then the poli-cies which have prevented direct entry of foreign labour via immigrationwould be expected to have been complemented by policies which pre-emptthe FPE theorem – and (since relative product prices have changed eitherbecause of liberal trade policies or changes in world prices) also pre-emptthe income-distribution consequences of the Stolper–Samuelson theorem.

We have here used the framework of the HO model as if the income-dis-tribution consequences of that economic model alone mattered for policy.The HO model comes naturally to mind as a framework to consider theinterface between trade and immigration policies, because of the natural

78 Arye L. Hillman and Avi Weiss

intuition that is offered for the relation between product-embodied anddirect international-factor movements, and because the skilled–unskilledincome differentials appear to be a reflection of the type of long-runchanges that are described by that model. One might, however, wonder alsoif any policy-maker or voter has ever believed or acted upon the predictionof the Rybczynski theorem (if he or she knew it). With some few exceptions(NAFTA in the USA, for example), policy questions have been formulatedin terms of economic interests as identified in the shorter-run specific-factors model of international trade. Factor owners with industry associa-tions have identifiable self-interested trade policy positions, butdomestically intersectorally mobile factors of production do not – anddomestically intersectorally mobile factors of production do not confrontthe indifference to labour immigration of the Rybczynski theorem but arerather exposed to real income declines by the immigration of competingforeign labour.

It seems reasonable that voters find appealing the idea, presented by thelong-run HO model, that the indirect entry (via embodiment in imports) offoreign low-cost labour into a country disadvantages domestic labour. Atthe same time, voters can also be expected to find appealing the idea, putforward in the short-run specific-factors model, that the direct entry ofimmigrants competing with domestic labour is also disadvantageous. Ifthere is a presumption one way or the other in the short-run specific-factorsmodel, it is that mobile labour benefits from free trade (see Ruffin andJones, 1977). Yet domestic mobile labour (which is of course mobile in boththe short and the long run) ostensibly bases its trade-policy position on thefactor-content propositions of the long-run HO model, and its position onimmigration policy on the short-run specific-factors model. Whether or notvoters are aware that their policy positions derive from two different eco-nomic models cannot be expected to affect self-interested voting behaviour.

In practice, the distinction between labour and capital owners is not asstark as presented in the traditional basic models. In modern marketeconomies, the Marxist demarcation has not been realised, and workers arealso capital owners to different degrees (see Mayer, 1984 on how both theHO and specific-factors models can incorporate diverse factor sources ofpeoples’ incomes). We can suppose that the distribution of ownership ofproductive wealth leaves the median voter not particularly well endowedwith capital, but the policy success of the median voter has been more inkeeping out or restricting access of foreign labour seeking to immigratethan in keeping out the same foreign labour via international trade.

We thus have a fundamental policy asymmetry. The policies adopted inthe developed world have allowed for liberal trade in goods and also liberalreal-capital exports which disadvantage low human-capital residents. The


same order of liberalism has, however, not been observed in policies towardlabour immigration.

Differences can be noted in the presumed basic rights. Liberal trade poli-cies reflect the right to transact in markets without prejudice due to thejurisdictional location of the buyers and sellers. Liberal policies with regardto international investment reflect property rights to capital and the rightof the owner to choose how and where his or her capital can most profit-ably be placed. The same liberalism with regard to international migrationwould presume the right of an individual to choose freely where he or shewishes to live and work without regard for the placement of nationalboundaries.

2 Cultural preferences and immigration policy

The policy asymmetry that is reflected in the restrictions on the immigra-tion of labour implies that we cannot view factors of production as inter-changeable if we wish to explain policy decisions. Labour is of coursedistinct from goods and capital, in that the latter are inanimate whereaslabour consists of people. People have cultures and cultural preferences, andthese cultural preferences can in turn affect population and immigrationpolicies. Besides determining immigration policies, own-culture preferencescan also lead a local population to implement policies of expulsion. Weneed only consider the history of the many expulsions of the Jews fromdifferent domains of Christian Europe, and in more contemporary timesthe expressions of anti-immigrant sentiment in election results in Franceand Austria and the behaviour of fringe groups in Germany and inDenmark.

What, then, are we to make of cultural preferences – if we do not wish toignore the influence of these preferences on immigration policies? Anapproach that an economist can offer to introduce cultural preferences is aTiebout-type locational model (see Hillman, 1994). Charles Tiebout (1956)proposed that locational choice among competing jurisdictions offers asolution to the preference-revelation problem for public goods. Jurisdictionsare able to exclude outsiders from consumption of the goods that are col-lectively provided to residents. Exclusion permits a price for access to becharged, so that the different offerings provided by jurisdictions have thecharacter of a market where consumers confront price offers for differentquantities and types of collective goods. Under the assumptions set out byTiebout, locational choice becomes a perfect substitute for a competitivemarket in collective goods, and the Pareto-efficiency of the market for publicgoods is restored. Applying the Tiebout model in the context of migrationpolicies, we perceive jurisdictions with cultural preferences and other


manifestations of attributes of people. People who wish to live with certaintypes of other people congregate together in the same jurisdiction.

The cultural preferences can be expressed in a federal system. Forexample, in the USA the Amish people choose to pursue their own valuesand live within their own culture within their closed communities, butunder the jurisdiction of the state of Pennsylvania and the US federalgovernment. A liberal view upholds the basic right of such congregationsof people to adopt a policy of cultural homogeneity, even if there aredifferences from mainstream values and culture – subject to evident qual-ifications regarding humaneness and external effects. These qualificationsare satisfied for the Amish. The qualifications are not satisfied, for example,if a community of people has a culture where there is predatory behaviourtowards neighbours and neighbouring communities and peoples.

Cultural preferences can also be expressed at the national level. TheAmish do not use immigration policy to restrict entry into their culturallyhomogeneous communities, since they have no jurisdiction over immigra-tion policy. Such preferences were, however, expressed in the immigrationpolicies of the American states before federation, and have been expressedin the immigration policies of the USA: the Lady of the Statue of Libertydid not welcome all peoples equally. Nor did the other New World coun-tries that sought immigrants practise culturally free policies.

A Tiebout model encompasses in principle preferences for collective fea-tures of life expressed in cultural preferences. Are there, however, limits toTiebout when culture is a locationally provided public good? That is, howfar can we take Tiebout as a justification for culturally- and nationally-based restrictive immigration policies? We will refrain from taking thesimple position that nationalism and cultural identity are necessarily asource of evil. Great evils have been committed because of nationalism andculture of some peoples – but other cultures have brought humane valuesto the world. Our question here in attempting to reformulate Tiebout is notwhether countries should be influenced in their immigration policies bydesires for cultural preservation and national identity – but given that coun-tries do base immigration policies on cultural preferences, can the Tieboutmodel assure the locational analogue to an efficient market?

The source of potential compromise of efficiency in a Tiebout world isthe need for competition among locations and voluntariness of locationalchoice. The locational analogue to a competitive non-spatial market sup-poses voluntary decisions by consumers, who choose the price andquality–quantity offer of a jurisdiction in the same manner as a consumerresponds to a market price in making a consumption decision. It is impor-tant that the locational decision be voluntary in the manner of the marketdecision. It is also important that there be no exclusion of people who are


willing to pay for admission to a jurisdiction – that is, the locational marketshould clear, so that in equilibrium no one is excluded from where they wishto be.

If personal cultural likes and dislikes are embedded in individual prefer-ences, some people may not wish to be with people who wish to be withthem, and vice versa. This is not an asymmetry that could be present in thestandard rendition of the Tiebout locational model, where all that mattersfor inclusion in a jurisdiction is willingness to pay, and not the identity ofthe prospective entrant. Some people may thus find themselves involuntar-ily excluded from a jurisdiction. Tiebout would assume that there exists aperfect substitute to the jurisdiction which excludes them, and that thisperfect substitute will accept them; or that the excluded people themselveshave the means and opportunity to create their jurisdiction – that is, com-petitive replication assures that there is no exclusion. Where competitivereplication is not possible, the problem of exclusion, or of facilitating dis-criminatory admission, remains. Discriminatory pricing for excludedpotential entrants resolves the problem of discriminatory admission (seeHillman and Swan, 1983): efficiency is maintained by permitting admissionby those people willing to pay a price which at least compensates those pre-viously present for their entry. If institutions to facilitate discriminatorypricing are not present, people who wish to pay the higher price withoutwhich they cannot have access to the jurisdiction cannot enter (or cannotavoid being expelled), and the equilibrium is inefficient.

Discrimination has been facilitated in some countries on the basis ofwealth. Australia and Canada, for example, have permitted immigrationcontingent on ownership of sufficient capital. This is, however, a rationingmechanism, and not a discriminatory pricing mechanism since the capitalis not transferred in payment (although the future returns may be taxed).

To find an example par excellence of a discriminatory pricing mechanism,we can again turn to the history of the Jews. Jews could immigrate toChristian jurisdictions (although in general not be citizens and not own landand often live only in designated areas), by being permitted to pay a specialtax to compensate for their presence. Under Islam, as well, Jews paid a specialtax. The mechanisms were discriminatory but efficient; the Jews could livetheir distinct communal and private lives within their culture and values bypaying the price which allowed their entry. Profound inefficiencies arosewhen a discriminatory pricing mechanism was not permitted to function.The market failure was immense when supreme-value ideologies denied thepossibility of a market solution (see Bernholz, 1991): the alternatives to dis-criminatory locational pricing have been expulsion and murder.

A final aspect of Tiebout that we wish to consider concerns propertyrights and theft in the context of migration. If people have different wealth,


then location is not based solely on preferences, and poorer people may findthemselves in jurisdictions where they do not wish to be. They then mayseek to exit, even if in so doing they break the law of the country of entryby being illegal immigrants. In non-spatial markets high-income popula-tions similarly fare better than the poor. We can compare the act of theft ina non-locational market with illegal immigration to a locational jurisdic-tion where the immigrant ‘has no right to be’. If basic rights are indeed thesame, illegal immigrants would then have the same moral deficiencies asthieves. There is no commandment which declares, ‘you shall not immigrateillegally into a country’, whereas there is a commandment that one shouldnot steal. The question here is whether failure to contribute to a publicgood from which one benefits is theft. Laws which deny the children ofillegal immigrants access to state-funded education have this connotation.

3 Models of the migration of people

We have thus far put forward a perspective on international migration withfoundations in the international-economics literature. This literature isapersonal, until we introduce people’s preferences for cultural and nationalhomogeneity and consider the possibilities in terms of Tiebout jurisdic-tions. An alternative perspective on international migration is provided bya point of departure from labour economics. In this literature, immigrantsare not so much factors of production as people with personal attributesand their own locational preferences. The push–pull distinction (seeZimmermann, 1994), for example, distinguishes between people ‘pushed’out of their countries as a result of adverse domestic conditions, and people‘pulled’ into richer countries by the attraction of higher wages. In eithercase, relative evaluations of circumstances influence the migration decision,and the international factor market income differences are present thatunderlie equilibrating international-factor movements. The push–pulldichotomy proposes, however, that we ask people why they left their home-land. If the answer is, ‘I left because of poverty and insecurity at home’, thisis push; if the answer is, ‘I left because of the better opportunities abroad’,this is pull. Under the conditions of push, people would not emigrate if onlyconditions were better at home, since higher incomes available abroadmight not compensate for their own-cultural preference. Under conditionsof pull, economic comparisons matter more. The push–pull framework ispersonal rather than apersonal, in asking why people are led to make inter-national migration decisions – in contrast to asking why internationalfactor movements take place.

The restrictions that countries place on international immigration alsobecome more personalised. A Tiebout re-characterisation of the models of


international economics sits uncomfortably with the models themselves,since we move from apersonal factors of production to consider thepossibility that people have likes and dislikes in social interaction and thatcultural preferences matter. In a push–pull dichotomy, we are confrontedat the onset with questions regarding who is pulled and who tends to bepushed. Whereas departure due to pull is voluntary, departure due to pushmay not be. Push–pull is therefore less politically correct in not avoiding ordownplaying contentious issues. Political correctness is indeed difficultwhen we consider the role of personal and cultural attributes of people ininfluencing migration policies and we investigate whether xenophobia orprejudice may be part of the explanation for immigration restrictions – orpart of the reason for expulsion from a jurisdiction (enforcement and expul-sion or deportation are of course also economic decision variables, seeEthier, 1986; Bond and Chen, 1987).

Restrictive immigration laws are enforced upon people rather thanfactors of production, and it is people who are apprehended and deported.One may therefore wish to personalise immigrants, and also introduceasymmetric information about their attributes (see, for example, Schmidt etal., 1994; Haque and Kim, 1995; Stark, 1995b). The ‘brain-drain’ modelshave described how differentiated populations in poorer countries aresubject to different emigration incentives (see, for example, Bhagwati, 1976;Kim, 1976; Kwok and Leland, 1982), either by criteria set by the receivingcountry’s government or by emigrants’ own self-selection (see Borjas,1987). Stark (1995b) has further proposed that immigrant remittances tothose remaining at home should be seen as a selection device to keep poten-tial low-skilled migrants from emigrating in a process of ‘strategic self-selection’. In the presence of asymmetric information about workers’ skills,it may be better for all workers from the emigrating country if low-skilledworkers do not migrate. That is, because of asymmetric information byemployers regarding immigrants’ skills, emigration by the low-skilledwould decrease the earnings ability of the high-skilled. If the loss to thehigh-skilled workers is sufficiently great, they may benefit by compensatingthe low-skilled workers for the damage they sustain by not migrating, thusmaking all better off. We can note that this requires substantial internalisa-tion by those who have emigrated (for an empirical evaluation, seeDocquier and Rapoport, 1997).

Further models by Stark (1984, 1991, 1995a) develop other personalaspects of migrants’ decision-making. Stark perceives people as havingpreferences biased in favour of remaining in their known home environ-ment rather than leaving for the foreign unknown, and immigration asoccurring on the basis of relative rather than absolute deprivation (see alsoStark and Taylor, 1991). That is, emigration takes place when people feel


sufficiently relatively deprived by reference to standards in their own coun-tries, and return emigration is predicted to take place once people haveaccumulated sufficient wealth for them no longer to feel deprivation rela-tive to their peer groups in their home neighbourhoods. Return emigrationoccurs, even in the face of reduced real income, and such reverse ‘interna-tional factor movement’ is contrary to the direction of movement predictedby factor-reward differences. Stark also uses the conception of ‘family’ toinvestigate migrant behaviour (see, for example, Stark, 1991). The notionof ‘family’ is another avenue to explain remittances and return migration,and also dispersed relocation to spread family risk and the role of networksin determining geographic clustering of immigrants.

The immigrants, whether legal or illegal, who have children (with them)require access to education, and they also benefit from other social services.In the USA there has been a debate against this background whetherimmigration to the USA is fiscally beneficial – that is, whether taxes paidby immigrants meet the cost of education and social welfare expenses.Huddle (1993) estimates that immigrants impose an annual cost of $40billion, while Passel and Clark (1994) estimate a net contribution of $27billion. Borjas (1994) estimates a $16 billion loss. At the same time, givendemographic trends, immigrants may in the longer term rescue the USsocial security system by providing a working population to sustain the old-age consumption of the population which has chosen low rates of repro-duction.

4 Endogenous immigration policy

Endogenous-policy models seek to explain the choice of observed policies.Enhanced efficiency is one reason for choice of a policy. Efficiency wouldimply the choice of free trade and, complementing the free movement ofgoods and services, free international movement of capital and labour asthese factors seek their highest return. Yet, as we have observed, countriesinvariably use their sovereignty to restrict immigration – which has led usto preferences for cultural and national homogeneity, and to domesticincome-distribution effects of migration. A political-economy modelexplicitly encompasses how collective decision-making institutions andoptimising policy-makers respond to the different sources of interest ofgainers and losers from policies and can explain why efficient policies arenot adopted. A public-choice model, in particular, encompasses awarenessof the possible imperfections of policy outcomes of representative democ-racy that arise as the consequence of the interaction between self-interestedpolitical agents who determine policy, and coalitions of likewise self-inter-ested economic agents who favour policies that may not be in the interest


of society at large. Such a model thus recognises that political agents maybe receptive to rent-seeking, so introducing an additional potential sourceof social cost. Direct-democracy models of endogenous policy, on the otherhand, exclude political agents from policy discretion; the median voterdetermines the chosen policy, but there is no basis for presumption that themedian voter’s preferred policy is also the socially optimal one.Endogenous-policy models do not, however, seek to recommend policies topolicy-makers. As Foreman-Peck (1992) succinctly states in reviewing nine-teenth- and early twentieth- century international immigration movements,the issues are simply who gains and loses from migration, and who is in aposition to influence policy. We proceed here to briefly review endogenous-policy models of international immigration. These models, again, tran-scend the economic analysis of international factor movements – to askwhy departures from efficient policies might be observed.

Thus, Benhabib (1996) considers a resident population which viewsimmigration policy as an instrument of income-maximisation, and derivesthe median voter’s preferred policy for capital and skill requirements of eli-gible immigrants, given the wealth holdings of a country’s population.There is of course no reason to suppose that the immigration policy whichis in the self-interest of the median voter is welfare-maximising for thecountry at large. However, under direct democracy the preferences of themedian voter nonetheless determine the policy equilibrium in the mannerdescribed by Benhabib. The median voter also determines the equilibriumpolicy under representative democracy if the issue of migration is dominantin the choice among whom to support among representatives offeringthemselves for election. Mazza and van Winden (1996) model an institu-tional framework which more directly reflects representative democracy, byspecifying policy as the outcome of political influence exercised by thegroups affected by immigration of labour. The weights which determine theability of a group to transform size into political influence are endogenousand changed by the presence of increased numbers of immigrants.Immigrants thereby affect endogenous domestic income distribution and,by the effects on political influence, can thereby over-compensate residentworkers for the decline in wages due to the presence of competing immi-grant labour. Mazza and van Winden also investigate policy coordinationamong different jurisdictions.

In Hillman and Weiss (1995), we observe that, despite immigration restric-tions, illegal immigrants are nonetheless openly present in many countries,and that the openly visible illegal immigrants also tend to be concentrated inselective industries. The median voter is a worker who would competedirectly with immigrants, and who would suffer a real-income decline fromtheir presence. Hence the equilibrium policy is that immigration is illegal.


This does not, however, mean that immigration is not permitted. Legalimmigration is not allowed; the median voter can, however, benefit fromillegal immigration. The illegality permits selective enforcement of restrictiveimmigration laws, which confines illegal immigrants to employment insectors where the median voter benefits from the immigrants’ presence. Inparticular, the median voter is happy to have illegal immigrants who areselectively confined to non-traded goods sectors. The ‘illegality’ of illegalimmigration has the effect of transforming immigrants from mobile factorsof production who would compete in domestic labour markets with themedian voter, to sector-specific factors of production whose presence is ben-eficial. Hence, while legal immigration is not allowed, illegal immigration isnonetheless an equilibrium endogenous policy.

A complementary perspective looks at the countries of emigration.Epstein et al. (1998) describe a country where spatial proximity of amember of the population or family to a ‘king’ determines how well theindividual or family fares from the king’s largesse in making gifts andplacing the burden of taxation. The population differs in productive andrent-seeking abilities – that is, people have different personal comparativeadvantages in production and in non-productive rent-seeking. Themembers of the population allocate their resources between productionand rent-seeking. The institutions which determine influence or proximityto the king establish whether more or fewer productive residents are locatedcloser to the king. The outcome of who is at the ‘end of the line’ of privi-lege then determines whether the best rent-seekers or the most productivesegments of the king’s population have an incentive to emigrate.Emigration unravels the king’s population, as people are pushed abroadbecause of their inferior positioning in the domain of privilege. However,in the absence of the redistribution determined by proximity to the king, noone has an incentive leave the country. All who leave are pushed.

Jahn and Straubhaar (1997) implicitly address the question of the natureof the preferences that underlie restrictive immigration laws. They point outthat immigration is illegal only because a law states it to be so. A wide spec-trum of immigration laws can in principle be legislated. Beyond the ques-tion of why people immigrate illegally, there is the more fundamentalquestion why the population of a country through its collective decision-making mechanisms has decided that entry into the country of foreignersis illegal.

Differences in immigration policies are studied by Buckley (1996), whocompares the policies of the USA and Canada, two New World recipientcountries. Canada has been more successful at attracting, on average,‘higher-quality’ immigrants than the USA, as reflected in higher educationalattainment, better language skills and more managerial and professional


skills. This is attributed by Buckley to the different criteria used to establisheligibility for immigration. The USA emphasises family reunification,whereas Canada specifies individual characteristics in a points system.Buckley’s endogenous-policy basis for the differences has three elements. Inthe USA, (1) employers and immigrants tend to form interest groups withpolitical influence, (2) politicians have a greater interest in furthering theirnatural constituencies, which include immigrants and (3) immigration policyas legislated and enforced by the federal government can conflict with thepolicies sought by individual states, so the policies which might be desired ata lower level of the federal structure are not feasible.1

Some countries have guest-worker programmes. Such programmes arealso open to interpretation as reflecting preferences for national culturalhomogeneity. The immigrants are designated as ‘temporary visitors’, and itis made clear by the conditions of their entry that they are expected in duecourse to depart rather than join the permanent resident population. Thetransience is consistent with the use of guest-workers as a policy responseto sectoral ‘labour shortages’. Yet the policy of rotation also seeks to ensurethat guest-workers continue to perceive themselves as transient, and in par-ticular do not bring their families with them. Often guests do not wish togo home; policy responses to ensure that the guests leave include contin-gent deferred payments (Simon, 1989), and placement of the obligation ofensuring departure on the employer who has been specifically granted thepermit to import the intended transient worker. Bonded employers ofguest-workers have limited means to ensure return emigration. Rather thanleave the country, the guest-worker can simply leave the legal employer forthe illegal market. Despite the policy intentions of guest-worker pro-grammes, such programmes are well known to be porous and to be thepoint of entry for illegal immigration (see Epstein et al., 1999). Suchattempts at cultural homogeneity by a policy of a revolving door have ingeneral not been successful.

5 Conclusions

This chapter has presented an overview of international migration oflabour which broadens the scope of questions and permissible answersbeyond the framework of market-determined responses to internationaldifferences and factor rewards and the characteristics of technologies. Weshould not be understood as suggesting diminished importance for thegeneral equilibrium framework of the models of international economics.On the contrary, no sound study of international immigration can beundertaken without an understanding of the resource-allocation andincome-distribution consequences of international-factor movements as


displayed in these models. We have, however, shown that, notwithstandingtheir fundamental analytical virtues, these basic models are limited in thescope of appreciation of the complexities that underlie peoples’ interna-tional migration decisions, and the policy decisions of others to attempt tothwart or regulate these decisions. Some aspects of behaviour are not exem-plary, in particular in chosen policies that reflect personal likes and dislikesof the population. However non-explanatory, these dimensions of behav-iour nonetheless are explanatory variables for migration-related phenom-ena beyond market responses to factor-reward differentials.

NOTE

1 Another reason is illegal immigration. The USA receives illegal immigrants fromMexico and Cuba, often with little human capital. Canada does not share aborder with these countries.

REFERENCES

Benhabib, J. (1996). ‘On the Political Economy of Immigration’, EuropeanEconomic Review, 40, 1737–43

Bernholz, P. (1991). ‘The Constitution of Totalitarianism’, Journal of Institutionaland Theoretical Economics, 147, 424–40

Bhagwati, J.N. (1976). ‘The Brain Drain’, International Social Science Journal, 28,691–729

Bond, E.W. and T.J. Chen (1987) ‘The Welfare Effects of Illegal Immigration’,Journal of International Economics, 23, 315–28

Borjas, G.J. (1987). ‘Self Selection and the Earnings of Immigrants’, AmericanEconomic Journal, 77, 531–53

(1994). ‘The Economics of Immigration’, Journal of Economic Literature, 32,1667–1717

Buckley, F.H. (1996). ‘The Political Economy of Immigration Policies’,International Review of Law and Economics, 16, 81–99

Card, D. (1990). ‘The Impact of the Mariel Boatlift on the Miami Labour Market’,Industrial and Labour Relations Review, 43, 245–57

Davies, J.B. and I. Wooton (1992). ‘Income Inequality and InternationalMigration’, Economic Journal, 102, 789–802

Docquier, F. and H. Rapoport (1997). ‘Are Migrant Minorities StrategicallySelf–Selected?’ SUITE Working Paper, 97/2

Epstein, G., A.L. Hillman and H. Ursprung (1998). ‘The King Never Emigrates’,Review of Development Economics,3(2)

Epstein, G., A.L. Hillman and A. Weiss (1999). ‘Creating Illegal Immigrants’,Journal of Population Economics,12, 3–21

Ethier, W.J. (1986). ‘Illegal Immigration: The Host-Country Problem’, AmericanEconomic Review, 76, 56–71


Foreman–Peck, J. (1992). ‘A Political Economy Model of International Migration,1815–1914’, The Manchester School, 60, 359–76

Haque, N.U. and S.J. Kim (1995). ‘Human Capital Flight: Impact of Migrationon Income and Growth’, International Monetary Fund Staff Papers, 42,577–607

Hillman, A.L. (1989). The Political Economy of Protection (Chur: HarwoodAcademic Publishers)

(1994). ‘The Political Economy of Migration Policy’, in H. Siebert, (ed.),Migration: A Challenge for Europe (Tübingen: J.C.B. Mohr, Paul Siebeck)

(1997). ‘The Social Consequences of Trade and Technology: Marginalizationand Policy Response in the Developed World’, Research Paper, World TradeOrganisation

Hillman, A.L. and P.L. Swan (1983). ‘Participation Rules for Pareto-OptimalClubs’, Journal of Public Economics, 20, 55–76

Hillman, A.L. and A. Weiss (1995). ‘A Theory of Permissible Illegal Immigration’,paper presented at CEPR Workshop on ‘Illegal Immigration’, Thessaloniki

Huddle, D. (1993). ‘The Costs of Immigration’, Rice University, manuscriptJahn, A. and T. Straubhaar (1997). ‘On the Political Economy of Immigration

Control’, paper presented at CEPR Workshop on ‘Illegal Immigration’,Athens

Kim, K.H. (1976). ‘The Economics of the Brain Drain: Pros, Cons and Remedies’,Journal of Economic Development, 1, 55–80

Kwok, V. and H. Leland (1982). ‘An Economic Model of the Brain Drain’, TheAmerican Economic Review, 72, 91–100

Mayer, W. (1984). ‘Endogenous Tariff Formation’, American Economic Review, 74,970–85

Mazza, I. and F. van Winden (1996). ‘A Political Economic Analysis of LabourMigration and Income Redistribution’, Public Choice, 88, 333–63

Mundell, R.A. (1957). ‘International Trade and Factor Mobility’, AmericanEconomic Review, 47, 85–95

Passel, J.S. and R.L. Clark (1994). ‘How Much Do Immigrants Really Cost? AReappraisal of Huddle’s “The Cost of Immigrants”’ (Washington, DC: UrbanInstitute)

Richardson, J. D. (1995). ‘Income Inequality and Trade: How to Think, What toConclude’, Journal of Economic Perspectives, 9, 33–56

Ruffin, R. and R. Jones (1977). ‘Protection and Real Wages: The Neo–ClassicalAmbiguity’, Journal of Economic Theory, 44, 337–48

Schmidt, C.M., A. Stilz and K.F. Zimmermann (1994). ‘Mass Migration, Unions,and Government Intervention, Journal of Public Economics, 55, 185–201

Schultze, G. (1998). The Political Economy of Capital Controls (Cambridge:Cambridge University Press)

Simon, J.L. (1989). The Economic Consequences of Immigration (Oxford: BasilBlackwell)

Slaughter, M.L. and P. Swagel (1997). ‘The Effect of Globalization on Wages in theAdvanced Countries’, Staff Studies for World Economic Outlook (Washington,DC: International Monetary Fund)


Stark, O. (1984). ‘Rural to Urban Migration in LDCs: A Relative DeprivationApproach’, Economic Development and Cultural Change, 32, 475–86

(1991). The Migration of Labour (Oxford: Basil Blackwell)(1995a). Altruism and Beyond: An Economic Analysis of Transfers and Exchanges

Within Families and Groups (Cambridge: Cambridge University Press)(1995b). ‘Return and Dynamics: The Path of Labour Migration when Workers

Differ in their Skills and Information is Asymmetric’, Scandinavian Journal ofEconomics, 97, 55–71

Stark, O. and J.E. Taylor (1991). ‘Migration Incentives, Migration Types: The Roleof Relative Deprivation’, Economic Journal, 101, 1163–78

Taylor, J.E. (1992). ‘Earnings and Mobility of Legal and Illegal Immigrant Workersin Agriculture’, American Journal of Agricultural Economics, 74, 889–96

Tiebout, C. (1956). ‘A Pure Theory of Local Expenditures’, Journal of PoliticalEconomy, 64, 416–24

Venables, A.J. (1998). ‘Trade Liberalisation and Factor Mobility: An Overview’,chapter 2 in this volume

Wood, A. (1995). ‘How Trade Hurts Unskilled Workers’, Journal of EconomicPerspectives, 9, 57–80

Wong, K. (1995). International Trade in Goods and Factor Mobility (Cambridge,MA: MIT Press)

Zimmermann, K. F. (1994). ‘European Migration: Push and Pull’, Proceedingsvolume of the World Bank Annual Conference on Development Economics,1994, Supplement to The World Bank Economic Review and The World BankResearch Observer, 313–42; reprinted in International Regional ScienceReview, 19 (1996), 95–128

Discussion

FRANCESCO DAVERI

Hillman and Weiss in chapter 4 effectively survey a large number of ideasembodied in the recent migration literature. Their survey succeeds in pro-viding both ‘the big picture’ and vivid examples of these ideas, and it istherefore a useful and enjoyable read.

The survey can be broadly split into three parts, each discussing a separ-ate question. The first part – inclusive of sections 1 and 2 – is about thecauses of asymmetries in trade and migration policies. It is argued that cul-tural differences play a major role in determining such policy asymmetries.


The second part (section 3) draws on theories of migration where culturalfactors play a role, unlike the standard theory based on wage differentials.The third part of the survey (section 4) concerns the endogenous determina-tion of immigration policies as a response to cultural heterogeneity.

The survey as a whole revolves around the idea that cultural heterogene-ity is an important determinant of the policy attitudes of governments. Thisis the main point I wish to take up in my discussion. While I agree thattaking heterogeneity seriously is important, I am not entirely convincedthat cultural differences across communities can go a long way in explain-ing migration policies.

Why should cultural heterogeneity matter for migration policies? Theauthors’ starting point is that factor movements are not interchangeable.Capital inflows, as well as the imports of goods, do not usually imply themixing up of people, unlike migration. Then, if natives dislike having toomuch variety of people around, liberalising labour flows is not the same asliberalising flows of capital or goods. Hillman and Weiss report clear exam-ples of how Jews painfully suffered the consequences of what can be termedthe ‘adverse cultural preferences’ of other people.

Yet understanding the observed asymmetry in trade and migration poli-cies in industrial countries does not require an appeal to culture. One caneasily accept that factors are not the same. One can also accept that asym-metries in policies towards migration and trade have to do with somespecificity of labour compared to capital (or goods). But factors of produc-tion are not the same for a number of reasons other than heterogeneity ofpreferences.

First of all, even non-racist voters dislike seeing the access to a congest-ible public good being diluted by the arrival of new entrants, not contrib-uting to the fiscal pool as much as they gain from it. When repatriatingAlbanians, Italians are not driven by resurrected Fascist sentiments.Instead, this occurs under the perception (or the fear) that unrestrictedflows of immigrants will congest the access to locally provided publicgoods, such as low-cost housing and other social services, such as healthcare and primary education.

Second, even when not congesting local public goods, immigration maybe harmful when illegal (i.e. when immigrants are constrained to look forjobs in non-traded sectors). If immigrants become street-vendors, they areoften perceived as competitors by small retailers, who also face severecompetition from large-scale retailers and malls. Moreover, illegal immi-grants tend to get involved in outright illegal activities as well, since this isa cheap, though risky, way of earning money. The operations of criminalorganisations (which levy implicit taxes on legal activities) often receive agood kick by the inflow of illegal immigrants.

92 Discussion by Francesco Daveri

Third, if immigrants are granted the same rights as the other citizens ina given community, political-economy reasons suggest that migration stilldiffers from capital inflows. All of these reasons have nothing to do withheterogeneity of preferences. First of all, Bhagwati and Rodriguez (1975),showed that the inflow of labour from abroad, unlike the inflow of capital,implies a reformulation of the social-welfare function of a given commu-nity. The welfare effects of migration were shown to be crucially dependenton the treatment of TLB (those left behind). In addition to that, supposethat voters have the same preferences but different endowments of labourand capital. In this case, if immigrants are endowed with less capital thannatives, the median voter theorem predicts that their arrival makes redis-tribution more likely.

None of these sketched rationales has anything to do with culturalpreferences. At the same time, in the absence of empirical work, I do notknow how to rule out these explanations as opposed to the rationalefavoured by Hillman and Weiss. This just tells us that, in principle, manyreasons may underlie the variety of trade and migration policies in place inindustrial countries. Further progress towards appropriately endogenisingmigration and trade policies would thus entail the specification of empir-ically meaningful ways of testing competing theories.

In any case, had cultural differences something to say on migration poli-cies, this ought to be in a cross-section of localities, rather than over time.Culture changes slowly compared to policies. In chapter 4, Hillman andWeiss quote the electoral success of chauvinist groups in Germany, Austriaand France as examples of ‘cultural’ distaste for foreigners. However, it isnot clear that the fortunes of Lepenists and the like are long-run in nature.If anything, as of today (December 1997), the electoral success of chauvin-ist movements in Europe seems waning already, while it was rampant onlyfew months ago. I simply do not see how such short-run policy swings canbe related to, presumably long-lasting, cultural preferences.

To sum up, taking heterogeneity of preferences seriously and under-standing their influence on the process of policy-making is an ambitiousand challenging task. A fruitful parallel route to take is then to formulatelocally verifiable predictions, over and above general propositions, whichare sometimes harder to test.

REFERENCE

Bhagwati. J. and C. Rodriguez (1975). ‘Welfare-theoretical Analyses of the BrainDrain’, Journal of Development Economics, 2, 196–221


5 Trade liberalisation and public-goodprovision: migration-promoting ormigration-deterring?

KONSTANTINE GATSIOS,

PANOS HATZIPANAYOTOU AND

MICHAEL S. MICHAEL

1 Introduction

Since its establishment in 1947, several rounds of GATT negotiations havepromoted the liberalisation of international trade in commodities. Mean-while, numerous preferential trading agreements (APEC, CARICOM,NAFTA) have deepened regional economic integration among countrieswilling to go further. Recognising the importance of barriers in servicestrade, including labour services, as costly as commodity trade barriers, theUruguay Round (1994) extended the principles of goods trade liberalisa-tion to service markets, through the General Agreement on Trade inServices (GATS). Moreover, in its constitution, the EU recognises the freemobility of labour as one of the ‘four–pillar freedoms’ that make theEuropean common market ‘an area without internal frontiers in whichthe free movement of goods, persons, services and capital are ensured’(EEC Treaty, art. 48 (2)).

Standard international trade theory has prescribed free trade as theregime that maximises the economic welfare of countries without marketpower, as well as global economic welfare. On the other hand, numerousstudies have examined the relationship between free commodity trade andfree international movements of factors of production. Within the tradi-tional Heckscher–Ohlin (HO) framework, it has been long established thatunder free trade and factor–price equalisation (FPE), commodity tradeand factor movements are perfect substitutes in a weak and strong versionof a quantitative and price-equalisation sense (see Mundel, 1957; Wong,1995).1 If under free trade factor prices do not equalise, however, then freetrade in goods and factors are substitutes in the weak price-equalisationsense and in the weak and strong quantitative sense.2 The formal treatmentof international trade and international migration in the literature has fol-lowed two basic approaches. First, that of the ‘brain drain’ associating

94

international migration with movements of human capital (Hamada andBhagwati, 1976). Second, that of including international labour mobilityinto the standard two-good, two-factor model of international trade (seeDixit and Norman, 1980; Brecher and Bhagwati, 1981).3

Despite the obvious advantages of free trade for most countries, and forthe world as a whole, tariffs, import quotas and voluntary export restraints(VERs), among other restrictive trade measures, have been used to protectdomestic markets. Within the neo-classical trade context when factor pricesequalise under free trade, such restrictive trade practices are required toinduce international factor (capital, labour) movements.4 On the otherhand, without FPE in free trade, trade restrictions are not required toinduce international-factor movements. Thus, according to standard tradetheory, liberalisation of commodity trade, by and large, generates a ten-dency for FPE across countries that is expected to dampen international-factor (e.g. labour) movements. Recent trends in the world economy,however, depict a broader liberalisation of commodity trade and increasedflows of international migration that do not comply fully with the tradi-tional theoretical prediction. But throughout the history of world migra-tion, international labour movements occurred not only because of wagedifferentials favouring the host countries. One prominent factor inducingworld-wide labour mobility is the level (and variety) of public goods immi-grants can consume in the host versus the source countries.

The literature on international migration has examined the impact ofinternational labour movements on the system of transfer payments andpublic-good provision in labour-exporting or labour-importing countries.5

How does migration affect a source or host country’s government taxrevenue and ability to provide public goods and services to its population?How extensive is the use of public goods and services by an average immi-grant family in a host country, and what is their contribution (throughincome taxes or other transfer payments) to their production cost? Doesthe pattern of public goods consumed by immigrants differ from that ofnatives in host countries, due to, for example, ethnic, age or householdcharacteristics?6 Such issues constitute the core on policy debates regard-ing the fiscal repercussions of international migration. To the best of ourknowledge, the issue of the ‘public-finance’ aspects of international migra-tion has not been adequately tackled within the literature of internationaland development economics.

The present chapter, using a general equilibrium trade model, reversesthe angle of analysis developed by the literature of fiscal repercussions ofinternational migration. Within a trade-theoretic context of a labour-exporting or a labour-importing country, we examine how trade liberalisa-tion affects domestic public-good provision, and the flows of international

Trade liberalisation and public-good provision 95

migration. To this end, it is assumed that the labour-exporting or labour-importing country is small in world commodity and factor (i.e. labour)markets, unable to affect through its actions the world prices of goods andthe world rate of return to labour. Under competitive conditions, thecountry produces two private traded – one exportable and one importable– goods, and a public-consumption good. Imports are subject to a tariff, ora quota or a VER constraint. Revenue from trade restrictions is lump-sumdistributed to domestic residents under a tariff or an import-quota regime,and to foreign residents (exporters or governments) under a VER regime.The government finances the provision of the public good through incometaxation and provides it to domestic residents free of charge. It is furtherassumed that migration is temporary in the sense that income earned byimmigrant workers in the host (labour-importing) country are remitted tothe source (labour-exporting) one.7 Moreover, labour mobility between thelabour-exporting or labour-importing country and the rest of the worldcontinues until the domestic and world net rate of return to the factor areequated. Within this framework, the chapter derives conditions underwhich trade liberalisation, through lower import restrictions reduce boththe provision of public good and domestic employment by causing emigra-tion. Three basic relationships emerge as the fundamental determinants ofour results. First, the relationship (i.e. complementarity–substitutability)between the importable and public-consumption goods in production,second, the impact of trade liberalisation on government net income taxrevenue and, third, the impact of international migration on the unit costof public-good production.

2 The model

Consider a small open economy with identical consumers producing twoprivate traded goods – one exportable and one importable – and one public-consumption good. Many primary factors (m), which are either intersectorallymobile or sector-specific, are used in the production of all three goods.

Domestic factor endowments, i.e. v, are fixed, i.e. dv50, but the country’ssupply of one factor, labour, is variable due to international migration. Thedomestic labour supply (L) is given by L5 1L*, where is the fixedendowment of labour, and L* (.0) denotes the number of immigrantworkers for a labour-importing (host) country, and L* (,0) denotes thenumber of emigrant workers for a labour-exporting (source) country. Weassume full employment so that dL5dL*. Let w and r (.0) denote, respec-tively, the domestic rate of return to labour and the income tax rate and letw* be the world net rate of return to labour.

No taxes or subsidies are levied on the exportable good, which is assumed

LL

96 Konstantine Gatsios et al.

to be the numeraire whose price is normalised to one. The domestic andworld relative price of the importable good is, respectively, p and p*. Thedifference between the two prices, denoted by t (5p2p*), may be due to atariff, an import quota or a VER. Since the country is assumed small in allworld markets, changing the import restrictions or the income tax rate r doesnot affect the world prices of goods and the world rates of return to factors.

Technologies in the production of private and public goods are homo-geneous of degree one and strictly concave in the primary factors. Theprivate sector produces the two traded goods and is assumed to behavecompetitively. The public sector is assumed to minimise the production costof the public-consumption good.

Let R(p, g, L) be the maximum value of the private gross domesticproduct (GDP) given the relative price of the importable, p, the supply ofpublic-consumption good g, and the domestic supply of labour L. The fixedendowments of the non-traded factors are omitted from the GDP functionsince they do not affect the analysis. The partial derivative of the GDP func-tion with respect to p (i.e. Rp 5R/p) is the country’s supply function ofthe importable, and with respect to L (i.e. RL5R/L) gives the marginalrevenue product of labour. Henceforth, all subscripts denote partial deriv-atives. The R(p, g, L) function is assumed to be strictly concave in L,concave in g (i.e. Rgg is non-positive), and to have, among others, the fol-lowing properties (see Abe, 1995):

Rg52Cg, and Rgg 52Cgg, (1)

where Cg(v) is the unit cost of the public good, assumed to be homogene-ous of degree one and concave in v (i.e. Cùùv50), and v is the vector offactor prices. The property Rg52Cg means that the unit increase in thepublic-good production causes a reduction in the value of private goodsproduced equal to its unit cost.

In equilibrium, the domestic rate of return to labour equals the factor’smarginal revenue product in the alternative domestic uses. That is, w5RL(p, g, L). Labour mobility between the labour-importing or labour-exporting country and the rest of the world continues until the domestic netrate of return to the factor equals the world net rate of return. That is:

(12r) w5 (12r) RL(p, g, L)5w*. (2)

The function E(p, g, u) represents the minimum expenditure required toachieve utility level u, at prices p, and at the level of the public-consump-tion good g. An increase in the consumption of the public good reduces theprivate expenditure required to achieve a given level of utility u. In thepublic-economics literature, 2Eg.0 is called ‘the consumer marginal will-ingness to pay for the public good’ (see King, 1986).


It is assumed that the government finances the production of the public-consumption good through income taxes, and that it provides it to house-holds free of charge. The government budget constraint requires thatincome tax revenue from the production of private- and public-consump-tion goods equal the cost of the public-good. That is,

B5r [R(p, g, L)1gCg (v)]2gCg (v)50, (3)

where B is the government net tax revenue.8 Differentiating (3), recallingthat L* is positive for a labour-importing country and negative for a labour-exporting one, and that dL5dL*, after rearranging terms we obtain:

dB5 [rRp 1 (12r)gRgp]dp1 [Rg 1 (12r)gRgg]dg1 [rRL1 (12r)gRgL]dL50. (4)

Observing (4) we note that changes in the domestic relative price of theimportable good, in the level of public-good provision and in the domesticlabour supply, all affect net government revenue through both a direct effectand an indirect effect attributed to the induced changes in the unit cost of thepublic good. Using the properties of the GDP function, (4) indicates that,other things being equal, an increase in the provision of the public goodreduces government net tax revenue – i.e. whether (B/g)5 [Rg1(12r)gRgg],0. Other things being equal, whether a decrease in the domestic relativeprice of the importable good reduces or raises government net tax revenue –i.e. (B/p)5 [rRp1(12r)gRgp] is positive or negative – in part depends onwhether Rgp (5Rpg) is positive or negative. We say that Rgp is positive whenthe importable and public-consumption goods are complements in produc-tion, and negative when they are substitutes.9 In the case of complementar-ity between the two goods in production, (B/p) is positive, while in the caseof substitutability (B/p) may be positive or negative. Hereafter, we call(B/p) the induced price effect of freer trade on government net income taxrevenue. Finally, recalling that Rg52Cg, gRgL is positive when changes in acountry’s domestic employment and payments to factors employed in theproduction of the public good are inversely related, and negative when theyare positively related.10 Thus, when gRgL is positive, international migrationincreases government net tax revenue for a labour-importing country, andreduces it for a labour-exporting one – i.e. (B/L)5 [rRL1(12r)gRgL].0.When gRgL is negative, international migration has an ambiguous impact ongovernment net tax revenue, regardless of whether the country is a labour-importing or a labour-exporting one. We call (B/L) the wage-migrationeffect of freer trade on government net income tax revenue.

We define the country’s trade-expenditure function, which equals theexcess of domestic expenditure over the country’s GDP, as Z(p, g, u, L)5E(p, g, u)-R(p, g, L). The properties of the trade-expenditure functionfollow those of the expenditure and GDP functions. For example, the


Z-function is strictly concave in p (i.e. Zpp,0). Its derivative with respect top (i.e. Zp) gives the country’s import-demand function, i.e. Zp(p, g, L, u)5Ep(p, g, u)–Rp(p, g, L). To indicate that imports under a quota or a VERregime are quantity-constrained we have:

Zp(p, g, L, u)5Ep(p, g, u)2Rp(p, g, L)5 p. (5)

We assume temporary migration (i.e. guest-workers), where immigrantwelfare is considered part of the source-country’s welfare. In the case of alabour-importing country, the budget constraint requires that privateexpenditure must equal net income from the production of private tradedand public goods, minus immigrant remittances, plus any revenue accruingto domestic residents from the prevailing trade restrictions.Correspondingly, for a labour-exporting country emigrant remittances areadded to its total income earnings. It is assumed that tariff revenue is lump-sum distributed to domestic households, that quota rents accrue to domes-tic residents (home government or importers), and that VER rents accrue toforeigners (foreign government or exporters). Then, the income-expenditureidentity for a labour-importing or labour-exporting country’s is given by:

E(p, g, u)5(12r) [R(p, g, L)2g Rg(p, g, L)]2w*L*1(12a)t Zp(p, g, L, u), (6)

where a, the fraction of revenue from trade restrictions accruing to foreign-ers, equals zero in the case of a tariff or an import quota, and one in thecase of a VER. Lastly, w*L* is positive for a labour-importing country andnegative for a labour-exporting one.

Under a tariff regime, (2), (3) and (6) can be solved for the three endoge-nous variables u, g and L as a function of the policy instruments t and r.Under an import quota or a VER regime, (2), (3), (5) and (6) can be solvedfor the four endogenous variables u, g, p and L as a function of the policyinstrument and r. Appendix 1 (p. 107) provides some details of thecomparative statics analysis.

In the remainder of the chapter we examine the effects of relaxing theimport restrictions (a decrease in the tariff rate, or an increase in the levelof imports under a quota or a VER regime) on the level of domesticemployment, and on the level of public-good provision.11 To acquire abetter understanding of the economic intuition of our results, we use thespecific-factors model as a special case.

3 Import restriction and international migration

An issue extensively examined in the international trade literature iswhether goods trade and international-factor movements are substitutes orcomplements. In the neo-classical framework with free trade and FPE, the

Z

Z


two are substitutes in a strong and weak version of a quantitative or price-equalisation sense (see Mundel, 1957). In such a case, trade impediments(e.g. a tariff), which discourage goods trade, are required to encouragefactor movements. The latter continue, in response to factor-pricedifferentials, until goods trade disappears. If, under free trade, factor pricesdo not equalise, goods trade and international-factor movements are sub-stitutes in the weak price-equalisation sense, and in the strong and weakquantitative sense. In this case, no trade impediments are required to induceinternational-factor movements.

In this section we examine, in the presence of public-good provision, theeffects of freer trade, due to lower import restrictions, on labour supply ina labour-exporting or a labour-importing country. Using (A1) and (A2) ofappendix 1, and recalling that a50 under an import tariff or a quota andthat a51 under a VER, the effects of lowering the import restrictions (i.e.dt,0, or d p.0) on the country’s employment are given by:

(dL/dt)5{2RLp1RLg d21 [rRp 1 (12r)gRgp]} d gt (7)

(dL/d p)i 5{2RLp1RLg d21 [rRp1 (12r)gRgp]} d gi, (8)

where i5q (quota), v (VER), gt5 (12r)(12 tZpu)Dt21.0, gq 5 (12r)(12

tZpu)Dq21,and gv 5 (12r)Dv

21,0. Furthermore, recall from the discussionof (4) that [rRp1 (12r)gRgp]5(B/p), and 5(B/g)5 [Rg1 (12r)gRgg],0.

Equations (7) and (8) differ by the respective term gj, j5t, q, v. Thecommon bracketed term in the two equations indicates that lowering thetariff rate (i.e. dt,0) or increasing the volume of imports (i.e. d p.0)under quantitative import constraints affects domestic employmentthrough direct (i.e. 2RLp dgj), and indirect (i.e. RLg (B/p)gj) changes inthe marginal revenue product of labour. The former represent the tradi-tional impact of altered import restrictions in the presence of internationalmigration. The latter, specific to the present context, is due to the inducedprice effect on government net income tax revenue and level of public-goodprovision. In examining these two effects on domestic employment wehave:

Lemma 1: Assume that the importable good is labour-intensive (i.e. RLp.0). Then, a decrease in its domestic relative price, due to lower importrestrictions, entails a direct negative impact on the level of domesticemployment. The opposite result obtains when RLp,0.

Lemma 1 is a well known conclusion in the international trade literature.Under the condition stated in this lemma freer trade, through lower importconstraints, entails a direct positive impact on the number of emigrants

Z

Z

Z


from a labour-exporting country and a direct negative impact on thenumber of immigrants into a labour-importing country.

Lemma 2: Sufficient conditions for lower import restrictions to induce anegative impact on either country’s domestic employment, through theinduced price effect, are either (1) RLg.0 and (B/p).0, or (2) RLg,0and (B/p),0.

Discussion: Consider, for example, the case where RLg.0 and (B/p).0.Then, a decrease in the domestic relative price of the importable, due tolower import restrictions, reduces government net tax revenue, and thus thepublic-good provision. Since also RLg.0, the induced decrease in public-good provision entails a negative impact on the country’s marginal revenueproduct of labour, and its levels of employment. That is, the induced priceeffect on government net income tax revenue entails a positive impact onemigration from a labour-exporting country, but a negative impact onimmigration into a labour-importing country. An analogous reasoningemerges when RLg,0 and (B/p),0.

Lemmas 1 and 2 suggest that in the present context of public-good provi-sion, the overall effect of freer trade, due to lower import restrictions, ondomestic employment in a labour-exporting or in a labour-importing countrydepends on (1) the factor intensity of the importable good, (2) the impact oflower import restrictions on government net tax revenue, which in partdepends on the complementarity or substitutability in production betweenthe importable and public-consumption goods and (3) the relationshipbetween the level of public-good provision and the marginal revenue productof labour. Sufficient conditions for lower import restrictions to overall reducedomestic employment in a labour-exporting or in a labour-importing countryare (1) that the importable good is labour-intensive, (2) that the importableand public goods are complements in production and (3) that public-goodprovision positively affects the marginal revenue product of labour. If one ofthese conditions does not hold, then lower import restrictions may raiseemployment in a labour-exporting or a labour-importing country.

To acquire a better understanding of these results, we apply the specific-factors trade model as a special case.12

3.1 A specific-factors model

Consider a small open labour-exporting or labour-importing country pro-ducing two private traded goods – one exportable, the numeraire, and oneimportable good – and one public-consumption good. Labour is interna-tionally and intersectorally mobile. The same sector-specific capital is used


in the production of the importable (M), and public-consumption good (g),while another type of capital, or land, is the specific factor in the productionof the exportable (A) good. Moreover, drawing on casual empiricism, wewould like to think of a labour-exporting country being a developingcountry where the capital–labour ratio in the production of the importablegood is greater than that in the production of the public-consumption good.The opposite could be true for a labour-importing developed country.

Appendix 2, in the tradition of sector-specific trade models (see Mayer,1974), shows that a decreased domestic relative price of the importablegood (due to lower import restrictions), reduces the wage rate (i.e. RLp.0),and the rate of return to the sector-specific capital in the production of theimportable good (i.e. rp .0). Using these results in the present context ofpublic-good provision, its unit cost of production falls (i.e. Cp

g.0) withtrade liberalisation. As a result, Rgp (52Cp

g,0, implying that the import-able and public-consumption goods are substitutes in production. We con-sider two cases, a case where the public-good is non-capital-intensive (i.e.kg,km) relative to the importable good, and thus RLg is positive, and thecase where kg .km, and thus RLg,0. The former case is expected to holdfor the developing labour-exporting country, but is less likely for the devel-oped labour-importing country, while the latter case is likely to hold for thelabour-importing country (see appendix 2).

Lemma 3, using the above specific-factors model, examines, underpublic-good provision, the overall effect of lower import restrictions on alabour-exporting and a labour-importing country’s employment when kg,km.

Lemma 3: Within this specific-factors model, lower import restrictionsreduce employment in a labour-importing or labour-exporting country if(B/p).0 (sufficient but not necessary) and may increase it if (B/p),0.

Discussion: A reduction in the tariff rate or an increase in the level of importsunder a quota or a VER regime reduces the domestic relative price of theimportable good. Since, within the specific-factors model, RLp.0, the mar-ginal revenue product of labour falls. Thus, through the first right-hand-sideterm of (7) and (8), i.e. 2RLpdgt, lower import restrictions exert a direct neg-ative impact on the country’s domestic employment. The second right-hand-side term in (7) and (8), i.e. RLg d21(B/dp)dgt, is the indirect impactof lower import restrictions on the marginal revenue product of labour,through the induced price effect on government net tax revenue.

Assuming that (dB/dp).0 means that, other things being equal, lowerimport restrictions which reduce the domestic relative price of the importablegood reduce government net tax revenue, and thus public-good provision.


When the public good is relatively non-capital-intensive compared to theimportable good (i.e. kg,km) – a more likely case in a labour-exportingcountry – demand for labour falls, thus affecting negatively the marginalrevenue product of labour, the domestic wage rate and the country’s domes-tic employment. In this case, freer trade has an overall negative effect on thecountry’s domestic employment, by causing emigration If, however (B/p),0, the consequent increase in public-good provision may increase domes-tic employment.

In the case where the production of the public-good is relatively capital-intensive compared to the imported good (i.e. kg .km) – a more likely casefor a developed labour-importing country – then RLg is negative. In thiscase, through the indirect effect and assuming that (B/p),0, implies that,other things being equal, lower import restrictions raise government net taxrevenue and public-good provision. Since the production of this good is rel-atively capital-intensive (i.e. kg .km), demand for labour falls, thus nega-tively affecting the marginal revenue product of labour, the domestic wagerate and the country’s domestic employment. As a result, freer tradereduces the country’s domestic employment by causing emigration. If,however, (B/p).0, the opposite result may emerge.

How do these results differ from those of trade liberalisation on interna-tional migration in the context of a standard (i.e. two-good, three-factor,constant-returns-to-scale production technologies identical across coun-tries) specific-factors model? In such a framework, assuming that labour isthe only internationally mobile factor (i.e. sector-specific factors are inter-nationally immobile), trade liberalisation tends to reduce internationaldifferences in wages and thus dampens its international mobility (seeVenables, 1997).13 Here, under public-good provision, aside of the impact oftrade liberalisation on international migration through the induced directchanges in the price of labour, two additional determinants must be consid-ered. First, the relative factor intensities between the public and the import-able (i.e. protected) goods and, second, the effect of trade liberalisation ongovernment net income tax revenue and the level of public-good provision.

4 Import restrictions and public-good provision

The present context of examining the effects of trade liberalisation raises aninteresting side-issue of international migration – that is, the issue of the pos-sible ‘public-finance’ repercussions of international labour movements. Inparticular, drawing again on casual empiricism, the impact of internationalmigration on government tax revenue and the provision of public goods is arecurring policy issue in labour-exporting and labour-importing countries.

In this section we examine how lower import restrictions, in part through


international migration, affect government net income tax revenue and theprovision of public goods in a labour-exporting or a labour-importingcountry. Using (A1) and (A2) of appendix 1, and recalling that a50 underan import tariff or quota and that a51 under a VER, the effect of lower-ing the import restrictions (i.e. dt,0, ord p.0) on the level of public-goodprovision is, respectively, given by:

(dg/dt)5{2 [rRp 1 (12r)gRgp]1 [rRL1 (12r)gRgL]RLL

21RLp}RLLgt (9)

(dg/d p)i5{2 [rRp1 (12r)gRgp]1 [rRL1 (12r)gRgL]RLL

21RLp}RLLgi, (10)

where i5q, v. Also from the discussion of (4), [rRp 1 (12r)gRgp]5 (B/p),and [rRL1 (12r)gRgL]5 (B/p).

Equations (9) and (10) differ again by their respective gj, j5 t, q, v, term.The common bracketed term in the two equations indicates that either low-ering the tariff rate (i.e. dt,0), or increasing the volume of imports (i.e.d p.0) under quantitative import constraints affects the provision of thepublic good through the induced price effect (i.e. 2 (B/p) RLLgj), andwage-migration effect (i.e. (B/L)RLLgj), on the government net income taxrevenue. In examining these two effects we have:

Lemma 4: If the importable and public-consumption goods are complementsin production (i.e. Rgp.0), then lower import restrictions, through the inducedprice effect, exert a negative impact on public-good provision. This result maynot hold if the two goods are substitutes in production (i.e. Rgp,0).

Discussion: A reduction in the tariff rate or an increase in the level ofimports allowed under a quota or a VER regime entails a decrease in thedomestic relative price of the importable good. As a result, its domestic pro-duction and government net income tax revenue decrease, negativelyaffecting the level of public-good provision. This is captured by the right-hand-side term – rRpRLLgj in (9) and (10). A further decrease in public-good provision, captured by the term 2 (12r)gRgpRLLgj, is induced if thiscommodity is a complement to the importable in production (i.e. Rgp.0).If, however, the two goods are substitutes in production (i.e. Rgp,0), thenthe reduction in import restrictions may positively affect the level of public-good provision through the induced price effect.

Lemma 5: A sufficient condition for lower import restrictions to induce anegative impact on public-good provision through the induced wage-migration effect is that RLp.0 and RLg.0.

Z

Z

Z


Discussion: A reduction in the tariff rate or an increase in the level ofimports allowed under a quota or a VER regime entails a decrease in thedomestic relative price of the importable good. If the importable good islabour-intensive (i.e. RLp.0) then the induced fall in the domestic relativeprice of the importable lowers the domestic wage, employment and, thus,net income tax revenue and the provision of the public good. This is cap-tured by the term rRLRLpgj in (7) and (8). A further reduction in incometax revenue and the provision of the public good, captured by the term(12r)gRgLRLpgj, is induced if the reduction in employment increases theunit cost of the public good. In such a case, the reduction in import restric-tions negatively affects, through the induced wage-migration effect, thelevel of public-good provision. If either of the two conditions does nothold, then freer trade, through lower import constraints, may positivelyaffect the public-good provision through the induced wage-migrationeffect.

Lemmas 4 and 5 suggest that sufficient conditions for lower importrestrictions to overall reduce public-good provision are (1) the importableand public goods are complements in production, (2) the importable goodis labour-intensive and (3) the outflow of labour increases the unit cost ofthe public good. The above results are further discussed within thecontext of the specific-factors trade model described in sub-section (3.1)and appendix 2, and by distinguishing between the case where kg ,km andkg .km.

4.1 The specific-factors model

Within the specific-factors model, it is well known that RLp is positive.Assuming that the same type of sector-specific capital is used in the produc-tion of the importable and public-consumption goods, and that kg,km,international emigration raises payments to factors employed in the pro-duction of the public good and its unit cost of production (i.e. 2gRgL5(dCg/dL),0). Trade liberalisation reduces the domestic price of the import-able, which in turn lowers the marginal revenue product of labour andresults in an outflow of labour from the country. The latter, other thingsbeing equal, raises the domestic rate of return to labour, and reduces therate of return to the specific factors (see appendix 2). Since kg,km, the unitcost of the public good increases, and international migration, through thewage-migration effect, reduces net income tax revenue and the provision ofthe public good.

Within the specific-factors model, Rgp is negative (see appendix 2) andthus the reduction in p (1) reduces the unit cost of the public good and (2)reduces the tax revenue. If the total effect is a reduction in net income tax


revenue, then trade liberalisation reduces the level of public-good provi-sion. If, however, net revenue increases, then trade liberalisation mayincrease the provision of the public good.

When, however kg.km – a likely case for a labour-importing developedcountry – then trade liberalisation that reduces the domestic price of theimportable and the marginal revenue product of labour causes an outflowof labour that raises payments to factors employed in the production of thepublic good and its unit cost of production (i.e. 2gRgL5 (dCg/dL).0). Insuch a case, international emigration, through the wage-migration effect,negatively affects net income tax revenue, and hence the provision of thepublic good if | (12t)gRgl |.tRgL. It is thus more likely that trade liberal-isation increases the provision of the public good when kg.km.

5 Conclusions

What are the links between international trade and incentives for interna-tional migration? In particular, how does trade liberalisation affect factorprices in labour-exporting and labour-importing countries, and thus theincentives for migration? These questions, among other related ones, havepreoccupied economists and policy-makers for a long time. In response, theinternational trade literature has invoked reasons – such as differences inendowments or production technologies among countries, imperfectcompetition and increasing returns – to establish the link between trade ingoods and factors of production. The simplest of these trade models (e.g.HO models), under rather restrictive assumptions, have concluded that thetwo are perfect substitutes.

In this chapter, departing from the standard trade approach, we considera prominent link between the two to be the level of public-goods provisionin labour-exporting and labour-importing countries. In the light of certainfiscal repercussions of international migration, we analyse the effect oftrade liberalisation on the provision of public goods and on internationallabour movements.

We build a general equilibrium trade model of a small (in world goodsand factor markets) open economy, producing one imported, one exportedand one domestic public-consumption good. The country’s imports arerestricted either with tariffs or quantitative restrictions (QRs) while it isassumed that there is perfect international labour mobility.

The effect of trade liberalisation on domestic employment and thus onemigration and on public-good provision depends on (1) the factor inten-sity of the importable good, (2) the relationship in production between theimportable and the public good and (3) the relationship between the levelof the public good and the marginal revenue product of labour. Sufficient


but not necessary conditions for trade liberalisation to reduce the level ofpublic-good provision and to cause emigration are (1) the importable goodis labour-intensive, (2) the importable and the public good are complementsin production and (3) an increase in the level of public-good provision toincrease the marginal revenue product of labour. We use the specific-factorsmodel, as an example, and we show that under plausible conditions it isvery likely that trade liberalisation will cause emigration and reduce thelevel of public-good provision.

The present analysis by no means exhausts the wealth of issues that arisein the context of trade, international migration and the provision of publicgoods. Further research will be required to answer questions such as: Howdo changes in the relative provision of public goods in labour-exporting andlabour-importing countries affect international migration? What are thefactors determining the size of the income tax rate and the desired level ofpublic-good provision? How does international migration respond tochanges in the (income) taxes imposed to finance the provision of publicgoods? What are the effects of international migration on the utility of arepresentative agent in a labour-exporting or a labour-importing country?Appropriate modifications of the present model can accommodate moreefficiently these and related questions.

APPENDIX 1

Under an import tariff, assuming that the labour-importing or labour-exporting country is a small open economy in world markets, (2), (3) and(6) constitute a system of three equations in terms of the endogenous vari-ables u, g and L, as a function of the policy instrument t. Totallydifferentiating the three equations we obtain:

(12tZpu)[Eg 1 (12r)gRgg1 tZpg] [(12r)gRgL1 tRpL] du0 [Rg1 (12r)gRgg] [rRL1(12r)gRgL] dg 50 (12r)gRLg (12r)gRLL dL

2 [rRp 1 (12r)gRgp2 tZpp] 2 (R2gRg)3 2 [rRp 1 (12r)gRgp] 4 dt1 32 (R2gRg)4 dr, (A1)2(12r)RLp] RL

where Dt, the determinant of the left-hand-side coefficients matrix, is posi-tive for stability.

Under an import quota or a VER, assuming that the labour-importingor labour-exporting country is a small open economy in world markets, (2),

4343


(3), (5) and (6) constitute a system of four equations in terms of the endoge-nous variables u, g, p and L, as a function of the policy instrument p.Totally differentiating the four equations we obtain:

where Dj, i5q,v, the determinant of the left-hand-side coefficients matrix isnegative for stability. Note that when a50 we get Dq (,0), and when a51we get Dv (,0).

APPENDIX 2

Within the context of the sector-specific model, we examine the effects ofinternational migration and lower trade restrictions (entailing a smallerdomestic relative price for the importable good), on the rates of return tofactors, and the unit cost of public-good production.

Profit-maximisation under perfect competition requires that the returnto each factor is equated to its marginal revenue product in the alternativeuses (see eq. (2)). That is:

w5pML(Km, Lm, g)r5pMK(Km, Lm, g)q5AK(Ka, 2Lm2Lg, g)pML(Km, Lm, g)5AL(Ka, 2Lm2Lg, g), (B1)

where w, r and q are the rates of return to the intersectorally mobile labour,to capital specific in the production of the importable good, and to capitalspecific in the production of the (numeraire) exportable good. The relativeprice of the former is given by p(5pM/pA), Mj (5M/j), j5Km, Lm and Ai(5Ai), i5Ka, La are the respective marginal products of factors in theproduction of the two private-traded goods.

We assume that it is the prices of traded goods and factor supplies thatdetermine the factor rewards, which in turn determine the unit cost of thepublic good (see Komiya, 1967, for a similar analysis in the context of two

LL

Z


1 [Eg1(12r)gRgg] [rRp1(12r)gRgp]1aZp] (12r)RgL du30 [Rg1(12r)gRgg] [rRp1 (12r)gRgp] [rRL1 (12r)gRgL]43 dg 45Zpu Zpg Zpp 2RpL dp0 (12r)RLg] (12r)RLp] (12r)RLL] dL

(12a)t 2 (R2gRg)3 0 4 dZp1 32 (R2gRg)4 dr, (A2)1 00 RL

traded and one private non-traded good). Thus, total differentiation of(B1), with respect to p, treating Lg and g as constant, yields:

(dw/dp)5V21pALL.0(dr/dp)5MK2V21pMLMKL.0(dq/dp)5V21AKLML,0(dLm/dp)52V21ML.0. (B2)

Total differentiation of eq. (B1), under the same assumptions, with respectto yields:

(dw/d )5V21pMLLALL,0(dr/d )5V21pMKLALL.0(dq/d )5V21pAKLMLL.0(dLm/d )5V21ALL.0, (B3)

where V5 (pMLL1ALL),0. Under constant returns to scale and theassumption of the sector-specific model, MKL (52MLL/km) and AKL (52AKK/ka), km5 (Km/Lm) and ka5(Ka/La) are both positive.

The unit cost of public-good production is given by Cg5wLg1rKg,where Lg and Kg, respectively, are the amounts of labour and sector-specificcapital used in the production of the public good. It is assumed that sector-specific capital used in the production of the public-consumption and theimportable goods is the same. Then, using eq. (B2), it is evident that reduc-ing import restrictions, which entails a reduction in the domestic relativeprice of the importable good, reduces the unit cost of public-good provi-sion. That is:

(dCg/dp)5Cgp5Lg (dw/dp)1Kg (dr/dp).0. (B4)

Recalling the properties of the GDP function, we have Rpg5Rgp (52Cgp)

,0. That is, within the context of the specific-factors model, the import-able and public-consumption goods are substitutes in production.

The effect of international migration on the unit cost of public-good pro-duction can be written as:

(dCg/d )52RgL5(Lg/km) (dw/d ) (km2kg), (B5)

where kg5 (Kg/Lg). Equation (B5) indicates that international migration,which entails a reduction in the source-country’s labour endowment,increases the unit cost of public-good production, i.e. (dCg/d ),0, if km.kg, if the production of the public good is relatively less capital-intensivethan that of the importable good. International emigration decreases theunit cost of public-good production, i.e. (dCg/d ).0, if km,kg, if the pro-duction of the public good is relatively more capital-intensive than that ofthe importable good.

L

L

LL

LLLL

L


NOTES

The authors gratefully acknowledge the constructive comments by R. Faini, I.Musu and participants in the CEPR conference on ‘Trade and Factor Mobility’. Wealone are responsible for any remaining shortcomings of the chapter.21 In the quantitative sense, goods trade and factor movements are substitutes

(complements) if an increase in the volume of trade reduces (raises) the levelof factor movements, or vice versa. In the price-equalisation sense, goodstrade and factor movements are substitutes if free trade in goods impliesFPE, or vice versa. The above definitions constitute the weak version of sub-stitutability or complementarity. In the strong version of the two definitions,the term ‘or vice versa’ is replaced by the term ‘and vice versa’ (see Wong,1995).

22 In 233 neo-classical trade models with endogenous effective supply of the thirdfactor, international trade in goods and factor movements are likely to becomplements rather than substitutes (see Findlay, 1995, ch. 5).

23 Ethier (1985) develops an alternative approach, which emphasises the embodi-ment of key characteristics of international migration (temporary nature, imper-fect substitutability and asymmetric variations in employment between migrantand native labour).

24 For tariff- or quota-induced international capital movements, see among others,Brecher and Diaz-Alejandro (1977); Brecher and Findlay (1983);Hatzipanayotou and Michael (1993).

25 For example, Grubel and Scott (1966) argue that an important way throughwhich emigration affects welfare in labour-exporting countries is through anyinduced changes in the cost of public-goods provision (public education, policeprotection, judicial services, etc.) to the remaining non-emigrant population.The authors note that governments in these countries may reduce the level ofsuch services by nearly the same proportion by which tax revenue declines due toemigration. Thus, the tax burden on non-emigrants, for providing these publicgoods, either remains unchanged or increases marginally. For a recent survey onmigration issues see Borjas (1994, 1995); Friedberg and Hunt (1995);Zimmermann (1995).

26 Simon (1984) maintains that because early immigrant populations are consti-tuted of more workers and fewer dependants, their contribution to the produc-tion cost of public goods and services is relatively high. Moreover, he estimatesthat in the USA, an average immigrant household whose head entered thecountry less than 10 years ago consumes 15–48 per cent fewer public servicesthan a native one. Greenwood and McDowell (1986) quote evidence whereby inthe USA, regardless of sex and duration of residence, immigrant families use upfewer public services compared to natives.

27 Remittances have been a major source of foreign exchange earnings for anumber of traditionally labour-exporting Southern European and Asian coun-tries. For example, during the 1980s, Greece’s average annual foreign exchangeearnings from remittances were approximately US$ 835 million, Spain andPakistan’s US$ 2billion, India’s US$ 2.5 billion and S. Korea’s US$ 1.4 billion(our own calculations, using OECD data).

28 The structure of the present model implies that the government sets the incometax at a rate r, which in turn determines the level of income taxes, and level ofpublic-good provision g. Alternatively, it could be assumed that it is the levelof g that the government chooses, and in turn adjusts the income tax rate to


generate the required income tax revenue for its financing. The process ofdetermining the optimal sizes of r and g is beyond the scope of the presentchapter.

29 From the properties of the GDP function Rgp52Cgp. Then, the importable and

public-goods are complements in production, if the induced decrease in thedomestic relative price of the importable increases the unit cost of public-goodprovision, and substitutes if it decreases the unit cost.

10 Using (1) and (2) we obtain gRgL 52gCgwRVL52VgRVL, where R 5R(pvp(p,g),

g), vp and vg, respectively, are the vectors of primary factors used in the produc-tion of the private traded and the public goods, and vp1vg5v since we areassuming full employment. International migration, which reduces domesticlabour supply, reduces (respectively, increases) payments to factors employed inthe production of the public-good if VgRVL (respectively, ,0).

11 Note that for the purposes of the analysis it is assumed that the income tax rateis constant (i.e. dr50). Thus, changes in the level of the public good are due tothe so-called price effect and wage-migration effect of trade liberalisation on taxrevenue and the cost of public-good provision.

12 In the specific-factors model, free trade in goods, alone, rarely leads to FPE. Butthe properties of the model ensure that only one factor (intersectorally mobileor sector-specific) needs to be internationally mobile for prices of all otherfactors to be equalised in free trade (see Findlay, 1995, ch. 1).

13 If the internationally mobile factor is a sector-specific one (e.g. capital), thentrade liberalisation induces international mobility of that factor and enhancestrade not only due to the lower trade restrictions, but also due to enlargeddifferences in relative factor endowments between countries following interna-tional factor mobility.

REFERENCES

Abe, K. (1995). ‘Tariff Reform in a Small Open Economy with Public-goodProduction’, International Economic Review, 33, 209–22

Borjas, G. (1994). ‘The Economics of Immigration’, Journal of Economic Literature,32, 1667–1717

(1995). ‘The Economic Benefits of Immigration’, Journal of EconomicPerspectives, 9, 3–22

Brecher, R. and J. Bhagwati (1981). ‘Foreign Ownership and the Theory of Tradeand Welfare’, Journal of Political Economy, 89, 497–511

Brecher, R. and C. Diaz-Alejandro (1997). Tariffs, Foreign Capital andImmiserizing Growth’, Journal of International Economics, 7, 317–22

Brecher, R. and R. Findlay (1983). Tariffs, Foreign Capital and National Advantagewith Specific Factors’, Journal of International Economics, 14, 277–88

Dixit, A. and V. Norman (1980). TheTheory of International Trade: A Dual GeneralEquilibrium Approach (Cambridge: Cambridge University Press)

Ethier, W. (1985). ‘International Trade and Labor Migration’, American EconomicReview, 75, 691–707

Findlay, R. (1995). Factor Proportions, Trade, and Growth (Cambridge: MA, MITPress)


Friedberg, R. and J. Hunt (1995). ‘The Impact of Immigrants on Host-countryWages, Employment and Growth’, Journal of Economic Perspectives, 9, 23–44

Greenwood, M. and J. McDowell (1986). ‘The Factor Market Consequences of USImmigration’, Journal of Economic Literature, 24, 1738–72

Grubel, H. and A. Scott (1966). ‘The International Flow of Human Capital’,American Economic Review, 56, 268–74

Hamada, K. and J. Bhagwati (1976). ‘Domestic Distortions, Imperfect Informationand the Brain Drain’, in J. Bhagwati (ed.), The Brain Drain and Taxation, vol.II: Theory and Empirical Analysis (Amsterdam, North-Holland)

Hatzipanayotou, P. and M. Michael (1993). ‘Import Restrictions, Capital Taxes,and Welfare’, Canadian Journal of Economics, 26, 727–38

King, M. (1986). ‘A Pigovian Rule for the Optimal Production of Public Goods’,Journal of Public Economics, 30, 273–91

Komiya, R. (1967). Non-traded Goods and the Pure Theory of InternationalTrade’, International Economic Review, 8, 132–52

Mayer, W. (1974). ‘Short-run and Long-run Equilibrium for a Small OpenEconomy’, Journal of Political Economy, 82, 955–67

Mundel, R. (1957). International Trade and Factor Mobility’, American EconomicReview, 47, 321–35

Simon, J. (1984). ‘Immigrants, Taxes and Welfare in the United States’, PopulationDevelopment Review, 10, 55–69

Venables, A. (1997). ‘Trade Liberalisation and Factor Mobility: An Overview’,chapter 2 in this volume

Wong, K. (1995). International Trade in Goods and Factor Mobility (Cambridge,MA: MIT Press)

Zimmermann, K. (1995). ‘Tackling the European Migration Problem’, Journal ofEconomic Perspectives, 9, 45–62; reprinted in M.N. Jovanovic (ed.),International Economic Integration. Critical Perspectives of the World Economy(London: Routledge, 1998)

Discussion

IGNAZIO MUSU

Chapter 5 examines two interesting related issues: the first is the effectof trade liberalisation on international migration; the second is the effect oftrade liberalisation on provision of public goods in different countries; thetwo effects are connected because the original contribution of this chapter

112 Discussion by Ignazio Musu

is to analyse the effects of trade liberalisation on migration through theeffects on provision of some public good.

The analysis is developed through a series of comparative static exerciseswith a general equilibrium model focusing on the production side of theeconomy and using duality theory. Although there are a number of analyt-ical results, depending on assumptions on labour intensity of the import-able good and on the degree of substitution between the importable and thepublic good, a general outcome seems to emerge from the whole analysis:this quite reasonable result being that trade liberalisation is likely to reduceimmigration or increase emigration not only directly by reducing the labourmarginal revenue, but also indirectly by reducing the tax revenue comingfrom trade restrictions, and hence by providing fewer funds to finance theprovision of a public good which is assumed to have a positive effect on thelabour marginal revenue. Hence the causal chain is from trade liberalisa-tion to lower taxes to lower provision of public good to lower demand forlabour.

A second interesting result more specifically concerns the relationbetween trade liberalisation and provision of public goods. The effect of alower provision of public goods from trade liberalisation comes from lowertax revenue; but this is not only the effect of lower trade restrictions, butalso of the lower level of domestic employment associated with fewer immi-grants and more emigrants. This result seems to me particularly relevant indeveloped countries, where immigrants may be important precisely becausethey are a source of government income to provide some goods which maybe attributed the feature of public goods, as it could be the case with a pay-as-you-go social security system in a country with serious demographicproblems. This result, however, cannot be fully appraised within theassumptions of the model, because among these assumptions we find thatof migration being temporary, which means that income earned by immi-grant workers in the host country are remitted back to the source one. Itwould be worth exploring the effects of relaxing this assumption.

On the other hand, it may be admitted that the effect of a lower provi-sion of public goods because of lower immigration can be a source of lessconcern when the public goods are those specifically required by immi-grants (public order, sheltering structures, health services), in which case,however, the feature of public goods is not fully present. To more satisfac-torily analyse this kind of problem, it would be worth extending the modelintroducing public goods in a utility function, and possibly distinguishingbetween residents and immigrants.

Finally, the chapter assumes full employment and hence directly relatesthe level of domestic employment to the level of migration. Analysing whathappens with disequilibrium in the labour market would be interesting,


because in many countries immigration is a source of concern precisely asa threat to employment of domestic residents.

A reading of this chapter confirms the extreme importance of furtherresearch on the relation between trade liberalisation and internationalmigration; the issue of public goods plays a crucial role in this relation, andin exploring it the chapter provides a relevant and original contribution.

114 Discussion by Ignazio Musu

Part TwoQuantifying the links betweentrade and migration

6 Trade and migration: a production-theory approach

ULRICH KOHLI

1 Introduction

Do guest-workers threaten the jobs of native workers? Do foreigners, willingto work for low wages, drive down domestic wages? These and similar ques-tions have long preoccupied workers, policy-makers and economists, inSwitzerland and elsewhere, and they have become particularly pressing at atime when globalisation is likely to increase exposure of workers fromindustrialised countries to competition from low-wage nations.

One line of research that has proved useful to address this type of ques-tion is what has become known as the ‘production-theory’ approach tomigration.1 This approach, introduced by Baldwin Grossman (1982), treatsforeign labour services as an input to the technology. Moreover, due todifferences in characteristics and attributes, foreign labour is treated asbeing conceptually different from domestic labour. This makes it possibleto determine whether immigrants and natives are substitutes or comple-ments in production. One can then also assess the job-displacement andincome-redistribution effects of international labour mobility.2

It is quite surprising, however, that none of the studies based on the pro-duction-theory framework has modelled international migration within anopen-economy setting. That is, no allowance has been made for possiblelinks between international factor movements and foreign trade.3 Manyadditional questions thus remain unanswered. Does immigration reduce orenhance a country’s foreign trade? Who benefits most from internationalmovements of labour in an open-economy context? How does a change inthe terms of trade affect the distribution of income in the presence of inter-national labour mobility? This lack of empirical evidence is all the moresurprising since the theoretical trade literature has examined the linksbetween trade and factor movements quite carefully without, however,coming to a firm conclusion. Thus, for some authors, foreign trade can beviewed as a substitute for international factor movements,4 while for othersit acts rather as a complement.5 The main objective of this chapter is to

117

attempt to give an empirical answer to some of these questions, and we shallsee that one has to be quite careful in defining in what sense trade and factormovements might be complements or substitutes. In what follows, we willmodel the production sector of the Swiss economy, taking into account theinput of foreign labour together with the input of intermediate products offoreign origin. This chapter can thus be viewed as a contribution towardsthe integration of the production-theory approach to migration and theproduction-theory approach to modelling foreign trade.6

The production-theory approach to modelling foreign trade viewsimports as an input to the technology.7 Imports are used together withprimary factor services to produce goods destined for domestic andforeign markets. This approach recognises the fact that most foreign tradeis in raw materials and non-finished products, and that even most so-called ‘finished goods’ must still go through a number of domestic chan-nels before meeting up with final demand, so that a significant proportionof the final price tag is accounted for by domestic value-added. An attrac-tive feature of the production-theory approach is that it rests on solidtheoretical foundations. Furthermore, it yields a wealth of results aboutthe substitution possibilities allowed for by the technology that cannot bematched by traditional methods which typically rely on single-equationmethods. Last but not least, unlike many models of international tradetheory, the production-theory approach can easily be implementedempirically since the data it requires are of the type contained in theNational Accounts.8 It is a simple matter then to extend the production-theory approach to import determination to allow for internationallabour mobility.

One unfortunate characteristic of the production-theory approach toimmigration is that it has involved a fair degree of confusion. Some of theempirical evidence is in terms of Allen–Uzawa elasticities of substitution,and some of it is in terms of Hicksian elasticities of complementarity.Analysts have often failed to fully appreciate this distinction when com-paring results drawn from different studies. Yet comparison of Allen–Uzawa elasticities of substitution and Hicksian elasticities of comple-mentarity is not a simple matter since the passage from one set of elastic-ities to the other is far from being trivial, and is not always well understood.Two inputs, such as domestic and foreign workers, could be complementsin the Hicksian sense and substitutes in the Allen–Uzawa sense. One objec-tive of this study is to clarify this point by reporting both Hicksian elastic-ities of complementarity and Allen–Uzawa elasticities of substitutionbetween resident and non-resident workers. By presenting our results forthese alternative settings, we will be better able to assess the effects ofimmigration on the income of domestic factors of production, and to

118 Ulrich Kohli

examine the effects of changes in the payments to immigrants on theemployment opportunities of resident workers. However, we will argue thatother sets of elasticities may be still better suited to analyse the impact ofimmigration in an open-economy context; this will make it possible toassess the impact of changes in domestic factor endowments and in tradedgood and service prices on immigration, foreign trade, production and thedistribution of income.

Immigration has long been an important issue in Switzerland where alarge proportion of the resident labour force is of foreign origin, and wherenon-resident workers have, at times, made up nearly a quarter of the labourforce. Non-resident workers either are holders of seasonal or yearlypermits, or they live in neighbouring countries and cross the border on adaily basis to work in Switzerland. As shown by figure 6,1, relative inputshave changed substantially over time. The use of resident labour does notexhibit much of a trend throughout the post-war period. Increases in thedomestic labour force have been offset to some extent by reductions in thelength of the work week. The use of non-resident labour, on the other hand,increased very substantially throughout the 1950s and much of the 1960s,but it fell dramatically in the mid-1970s. The input of capital increasedsteadily, whereas the use of imports, on average, increased at an even faster

A production-theory approach 119

Figure 6.1 Input quantities

pace. The sharp reduction in the number of non-resident workers in themid-1970s is apparent in figure 6.2, which shows an implicit Tornqvistindex of aggregate inputs (dashed line), and it has had a marked effect onthe level of gross output (solid line). The difference between the two linescan be interpreted as total factor productivity (TFP), and it has increasedthroughout much of the period.

Concern has been expressed recently that flexible functional forms maynot be flexible enough to adequately model technological change and pro-ductivity growth, see Diewert and Wales (1992). This may be particularlyrelevant here, given the shock which apparently – judging from figure 6.2 –affected the Swiss economy in the mid-1970s. As suggested by Diewert andWales, one way of handling the problem is with the help of spline functions.In this chapter, we will innovate and experiment along an alternative route,by increasing the level of flexibility of the aggregator function with respectto time.

The remainder of this chapter is organised as follows. Section 2 brieflyreviews alternative descriptions of the aggregate technology. Section 3 dis-cusses the comparative statics of the model, while section 4 examines itsempirical implementation. Sections 5 and 6 present our main empiricalresults, and section 7 draws some conclusions.

120 Ulrich Kohli

Figure 6.2 Gross output and aggregate input

2 The production-theory approach to modelling the demand forimports and foreign-labour services

The production-theory approach to migration treats immigrants as an inputto the technology. Similarly, the production-theory approach to foreigntrade views imports as intermediate products. We will therefore considerthat aggregate output is produced with the use of four inputs: imports (M),non-resident labour (N), resident labour (L) and capital (K). The aggregatetechnology can be represented by the following production function:

y5f(x), (1)

where y is the quantity of gross output, x; [xj] (j[M, N, L, K) is the vectorof input quantities. We assume that f(x) is increasing, quasi-concave, andlinearly homogeneous with respect to the components of x.

Variable returns to scale have become an important ingredient of manymodels, in international trade theory as well as in growth theory. However,under variable returns to scale, aggregation over agents in order to get arepresentation of the country’s technology may become next to impossible,due to departures from perfect competition, and because intermediategoods no longer necessarily net out. We do not believe that variable returnsto scale at the plant, firm or industry level can be adequately accounted forby simply relaxing the linear homogeneity assumption at the aggregatelevel. For the sake of coherence, we therefore prefer to proceed with con-stant returns to scale as our maintained hypothesis.

Our treatment clearly rests on some heroic assumptions about aggrega-tion. For a start, we assume that all outputs can be aggregated into a singlecomposite good, which can be either absorbed at home or exported to therest of the world.9 Furthermore, considering four inputs only may seemrestrictive as well. The reader may feel that it would important to dis-tinguish between skilled and unskilled, resident and non-resident workers,and perhaps that one should disaggregate capital and/or imports. In truth,the theoretical model which we will develop in this section and the next caneasily accommodate disaggregation almost ad infinitum. It suffices to inter-pret y and the xjs as vectors, and to use a transformation function insteadof a production function as a starting point. To keep the empirical applica-tion manageable, however, one must obviously draw the line somewhere.Degrees of freedom rapidly vanish as the number of inputs and/or outputsincrease, and difficulties meeting the required regularity conditions withoutcompromising flexibility rapidly becoming overwhelming. In reference tothe literature, our model of the aggregate technology, by considering fourinputs, is already more general than most.

It would of course also be of interest to disaggregate labour between


native and foreign-born workers, or between Swiss citizens and foreigners.This does not seem to be possible, however, since the income data are notavailable. One can argue, however, that the distinction between non-resi-dent and resident workers, which is dictated by data availability, does makesense, since it amounts to distinguishing between recent and temporaryimmigrants (‘guest-workers’), on the one hand, and non-recent and long-term immigrants on the other, the latter being well integrated and similarto Swiss workers in most respects.

The use of non-resident labour services is somewhat concentrated, both interms of industries and geographically. The number of transborder workersis naturally largest in border areas, particularly so in Geneva, Basle and theTicino. The ‘frontaliers’ tend to be employed in a large variety of industries,and in a wide range of skills. Holders of seasonal and yearly permits, on theother hand, tend to be more evenly distributed across the country, but theytend to be low-waged and concentrated in particular industries, such asconstruction or food and lodging services. This might militate in favour ofusing a two-sector model. However, as argued previously, even though dis-aggregation of output could easily be handled by the model, it would makethe empirical application that much more complex, and it is not clear that ifsome further disaggregation were indeed contemplated, this would be themost urgent direction in which it should be carried out.10

While international labour mobility is an important issue in the Swisscase, one could argue that capital mobility is even more relevant. Noattempt is made here to model international capital movements. There areseveral reasons for this. Some practical ones first: distinguishing betweendomestically and foreign owned capital would increase the size of themodel, and data on the quantity and the price of both types of capitalwould be difficult to obtain. In fact, it is likely that as far as productionpossibilities are concerned, the question of ownership is largely irrelevant.This suggests that the return to both types of capital might be same, inwhich case aggregation can be justified on Hicksian grounds. There are alsosome conceptual reasons why the issue of international capital mobility isnot tackled here. By focusing on current production decisions exclusively,we do not attempt to model the investment process. In any case, interna-tional capital flows typically take the form of movements in equity andfinancial assets: it is the ownership of the capital stock that is traded, ratherthan the physical capital stock itself. Modelling international capital flowsas exogenous changes in the domestic endowment of capital seems veryunsatisfactory. The dismantling of a machine or a factory in order torebuild it in a different country is the exception, not the rule. Moreover, tothe extent that international capital flows can have a direct effect on thedomestic endowment of capital, they should first impact on the trade

122 Ulrich Kohli

account – as part of imports or exports of capital goods – and on grossoutput – as part of domestic investment or disinvestment.

Under cost-minimisation, the country’s technology can also be describedby the unit cost function defined as:

c(w);minx

wjxj : f(x)$1 (2)

where w; [wj] is the vector of input prices. Under competitive conditions,unit cost is equal to the price of aggregate output (p).

By Shephard’s (1953) lemma, the unit-output cost-minimising demandfor inputs can be obtained by differentiation of (2):11

5 , j[{M, N, L, K}. (3)

It is noteworthy that the demand for all inputs, including imports andforeign labour services, will generally depend on all four input prices. Achange in the price of imports, for instance, is liable to affect the demandfor foreign labour services, just as a change in the wage rate of non-residentworkers may impact on the demand for imports.

To assess these price effects – and, more generally, the substitutionpossibilities allowed for by the technology – it is useful to compute theAllen–Uzawa elasticities of substitution between input j and input k, sjk. Itis of course well known that:12

sjk5 , j, k[{M, N, L, K}, (4)

where cj(w);c(w)/wj, and cjk(w);2c(w)/(wjwk); sjk is positive if inputsj and k are substitutes in the Allen–Uzawa sense and it is negative if the twoinputs are complements. These elasticities are relevant if one wishes toassess the impact of an increase in the price of one input on its quantitydemanded, and on the quantities of all other inputs. Thus, if sLN is positive,one can assert that a reduction in the wage paid to guest-workers will reducethe demand for native labour services, and could thus lead to increasedunemployment among resident workers.

The Allen–Uzawa elasticities of substitution given by (4) are defined forgiven input prices, and they are not to be confused with Hicksian elastic-ities of complementarity, cjk, defined for given input quantities.13 If the pro-duction function is known, the elasticities of complementarity can beobtained directly as:

cjk5 , j, k[{M, N, L, K}, (5)f(x)fjk(x)fj(x)fk(x)

c(w)cjk(w)cj(w)ck(w)

c(w)wj

xj

y

65oj


where fj(x);f(x)/xj, and fjk(x);2f(x)/(xjxk}); cjk is positive if inputs jand k are q-complements in the Hicksian sense, and it is negative if they areq-substitutes. The sign of cLN is crucial in determining whether an increasein immigration raises or reduces the return to domestic labour.

The Hicksian elasticities of complementarity can also be derivedfrom estimates of the cost function. Let us define S; [sjk] as theAllen–Uzawa substitution matrix and C; [cjk] as the Hicksian q-com-plementarity matrix. As shown by Kohli (1991), the link between C andS is given by:

C5 S21S21S21, (6)

where:

C uC; 3 4 (7)

u9 0

S uS ; 3 4 (8)

u9 u

S is a 535 diagonal matrix with the four cost shares and a one as ele-ments, and u is a four-dimensional unit vector. Thus, the passage fromone set of elasticities to the other is not trivial, and it involves the inver-sion of a bordered matrix. Note that as soon as the number of inputsexceeds two, one cannot infer anything about the sign of sjk from thesign of cjk alone.14 Thus, two inputs could be Hicksian q-complements,and yet they could equally well be Allen–Uzawa substitutes or comple-ments.

3 Comparative statics

3.1 The cost-function setting

While both the Allen–Uzawa substitution matrix S and the Hicksian q-complementarity matrix C are informative as to the relationships betweenthe various inputs, their elements – particularly their size – are somewhatdifficult to interpret. It is often more convenient to resort to ordinary priceand quantity elasticities. Consider demand system (3). The comparativestatics of that system can be captured by a matrix of price elasticities ofdemand E; [«jk] defined as follows:

124 Ulrich Kohli

«jk; . (9)

It is well known that the elements of E can be obtained directly from theAllen–Uzawa elasticities of substitution:15

«jk5sjksk, (10)

where sk is the cost share of the kth input, sk;wkxk/(Swjxj). Moreover, itfollows directly from (2) and (3) that ln xj(w,y)/ln y51, that ln c(w)/lny50, and that ln c(w)/ln wj5sj. The full comparative statics of the modelcan therefore be summarised by the following system:

where the hats (^) denote relative changes.

3.2 The production-function setting

Alternatively, if one departs from the Hicksian elasticities of comple-mentarity, one would naturally be led to the matrix of quantity elasticitiesof inverse demand, H; [h jk], to describe the comparative statics of themodel. These elasticities are defined as:

h jk; . (12)

The elements of H can be obtained directly from the Hicksian elasticitiesof complementarity as follows:

h jk5cjksk, (13)

where sk is again the cost share of the kth input. The full comparative staticsof the model can now be summarised in the following way:

ln wj(x,p) ln xk

lnxj(w,y) lnwk


«MM «MN «ML «MK : 1xM wM

«NM «NN «NL «NK : 1xN wN

«LM «LN «LL «LK : 1 , (11)xL 5 wL

«KM «KN «KL «KK : 1xK wK… … … … … … …p y

sM sN sL sK : 0

3 43 43 4

where we have taken account of the first-order conditions for profit-maximi-sation which imply that ln f(x)/ln xj5sj. Moreover, ln wj(x,p)/ln p51, andln f(x)/ln p50. It is immediately visible that the square matrix in (14) is simplythe inverse of the square matrix in (11); this illustrates our earlier result (6).

3.3 The GNP-function setting

While both the price elasticities of demand (matrix E) and the quantityelasticities of inverse demand (matrix H) are relevant in some circumstances,one can argue that in an open-economy context, yet another set of elastic-ities is necessary. Indeed, it is reasonable to assume that the endowments ofdomestic factors (capital and resident labour) are given. At the same time,considering Switzerland as a small open economy, we can view the prices ofoutput, imports and non-resident labour services as pre-determined aswell.16 In such a setting, which is known as the GNP-function framework,17

the substitution possibilities allowed for by the technology can be describedby the following matrix of price and quantity elasticities, «; [ejk]:

126 Ulrich Kohli

hMM hMN hML hMK : 1wM xM

hNM hNN hNL hNK : 1wN xN

hLM hLN hLL hLK : 1 , (14)wL 5 xL

hKM hKN hKL hKK : 1wK xK

… … … … … … …y p

sM sN sL sK : 0

3 4 433 4

eYY eYM eYN : eYL eYKy p

eMY eMM eMN : eML eMKxM wM

eNY eNM eNN : eNL eNKxN 5 wN . (15)… … … … … … …wL xK

eLY eLM eLN : eLL eLKwK xK

eKY eKM eKN : eKL eKK

4 43 33 4

These elasticities are defined for given prices of gross output, imports andforeign labour services, and for given domestic factor endowments. Theelasticities in the northwest corner of (15) are the price elasticities of outputsupply and variable input demand. The elasticities in the southwest cornerindicate the impact of changes in the prices of output and variable inputson domestic factor rental prices (the Stolper–Samuelson (SS) effects). Theelasticities in the northeast corner of (15) capture the effect of changes indomestic factor endowments on the supply of output and on the demandfor variable inputs (the Rybczynski effects). In the southeast corner, finally,we find the quantity elasticities of the inverse demand for the domesticfactors.

It is noteworthy that the homogeneity properties of the GNP functionthat is dual to f(x) imply the following restrictions:

Sheih 5 0, i, h[{Y,M,N}

Skejk 5 0, j, k[{L,K}

Sjeij 5 1, i[{Y,M,N}, j[{L,K}

Sieji 5 1, j[{L,K}, i[{Y,M,N}.

The signs of eLN and eKN are particularly important in determining theimpact of increased immigration on domestic factor payments; thus, ifeLN.0 a reduction in wN, which leads to an increase in the demand forguest-workers, will tend to reduce the rental price of resident labour. Thesign of eMN, on the other hand, will indicate the impact of changes in thewage rate paid to non-resident workers on the demand for imports. IfeMN.0, immigration will tend to act as a substitute for foreign trade in thesense that an increase in the cost of non-resident labour would tend tofavour imports.

The elasticities contained in (15) can easily be calculated from the esti-mates of the price elasticities of input demands contained in (11). That is,it is a simple matter to solve the system of (11) for y, xM, xN, wL and wK ,as functions of p, wM, wN, xL and xK.18 Of course, the GNP-function elastic-ities could be obtained directly if one chose to use the GNP-function todescribe the aggregate technology, just as the elasticities in (14) could beobtained directly from the production function, without having to rely on(6). In order to best illustrate the difference between the various sets ofelasticities, however, we want to derive them all from the same set ofeconometric parameter estimates. That is, we do not want to change theeconometric specification as we move from one set of elasticities to thenext.


3.4 Immigration quotas

One might argue that the GNP-function setting is inappropriate in that ittreats foreign labour services as a variable input, whereas, in practice,working permits are not issued freely. This militates in favour of treating xNas a fixed, rather than as a variable, input. Thus, compared to (15) therewould now be three – rather than just two – fixed inputs. The comparativestatics of the model could then be summarised by the following matrix ofprice and quantity elasticities, F; [fjk]:

Matrix F can easily be computed from estimates of E. The elements ofF will indicate the impact of higher immigration on domestic factor rentalprices, as well as on imports and gross output. At the same time, they willyield information about the impact of a change in the terms of trade ondomestic as well as foreign factor rental prices. A negative value of fMNwould suggest that immigration and trade are substitutes in the sense thata larger contingent of non-resident workers tends to reduce the demand forimports.

3.5 Domestic workers’ displacement effects

One of our original questions was: does immigration tend to reduce employ-ment opportunities for native workers? None of the settings examined so farmakes it possible to directly answer this question. Indeed, to address thedomestic employment issue, one should treat the input of resident workersas variable. This would be legitimate in the short run in the presence ofdomestic wage rigidities, for instance. Thus, it would be of interest toexamine the effect of higher immigration when resident labour, togetherwith imports and gross output, are treated as variable. The comparativestatics of the model could then be described by the following matrix Q; [u]:

128 Ulrich Kohli

fYY fYM : fYN fYL fYKy p

fMY fMM : fMN fML fMKxM wM… … … … … … …wN 5 xN . (16)

fNY fNM : fNN fNL fNKwL xL

fLY fLM : fLN fLL fLKwK xK

fKY fKM : fKN fKL fKK

4 43 33 4

The price and quantity elasticities are defined for given output, import,and resident labour prices, and for given quantities of non-resident labourand capital. A negative value of fMN would indicate that immigration andtrade are substitutes, in the sense that access to a larger pool of non-resi-dent workers would tend to reduce the demand for imports, given variabledomestic employment.

At this stage, the reader might wonder which one of the five settingsreviewed here best describes reality, and which one is the most useful.19 Infact, all five settings describe the same technology, and thus are equally‘correct’ or ‘incorrect’. Yet, some settings may be more relevant thanothers, depending on the circumstances and on the problem which onewants to analyse. The cost-function setting is probably the one that is bestknown, and it is useful in the short run, when input prices may be stickyand the level of output pre-determined. In the long run, on the otherhand, the production function and the GNP-function settings may bemore useful. If immigration quotas are the issue, then clearly the elastic-ities contained in (16) and (17) are more relevant. Thus, there does notexist a set of elasticities that is to be favoured over all other ones. Ourpoint here was rather to show that a custom-built set of elasticities thatbest suits the problem at hand can easily be obtained. Furthermore, greatcare should be taken when comparing elasticities drawn from differentsources, since elasticities typically only show partial – or ceteris paribus –effects, and it is essential to always keep in mind what the other thingsbeing held constant actually are. The obvious implication of this is thatempirical estimates of a particular effect – the impact of an increase in thenumber of guest-workers on the earnings of resident workers – mightdiffer greatly depending on which other variables are allowed to adjust,and which ones are held constant. Finally, one should be aware of the factthat our approach, by focusing on production decisions exclusively,


uYY uYM uYL : uYN uYKy p

uMY uMM uML : uMN uMKxM wM

uLY uLM uLL : uLN uLKxL 5 wL . (17)… … … … … … … …wN xN

uNY uNM uNL : uNN uNK

wK xK

uKY uKM uKL : uKN uKK

4 43 33 4

merely offers a partial equilibrium treatment. Imaginative use of this tech-nique makes it possible to deal with a large variety of circumstances, andto answer many questions, but there is no doubt that this approach has itslimits, and that ultimately only a general equilibrium approach, endogen-ising the prices as well as the quantities of all inputs and outputs, can givecomplete answers. This is particularly true in the immigration debatesince, for instance, the supply of migrants should probably be explainedas well. Similarly, one could argue that the supply of foreign goods, thework–leisure decision of domestic workers, the level of capacity utilisa-tion, domestic investment decisions, and intertemporal consumptionbehaviour should be modelled too.

4 Empirical implementation

4.1 Functional form

Upon specification of a functional form, cost function (2) can be estimated.However, it is statistically more efficient to estimate instead the system ofderived demand functions (3). The estimates can then be used to recover(2), and to calculate estimates of (4), (6), (10), (13) and (15)–(17).

In what follows, we will use the Translog functional form (Christensen etal., 1973). It can be written as:

ln c(w,t)5a0 1 aj ln wj1 gjk ln wj ln wk 1

aTt1 djT ln wjt1 dTTt2, j, k [{M, N, L, K}, (18)

where t is a time trend intended to capture the effects of technologicalchange. Symmetry implies gjk5gkj. Linear homogeneity requires Sjaj51,Sgjk50, and Sjdj T50.

As indicated by (3), the input demand functions can be obtained bydifferentiation of (18). In share form:

sj 5aj 1 gjk ln wk1djTt, (19)

where sj is once again the cost share of the jth input.

4.2 Smoothing

Some concern has been expressed that flexible functional forms may not beflexible enough to pick up the ups and downs of technological change and

ok

12o

j

12 o

jo

ko

j

130 Ulrich Kohli

productivity growth (see Diewert and Wales (1992), for instance). TheTranslog cost function (18) provides a second-order approximation to anarbitrary cost function, with respect to prices as well as with respect to time.This means that TFP is modelled as a quadratic function of time, or thatthe rate of technological change, which can be obtained by differentiating(18) with respect to t, is a linear function of time. That is, except for theinfluence of changing relative prices, the rate of technological change iseither monotonically increasing or decreasing throughout the entireestimation period. This may be too restrictive an assumption, and someadditional flexibility may be needed. One way of handling this problem iswith the help of spline functions, see Diewert and Wales (1992); Fox (1994).An alternative may be focusing on TFP and applying a number of filters,see Fox (1994); Fox and Kohli (1998). In this chapter, we will experimentalong an alternative route, by increasing the level of flexibility of the aggre-gator function with respect to time. That is, while we continue to assumethat the cost function provides a second-order approximation with respectto prices, we increase the degree of the approximation with respect to timeto a degree t.2. In what follows, we consider values for t equal to 3 and 4.In the t54 case, the Translog cost function becomes as follows:

ln c(w,t)5a0 1 aj ln wj1 gjk ln wj ln wk 1

aTt1 djT lnwj t1 djTT lnwj t21 djTTT lnwj t

3 1

dTTt2 1 dTTTt3 1 dTTTTt4, j,ke{M,N,L.K}, (20)

where SjdjTT50 and SjdjTTT50. The case t53 requires djTTT5dTTTT50,;j[M,N,L,K. If, moreover, one imposes djTT 5dTTT50, ;j[M,N,L,K,one is back to the t52 standard flexible specification (18). In the generalcase, the input demand-share equations become:

sj5aj 1 gjk ln wk1djTt1djTTt21djTTTt3. (21)

One could obviously keep increasing the flexibility with respect to timealmost ad infinitum, or at least until all degrees of freedom are exhausted.However, there seems to be little point in doing so. For one thing, while onewishes to make the model sufficiently flexible with respect to time, one alsowishes to keep it as parsimonious as possible. The t54 version of the modelallows for two inflection points in the path of TFP, as opposed to none withthe standard specification. This should be enough to pick up the major

ok

124

16

12

oj

oj

oj

12 o

jo

ko

j


trends in factor-productivity developments over a 30- or 40-year period,and it is precisely the number advocated by Diewert and Wales (1992) intheir study. Another reason why one may not want to increase the value oft beyond reason is that, while it will tend to increase the goodness of fit, itwill gradually squeeze out any role for economic variables such as relativeprices. One knows, indeed, that any economic time series can be exactlyexplained by a polynomial of time of sufficiently high degree.

4.3 Data

The cost function is estimated for Switzerland with annual data coveringthe period 1950–86. We require price and quantity series for all four inputsand for gross output. Gross output and import figures are derived from theNational Income and Product Accounts, together with the income sharesof labour and capital. The price of imports is inflated by import duties,whereas the price of gross output is net of indirect taxes, but includes sub-sidies.

The quantity of labour is obtained by multiplying employment figures bythe average length of the work week. The quantity of capital is calculatedas a Tornqvist quantity index of structures and equipment. Both stocks areobtained by cumulating the corresponding real investment series, subject toexogenous rates of depreciation; this was accomplished over an extendedperiod of time (starting in 1914) so that by the beginning of the sampleperiod the figures are essentially independent of initial values. The price oflabour and capital services were obtained by deflation.

The disaggregation of employment between resident and non-residentlabour is based on Favarger (1992). The resident-worker category com-prises natives as well as foreign workers who are residents of Switzerland.Non-resident workers are holders of seasonal permits, annual permits ortransborder permits. Data on the number of non-resident workers arereadily available, whereas their income can be derived from the Swissbalance of payments which records payments to non-resident workers. Allprices are normalised to unity for 1980. Quantities are expressed in million1980 Swiss francs; t is defined as a time trend with unit annual incrementsand normalised to zero for 1980. Summary statistics of the input data areshown in table 6.1; additional details, and the series themselves, are avail-able upon request.

4.4 Stochastic specification and estimation techniques

We assume that demand functions (21) are exact, except for errors in optim-isation. We specify a vector of additive disturbances which we assume to be

132 Ulrich Kohli

identically distributed, serially independent, normal random vectors withmean vector zero. The model is estimated by using the algorithm of Berndtet al. (1974); this is essentially an iterative version of Zellner’s (1962)method for seemingly unrelated regressor equations, and it is numericallyequivalent to maximum likelihood.

5 Empirical results

Estimates of (19) are reported in table 6.2, column (1). We verified that theestimated cost function satisfies all required regularity conditions (mono-tonicity and concavity) for all observations.20 Asymptotic t-values areshown in parentheses. We next re-estimated the t53 and t54 versions ofthe cost function – see (20) above. The corresponding parameter estimatesare shown in columns (2) and(3) of table 6.2; again, we verified that all regu-larity conditions are met for all observations. It is apparent that the addi-tional flexibility with respect to time leads to a significant improvement inthe goodness of fit: conducting a likelihood-ratio test for the hypothesist52, conditional on the alternative hypothesis t53, produces a test statis-tic of 61.96 for a x2 critical value of 9.21, at the 99 per cent confidence levelwith two degrees of freedom. The test for t53, conditional on t54, yieldsa test-statistic of 30.64 for the same x2 critical value. The lower levels of flex-ibility with respect to time are therefore decisively rejected, and we proceedwith the t54 estimates shown in table 6.2, column (3), as our preferred setof estimates.21


Table 6.1. Input data – summary statistics

Mean 1980 value Stand. dev. Min Max(1) (2) (3) (4) (5)

wM 0.77652 1.00000 0.17273 0.56926 1.12891wN 0.59952 1.00000 0.40951 0.11137 1.42308wL 0.60442 1.00000 0.40484 0.13611 1.46850wK 0.80162 1.00000 0.18901 0.58467 1.25165

xM 40214.2 71760.7 23703.0 29707.8 86606.9xN 12520.9 28963.9 25075.9 24500.2 20925.9xL 80794.0 91691.4 28017.2 67127.4 91691.4xK 30126.0 43794.3 13832.5 28918.7 50798.1

sM 0.27931 0.33190 0.02498 0.22901 0.33190sN 0.06451 0.04146 0.02651 0.02523 0.11093sL 0.42359 0.42408 0.03667 0.36734 0.50230sK 0.23259 0.20255 0.01688 0.20114 0.25374

134 Ulrich Kohli

Table 6.2. Parameter estimates (asymptotic t-values inparentheses)

t52 t53 t54(1) (2) (3)

aM 0.30889 0.30778 0.30748(71.22) (67.28) (62.49)

aN 0.04511 0.05015 0.04089(11.01) (14.27) (15.42)

aL 0.43342 0.42502 0.43595(54.45) (72.68) (77.65)

gMM 20.08138 0.07127 0.07439(2.82) (1.86) (2.03)

gMN 20.10102 20.05914 20.07482(24.67) (22.41) (24.08)

gML 0.07754 0.03797 0.04292(2.42) (0.98) (1.17)

gNN 20.10114 20.04785 20.05360(23.50) (21.68) (23.10)

gNL 0.25804 0.12210 0.16097(6.72) (3.17) (6.45)

gLL 20.42431 20.11577 20.18427(26.52) (21.86) (23.78)

dMT 0.00345 0.00292 0.00355(3.78) (3.48) (3.68)

dNT 20.00793 20.00724 20.00604(29.14) (211.09) (2 11.44)

dLT 0.00732 0.00586 0.00379(4.31) (5.37) (3.42)

dMTT .— 20.00001 20.00003(20.39) (20.30)

dNTT .— 20.00014 0.00022(25.48) (4.09)

dLTT .— 0.00029 20.00018(6.67) (21.61)

dMTTT .— .— 0.00000(0.63)

dNTTTT .— .— 0.0001(6.92)

dLTTT .— .— 20.00001(24.49)

LL 351.01 381.99 397.31

The parameter estimates in table 6.2 can be used to calculate Allen–Uzawaelasticities of substitution (sjk) and Hicksian elasticities of complementarity(cjk); these are reported for selected years in tables 6.2 and 6.3, respectively.It is apparent from table 6.3 that non-resident workers are Allen–Uzawacomplements for imports as well as for capital.22 All other pairs of inputs areAllen–Uzawa substitutes for each other. Judging from table 6.4, non-residentworkers and resident workers are Hicksian q-substitutes for each other,23

whereas all other pairs of inputs are Hicksian q-complements for each other,


Table 6.3. Allen–Uzawa elasticities of substitution, selected years

1950 1960 1970 1980 1986

sMM 21.906 21.721 21.613 21.465 21.432sMN 213.810 22.734 22.100 24.951 24.078sML 1.353 1.401 1.390 1.320 1.316sMK 0.202 0.368 0.377 0.359 0.344sNN 2161.943 221.471 217.976 255.513 244.574sNL 15.506 6.232 5.813 10.030 8.914sNK 25.709 20.683 20.569 22.692 22.358sLL 21.623 22.589 22.770 22.263 22.293sLK 0.832 0.810 0.793 0.791 0.780sKK 21.571 21.466 21.515 21.601 21.623

Note:These estimates are based on the parameter values shown in table 6.1, column (3).

Table 6.4. Hicksian elasticities of complementarity, selected years

1950 1960 1970 1980 1986

cMM 221.928 26.243 25.266 25.529 24.984cMN 44.668 5.982 4.821 8.970 7.368cML 9.752 1.806 1.572 2.620 2.238cMK 23.712 1.809 1.885 0.885 1.203cNN 2118.381 215.491 213.190 233.162 227.604cNL 224.268 23.225 22.776 27.561 26.245cNK 20.226 3.547 3.533 8.782 8.188cLL 25.815 21.824 21.790 22.796 22.495cLK 5.447 1.991 2.039 3.350 3.255cKK 210.510 26.095 26.728 29.698 210.519

Note:These estimates are based on the parameter values shown in table 6.1, column (3).

except early on in the sample when one finds some evidence of q-substi-tutability between imports and capital.24 Thus, trade and labour mobility arecomplements in the Swiss case, both in the Allen–Uzawa sense and in theHicksian sense. Non-resident and resident workers, on the other hand, aresubstitutes in both settings. Imports and capital, as well as resident workersand capital, are substitutes in the Allen–Uzawa sense, but complements inthe Hicksian sense for most of the sample period.

To get a better feeling of the magnitudes involved, we move to tables 6.5and 6.6, which report 1986 estimates of matrices E and H. Consider the

136 Ulrich Kohli

Table 6.5. Price elasticities of input demand (cost-functionsetting, matrix E, 1986 estimates)

«mn; lnhm(y,wM,wN,wL,wK) / lnzn

hme{xM,xN,xL,xK},zne{wM,wN,wL,wK}

zn5wM zn5wN zn5wL zn5wK

«Mn 20.449 20.192 0.570 0.071«Nn 21.279 22.094 3.860 20.487«Ln 0.413 0.419 20.993 0.161«Kn 0.108 20.111 0.338 20.335

Note:These estimates are defined for a given quantity of output andgiven input prices. They are based on the parameter values shownin table 6.1, column (3).

Table 6.6. Price elasticities of input demand (production-function setting, matrix H, 1986 estimates)

hmn; lnhm(p,xM,xN,xL,xK) / lnzn

hme{wM,wN,wL,wK},zne{xM,xN,xL,xK}

zn5xM zn5xN zn5xL zn5xK

hMn 21.563 0.346 0.969 0.248hNn 2.311 21.297 22.704 1.690hLn 0.702 20.293 21.080 0.672hKn 0.377 0.385 1.409 22.171

Note:These estimates are defined for a given price of output and giveninput quantities. They are based on the parameter values shownin table 6.1, column (3).

price elasticities of input demands shown in table 6.5 to begin with. Theown-price elasticity of the demand for imports is estimated to be 20.449 –which may seem quite low – while the own-price elasticity of the demandfor non-resident labour is found to be rather large in absolute value, at2 2.094. Remember though that these price elasticities are defined for giveninput prices and a given level of gross output. The estimate of «MN showsthat a 1 per cent increase in the cost of non-resident labour services wouldreduce the demand for imports by about 0.2 per cent, but increase thedemand for resident labour by about 0.4 per cent. A 1 per cent increase inthe price of imports, on the other hand, would reduce the demand forforeign labour services by about 1.3 per cent.

Looking at the estimates of the hjks shown in table 6.6, one sees that a1per cent increase in the number of non-resident workers would increasethe marginal product of imports by about 0.3 per cent, the marginalproduct of capital by about 0.4 per cent, but reduce the marginal productof resident workers by about 0.3 per cent. An increase in the endowment ofcapital, on the other hand, would enhance the marginal product of all threeother inputs, not least non-resident workers.

As argued earlier, it might be more appropriate to consider the prices ofgoods and of non-resident labour services as pre-determined, together withthe endowments of domestic factors. This brings us to the GNP-functionsetting, and to the corresponding elasticity estimates (matrix «) shown intable 6.7. The own-price elasticity of the demand for imports is now found


Table 6.7. Price and quantity elasticities (GNP-function setting, matrix «,1986 estimates)

emn; lnhm(p,wM,wN,xL,xK) / lnzn

hme{y,xM,xN,wL,xK},zne{p,wM,wN,xL,xK}

zn5p zn5wM zn 5wN zn5xL zn5xK

eYn 0.568 20.420 20.148 0.439 0.561eMn 1.338 21.056 20.282 0.261 0.739eNn 3.157 21.883 21.274 21.620 2.620eLn 1.014 20.189 0.176 20.422 0.422eKn 2.719 21.123 20.596 0.885 20.885

Note:These estimates are defined for a given output, import and non-resident labour service prices, and for given domestic-factorendowments. They are based on the parameter values shown intable 6.1, column (3).

to be 21.056, while the own-price elasticity of the demand for foreignlabour services is 21.274. These elasticities cannot be compared directly tothe ones reported in table 6.5, since eMM and eNN – contrary to «MM and «NN– are defined for variable gross output and fixed endowments of capital anddomestic labour. Of considerable interest are the so-calledStolper–Samuuelson elasticities (eij, i[{Y,M,N}, j[{L,K}) andRybczynski elasticities (eji, j[{L,K}, i[{Y,M,N}). These show, forinstance, that an increase in the price of imports hurts capital relativelymuch more than labour, whereas an increase in the price of non-residentlabour actually favours resident labour. An increase in the endowment ofcapital will tend to heavily increase the demand for non-resident labour ser-vices. Looking at the negative signs of the cross-price elasticities eMN andeNM, one can conclude that trade and labour mobility are complements inthe sense that, for given domestic factor endowments, an increase in theprice of either foreign input will tend to reduce the demand for both.

We next turn to the question of the effects of exogenous changes in thenumber of non-resident workers. The relevant elasticities (matrix F, 1986figures) are shown in table 6.8. If the number of non-resident workers is heldconstant, the own-price elasticity of imports is now found to be quite a bitcloser to zero than otherwise (compare the estimate of fMM with that ofeMM shown in table 6.7), which is an illustration of the Le Châtelier princi-ple. Of special interest is the estimate of fLN which shows that, as one mighthave expected it, an increase in the number of non-resident workers

138 Ulrich Kohli

Table 6.8. Price and quantity elasticities (immigration-quotas setting, matrixF, 1986 estimates)

fmn; lnhm(p,wM,xN,xL,xK) / lnzn

hme{y,xM,wN,wL,wK},zne{p,wM,xN,xL,xK}

zn5p zn5wM zn 5xN zn5xL zn5xK

fYn 0.201 20.201 0.116 0.627 0.256fMn 0.640 20.640 0.221 0.620 0.159fNn 2.478 21.478 20.785 21.272 2.057fLn 1.449 20.449 20.138 20.645 0.783fKn 1.241 20.241 0.468 1.643 22.111

Note:These estimates are defined for a given output and import prices,and for given quantities of labour (non-resident and resident) andcapital. They are based on the parameter values shown in table6.1, column (3).

depresses the wages of resident workers, but the impact is very small.25

Looking at the Stolper–Samuelson effects, it is interesting to note that non-resident workers are relatively more hurt than the domestic factors of pro-duction by a worsening in the terms of trade. Once again, we find thatimports and foreign labour services seem to work hand in hand.

We now turn to the estimates of matrix Q shown in table 6.9. Rememberthat these price and quantity elasticities are valid for given prices of output,imports and resident labour services, and given quantities of non-residentlabour and domestic capital. They are thus relevant in the short run, ifdomestic wages are sticky. With resident labour services variable, thedemand for imports is found to be quite price-elastic.26 The demand for res-ident labour is found to be quite price-elastic as well. The estimate of uLNindicates the effect of an increase in the number of non-resident workers onthe demand for resident workers. Its negative value shows that, as expected,foreign workers displace native workers. The effect is not as small as it seemssince the number of resident workers is about five times larger than thenumber of non-resident workers. An elasticity of 20.214 thus indicates thatthe replacement effect is almost one for one. However, what is crucial hereis the assumption of a fixed domestic wage rate. The relatively large priceelasticity of the demand for resident labour reveals that not much of areduction in domestic wages would be needed to restore full employment.Indeed, this is confirmed by the very low estimate of fLN shown in table 6.8.Thus, if immigration causes unemployment, then the prime culprit will be


Table 6.9. Price and quantity elasticities (variable-resident-employment setting, matrix Q, 1986 estimates)

umn; lnhm(p,wM,wN,yN,xK) / lnzn

hme{y,xM,xL,wN,wK},zne{p,wM,wL,xN,xK}

zn5p zn5wM zn5wL zn5xN zn5xK

uYn 1.610 20.637 20.972 20.018 1.018uMn 2.032 21.071 20.961 0.089 0.911uLn 2.246 20.696 21.550 20.214 1.214uNn 20.378 20.593 1.971 20.513 0.513uKn 4.931 21.385 22.547 0.117 20.117

Note:These estimates are defined for a given output, import andresident labour service prices, and for given quantities of non-resident labour and capital. They are based on the parametervalues shown in table 6.1, column (3).

domestic wage rigidity. One again, we find that trade and labour mobilityare complements, in the sense this time that an increase in the number ofnon-resident workers will tend to increase the demand for imports.

Comparing our results with those reported in Kohli (1993), we find someinteresting similarities, in spite of the differences in specification, functionalform and modelling of technological change. Thus, both studies found thatnon-resident workers are substitutes for resident workers, and complementsfor capital in the Allen–Uzawa sense. Both studies also found that importsare Hicksian complements for capital and both types of labour, and thatnon-resident workers are Hicksian substitutes for resident workers. Thereare some differences, however. Thus, the earlier study suggested that importsmight be Allen–Uzawa substitutes for foreign labour, and that non-residentlabour might be a Hicksian substitute for capital. However, it is ratherremarkable that in spite of these differences all ejk elasticities – i.e. those con-sistent with the GNP-function setting – have the same signs. In particular,both studies concur that imports and non-resident labour services arecomplements in the sense that an increase in one input will reduce thedemand for the other, and that an increase in the cost of non-resident labourservices will favour resident workers, but penalise capital owners.

6 Total factor productivity

Our econometric results can also be used to model TFP. A state-of-the artmeasure of the change in TFP is given by the following index.27

Rt,t21; , (22)

where the subscripts indicate the time period. This index shows the costreduction that results from the passage of time, using either period-t orperiod-t-1 input prices as weights. It can be shown that if the cost functionis Translog, then Rt,t21 can be calculated from knowledge of the data alone,as follows:

Rt,t215 , (23)

where Wt,t21 is a Tornqvist input price index:28

Wt,t215exp (sjt 1sjt21) (ln wjt – ln wjt21) , (24)

and Pt,t21 is the output price-change factor:

Pt,t21 5 . (25)pt

pt21

43oj

12

Wt,t21

Pt,t21

Îc(wt21,t 2 1) c(wt,t 2 1)c(wt21,t) c(wt,t)

140 Ulrich Kohli

Since the construction of the data ensures that the value of gross outputequals the total cost of inputs:

ptyt 5 wjtxjt, (26)

it follows immediately that Rt,t21 can also be calculated as:

Rt,t215 , (27)

where Yt,t21 is the output-change factor:

Yt,t215 , (28)

and X*t,t21 is an implicit Tornqvist index of input quantities:

X*t,t21 ; . (29)

Thus, as suggested in the Introduction, TFP growth can be deduced fromfigure 6.2 by looking at the difference between the two lines. The cumulatedindex of TFP is also shown in figure 6.3. One sees that TFP has increased

Pt,t21Yt,t21

Wt,t21

yt

yt21

Yt,t21

Xt*,t21

oj


Figure 6.3 Total factor productivity

substantially during the post-war period, although the index fell verydramatically in the mid-1970s. An entire decade had elapsed before it fullyrecovered.

We can also follow an econometric approach to measuring TFP growth.That is, we can obtain Rt,t21 from the estimates of the structural parame-ters of the cost function. We begin by introducing (20) into the logarithmicversion of (22). This yields:

ln Rt,t215aT1 djT (ln wjt2 ln wjt21)1

djTT(ln wjt 2 ln wjt21)(2t21)1

djTTT(ln wjt2 ln wjt21)(3t2 23t11)1

djTT(2t21)1 dyTTT(3t22311)1

djTTTT(4t3 26t2 14t21). (30)

Computation of (30) requires estimates of aT, dT, and – in the higher-order flexibility cases – dTTT and dTTTT. These estimates can be obtained byregressing the so-far unexplained portion of the price of output on t, t2, t3

and t4. For this purpose, we have estimated the following equation:

ln ut 5a01aTt1 dTTt21 dTTTt31 dTTTTt4, (31)

where

ln ut ; ln pt 2 aj lnwjt2 gjk ln wjt ln wkt 2

djT ln wjtt 2 djTT ln wjtt2 2 djTTT ln wjtt

3. (32)

Our results are summarised in figure 6.3, which shows the econometric esti-mate of the index, based on (30), using the second-order, third-order andfourth-order approximations, respectively. One sees that the second- andthe third-order approximations do a rather poor job at tracking the TFPindex. Both curves are concave, and they indicate a severe slowdown in therate of technological progress. The second-order approximation actuallysuggests technological regress towards the end of the sample period. Thefourth-order approximation, on the other hand, performs much better, and

oj

oj

oj

ok

oj

12o

j

124

16

12

124

16

12

12 o

12 o

12 o

142 Ulrich Kohli

it indicates a resurgence of technological progress towards the end of thesample period, after the downturn of the mid-1970s.

The two-stage approach which we have followed here to estimate the rateof technological progress is somewhat similar to the one used by Fox andKohli (1998). There is one major difference, however. Thus, in the presenttreatment, the first stage already takes account of the type of smoothingwhich will be applied in the second stage. That is, if a fourth-orderapproximation is selected, this impacts on the estimation of the input-shareequations since, as shown by (21), these will then be functions of t, as wellas of t2 and t3.

7 Conclusions

The main purpose of this chapter was to combine the production-theoryapproach to modelling immigration with the production-theoryapproach to modelling foreign trade. Indeed, it is rather odd that untilnow international labour mobility and international trade were essen-tially modelled separately in empirical work. Yet immigration is clearlyan international phenomenon; the production-theory approach toimmigration must thus be compatible with the production sector of anopen economy.

An attempt was made here to increase the flexibility of flexible functionalforms with respect to time. A second-order approximation with respect totime does not allow for any inflection points in the path of TFP. Given theease with which our approach can be implemented, it may be viewed as anattractive alternative to the spline method proposed by Diewert and Wales(1992).

Among our main empirical results, we have found that the displacementeffects of immigration into Switzerland could be very severe in case ofdownward wage rigidity. If wages are flexible, increased immigration willlower the income of domestic workers, but only weakly so. Capital ownersare the beneficiaries of immigration; it is therefore not surprising that thecall for increased access to foreign labour markets generally emanates frombusiness leaders.

Imports and non-resident labour are found to be complements for eachother, in many different meanings of the word. They are Allen–Uzawacomplements, and they also are Hicksian q-complements. Furthermore,they are also complements in the sense that more non-resident workers tendto lead to more imports; this is true whether or not resident labour servicesare held constant. Moreover, in the GNP-function setting, a reduction inthe price of non-resident labour services, which tends to favour the demandfor such services, also pulls along the demand for imports.


NOTES

Originally presented at the Venice conference (January 25–27 1997). I wish to thankMarzio Galeotti, Klaus F. Zimmermann and an anonymous referee for many usefulcomments and suggestions. Financial assistance from the Swiss National ScienceFoundation under grant no. 12-45777.95 is gratefully acknowledged.11 See Greenwood and McDowell (1986) for a survey of the economic effects of

immigration.12 The production-theory approach has been applied to Swiss data by Butare and

Favarger (1992); see Bürgenmeier et al. (1992) for a condensed version in English.13 Wong (1988) does discuss the impact of international labour mobility on the

volume of trade. However, he treats foreign labour as a perfect substitute fordomestic labour. Hence, there is no difference in his model between migrationand a change in the domestic endowment of labour.

14 See Mundell (1957), for instance.15 See Markusen (1983) among others.16 This chapter differs from our earlier work (Kohli, 1993) in several important

respects, particularly the choice of specification and functional form, the deriva-tion of the comparative statics of the model and the modelling of technologicalchange.

17 See Burgess (1974a, 1974b); Woodland (1982); Kohli (1991).18 See Kohli (1982, 1992) for estimates for Switzerland.19 For an alternative treatment, albeit without international labour mobility, see

Kohli (1991), for instance.10 The question would then arise as to whether the two (or more) outputs are pro-

duced jointly or not; see Kohli (1991) for estimates of labour- demand functionsunder both assumptions. It is also noteworthy that there seems to be very little dis-cussion of multiple-output models in the labour-demand literature; for instance,this topic is not even raised in Hamermesh’s (1993) otherwise very detailed book.

11 See Diewert (1974).12 Uzawa (1961).13 See Hicks (1970). We use the terminology of Sato and Koizumi (1973) and

Syrquin and Hollender (1982).14 This point is also made by Hamermesh (1993).15 See Kohli (1991), for instance.16 This differential treatment of resident and non-resident labour is made possible

by our assumption that the two types of workers have different attributes, andhence are separate inputs.

17 See Kohli (1978, 1991), Woodland (1982); the GNP-function label comes fromthe fact that this setting corresponds to the description of the technology by avariable-profit function which is the solution of maximising GNP, for givenfactor endowments and given prices of traded goods and services. In truth,because net labour payments abroad are taken into account, but net capital pay-ments are not, the function is a hybrid of GNP and GDP.

18 See Kohli (1991) for additional details.19 The list is not exhaustive: there are many more combinations of endogenous

versus exogenous treatments which we could have considered.20 Monotonicty and concavity are an integral part of our theoretical framework,

and they must be satisfied for the econometric estimates to be economicallymeaningful. Regularity-condition failures have plagued previous work in theimmigration literature; thus, a little known fact about Baldwin Grossman’s

144 Ulrich Kohli

(1982) widely cited study is that her estimates fail to satisfy concavity. In theTranslog case, curvature conditions can be imposed locally if needed. However,there is no guarantee that this will be sufficient for the conditions to be met forall observations. Moreover, imposition of curvature conditions often leads to areduction in the rank of the Hessian matrix. It is therefore fortunate that theseconditions were met here at the outset, since a rank reduction would rule outsome of the manipulations described in section 3.

21 One can see that the estimates of some of the gjrs change substantially as t isallowed to increase; this is not surprising since our test results indicate thatrestricting the value of t does violence to the data.

22 Butare and Favarger (1992), on the basis of a three-input Translog productionfunction that excludes imports and for which t52, also find that non-residentworkers are Allen–Uzawa complements for capital.

23 This is in contrast with the findings of Butare and Favarger (1992).24 The finding that the Hicksian elasticity of complementarity between imports

and capital changes sign, even though the Allen–Uzawa elasticity of substitu-tion does not, once again underscores the fact that the passage from one set ofelasticities to the other is not trivial.

25 This result confirms the findings obtained for other countries; see Greenwoodet al. (1996, 1997) for recent findings for the USA.

26 The comparison of uMM with fMM provides another illustration of the LeChâtelier principle.

27 See Caves et al. (1982); Diewert and Morrison (1986), for instance.28 Given that the cost function is Translog, Wt,t21 can also be interpreted as the

geometric mean of two indexes, both measuring the cost implications of changesin input prices, but one doing it with reference to time-t-1’s technology, and theother one refering to time-t’s technology:

Wt,t21 ; .

REFERENCES

Baldwin Grossman, J. (1982). ‘The Substitutability of Natives and Immigrants inProduction’, Review of Economics and Statistics, 64, 596–603

Berndt, E. R., B. H. Hall, R. E. Hall and J. A. Hausman (1974). ‘Estimation andInferences in Nonlinear Structural Models’, Annals of Economic and SocialMeasurement, 3/4, 653–65

Bürgenmeier, B., T. Butare and P. Favarger (1992). ‘Effects of Foreign Labour onthe Production Pattern: The Swiss Case’, Swiss Journal of Economics andStatistics, 128, 103–24

Burgess, D. F. (1974a). ‘A Cost Minimisation Approach to Import DemandEquations’, Review of Economics and Statistics, 56, 225–34

(1974b). ‘Production Theory and the Derived Demand for Imports’, Journal ofInternational Economics, 4, 103–17

Butare, T. and P. Favarger (1992). ‘Analyse empirique du cas de la Suisse’, in B.Bürgenmeier (ed.), Main d’oeuvre étrangère: Une analyse de l’économie suisse(Paris: Economica)

Î c(wt,t 2 1) c(wt,t)c(wt21,t 2 1) c(wt21,t)


Caves, D. W., L. R. Christensen and W. E. Diewert (1982). ‘The Economic Theoryof Index Numbers and the Measurement of Input, Output, and Productivity’,Econometrica, 50, 1393–1414

Christensen, L. R., D.W. Jorgenson, and L. J. Lau (1973). ‘TranscendentalLogarithmic Production Frontiers’, Review of Economics and Statistics, 55,28–45

Diewert, W. E. (1974). ‘Applications of Duality Theory’, in M. D. Intriligator andD. A. Kendrick (eds.), Frontiers of Quantitative Economics, vol. 2 (Amsterdam:North-Holland)

Diewert, W. E. and C. J. Morrison (1986). ‘Adjusting Output and ProductivityIndexes for Changes in the Terms of Trade’, Economic Journal, 96, 659–79

Diewert, W. E. and T. J. Wales (1992). ‘Quadratic Spline Models for Producer’sSupply and Demand Functions’, International Economic Review, 33, 705–22

Favarger, P. (1992). ‘Evolution de l’économie suisse depuis 1948’, in B. Bürgenmeier(ed.), Main d’oeuvre étrangère: Une analyse de l’économie suisse (Paris:Economica)

Fox, K. J. (1994). ‘Model Selection Criteria and the Non-Parametric Estimation ofTechnical Progress’, School of Economics Discussion Paper, 94/17, TheUniversity of New South Wales

Fox, K. J. and U. Kohli (1998). ‘GDP Growth, Terms-of-trade Effects, and TFP’,Journal of International Trade and Economic Development, 7, 87–110

Greenwood, M. J. and J. M. McDowell (1986). ‘The Factor Market Consequencesof US Immigration’, Journal of Economic Literature, 24, 1738–72

Greenwood, M. J., G. L. Hunt and U. Kohli (1996). ‘The Short-run and Long-runFactor-market Consequences of Recent Immigration to the United States’,Journal of Regional Science, 36, 43–66

(1997). ‘The Factor-market Consequences of Unskilled Immigration to theUnited States’, Labour Economics, 4, 1–28

Hamermesh, D. S. (1993). Labor Demand (Princeton: Princeton University Press)Hicks, J. R. (1970). ‘Elasticity of Substitution Again: Substitutes and

Complements’, Oxford Economic Papers, 22, 289–96Kohli, U. (1978). ‘A Gross National Product Function and the Derived Demand

for Imports and Supply of Exports’, Canadian Journal of Economics, 11,167–82

(1982). ‘Production Theory, Technological Change, and the Demand forImports: Switzerland, 1948–1976’, European Economic Review, 18, 369–86

(1991). Technology, Duality, and Foreign Trade: The GNP Function Approach toModeling Imports and Exports (Ann Arbor: University of Michigan Press)

(1992) ‘Production, Foreign trade, and Global Curvature Conditions:Switzerland, 1948–1988’, Swiss Journal of Economics and Statistics, 128, 3–20

(1993). ‘International Labor Mobility and the Demand for Imports’, SwissJournal of Economics and Statistics, 129, 547–61

Markusen, J. (1983). ‘Factor Movements and Commodity Trade as Complements’,Journal of International Economics, 13, 341–56

Mundell, R. (1957). ‘International Trade and Factor Mobility’, American EconomicReview, 47, 321–35

146 Ulrich Kohli

Sato, R. and T. Koizumi (1973). ‘On the Elasticities of Substitution andComplementarity’, Oxford Economic Papers, 25, 44–56

Shephard, R. W. (1953). Cost and Production Functions (Princeton: PrincetonUniversity Press)

Syrquin, M. and G. Hollender (1982). ‘Elasticities of Substitution andComplementarity: The General Case’, Oxford Economic Papers, 34, 515–19

Uzawa, H. (1961). ‘On a Two-sector Model of Economic Growth’, Review ofEconomic Studies, 29, 40–7

Wong, K. (1988). ‘International Factor Mobility and the Volume of Trade: AnEmpirical Study’, in R. C. Feenstra (ed.), Empirical Methods for InternationalTrade (Cambridge, MA: MIT Press)

Woodland, A. D. (1982). International Trade and Resource Allocation (Amsterdam:North-Holland)

Zellner, A. (1962). ‘An Efficient Method of Estimating Seemingly UnrelatedRegressions and Tests for Aggregation Bias’, Journal of the AmericanStatistical Association, 57, 348–68

Discussion

MARZIO GALEOTTI

Labour, whatever its geographical origin or its other characteristics, is aninput to the production process. Therefore, the issue of the impact of migra-tion upon the utilisation of native workers and/or the rate of return of theirservices has to be studied within a production-theoretic approach. UlrichKohli has done this using more disaggregated data then those used in thischapter (Greenwood et al., 1996). In chapter 6, however, the author startsfrom the observation that several researchers have indeed studied the migra-tion issue using a production-theoretic approach, but they have surprisinglynot taken into account the fact that, rather than ‘importing’ workers, thedomestic economy can import the goods that those workers contribute toproduce in the home country. This remark brings up the issue of trade as apossible alternative (or actually as a complementary activity) to labour-factor mobility. Hence the interest in characterising the substitutionpossibilities in production between ‘imported’ labour and imported goods.

Two brief remarks come to mind at this point. First, the production-based analysis of migration invariably holds the number of hours worked


per worker constant and concentrates on changes along the ‘extensivemargin’ of labour. It is important to bear in mind that possibilities of sub-stitution of hours with workers are ruled out in this case, and this restric-tive assumption certainly affects the results concerning the substitutionpatterns of native workers, immigrant workers, imported goods and otherinputs. Secondly, another interesting issue would be the extension of theanalysis of labour mobility vis-à-vis trade flows to also account for theinternational mobility of capital in the form of technology transfers, viaimports of capital goods and foreign direct investment (FDI). Chapter 9 byCollins et al. in this volume tackles this issue from a historical perspective(Collins et al., 1998).

Ulrich Kohli is one of the first authors to have pioneered the applicationof modern microeconomic theory of production and of the duality betweenproduction, cost and profit functions to international trade. He has doneso in several papers, and the present chapter is a nice example of the flex-ibility of this analytical tool for both theoretical and empirical purposes.Experience suggests that some people, faced with these applications for thefirst time, remain somewhat perplexed considering the aggregation assump-tions on which they are built: a story developed with the individual deci-sion-maker in mind is applied to data covering the whole economy. As amatter of fact, one should not be surprised: in the traditional approach tointernational trade analysis, in the study of labour demand and in manyother instances such aggregation assumptions are often implicitly made.The present approach makes it clear what are the benefits against the costsof the aggregation assumption: the ability to generate a wealth of empiricalresults that only enough structure imposed on the data can yield.

Ulrich Kohli in this chapter shows how the many relevant issues andquestions which are at the heart of the migration debate can be empiricallyaddressed and answered within a production-theoretic framework byappropriately treating the various inputs to the production process and thecorresponding prices. The basis for the several empirical exercises that arecarried out is the estimation of (the input share equations associated with)a cost function treating all inputs as variable. On this basis the author is ableto characterise p-substitution (Allen–Uzawa elasticities of substitution) aswell as, through appropriate manipulations, q-substitution (Hicks’ elastic-ities of complementarity). With further manipulations, he then proceeds tocompute similar elasticities by assuming that: (1) domestic labour andcapital are fixed inputs (in the short run), and (2) foreign labour is alsofixed. Now, this information could have been obtained more directly byestimating the GNP function or variable profit function. In fact Lau (1976)shows that the (normalised) restricted profit function is the most general(dual) representation of the firm’s technology that ecompasses unrestricted

148 Discussion by Marzio Galeotti

profit, revenue, restricted and unrestricted cost functions. Perhaps it wasmore natural to base the estimation on this more general representationand on that basis to infer elasticities for specific cases, instead of followinga ‘specific- to-general’ procedure as done by Kohli. Obviously also in thismore general case one would have used the same set of parameter estimatesto compute all the elasticities of interest. As a matter of fact, working witha GNP function grants an extra degree of freedom, in that it allows deriva-tion and estimation of supply equations for domestic output and exports,along with input demand equations. This would bring increased efficiencyof the estimates and dispense with the assumption made in the chapter ofperfect substitutability between domestic and exported outputs.

The chapter uses a flexible functional form for the cost function – theTranslog – which successfully goes through the difficult passage of satis-fying unconditionally the curvature conditions required for a wellbehaved cost function. Those who are familiar with applied productionanalysis making use of flexible functional forms will find this fact quiteremarkable. One aspect, however, with respect to which the chapter issomewhat deficient is the description of the data. While it is true that, asthe author himself notes, one of the advantages of the production-theo-retic approach to trade is that it uses easily available data from theNational Income and Product Accounts (this is not entirely true sincedata for the user cost of capital and its stock need to be constructed), nodistinction is made among employment by country of residency. I wouldhave liked to know how the price of the two types of labour was computedand where the necessary data came from. I have always believed that themain difficulty here was to obtain separate labour cost data for residentand non-resident workers.

The data available for investigating the trade-/factor-mobility issue areclearly a major constraint to the depth with which the problem can bestudied. In fact, not only must one clearly define which notion of elasticityof substitution/complementarity is used, but must also be careful in defin-ing the criterion used to distinguish various types of labour inputs. And herethe data available can play an important role. Ethnicity is the criterionadopted in Kohli’s chapter. Of course, if the data permitted other potentiallymore relevant criteria could be used, chiefly education and skill/experience.Riccardo Faini has brought to my attention, for instance, a paper by Gangand Rivera-Batiz (1994) which studies the q-complementarity issue (only)by concentrating on those two attributes of the labour input. In general, allthese procedures amount to hedonic adjustments to measured inputs andoutputs and one can in principle operate as many adjustments as the datapermit. Incidentally, let me note that ethnicity (perhaps correlated with dis-tance) may have independent explanatory power as an attribute, besides that


of skill and education, so that ideally one would like to take into account allthese three dimension in the analysis.

A further remark is that the production approach to trade typicallyassumes that the supply of inputs is perfectly elastic, so that the ‘wholeeconomy’ takes input prices as given. It is, however, difficult to accept thisassumption for factors such as labour, even if it may be reasonable forimports in a small open-economy context. Actually, my non-expert impres-sion is that supply considerations play an important role in many instancesof the migration debate. I am thinking of Italy, where certain low-skilledjob opportunities are not picked up by native workers even in the presenceof high unemployment rates.

Finally, from an econometric standpoint I am always a bit surprised bythe fact that applied production studies generally do not worry about theendogeneity of the right-hand-side variables of the estimated equations.Taken literally, this fact implies that estimated elasticities are not reliable.

In summary, the reason I was asked to comment on this chapter isbecause I have some background in applied production analysis. I thereforecannot but appreciate it because Kohli has once again shown how produc-tion theory is a tool of analysis that lends itself to interesting extensions andapplications and that can be taken to relate to very relevant policy issuessuch as the trade/migration debate.

REFERENCES

Collins, W.J., K.H. O’Rourke and J.G. Williamson (1998). ‘Were Trade and FactorMobility Substitutes in History?’, chapter 9 in this volume

Gang, I.N. and F.L. Rivera-Batiz (1994). ‘Labour Market Effects of Immigrationin the United States and Europe’, Journal of Population Economics, 7

Greenwood, M.J., G. Hunt and U. Kohli (1996). ‘The Short-run and Long-runFactor-market Consequences of Recent Immigration to the United States’,Journal of Regional Sciences, 36, 43–66

Lau, L.J. (1976). ‘A Characterization of the Normalized Restricted Profit Function’,Journal of Economic Theory, 12

150 Discussion by Marzio Galeotti

7 Migration, dual labour markets andsocial welfare in a small openeconomy

TOBIAS MÜLLER

1 Introduction

In most European countries, liberal immigration policies came to a halt inthe 1970s. More recently, hostility towards immigrants seems to be growingagain and immigration is increasingly seen as having negative consequencesfor the host country, especially if immigrants are unskilled. This attitude isin contrast to the ‘guest-worker’ policy of the 1950s and 1960s when largenumbers of unskilled workers migrated to North European countries. Atthat time, their arrival was seen as largely beneficial for the host countries.Entrepreneurs in labour-intensive industries saw immigration as an alter-native to protection, in the context of increasing import competition fromdeveloping economies (Bhagwati, 1982). The shift towards more restrictiveimmigration policies in the 1970s was accompanied by a resurgence of pro-tection. The steady reduction in tariff protection since 1947 was offset in the1970s and 1980s by the growth of non-tariff barriers (Bhagwati, 1988).

How can these changes in migration and trade policies be explained?They are often imputed to macroeconomic difficulties and to risingunemployment in Europe. However, there is no obvious relation betweenprotection and unemployment and most empirical analyses fail to find alink between immigration and unemployment. In this chapter, I explore analternative explanation of these developments. Starting from Bhagwati’s(1982) discussion of the choice between protection and immigration policy,I extend his analysis by focusing on a more complete representation of the‘guest-worker’ system,1 and by taking income distribution explicitly intoaccount.

Indeed, the standard welfare analysis of migration fails to explain theobserved attitudes towards immigration. According to the utilitarianwelfare criterion, immigration yields a surplus to the host country. Thus,utilitarianism cannot justify restrictive immigration quotas. From adifferent perspective, it is often acknowledged that unskilled immigration

151

increases income inequality among natives. According to this argument,the guest-worker policy, favouring the immigration of unskilled workers,would be rejected. As a description of the guest-worker system, the stan-dard analysis is deficient in two respects. First, it neglects the existence ofdiscrimination against immigrants, which is common especially in coun-tries favouring temporary immigration. Second, the impact of immigrationon income distribution is generally analysed separately from efficiencyconsiderations (see also chapter 6 in this volume). Therefore, the questionwhether the aggregate gain from unskilled immigration prevails over itsunfavourable influence on income distribution cannot be answered. In thischapter, these two issues are addressed. Efficiency and distributionalaspects are considered in an integrated framework by using Atkinson’s(1970) social welfare function. Discrimination is modelled in the frame-work of a segmented labour market, and the legal restrictions faced byimmigrants in the labour market are taken explicitly into account.

In most European countries, discrimination against immigrants mani-fests itself as ‘equal pay for equal work, but unequal work’ (Hammar,1985). Trade unions usually demand that all workers, regardless ofnationality, be offered the same wage. However, in countries having imple-mented a guest-worker system, immigrants are directed towards certainoccupations, characterised by low wages and bad work conditions. Theimmigrants’ access to stable jobs with good work conditions is oftenlimited by legal discrimination, since some countries restrict work permitsto certain occupations or sectors, or allow firms to hire immigrants onlyif no native worker can be found. The behaviour of employers often alsoappears to be discriminatory against immigrants, who usually have lesscountry-specific human capital and higher expected quit rates thannatives. In quantitative studies, however, the evidence on discriminationagainst immigrants is not as clear-cut. Dustmann (1993) reports forGermany that earnings of foreign workers are not only initially lower (by13–19 per cent) than those of natives; he does not find any convergenceover time towards the natives’ earnings level. However, not all studies onGerman data confirm these results (see Zimmermann, 1993, for a survey).For Switzerland, preliminary results by de Coulon and Flückiger (1995)indicate that immigrants with the same characteristics than nativesreceive 10 per cent lower wages.

In the model presented below, I assume that discrimination againstimmigrants takes place in a dual labour market (Piore, 1979). Wages in theprimary sector are determined by efficiency-wage considerations.Therefore, primary-sector jobs are rationed and immigrants have onlylimited access to them. However, immigrants can always find a job in thesecondary sector where the wage rate is set at the market-clearing level.

152 Tobias Müller

Discrimination implies that immigrants receive the same wage as natives fora given job, but they have a smaller probability of finding a job in theprimary sector. In this case, immigration increases the natives’ chances offinding a primary-sector job. From the viewpoint of social welfare,however, this gain is tempered with an increase in income inequality, sinceimmigration tends to reduce the wage rate relative to the return to capital.On the other hand, protection of sectors that are intensive in unskilledlabour expands the output of the secondary sector and therefore diminishesthe natives’ chances of finding a primary-sector job.

This model is closely related to those developed by Ethier (1985),Schmidt et al. (1994) and Winter-Ebmer and Zweimüller (1996), who takeinto account the specific position of immigrants in the labour market. Theirapproaches are complementary to the efficiency-wage model used here.Ethier (1985) shows how the hiring of immigrants can insulate nativeworkers from employment fluctuations. There is discrimination againstimmigrants in the sense that only natives have long-term, implicit labourcontracts, whereas immigrants are hired freely at the current wage rate.Schmidt et al. (1994) analyse the impact of immigration in the presence oftrade unions. There is discrimination against immigrants in the sense thatimmigrant welfare is not given the same weight in the union’s objectivefunction as native welfare. In this model, immigration might lead to higherunskilled employment if skilled and unskilled labour are q-complements.Winter-Ebmer and Zweimüller (1996) use an insider–outsider model ofwage bargaining. They assume the existence of a two-tier wage system,where immigrants (outsiders) receive lower wages than native workers(insiders). Because of discrimination, immigration has an ambiguous effecton native wages.

The remainder of the chapter is organised as follows. In section 2, theimpact of immigration and protection on social welfare is discussed,assuming that there is no discrimination against immigrants. An open-economy efficiency-wage model of a dual labour market is presented insection 3 and the impact of immigration and protection analysed from atheoretical standpoint in section 4. Finally, in section 5, the dual labourmarket model is integrated into a simulation model for Switzerland and theimpact of immigration and protection on social welfare is simulated fordifferent labour market closures. The evolution of Swiss immigration policyis interpreted in that context. Section 6 draws some conclusions.

2 Social welfare effects of migration and protection

Standard welfare analysis shows that natives, as a group, gain fromimmigration (Berry and Soligo, 1969). The gain is higher, the more the

Migration, dual labour markets and social welfare 153

immigrants’ relative factor endowment differs from the natives’. However,by contrast with the debate on trade liberalisation, nobody seriously pro-poses to allow free international migration. One of the main reasons for thisattitude is that immigration, especially unskilled immigration, tends toincrease income inequality, since it reduces the wage rate and raises thereturn to capital. The widespread intuition assuming that the redistributiveimpact of unskilled immigration is stronger than the efficiency gain is con-firmed by the following result (Müller, 1998a): if (physical or human)capital is unequally distributed among natives, and if immigrants hold lesscapital than natives, infinitesimal immigration decreases social welfare.2

This implies that the generalised Lorenz curve representing the income dis-tribution before immigration dominates the generalised Lorenz curve afterimmigration.

At first sight, this result seems to contradict the standard analysisshowing the welfare gains from immigration. In fact, the standard analysisimplicitly uses a particular form of the social welfare function, often calledthe ‘utilitarian’ or ‘Benthamite’ social welfare function, which takes onlymean income into account. It is the only form of the social welfare functionfor which infinitesimal immigration has no impact on social welfare. Thisis due to the fact that the efficiency gain, which is only of second order, dis-appears when immigration is infinitesimal. In order to avoid confusion, Iwill hereafter denote the utilitarian welfare criterion by ‘U-welfare’,whereas the term ‘social welfare’ will refer to the more general socialwelfare function of the Atkinson (1970) type.

The social welfare functions underlying the result quoted above are oftencriticised because of the assumption that utility is cardinal and that inter-personal comparisons of utility levels are possible. These restrictiveassumptions are, however, necessary if one wants to be able to rank situa-tions for which the criterion of Pareto-optimality is not conclusive. On theother hand, it is important to stress that the utilitarian social welfare criter-ion, which is routinely used in most discussions of welfare effects, is evenmore restrictive than the class of social welfare functions defined above.The utilitarian criterion would be pertinent only if any change in immigra-tion policy were accompanied by non-distortionary redistribution mea-sures which would compensate the losers. This is not a very realisticassumption.

In view of these considerations, what kind of immigration policies wouldgovernments be expected to establish? If the objective of the governmentwere to maximise social welfare, it would not allow any unskilled immi-grants to enter the country, but might welcome skilled immigrants, for tworeasons. First, unskilled immigration tends to increase income inequality,whereas the impact of skilled immigration on inequality is ambiguous.

154 Tobias Müller

Second, the efficiency gain is probably greater in the case of skilledimmigration, because of the complementarity between capital and skilledworkers (Borjas, 1995).

In short, the guest-worker system of immigration does not look at allattractive from the perspective of social welfare. Of course, the view thatgovernments act according to a social welfare criterion is simplistic.Nevertheless, a social welfare function is a convenient and rigorous way ofconsidering efficiency and equality aspects in a common framework.Moreover, more elaborate theories of political economy seem to lead tosimilar conclusions. For example, Benhabib’s (1992) direct-democracymodel of immigration policy shows that unskilled immigration would beaccepted in a national referendum only if the median capital–labour ratiois greater than the capital–labour ratio of the person who is indifferent toimmigration. If immigration is infinitesimal, the latter is equal to theaverage capital–labour ratio. Therefore, if the distribution of capital isasymmetric and skewed to the right (a plausible assumption), unskilledimmigration would be rejected in a referendum.

Now turn to the social welfare impact of protection. On the one hand, ifimportables are labour-intensive, a tariff will in general increase the wagerate relative to the return to capital, and therefore reduce income inequal-ity. On the other hand, the efficiency loss induced by a tariff is small for lowtariff levels, since a tariff has no first-order effect on U-welfare if the initialsituation is free trade (Neary, 1988). Consequently, in terms of socialwelfare the efficiency loss of protection tends to be outweighed by reducedinequality if the initial tariff level is not too high.

To sum up, a government acting according to the social welfare criterionwould hardly implement a policy of free trade combined with large-scaleimmigration of unskilled workers. Then why did some European countriesallow immigration in the form of the guest-worker system? As manyobservers have suggested, this question cannot be answered without takinginto account the fact that there is discrimination against immigrants in thedual labour markets of the host countries. This issue is taken up below.

3 A model of dual labour markets and discrimination

This section describes the model of a dual labour market which is usedbelow to reconsider the impact of immigration and of trade policy on thewelfare of natives. In this analysis, the role of discrimination against immi-grants is highlighted.

The dual labour market is modelled in a standard efficiency-wage frame-work.3 Work conditions in the primary and the secondary sectors are notidentical. The primary sector offers jobs with good working conditions,


stable employment relationships and good chances for internal promotion.By assumption, workers in this sector cannot be perfectly monitored. Thus,firms in the primary sector prefer to pay wages above market-clearing levelsin order to induce workers to supply effort. As a consequence, jobs arerationed in the primary sector and workers are queueing up for them.However, they can always find jobs in the secondary sector. These jobs aremuch less attractive and consist in repetitive tasks that can be monitoredwithout cost. In this sector, the wage rate is set competitively. Thus, there isno unemployment.4

For the sake of simplicity, the instantaneous utility function of a repre-sentative worker is assumed to have the following separable structure:

u(c1,c2,e)5m(c1,c2) 2 e, (1)

where c1 and c2 are the consumption levels of the two traded goods, m is ahomothetic quasi-concave function, and e denotes effort. The variable e cantake only two values: 0 if the worker does not make an effort (i.e. if she‘shirks’), and e.0 if she does not shirk. A worker’s indirect utility function,derived from (1), is given by:

v(p1,p2,w,yo,e)5 2 e, (2)

where p is a price-index dual to m, p1 and p2 are goods prices, w is the wagerate and yo is income from other sources (capital income, transfers).

Natives are assumed to maximise expected utility over their infinite lifehorizon:

U5E v(p1,p2,w,yo,e)exp(2rt)dt , (3)

where r is the natives’ discount rate. A worker who shirks faces a probabil-ity d of being discovered and fired. Moreover, an exogenous proportion qof workers quit primary jobs in each period.

All native workers in the secondary sector have the same probability offinding a primary-sector job. This is not necessarily the case for immigrants:different cases will be examined below. The problem of a native worker inthe primary sector, who has to decide whether to shirk or not, can beanalysed by relating the utility levels that she can attain in the two cases. LetV1

n denote the expected present value of utility of a non-shirking nativeworker holding a primary-sector job. If she shirks, the expected presentvalue of utility is V1

s. Let V2 denote the present value of utility of a sec-ondary-sector job. To relate these three situations, I follow the asset-equa-tion approach introduced by Shapiro and Stiglitz (1984). If a worker has ajob in the primary sector, she will receive wage w1. She has an exogenous

21E`

0

w 1 yo

p(p1,p2)

156 Tobias Müller

probability q of quitting that job. In that case, she will lose, in terms ofutility, the difference between V1

n and V2. If a worker does not shirk, thereturn to a primary-sector job is therefore equal to:

rV1n5 2e2q(V1

n2V2). (4)

If the worker decides to shirk, her instantaneous utility is greater becauseshe does not supply any effort. However, she faces a higher probability oflosing her job in the primary sector since she might be detected as a shirkerand fired. For a shirking worker, the return to a primary sector job is there-fore given by:

rV1s 5 2 (q1d) (V1

s 2 V2). (5)

A worker in the primary sector does not shirk if V1n$V1

s. Using (4) and(5), this condition can be rewritten as follows:

d (V1n2V2)$e. (6)

The term on the left represents the cost of shirking, equal to the expectedutility loss of a shirker whose probability of being detected and fired is equalto d. A worker does not shirk if this cost is greater than the immediatebenefit of shirking, which consists in avoiding any effort.

In the steady state, if a worker decides not to shirk, she will never shirkin a primary-sector job. For such a worker, the return to a job in the sec-ondary sector is equal to:

rV2 5 2 e1a (V1n 2 V2), (7)

where a is the probability of moving from a secondary-sector to a primary-sector job. (Conversely, for a worker who always shirks, V1

n must be replacedby V1

s in eq. (7).Using (4) and (7), the no-shirking condition (6) can also be expressed as:

$ (r1a1q). (8)

At equilibrium, there is no shirking and (8) holds with equality, since thereis no reason for a primary-sector firm to pay a higher wage.

To derive an expression for the probability of moving from a secondary-sector to a primary-sector job, a, assume first that no immigrants arepresent in the country. The flow out of the primary sector is qL1, where L1is native employment in the primary sector. The flow into the primarysector is a (L2L1), where L is total native employment. At equilibrium,

ed

w12w2

p(p1,p2)

w2 1 yo

p(p1,p2)

w1 1 yo

p(p1,p2)

w1 1 yo

p(p1,p2)


these two must be equal. Thus, if all native workers have the same proba-bility of finding a primary-sector job, a is given by:

a5q L1/(L2L1). (9)

When immigrants arrive in a dual labour market, the economic outcomedepends on social and institutional arrangements. In order to highlightthese differences, I will discuss two extreme cases.

First, I assume that the law does not allow employers to recruit immigrantsif suitable native candidates can be found. Consequently, all immigrants areforced to accept jobs in the secondary sector. This can be seen as a simplifiedview of a guest-worker system. In this case, all primary-sector jobs are heldby natives. Therefore, the natives’ probability of finding a primary-sector job(a) depends only on native employment and is given by eq. (9).

Alternatively, I assume that immigrants are identical to natives in allrespects. In that case, immigrants cannot be distinguished from natives andthere is no discrimination against them. In a certain sense, this representsthe ideal case of a ‘melting pot’, where immigrants are expected to stay inthe host country and where they have the same rights as natives. Thus, immi-grants have the same probability as natives of finding a job in the primarysector, i.e. L1/L5L1*/L*, where L* is total employment of immigrants andL1* is primary-sector employment of immigrants. Therefore a is given by:

a5 5 .

Obviously, this expression is identical to the guest-worker case. However,the structure of employment is different in the two cases. Indeed, in themelting-pot case, L1*5L* (L1/L) is immigrant employment in the primarysector, whereas no immigrants are employed in the primary-sector in theguest-worker system. Secondary employment of immigrants is L2*5L* (12L1/L) in the melting-pot case, and L2*5L* in the guest-worker system.

In order to compare the effects of immigration and protection in a smallopen economy, the relation between the dual labour market and goodstrade must be specified. I will procede in two steps. First, I discuss the effectsof immigration and protection on U-welfare from a theoretical viewpoint,using a simple two-sector model. However, in the choice between tariffs andimmigration, there remains a trade-off between efficiency and equity. Thisissue is taken up in section 5, with the help of a multi-sector, three-factorsimulation model.

4 Labour market effects of immigration and protection

In a standard small-country model, immigration yields a U-welfare gain,whereas protection induces a loss. Does the segmentation of the labour

qL1

L 2 L1

q(L1 1 L1*)L 1 L* 2 L1 2 L*1

158 Tobias Müller

market change this result? In this section, I discuss this question by con-sidering small (infinitesimal) immigration flows or tariff changes and byusing the utilitarian welfare criterion. Distributional issues and socialwelfare considerations are postponed until section 5.

I assume that capital is specific to the primary and the secondary sectors.4

Both sectors exhibit constant returns to scale and produce traded goods.The country is a price-taker on goods markets. The primary sector offersonly ‘good’ jobs, paying efficiency wages w1, whereas the secondary sectoroffers only ‘bad’ jobs, paying the competitive wage w2. Firms are assumedto maximise profits, so that wage rates are equal to the marginal product oflabour in each sector. The relation between wage rates in the two sectors isgiven by eq. (8), holding with equality, and (9). Assuming to begin with thatimmigrant employment is exogenous, equilibrium in the dual labourmarket can thus be described by the two following equations, which arerepresented diagrammatically in figure 7.1:

w15p(p1,p2) 1p2 f 2L(K2,L1L*2L1 2L1*) (10)

w15p1 f1L(K1,L1 1L1*), (11)

where f i is the production function of sector i and f iL denotes the partial

derivative of f i with respect to L. Equation (10) reflects both the no-shirk-ing constraint and the marginal product of labour in the secondary sector.Since the derivative of the right-hand side with respect to L1 is positive, thisequation is represented by the upward-sloping curve, labelled NSC1MPL2, in figure 7.1. The marginal product of labour in the primary sectoris depicted by the downward-sloping curve, labelled MPL1.

In order to simplify the welfare analysis, I assume that the capital stocks

ed 1r 1

qLL 2 L1

2


Figure 7.1 Impact of immigration on native employment: the ‘melting-pot’ case

of both sectors are entirely owned by natives and that the new immigrantsdo not bring any capital with them. I assume furthermore that there are ini-tially no tariffs and no immigrants present in the host country. It is wellknown since Grubel and Scott (1966) that infinitesimal immigration has noimpact on the natives’ U-welfare in a model without distortions. Thiswould also be true in the present model if the proportion of natives workingin the primary sector remained constant after the arrival of immigrants.6

Consequently, the equivalent variation of native U-welfare caused byimmigration is equal to:

EV5 5 dL1.

Native U-welfare improves if the probability for natives of finding aprimary-sector job increases as a consequence of immigration. Thus thequalitative U-welfare consequences of infinitesimal immigration can beanalysed simply by determining the sign of the change in native primary-sector employment. The case of tariffs is similar. As mentioned above, atariff has no first-order effect on U-welfare if the initial situation is freetrade. Therefore, the U-welfare effect of a small tariff on imports dependscrucially on the variation of native employment in the primary sector.

Now turn to the impact of immigration on U-welfare of natives. Assumefirst that immigrants cannot be distinguished from natives, so that they arenot discriminated against in the host country (the melting-pot case). Figure7.1 illustrates the ambiguous impact of immigration on native employmentin the primary sector. Immigration shifts the MPL1-curve to the left andthe NSC1MPL2-curve to the right. The primary-sector wage unambigu-ously falls, but the impact of immigration on native employment in theprimary sector, and thus on native U-welfare, is ambiguous. It can be shownthat the natives’ probability of finding a primary-sector job rises only if theelasticity of labour demand is (much) greater in absolute value in theprimary sector than in the secondary sector (see Müller, 1998b).

This uncertain outcome of immigration would obviously be changed ifimmigrants could be prevented from penetrating the primary sector. Incountries that have implemented the guest-worker system, there is occupa-tional segregation because immigrants do not have the same chances asnatives of finding a primary-sector job. Discrimination might be explicit, asin legal dispositions limiting the rights of immigrants, or it might be due tothe fact that employers perceive immigrants as a distinct group withcharacteristics that differ from the natives’ (e.g. different quit rates). Thelatter case will be discussed below in section 5. Here I simply assume thatthe host country does not grant the same rights to immigrants as to natives.In Switzerland, for example, employers who want to obtain a work permit

w1 2 w2

p(p1,p2)w1dL1 1 w2dL2

p(p1,p2)

160 Tobias Müller

for an immigrant must prove that they are unable to recruit a native worker(or a foreigner with a permanent residence permit). In the present model ofa dual labour market, the consequence of such a regulation is that immi-grants are not able to obtain primary-sector jobs, since natives alwaysprefer those to secondary-sector jobs.

The U-welfare impact of guest-worker immigration can be seen in figure7.2. Since immigrants are confined to the secondary sector, immigrationdoes not affect the MPL1-schedule, but shifts the NSC1MPL2-curve to theright. Therefore, native employment in the primary sector rises unambigu-ously and primary-sector wages decline.

As in a standard specific-factors model, returns to capital rise in bothsectors and the two wage rates fall. However, the wage differential increases.Indeed, since the probability of being hired in the primary sector has risenfor natives, the wage differential must be higher in order to prevent themfrom shirking. Interestingly, it is possible in this model that immigrationincreases the natives’ expected labour income even if both wage rates fall(see Müller, 1997b). However, natives who work in the secondary sector andwho do not receive any capital income are necessarily worse off ex post.Therefore, ex post there is necessarily a conflict between the aggregate gainfrom immigration and a more unequal income distribution.

What is the welfare impact of a tariff if the labour market is segmented?The implementation of a tariff increases output in the secondary sector andmoves labour from the primary to the secondary sector. Indeed, the NSC1MPL2-schedule shifts to the left; the MPL1-curve is not affected by the tariff.Obviously, the tariff has the opposite effect of guest-worker immigration. Inparticular, primary-sector employment diminishes and the tariff induces a U-welfare loss for natives. The impact of a tariff on income distribution is


Figure 7.2 Impact of immigration on native employment: the ‘guest-worker’ case

ambiguous in this model. Protection increases the return to capital in theprotected sector and lowers it in the other sector. The primary-sector wagerises, but proportionally less than the price of the importable good.

5 Immigration, protection and social welfare: a simulation analysis

From the discussion in the preceding sections it follows that in the case ofprotection and immigration, the government faces a trade-off betweenefficiency and equality. On the one hand, guest-worker immigration entailsa first-order efficiency gain, since it increases the proportion of nativesholding primary-sector jobs, but it also leads to a more unequal income dis-tribution. On the other hand, protection produces a first-order efficiencyloss, but might well reduce income inequality. Under what conditions doesthe efficiency gain of immigration outweigh its adverse distributionalimpact? When is the efficiency loss due to protection compensated by itsfavourable impact on income distribution? Which option would be pre-ferred by a decision-maker motivated by social welfare? In this section, Iconsider these questions from an empirical perspective with the help of asimulation model which is calibrated on Swiss data. Of course, the answersto the questions above depend strongly on value judgements, in particularon the degree of ‘inequality-aversion’ of decision-makers.

5.1 The simulation model

In the simulation model, capital is sector-specific7 and there are two skillcategories of labour, mobile between sectors. Since the simulations focus onthe immigration of unskilled workers, I assume, for simplicity, that theskilled labour market is competitive (for a similar hypothesis in a modelwith trade unions for unskilled workers, see Schmidt et al., 1994). By con-trast, the unskilled labour market is segmented and there are good and badjobs, as described above in section 3. Markets for goods are assumed to becompetitive and the firms’ production functions are nested CES functions.Because of the small-country assumption, the domestic prices of tradedgoods are fixed. Some sectors, however, produce non-traded goods. Sincetheir prices are endogenous, the demand side matters in the simulationmodel (the equations of the model are given in the appendix, p. 175).

Obviously, the simulation model differs only very little from the theoret-ical model used in section 4. The main differences are the presence of twolabour categories and of non-traded goods. Furthermore, in the simulationmodel all industries offer good and bad jobs; however, the proportion ofgood jobs varies from one industry to the other.

Preferences of all households are described by Cobb–Douglas utility

162 Tobias Müller

functions. Moreover, I assume that all households have identical utilityfunctions. Since the conditions of exact linear aggregation are satisfied inthis case, domestic demand depends only on aggregate income and is notaffected by a change in the distribution of income. As a consequence,income distribution issues can be considered independently from thedetermination of equilibrium.

In order to model immigration as well as income distribution amongnatives, three types of households are distinguished in the model: (1) immi-grant households, who do not own any capital; (2) native householdsendowed with unskilled labour; (3) native households endowed with skilledlabour. All native households receive some income from capital. Forsimplicity, I assume that the share of capital income received by a nativehousehold is equal to its share of native labour income. Tariff revenues areredistributed to all households (including immigrants) according to theirshare in total income. Thus, a change in tariffs does not affect incomeinequality.

These assumptions imply that native income inequality depends only onrelative income from skilled and unskilled labour. In particular, a rise in thereturn to capital does not increase income inequality. This is, of course, aconservative assumption since it tends to under-state the rise in incomeinequality due to guest-worker immigration. However, this treatment ofincome distribution captures the main source of income inequality, asthe following result in Flückiger and Silber (1995) shows. Decomposing theoverall Gini index by income source, these authors conclude thatthe contribution of labour income to an overall Gini index of 0.40 is esti-mated to be equal to 0.24.

The main indicator used to evaluate the impact of immigration and tariffpolicies is the social welfare of natives. I assume thereby that the govern-ment does not redistribute income in order to compensate native house-holds for any losses due to those policies. Social welfare depends on theindividual utility levels of natives. Because of the Cobb–Douglas specifica-tion, indirect utility is equal, by appropriate normalisation, to real income– i.e. nominal income deflated by a geometric price index. The measure ofsocial welfare used in the simulations is Atkinson’s (1970) equally-distrib-uted equivalent income, ye. It defines the level of per capita real incomewhich if equally distributed would provide the same level of social welfareas the actual distribution. The degree of inequality-aversion is captured bya parameter, «, which allows us to cover a large range of value judgements.If «50, the decision-maker is completely insensitive to distribution issues,since in this case ye is equal to average income (the utilitarian case). Withincreasing «, the decision-maker attaches more and more weight to lowerincomes.


5.2 Policy experiments and labour market specifications

To analyse the effects of immigration and protection on social welfare inSwitzerland, three main policy experiments are carried out using fourdifferent specifications of the labour market (see table 7.1). In a first set ofsimulations, immigration quotas and tariff rates are fixed. Then, the conse-quences of varying immigration and protection levels and of different skilllevels of immigrants are analysed.

In the first set of simulations, the impact of migration policy is simulatedby assuming that 200,000 unskilled immigrants arrive to Switzerland (M).This can be interpreted as an estimation of the number of ‘guest-workers’in Switzerland since it is approximately the number of low-skill foreignerswho had temporary working permits in 1985, the base year of the model(foreigners holding a permit of residence are treated as natives).

The effects of protection are captured through two policy experiments.First, I consider the effects of sector-specific protection by simulating atariff on imports of Textile and Clothing (P21). The ad valorem equivalentof tariff barriers in these sectors is not far from 10 per cent in Switzerland.It is, however, difficult to compare simulations M and P21 because of theirdifferent sectoral impact. In the two-sector framework of Bhagwati (1982),immigration and protection can be considered as being equivalent in thesense that a same output level of the importable good can be achieved byboth policy instruments. Simulations P21 and M are not equivalentaccording to that definition since the former increases the output level ofthe two protected sectors in very large proportions, whereas the latterincreases moderately the output of all low-skill sectors.

In experiment P22, the tariff structure is designed in such a way that theoutput levels of all low-skill traded goods are equal to their output levels in

164 Tobias Müller

Table 7.1. Policy experiments and labour market specifications

Policy experiment

M Immigration of unskilled workers (quota of 200,000 immigrants)P-1 Protection (10 per cent tariff on Textiles and Clothing)P-2 Protection (tariffs on all low-skill traded goods)

Labour market specification

STD Integrated labour market, no efficiency wages, no discriminationM-P Dual labour market, no discrimination against immigrantsG-W Dual labour market, legal discrimination against immigrantsS-D Dual labour market, immgrants and natives have different quit rates

experiment M. Thus, comparability between immigration and protection inthe sense of Bhagwati (1982) is ensured (however, firms in skill-intensivesectors prefer option M to P-2, since their production rises in the first caseand falls in the second). The tariff rates in experiment P-2 range from 1.0per cent (Beverages) to 2.6 per cent (Agriculture).

In order to highlight the role of a segmented labour market and of dis-crimination against immigrants, the experiments are carried out using fouralternative labour market specifications. The first case represents the stan-dard specific-factors model (STD) where the labour market is integratedand wages are set competitively. There is no discrimination against immi-grants. This specification is useful as a benchmark for the other simulations.In the second case, the labour market is assumed to be segmented, withefficiency wages in the primary sector, but discrimination does not takeplace because immigrants and natives are indistinguishable. In particular,their quit rates are perceived to be identical by employers. This is themelting-pot (M–P) specification discussed above. In the third case, I assumethat immigrants face legal discrimination, preventing them from enteringthe primary sector. Since employers must prove that they cannot find anative worker to fill a job vacancy and since primary jobs are preferred tosecondary jobs, immigrants are stuck in the secondary sector. This is theguest-worker (G–W) specification.

In the fourth case, discrimination originates in the fact that immigrants areperceived as having, on average, higher quit rates than natives. Empirically,this assumption can be justified by the observation that many immigrantsintend to return to their home country in the near future. This is true evenfor immigrants who arrived in the host country a long time ago. ForGermany, Dustmann (1993) reports that 55 per cent of all immigrants intendto return to their country of origin within the next 10 years. Among them, 85per cent have been living in Germany for more than 10 years. If quit rates aredifferent, discrimination occurs even if immigrants have the same legal rightsas natives in the labour market. It is a form of statistical discrimination (S–D)because the membership in a group (natives, immigrants) determines an indi-vidual’s probability of finding a primary-sector job. This can be seen from theno-shirking conditions for natives and immigrants (see (40) and (41) inappendix table 7A.1). Because of competition between firms, natives andimmigrants are paid the same wage in the primary sector. Since immigrantsexpect to stay a shorter time in primary-sector jobs than natives, primary-sector employers will hire proportionately fewer immigrants in order toinduce them not to shirk. Note that this treatment of discrimination followsvery closely Bulow and Summers (1986), who apply it to the case of dis-crimination against women. In the simulations, I assume that the immi-grants’ quit rates are three times higher than the natives’.


The simulation model is calibrated on Swiss data for the year 1985 (seethe appendix), taking into account the presence of immigrants and the factthat the Textile and Clothing sectors are protected. Then a hypothetical freetrade, no-immigration situation is simulated; this is the starting point forall policy experiments.

5.3 Simulation results

The policy experiments described above shed new light on Bhagwati’s(1982) discussion of tariff and immigration policies in the context ofincreased import competition. First, I follow his analysis by assuming thatthe two policy options have the same objective: to achieve a given outputlevel in low-skill sectors. Second, the consequences of different protectionand immigration levels are discussed. Finally, an informal sensitivity analy-sis of the model is carried out.

5.3.1 The choice between protection and immigration The results of thepolicy experiments described above suggest that, from the perspective ofnative social welfare, immigration might be preferred to protection only ifthere is discrimination against immigrants or if the government is com-pletely insensitive to the distribution of income among natives (see table7.2).

Consider first the case of an integrated, competitive labour market(STD). In this context, unskilled immigration (M) is not an attractivepolicy option from the viewpoint of social welfare. As expected from the

166 Tobias Müller

Table 7.2. Impact of immigration and protection on native social welfare(percentage changes in equally distributed equivalent income, ye )

Policy Labour marketInequality-aversion parameter («)

experiment specification 0 0.2 0.5 1.0 2.0 3.0

M STD 0.05 20.03 20.15 20.35 20.72 21.03M-P 0.04 20.04 20.17 20.37 20.75 21.07G-W 0.23 0.16 0.07 20.10 20.42 20.70S-D 0.11 0.03 20.08 20.27 20.63 20.94

P-1 STD,M-P,G-W,S-D 20.05 20.04 20.04 20.02 0.01 0.04

P-2 STD 20.02 20.01 0.02 0.06 0.14 0.21P-2 M-P,G-W,S-D 20.03 20.02 0.01 0.05 0.13 0.19

discussion in section 2, the U-welfare gain from immigration is very smalland the adverse impact on income distribution prevails even for low levelsof inequality-aversion. On the other hand, protection of low-skill sectorsincreases the unskilled wage relative to the skilled wage (see table 7.3) andthus has a favourable impact on income distribution. The efficiency lossfrom protection is relatively small, especially in the broad-based approachwith moderate tariffs applying to many goods (P–2). Consequently, protec-tion is preferred to immigration unless the degree of inequality-aversion isvery small.

What if the labour market is segmented? If there is no discriminationagainst immigrants (M–P), both immigration and protection induce a shiftof native unskilled employment from the primary to the secondary sector.However, this effect is so small that it hardly influences the social welfareoutcome.


Table 7.3. Impact of immigration and protection on output and the labourmarket (percentage changes from the base values)

Unskilled immigration Protection

Experiment M P-1 P-2Labour market specification STD M-P G-W S-D STD G-W G-W

Output: aggregate and selected low-skill sectorsAggregate output 2.0 1.9 1.9 1.9 0.1 0.1 0.1Food 7.0 7.0 5.7 6.5 21.4 21.4 5.7Textile 3.9 3.9 3.9 3.9 79.1 78.9 3.9Clothing 3.7 3.7 5.5 4.4 30.1 29.7 5.5Construction 5.7 5.7 9.1 7.0 21.0 21.1 9.1Hotels, restaurants 2.9 3.0 6.5 4.3 21.4 21.4 23.2

Employment, wages and capital incomeTotal unskilled employment 6.0 6.0 6.0 6.0 0 0 0Native unskilled employment 0 0 0 0 0 0 0– Primary sector — 20.5 4.1 1.5 — 20.1 20.3– Secondary sector — 2.0 217.7 26.5 — 0.5 1.2Unskilled wage ratea 22.4 22.4 21.6 22.1 0.1 0.0 0.2– Primary sector — 22.3 21.6 22.0 — 0.0 0.2– Secondary rate — 22.4 24.7 23.3 — 0.1 0.4Skilled wage rate 1.0 1.0 1.0 1.0 20.2 20.2 20.5Capital income 1.1 1.1 1.0 1.1 20.2 20.2 20.0

Note:a In the case of dual labour markets, average unskilled wage rate for natives.

If there is discrimination against immigrants, natives will find unskilledimmigration more profitable. In the guest-worker system (G–W), sec-ondary-sector employment of natives declines by more than a sixth as aconsequence of immigration. By contrast, the impact of immigration onaverage native income is surprisingly small (it rises by only 0.2 per cent).This is due to the fact that only a small share (9 per cent) of native workershold secondary-sector jobs. Moreover, the sign of the social welfare effectdepends on the inequality-aversion parameter. Therefore, if the govern-ment is insensitive to distribution issues, it would favour guest-workerimmigration. On the other hand, if the government is characterised by aninequality-aversion parameter greater than 0.7, it would be hostile tounskilled immigration and prefer protection.

If discrimination has its source in different quit rates (S–D), the outcomelies in between the M–P and the G–W cases. Indeed, the U-welfare gain fornatives is smaller than in the guest-worker case since only 7 per cent ofnative workers in the secondary sector succeed in moving to the primarysector.

5.3.2 The welfare impact of different immigration or protection levels In thesimulations above, it is assumed that the government considers only twopolicy options with pre-determined levels of immigration or protection.Obviously, these levels might also be chosen by the government. In partic-ular, to counter the continued pressure of import competition from devel-oping countries, the government might envisage large-scale immigration orhigh tariff rates.

The impact of varying immigration levels on the social welfare of nativesis depicted in figure 7.3. As expected from the simulations above, large-scaleimmigration is beneficial for natives only if the degree of inequality-aver-sion is low. But figure 7.3 reveals two more striking features. First, succes-sive immigration waves become increasingly beneficial, or lessunfavourable, for natives (the curves are concave). Second, the differencebetween the G–W and the M–P cases increases with rising immigrationlevels. An intuitive understanding of the first observation can be obtainedby considering the impact of immigration when some immigrants arealready living in the host country. Each new arrival of unskilled immigrantsprompts, over and above the second-order U-welfare gain, a redistributionof income from earlier unskilled immigrants towards natives, because of thefall of unskilled wages relative to skilled wages and to the return to capital.Thus, the more unskilled immigrants reside in the country the more bene-ficial a new arrival of unskilled immigrants is for natives. Obviously, theincreasing gain for natives is obtained at the expense of immigrants.

However, in a M–P system, this redistribution effect is quite small. In

168 Tobias Müller

terms of native social welfare, it outweighs the increase in income inequalityamong natives only if the degree of inequality-aversion is small. If «51~or~2, the social welfare effect remains negative even for high levels ofunskilled immigration. Compared to the M–P case, immigration is doublybeneficial for natives in a G–W system. On the one hand, immigration enablesmore natives to find a job in the primary sector and, on the other, it increasesthe wage gap between the primary and the secondary sectors. The influenceof the first effect gradually diminishes with rising immigration levels, sincethere are fewer and fewer natives in the secondary sector who could gain fromshifting to the primary sector.8 The second effect, however, becomes increas-ingly favourable to natives, because it reinforces the redistribution effect men-tioned above. Indeed, the drastic fall in secondary-sector wages hurtsespecially immigrants, whereas primary-sector wages, which most nativesreceive, fall less than in the M-P case. In other words, the segregation of thework force is reinforced with rising immigration: almost all natives have aprimary-sector job, whereas all immigrants have secondary-sector jobs.

Assuming that the government aims at maximising social welfare,9 whatare the implications of these observations for migration policy? If no otherpolicy instruments are available, the government would choose either freeimmigration or no immigration, depending in particular on the degree ofinequality-aversion. Furthermore, if the government pursues only the inter-ests of natives, it would opt for a guest-worker system, which is more advan-tageous than the melting-pot system in two respects. First, its impact onsocial welfare is more favourable for natives for a given level of immigra-tion, as shown above. Second, the immigration pressure is likely to be


Figure 7.3 Impact of different immigration levels on native social welfare: the‘melting-pot’ Ïê(M–P) and ‘guest-worker’ Ïê(G –W) cases

smaller than in the melting-pot case, since the immigrants’ wages decreasemore rapidly with rising immigration levels, the immigrants being confinedto a small labour market segment. Moreover, the bad working conditionsin the secondary sector discourage many potential immigrants.

It might seem unrealistic that the model contains no mechanism whichwould limit the usefulness of increasing immigration levels for natives.However, if a low degree of inequality-aversion is assumed, the guest-worker specification of the model reflects Swiss immigration policy quitewell until the first half of the 1960s (Hoffmann-Nowotny, 1985): this policywas liberal towards the exterior and restrictive towards the interior. Borderswere open to migrants, but migrants met hard restrictions once inside thecountry. As a result of that policy, the number of foreigners in Switzerlandincreased from 285,000 in 1950 to 810,000 in 1965.

Swiss immigration policy underwent an important turnaround in 1970when restrictive immigration quotas were introduced. According toHoffmann-Nowotny (1985), this change in policy was ‘not dictated by eco-nomic interests at all, it was instead the result of grassroots pressure basedmainly on the issue of “over-foreignization”’. The fear of ‘over-foreigniza-tion’ (Überfremdung – an unpalatable German word which evokes theSwiss’ fear of losing their cultural identity because of the presence offoreigners) can be linked to the failure of the policy of ‘rotation’. Instead ofreturning to their home country, as had been expected, many immigrantsdecided to stay on in Switzerland. This observation suggests an alternativeinterpretation, which does not refer to the concept of cultural identity, ofthe shift towards a restrictive immigration policy. As discussed above, theG–W system was advantageous for natives because of the legal discrimina-tion against immigrants. By contrast, there is no legal discriminationagainst foreigners holding a permit of residence, since they have the samerights in the labour market as natives. Thus, when it became clear that manyimmigrants would in fact become permanent residents, unskilled immigra-tion appeared much less attractive from the perspective of native socialwelfare (see the M–P specification in figure 7.3). It is therefore likely thatfurther immigration would be halted in these conditions.

Moreover, if immigrants are expected to become permanent residents, itis difficult for the government to argue that it is completely indifferent totheir welfare. Yet, if immigrants are included in the government’s socialwelfare objective, the social welfare implications of immigration are verydifferent from that outlined above. On the one hand, the arrival of unskilledimmigrants is likely to decrease social welfare because it diminishes averageincome (of natives and immigrants) and increases income inequality. Thus,further unskilled immigration will be stopped a fortiori. On the other, fromthe perspective of social welfare it is not advisable to discriminate against

170 Tobias Müller

immigrants who are expected to become permanent residents (interestingly,since 1970 the Swiss government has improved the legal position of immi-grants and lifted some of the discriminatory restrictions). Indeed, the M–Ppolicy performs better than the G-W system with respect to the socialwelfare of natives and immigrants, for any degree of inequality-aversion. If1 million immigrants are present in Switzerland (the actual number offoreigners in 1985), social welfare is 0.5 per cent higher if «50 (2.4 per centif «51) in the M–P system. This is hardly surprising. Since in the G–W caseimmigrants are confined to the secondary sector, total secondary-sectoremployment expands more, relative to primary-sector employment, than inthe M–P case. Moreover, overall income inequality is greater in the G–Wsystem because immigrants receive lower wages.

Now turn to the consequences of rising tariff levels. It is well known thatthe efficiency cost of protection increases with initial tariff levels (Neary,1988). In terms of social welfare, the improvement in income distributioninduced by protection is therefore less likely to compensate for the dead-weight loss when tariff rates reach high levels. This can be seen in figure 7.4,which depicts the social welfare effects of a uniform proportional increasein tariffs on low-skill goods (the tariff structure is the same as in experimentP–2). Optimal tariff rates are quite low for moderate levels of inequality-aversion. Would protection of low-skill sectors have been an alternative tolarge-scale immigration in Switzerland? The arrival of almost 1 millionimmigrants in the 1950s and 1960s had an important impact on the outputlevels of low-skill sectors. If the government had tried to achieve a similar


Figure 7.4 Impact of different tariff levels on social welfare (the G –W case)

impact by tariff policy, the required tariff rates would have exceeded 10 percent on average. Figure 7.4 shows that, for moderate levels of «, such apolicy would have been less beneficial than guest-worker immigration forthe social welfare of natives.

Discrimination hardly affects the impact of protection on social welfare(see table 7.2). Even if immigrants are included in the social welfare func-tion, the impact of protection is not altered significantly, since all unskilledworkers, natives and immigrants, gain from the redistribution of incomeresulting from a rise in tariffs. Thus, if the failure of the policy of rotationis acknowledged, protection appears more advantageous than immigra-tion.

5.3.3 Sensitivity analysis Since the model has been calibrated using manysimplifying assumptions, it is useful to test the sensitivity of the results tosome of them. For comparability, all simulations are carried out using theG–W specification (see table 7.4). Of course, this is no full-fledged sensitiv-ity analysis since every change in parameters or in model structure is sim-ulated separately. Overall, the qualitative welfare results of the model arequite robust and all model variants give similar estimates of the variationof average native income (U-welfare). By contrast, results differ signifi-cantly as to the impact of immigration on native income inequality.

5.3.4 Income distribution (I-D) The assumption that capital income isdistributed to natives in proportion to their labour income tends to under-estimate the impact of immigration on income inequality. In order to testthe sensitivity of results to this assumption, I re-calculated the social

172 Tobias Müller

Table 7.4. Social welfare effects of immigration: sensitivity analysis(percentage changes in equally distributed equivalent income, ye )

Sensitivity LabourInequality-aversion parameter («)

analysis market 0 0.2 0.5 1.0 2.0 3.0

I-D G-W 0.23 0.09 20.15 20.57 21.25 21.58LMS-1 G-W 0.23 0.18 0.09 20.05 20.32 20.57LMS-2 G-W 0.30 0.24 0.15 20.02 20.36 20.73LMS-3 G-W 0.26 0.20 0.10 20.05 20.35 20.61S-25 G-W 0.16 0.17 0.19 0.22 0.26 0.28S-25 and I-D G-W 0.16 0.09 20.04 20.27 20.66 20.85S-50 G-W 0.12 0.20 0.33 0.53 0.91 1.23S-50 and I-D G-W 0.12 0.11 0.08 0.04 20.04 20.09

welfare measure using the extreme alternative assumption that all incomefrom capital is paid to native skilled households. In this case, which cer-tainly over-estimates the impact of immigration on inequality, the level ofinequality-aversion at which a decision-maker would be indifferent to G–Wimmigration is around 0.3, down from 0.7 in the original simulation.

5.3.5 Labour market segmentation (LMS) In the treatment of the primaryand secondary sectors, several crucial assumptions rely on very littleinformation. I will test the sensitivity of results to four of them. First, someauthors interpret the dual labour market hypothesis as implying that ‘good’and ‘bad’ jobs are located in different firms. To test this assumption, I re-calibrate the model assuming that three industries (i.e. Construction,Arrangement and Hotels and restaurants) offer only ‘bad’ jobs forunskilled workers, whereas the others offer only ‘good’ jobs. The results aresurprisingly similar to the original version of the model (LMS–1).

Second, the wage differential between jobs in the primary and secondarysectors is not estimated directly. In the model it is calibrated using theaverage wage differential between unskilled natives and immigrants (15 percent). Doubling this latter number increases the U-welfare gain fromimmigration by only a third (LMS–2).

Third, there is no reliable information on the share of unskilled nativesworking in the secondary sector. Assuming that this share is only half aslarge as in the original model (where it is 16 per cent) yields results thatdiffer very little from the original model (LMS–3). This is due to the factthat the wage differential between unskilled natives and immigrants is given(see above).

5.3.6 Skill-mix of immigrants (S) In the simulations above, all immigrantswere assumed to be unskilled. Yet, in recent years, the Swiss governmenthas tried to facilitate the immigration of highly skilled workers. Such apolicy has many advantages. Skilled immigrants pay higher taxes and areless likely to be unemployed than unskilled migrants. Moreover, theirhuman capital may have external effects in production. Here I focus onanother issue: their impact on income distribution and social welfare.

I simulate this immigration policy by assuming that either 25 per cent(S–25) or 50 per cent (S–50) of immigrants are skilled. Since skilled labourand capital are complements in production, the immigration of skilledworkers increases the return to capital even more than the arrival ofunskilled workers. Hence, the assumption on the distribution of capitalincome among natives becomes a more sensitive issue. Thus, I report alsothe social welfare effect under the alternative hypothesis that all capitalincome is paid to skilled workers (I–D). The results show that the presence


of skilled immigrants attenuates the adverse impact of immigration onincome inequality. On the other hand, the U-welfare gain from immigra-tion is reduced because there is no discrimination against skilled workers.

Therefore, a government entirely insensitive to income distributionwould admit only unskilled ‘guest-workers’. By contrast, a governmentconcerned with income inequality would rather try to attract skilled immi-grants. Nevertheless, it would not abolish legal discrimination against theunskilled immigrants if it expects them to return to their country of origin.Recent Swiss immigration policy bears a strong resemblance to the latteroption. Indeed, an important share of recently arrived immigrants arehighly skilled (see table 7.5). However, the proportion of immigrants withelementary education remains high and intermediate education levels areunder-represented. This suggests that recent policy is a typical Swiss com-promise: part of the immigration quota is allocated to highly skilledworkers, while the remainder is used to pursue the traditional guest-workerpolicy.

6 Conclusions

From the results above, it appears that immigration of unskilled workersseems advantageous to natives only if immigrants ultimately return to theirhome country. If, however, immigrants acquire the status of permanent res-idents and thereby the same rights as natives in the labour market, unskilledimmigration is likely to meet with much more resistance. In that case,natives might find protection the better response to increased importcompetition, from the perspective of social welfare. In this view, the failureof the rotation system might be one of the causes of the switch to restrictive

174 Tobias Müller

Table 7.5. Education level of the work force in Switzerland, by nationality(percentage of the work force of each category)

Foreigners

PermanentEducation level Swiss residents Others

Elementary education, compulsory school 14.8 40.7 29.0Apprenticeship, professional or high school 74.5 50.7 36.1University 10.1 27.6 26.0

Other training, no answer 20.6 21.0 28.9

Source: Swiss Labour Force Survey (1991).

immigration policies in the 1970s. In turn, this change in immigrationpolicy may have contributed to the resurgence of protectionism in Europe.

Swiss immigration policy today seems to waver between a narrow viewof national interest, which favours a guest-worker system, and a broaderview of social welfare, focusing on the economic and social integration ofimmigrants. The fact that an increasing share of foreigners hold permanentresidence permits shows that the guest-worker system has not worked asexpected. Nevertheless, many discriminatory aspects of this system havebeen maintained. As a consequence, further immigration of unskilledworkers is still perceived as being advantageous for natives, at least in theshort run. This would probably change if the government put more empha-sis on the better integration of foreigners. However, it is clear that if the eco-nomic consequences of immigration shape the natives’ attitudes towardsimmigrants, the choice between a discriminatory guest-worker system anda policy oriented towards the integration of immigrants is largely deter-mined by ethical values.

APPENDIX

A.1 The simulation model

The model is disaggregated into 28 production sectors, indexed over i or j.Some of the goods and services that are produced are not traded internation-ally (see table 7A.3). The production functions are nested CES functions.Their structure is depicted in figure 7A.1. Skilled labour and capital, whichis specific to each sector, are assumed to be separable from other inputs sinceempirical evidence seems to indicate that these factors are p-complementary


Figure 7A.1 Structure of nested production functions in the simulation model

(see Hamermesh, 1993). Unfortunately, such evidence is completely lackingfor the relation between (unskilled) primary and secondary jobs, which Iassume to be separable from other production factors.

Firms minimise costs subject to the production function. The deriveddemand equations resulting from this problem are equations (12)–(16).Equation (17) is marginal cost, table 7A.1 contains all equations of themodel and table 7A.2 lists the variables of the model.

Equations (18)–(27) define the distribution of income to households.Subscripts s and u designate skilled and unskilled labour, superscripts (orsubscripts) n and f designate native and foreign (immigrant) households orlabour supply. Subscripts 1 and 2 refer to primary and secondary-sectorjobs. Native households own all domestic capital and also receive somecapital income from abroad (18). The variable fs determines the share ofcapital income distributed to skilled households, which is equal to theirshare in native labour income. The variable cn determines the share of nettariff income distributed to (skilled and unskilled) native households. It isequal to the natives’ share in capital and labour income. The variable s1 des-ignates the share of (native) unskilled labour income that (native) primary-sector workers receive.

Equation (28) is Atkinson’s (1970) measure of equally distributed equiv-alent income. The domestic final demand equation (29) is derived from aCobb–Douglas utility function. Because of the assumption of identicalpreferences, domestic demand depends only on aggregate income.Equation (31) is the small-country assumption. The current accountbalance (B) is exogenous and the real exchange rate adjusts to ensure exter-nal equilibrium. Labour supply by immigrants is proportional to thenumber of immigrants who live in the host country ((25) and (30)). Theequilibrium condition (39) for secondary-sector jobs includes also exoge-nous labour supply by ‘border workers’ – i.e. commuters who work inSwitzerland but are not allowed to live there. Equations (40) and (41) arethe efficiency-wage equations determining the relation between the wagegap and employment in the primary and secondary sectors. They arederived from (8) and (9) in the text. Equation (42) defines the geometricconsumer price index, which is dual to the Cobb–Douglas utility function,as the numeraire. Thus, all income variables can be interpreted in realterms.

A.2 Data and calibration

The simulation model is calibrated on a social accounting matrix forSwitzerland in 1985 (see Antille et al., 1991). The breakdown of wageincome by skills is taken from Gaillard et al. (1991). In 1985, the foreign

176 Tobias Müller


Table 7A.1. Equations of the simulation model

ProductionKj.5Kj(rj,ws,w1,w2,P1, … PM,Qj) (12)

LSj.5LSj(rj,ws,w1,w2,P1, … PM,Qj) (13)

LU j1.5LUj

1(rj,ws,w1,w2,P1, … PM,Qj) (14)

LUj2. 5LUj

2(rj,ws,w1,w2,P1, … PM,Qj) (15)

Iij. 5Iij(rj,ws,w1,w2,P1, … PM,Qj) (16)

Pj.5MCj(rj,ws,w1,w2,P1, … PM) (17)

IncomeYK.5SiriKi1YKF E (18)

YL.5ws(Lsn1L f

s )1w1(L1n1L 1

f )1w2(L2n1L 2

f ) (19)

YT.5SitiMNiPWiE (20)

cn.5 (wsLsn1w1L1

n1w2L2n1YK)/(YL1YK) (21)

fs.5wsLsn/(wsLs

n1w1L1n 1w2L2

n) (22)

s1.5w1L1n/(w1L1

n1w2L2n) (23)

Nsys.5fscn(YL1YK1YT2B)

.5wsLsn1fsYK 1fscn(YT2B) (24)

(L1n/Lu

n)Nuy1.5s1(12fs)cn(YL1YK1YT2B)

.5w1L1n1s1(12fs)YK1s1(12fs)cn(YT2B) (25)

(L2n/Lu

n)Nuy2.5 (12s1)(12fs)cn(YL1YK1YT2B)

.5w2L2n1(12s1)(12fs)YK1 (12s1)(12fs)cn(YT2B) (26)

Nfyf.5 (12cn)(YL1YK1YT2B)

.5wsLfs 1w1L 1

f 1w2L 2f 1(12cn)(YT2B) (27)

Natives’ social welfare (equally distributed equivalent income)

ye5[ fs ys

12« 1 (L1n/Lu

n)fuy112« 1(L2

n /Lun)fuy2

12«] /PiPibi if «?1

(28)ye.5 h ys

fsy1(L1

n/Lun)fuy2

(L2n /Lu

n)fu/PiPibi if «51

where fs5Ns/(Ns 1Nu) and fu5Nu/(Ns1Nu)

Domestic final demand

PiCi.5bi(YnL1YK1YT2B) (29)

11 2 «

178 Tobias Müller

Table 7A.1. (cont.)

Net imports and equilibrium on goods markets

MNi.5Ci1SjIij 2Qi (30)

Pi.5 (11 ti)PWiE if i is a traded good (31)

MNi.50 if i is a non-traded good (32)

Unskilled employment and immigrant labour supply

Lun..5L1

n1L2n (33)

Luf .5L 1

f 1L 2f (34)

Luf .5a f

uNf (35)

L fs .5a f

sNf (36)

Equilibrium on labour markets

Lsn.1 .Ls

f .5SjLSj (37)

L1n.1L 1

f 5SjLUj1 (38)

L2n..1L 2

f 1 L2b5SjLUj

2 (39)

Efficiency wages

. (40)

. (41)

NumerairePiP i

bi.51 (42)

Alternative labour market specifications• Standard (STD): e/d50• Melting-pot (M-P): qn 5qf

• Guest-worker system (G-W): replace (41) by L 1f 50

• Statistical discrimination (S-D): qf.qn

w1 2 w2

PiPbii

5ed 1r 1

qfL fu

Lf22

w1 2 w2

PiPbii

5ed 1r 1

qnLnu

Ln22


Table 7A.2. Variables of the simulation model

Endogenous variablesPj domestic price of good jrj return to capital in industry jws skilled wage ratew1 unskilled wage rate (primary-sector jobs)w2 unskilled wage rate (secondary-sector jobs)Qj gross output of industry jLSj skilled labour used in production of industry jLUj

1 primary-sector employment (of unskilled workers) in industry jLUj

2 secondary-sector employment (of unskilled workers) in industry jIij intermediate good i used in production of industry jL1

n primary-sector employment of unskilled nativesL2

n secondary-sector employment of unskilled nativesLs

f skilled labour supply, by immigrantsLu

f unskilled labour supply, by immigrantsL1

f primary-sector employment of unskilled immigrantsL2

f secondary-sector employment of unskilled immigrantsYK total income from capitalYL total labour income accruing to residents (natives and immigrants)YT tariff incomecn natives’ share in total capital and labour incomefs skilled natives’ share in native labour incomes1 share of primary-sector income in native unskilled labour incomeys per capita income of skilled nativesy1 per capita income of unskilled natives working in the primary-sectory2 per capita income of unskilled natives working in the secondary-sectoryf per capita income of immigrantsye equally-distributed equivalent income of nativesCi total domestic final demandMNi net imports of good iE real exchange rate

Exogenous variablesKj capital stock in industry jNf number of immigrantsNs number of skilled nativesNu number of unskilled nativesLs

n skilled labour supply, by nativesLu

n unskilled labour supply, by nativesLb

2 unskilled labour supply, by border workersYKF income from capital abroad, held by nativesB current account balanceti tariff rate on good iPWi world price of good iqn quit rate of nativesqf quit rate of immigrants

population in Switzerland was 1.02 million or 15.6 per cent of total popula-tion. However, the majority of foreigners were in the possession of a permitof residence, giving them the same rights on the labour market as Swiss cit-izens. (In the administrative jargon, Swiss citizens and foreigners with res-idence permits are called the ‘indigenous work force’.) However, foreignerswithout a residence permit face various restrictions. Therefore, it seems rea-sonable to define as ‘guest-workers’ all unskilled foreigners who do not holda permit of residence. In 1985, the number of foreigners without a permitof residence was 283,000, or 4.3 per cent of total population (borderworkers are not included in this number). Of these, 200,000 are estimatedto have low skills.

It is difficult to find an operational definition of the secondary sector inthe literature on dual labour markets. Trying to quantify the number andthe sectoral distribution of secondary-sector jobs is even more hazardous.In their test of the dual labour market hypothesis for the US, Dickens andLang (1985) conclude that 12 per cent of working male heads of householdsare employed in the secondary sector. They also note that this proportionis likely to be higher for women. Unfortunately, no such estimates exist forSwitzerland. So I assume that 12 per cent of the total population (includ-ing foreigners) hold a secondary-sector job in the initial equilibrium. As tothe sectoral distribution of these jobs, there is no quantitative evidence tomy knowledge. I assume that the number of secondary jobs in each indus-try is proportional to the number of foreign workers without a permit ofresidence (see table 7A.3). This assumption ensures that there is no sectoralbias in the distribution of foreigners holding secondary-sector jobs.

In order to calibrate the no-shirking conditions, one needs to know thewage differential between the primary and secondary sectors. The latter canbe calculated from the wage differential between indigenous workers and‘guest-workers’, using the shares of natives and immigrants working in thesecondary sector (since the proportion of immigrants holding a secondary-sector job is higher than the proportion of natives, a positive wagedifferential between the primary and the secondary sectors is reflected inlower average wages for immigrants). De Coulon and Flückiger (1995) esti-mate the wage differential between Swiss and foreign workers to equalapproximately 10 per cent. This number probably under-estimates the wagedifferential of the present model for two reasons. First, recently arrivedimmigrants are under-represented in their database (the Swiss labour forcesurvey). Second, all foreigners are lumped together in their estimation, incontrast with my definition of immigrants. Foreigners holding a residencepermit (and having arrived a long time ago) can be expected to receivehigher wages than more recent immigrants. Therefore, I chose to fix thewage differential between indigenous workers and recent immigrants at a

180 Tobias Müller


Table 7A.3. Employment, Switzerland, 1985, by sector and by nationalityof workers

Share of Foreignersforeign without permit

Employmentworkers of residencea

(000) Percentage of employment)

Agriculture, forestry 222 27.9 26.2Electricity, gas, water 224 24.6 21.0Food 260 27.1 10.6Beverage 228 16.9 25.3Tobacco 224 32.5 28.7Textiles 232 39.7 13.0Apparel 237 46.7 21.3Lumber, furniture 268 20.8 28.1Paper 217 32.7 29.0Printing, Graphic arts 263 16.7 23.3Leather, Shoes 211 43.0 11.2Chemical industry 270 31.9 15.6Oil refineries 220 31.9 15.6Plastics 222 36.1 15.0Non-ferrous minerals 232 37.3 17.2Metal 299 34.3 11.9Machinery 157 26.7 26.3Electrical machinery, watches, jewelry 257 26.2 29.1Construction 210 51.0 32.7Arrangement, installationb 139 23.9 15.3Wholesale and retail trade 494 14.8 24.9Hotels, restaurants 214 40.1 27.9Transport 137 13.0 24.6Communication 265 24.2 21.5Banking 109 11.6 23.2Insurances 251 27.6 21.6Other servicesb 322 17.1 26.4Government, social securityb 495 13.2 24.6

Total 3418 21.8 10.1

Notes:a Foreigners holding seasonal permits, permits of abode and border workers.b Sectors producing non-traded goods or services.

slightly higher level, more consistent with the empirical results byDustmann (1993) for Germany (15 per cent).

The quit rate and the discount rate of natives were quantified as indicatedin table 7A.4. In fact, only their ratio matters, since e/d is calibrated frombase year data by using (40). The substitution elasticities in production (seetable 7A.4) were chosen on the basis of the survey by Hamermesh (1993).The calibration of all other parameters is standard.

NOTES

I would like to thank Marc Bacchetta, Bernard Decaluwé, Jaime de Melo and RudolfWinter-Ebmer for very helpful suggestions and comments. This work was supportedby the Swiss National Fund for Scientific Research (Grant no. 12-42011.94). Themethodological aspects were developed under the auspices of the Programmed’analyses et de recherches économiques appliquées au développement international(PARADI), which is funded by the Canadian International Development Agency.1 Unfortunately, the euphemistic term ‘guest-worker’ has been used too often to be

ignored. Would you ask your guests to clean your bathroom?2 The social welfare function is assumed to be of the Atkinson (1970) type: it is

individualistic, symmetric, additively separable and inequality-averse. It can berepresented as W5eu(y) f(y)dy, where y is income, f is the frequency densityfunction of income and u is an increasing, strictly concave function.

3 The basic structure of this model follows Shapiro and Stiglitz (1984). Bulow andSummers (1986) and Jones (1987) adapted it to a dual labour market. Morerecently, efficiency wages have been integrated into international trade models.Copeland (1989) analyses trade policy issues using a Ricardian trade model witha dual labour market and efficiency wages. Brecher (1992) develops an efficiency-wage model with unemployment and a non-traded good.

4 Unemployment can be added to this model by assuming that primary-sectorfirms hire only unemployed workers. In that case, some workers prefer to remainunemployed and to wait for a job in the primary sector, rather than work in thesecondary sector. See Bulow and Summers (1986).

182 Tobias Müller

Table 7A.4. Parameters of the simulation model

Parameter

Elasticity of substitution: primary jobs – secondary jobs sUj 2.00

Elasticity of substitution: skilled labour – capital sKSj 0.50

Elasticity of substitution: unskilled labour – capital/skill aggregate s jVA 1.50

Elasticity of substitution: value added – intermediate inputs s jZ 0.10

Discount rate (efficiency wages) r 0.05

Native quit rate qn 0.10

5 The Heckscher–Ohlin (HO) case with intersectoral capital mobility is analysedin Müller (1997b).

6 To see this, differentiate total income of natives, Yn, holding L1 constant. Then,because of constant returns to scale:

dYn5L1dw11K1dr11 (L2L1)dw21K2dr25p1(L1 f 1

LL1K1 f 1KL)dL1*1p2 [(L2L1) f

2LL1K2 f 2

KL](dL*2dL1*)50,

where ri is the return to capital in sector i. Note that this property does not holdif immigrants are present at the initial equilibrium (L*.0) or if some capital isowned by non-residents.

7 This choice is motivated by empirical evidence. Indeed, Kohli (1993) shows thatthe specific-factors model performs substantially better than the HOS produc-tion structure in explaining the US experience.

8 However, some natives always remain in the secondary sector, which is expand-ing with immigration; otherwise natives working in the primary sector would losethe incentive not to shirk.

9 This is obviously a naive view of policy-making. Social welfare considerations donot constitute a theory of political economy. However, since income distributionissues are prominent in the immigration debate, the concept of social welfare iscertainly a better description of a government’s objective than the traditionalcriterion of U-welfare.

REFERENCES

Antille,G., M. Bacchetta, F. Carlevaro, C. Maranon, T. Müller and N. Schmitt(1991).Effets d’équilibre général de l’intégration de la Suisse à l’Europe (Coire:Editions Rüegger)

Atkinson, A. B. (1970). ‘On the Measurement of Inequality’, Journal of EconomicTheory, 2, 244–63

Benhabib, J. (1992). ‘A Note On the Political Economy of Immigration’, New YorkUniversity, mimeo

Berry, R. A. and R. Soligo (1969). ‘Some Welfare Aspects of InternationalMigration’, Journal of Political Economy, 77, 778–94

Bhagwati, J. N. (1982). ‘Shifting Comparative Advantage, Protectionist Demands,and Policy Response’, in J. N. Bhagwati (ed.), Import Competition andResponse (Chicago: University of Chicago Press)

(1988). Protectionism (Cambridge, MA: MIT Press)Borjas, G. J. (1995). ‘The Economic Benefits from Immigration’, Journal of

Economic Perspectives, 9, 3–22Brecher, R. A. (1992). ‘An Efficiency-wage Model with Explicit Monitoring’,

Journal of International Economics, 32, 179–91Bulow, J. I. and L. H. Summers (1986). ‘A Theory of Dual Labor Markets with

Application to Industrial Policy, Discrimination, and KeynesianUnemployment’, Journal of Labor Economics, 4, 376–414

Copeland, B. R. (1989). ‘Efficiency Wages in a Ricardian Model of InternationalTrade’, Journal of International Economics, 27, 221–44


De Coulon, A. and Y. Flückiger (1995). ‘Différences de salaire entre Suisses etétrangers’, Université de Genève

De New, J. P. and K. F. Zimmermann (1994). ‘Native Wage Impacts of ForeignLabor: A Random Effects Panel Analysis’, Journal of Population Economics,7, 177–92

Dickens, W. T. and K. Lang (1985). ‘A Test of Dual Labor Market Theory’,American Economic Review, 75, 792–805

Dustmann, C. (1993). ‘Earnings Adjustments of Temporary Migrants’, Journal ofPopulation Economics, 6, 153–68

Ethier, W. J. (1985). ‘International Trade and Labor Migration’, AmericanEconomic Review, 75, 691–707

Flückiger, Y. and J. Silber (1995). ‘Income Inequality Decomposition by IncomeSources and the Breakdown of Inequality Differences Between Two PopulationSubgroups’, Swiss Journal of Economics and Statistics, 131, 599–615

Gaillard, S., R. Salzgeber and J. Schütz (1991). Europäische Integration:Arbeitsmarktliberalisierung und Strukturwandel in der Schweiz (Chur: VerlagRüegger)

Grubel, H. B. and A. D. Scott (1966). ‘The International Flow of Human Capital’,American Economic Review, 56, 268–74

Hamermesh, D. S. (1993). Labor Demand (Princeton: Princeton University Press)Hammar, T. (1985). European Immigration Policy: A Comparative Study

(Cambridge: Cambridge University Press)Hoffmann-Nowotny, H.-J. (1985). ‘Switzerland’, in T. Hammar (ed.), European

Immigration Policy: A Comparative Study (Cambridge: Cambridge UniversityPress)

Jones, S. R. G. (1987). ‘Minimum Wage Legislation in Dual Labor Market’,European Economic Review, 31, 1229–46

Kee, P. (1995). ‘Native-immigrant Wage Differentials in the Netherlands:Discrimination?’, Oxford Economic Papers, 47, 302–17

Kohli, U. (1993). ‘US Technology and the Specific-factors Model’, Journal ofInternational Economics, 34, 115–36.

Müller, T. (1998a). ‘Immigration, Protection, and Social Welfare’,,at proof.(1998b). ‘Dual Labor Markets and Discrimination against Immigrants’, Working

Paper, Department of Economics, University of GenevaNeary, J. P. (1988). ‘Tariffs, Quotas, and VERs with and without Internationally

Mobile Capital’, Canadian Journal of Economics, 21, 714–35Piore, M. J. (1979). Birds of Passage: Migrant Labor and Industrial Societies

(Cambridge: Cambridge University Press)Schmidt C. M., A. Stilz and K. F. Zimmermann (1994). ‘Mass Migration, Unions,

and Government Intervention’, Journal of Public Economics, 55, 185–201Shapiro, C. and J. E. Stiglitz (1984). ‘Equilibrium Unemployment as a Worker-dis-

cipline Device’, American Economic Review, 74, 433–44Winter-Ebmer, R. and J. Zweimüller (1996). ‘Immigration and the Earnings of

Young Native Workers’, Oxford Economic Papers, 48, 473–91Zimmermann, K. F. (1993). ‘Ökonomische Konsequenzen der Migration für den

heimischen Arbeitsmarkt’, Schweiz. Zeitschrift für Volkswirtschaft undStatistik, 129, 283–301

184 Tobias Müller

Discussion

RUDOLF WINTER-EBMER

Chapter 7 is very interesting, and a contribution I certainly enjoyedreading. Looking at the topic, I even wondered why nobody else hadtackled this problem earlier. Tobias Müller starts from the observation thatthe assimilation process of immigrants is different in, say, the USA fromWestern Europe. The famous Chiswick (1978) study suggests that foreign-born workers catch up in earnings with natives in 13 years. In the typical‘guest-worker’ country, Germany, this does not seem to be the case at all:foreigners start at lower levels and have lower rewards to job experiencethan Germans (Licht and Steiner, 1994).

This differential treatment of immigrants gives rise to the suspicion ofsegmented labour markets for natives and foreigners. Therefore, the ques-tion arises: if natives can discriminate against foreigners, does this changethe welfare assessment of immigration? One way towards this end is to lookat migration policies having the potential of a Pareto-improvement – pro-vided some sort of redistributive scheme is applied. Chapter 7 uses adifferent framework: welfare of natives is expressed by an explicit socialwelfare function of the Atkinson type, so that inequality-aversion of thesocial planner can be explicitly considered. The profitability of a specificmigration scenario can thus be judged with respect to the inequality-aver-sion of the society. It turns out that for low levels of inequality-aversionimmigration is beneficial for the host country, even more so if the immi-grants have no access to the primary sector of the economy. The moreimportant inequality concerns get, the less favourable immigration will be.

This contrasts with the impact of protectionist measures: here we havean efficiency loss, but this loss will be over-compensated for by the (welfare-enhancing) impact of less inequality. Furthermore, the more inequality-averse the society is, the more favourable the protection scenario gets. Thechoice of the best policy could be resolved if we knew the inequality-aver-sion of the society, measured by Atkinson’s «. But we don’t. One indicationcould be the interpretation of «: a value of «50 (0.2, 1, 2) would mean thatredistributing one dollar to the poor would be considered as socially worth-while, (even) if the rich would have to pay one dollar (1.14, 2, 4 dollars) intaxes. This could also be interpreted in terms of the bureaucratic losses thepolicy-maker is willing to incur in order to redistribute. If law-makers hada social welfare function in mind with «50.2, they should be prepared toaccept bureaucratic losses of up to 14 per cent of transfers given. Such a low


value of « would certainly speak more in favour of immigration ascompared to protectionist measures. Bhagwati (1982) discusses two recipesto counter a shift in comparative advantage in the production of a labour-intensive good, say, footwear: protection of these industries from importcompetition or an increase in immigration quotas. In fact, for Bhagwatiimmigration and protection can be considered as being equivalent in thesense that they allow the same level of the import-competing good. Inchapter 7, Müller tries to understand why different policy options have beenused in the 1960s and in the 1980s: a guest-worker system with discrimina-tion proved profitable in the 1960s because it increased welfare of natives(and the welfare of the immigrants was not included in the social welfarefunction anyway). As the temporary immigrants became more and morepermanent, discriminating against them became less of an option: there-fore, large-scale immigration became less favourable from the income dis-tribution point of view, as compared to protection. However, otherinterpretations are possible. The 1960s were a period of labour shortage,whereas the 1980s were characterised by chronic unemployment in WesternEurope. This example illustrates nicely the tension between long-run equi-librium trade theory and short-run labour market concerns. The currentpolitical discussion sets very different alternatives: trade restrictions orimmigration restrictions. This reflects the overwhelming importance ofshort- or medium-term unemployment problems in the realm of stickywage adjustments and high unemployment persistence.

The dual labour market economy here is modelled as an efficiency-wageframework of the shirking type. Workers in the primary sector work hard,because they do not want to fall back to low (market-clearing) wages of thesecondary sector. It would be interesting to see if a different modelling ofthe efficiency wage mechanism changed matters greatly. Suppose, forexample, that there is unemployment which acts as the threat mechanismin the non-shirking condition. More immigration would prima facieincrease unemployment and reduce the necessary efficiency-wage premium,leading to falling wages in the primary sector, which might possiblydecrease or increase inequality, depending on the share of skilled versusprimary-sector unskilled workers in the work force.

In the specific-factors model used here, capital is sector-specific and themain gain from immigration (apart from higher returns to capital) is thepossibility for natives to rise into the primary sector. In the case of non-dis-crimination, this advancement is highly improbable, because it wouldrequire the labour elasticity in the primary sector to be much higher thanin the secondary sector. Moreover, as stressed by Müller in a previousversion of his study, if capital were mobile between sectors in aHeckscher–Ohlin (HO) world, a similar result would restrain welfare gains

186 Discussion by Rudolf Winter-Ebmer

from immigration: immigration would increase native employment only inthe primary sector, if the secondary sector is relatively capital-intensive.Considering these cases, the present one seems restrictive.

The calculation of U-welfare for natives assumes that capital income isdistributed to native households in proportion to their wage incomes. Inthis framework inequality considerations are confined exclusively towithin-workers’ redistribution, distribution issues between capital ownersand employees are therefore neglected. The sensitivity check on p.000shows that competing assumptions, like all capital income to the skilled,reduce the admissible «-values for welfare-improving migration consider-ably. I suppose that this would even be more the case if all capital incomewere to go to a separate capitalist class.

Finally, it might be interesting to compare the results of this chapter withthe literature on Pareto-gains from trade liberalisation. Whereas generallyin models with no distortions Pareto-gains from trade can be achieved vialump-sum transfers, Brecher and Choudri (1994) extend the analysis toallow unemployment in the form of an efficiency-wage mechanism.Individual workers losing jobs would require some sort of compensation tomaintain their pre-liberalisation welfare levels. This compensation wouldweaken the incentive to work hard unless it were countered by risingefficiency wages. As Brecher and Choudri show, rising wages may lead viarising prices to the infallibility of Pareto-gains from trade liberalisationusing any transfer system. The question of compensation for losers is notaddressed in Müller’s chapter, but it is certainly important for politicaleconomy aspects of migration legislation.

Coming to the empirical implementation, I am a little bit worried by thenested CES specification. Basically, the set-up restricts the substitutionpossibilities – there is no room for complementarity relations because twofactors each form a composite factor with separability assumptions. Aboveall, the separability between skilled and unskilled workers seems question-able. In fact, it is assumed that secondary, primary and skilled workersinteract in the production process (in a representative firm). This runssomewhat counter to my intuition of segmentation, which goes primarilyacross firms, possibly along firm size categories – e.g. good jobs in big firmswith internal labour markets, etc. Interestingly, the author gets very similarresults if he reserves the good jobs to three sectors only. The categoriesnatives and immigrants – excluding those with a permanent residence –seem not particularly evident to me, unless there is evidence of differentialassimilation processes of the two immigrant groups (of course, the finalquestion concerns the inclusion in the social welfare function, as is men-tioned at the end of the chapter!). Similarly, the introduction of primaryand secondary labour markets for unskilled workers only is somewhat


peculiar, whereas no efficiency-wage structure is assumed for the skilledlabour market.

Notwithstanding these critical observations, the empirical results aresensible and very interesting. The mechanism of the upgrading of nativeworkers’ jobs caused by increasing immigration is an idea which is oftenheard from practitioners of the labour market. In an empirical study forAustria, we found related evidence: native workers employed in firmswithout foreigners had flatter earnings–experience profiles than those beingemployed in firms with a considerable presence of immigrants (Winter-Ebmer and Zweimüller, 1995). This suggests that in the presence of immi-grants native workers profit more from internal labour markets. Theimplementation of higher quit rates for immigrants is a good extension,which comes from the discrimination literature, where job segregation anddifferential job promotion is explained efficiently by differentials inexpected quit rates. Again, these simulation patterns confirm econometricresults of statistical discrimination in job advancement (Winter-Ebmer andZweimüller, 1997) owing to expected quit rates.

I have some problems with the variation in the number of migrants.The general result is that immigration is welfare-enhancing as long asinequality-aversion is not too great. Moreover, immigration gets evenmore favourable the higher the stock of immigrants is already.Immigration in the guest-worker system is doubly beneficial for natives:(1) it enables more natives to find a job in the primary sector and (2) itincreases the wage gap between the primary and the secondary sector. Ofcourse the first effect vanishes as more immigrants arrive, but the secondremains. This positive welfare effect of redistribution is caused by the spe-cific situation chosen: wages for skilled labour and returns to capital rise,but wages in the secondary sector fall drastically (where no natives workin the end).

To conclude, I consider the inclusion of a dual labour market togetherwith more explicit treatment of inequality-aversion a major topic in theassessment of the consequences of immigration policies. Müller has gone aconsiderable way down this road. I would guess – given the great manyextensions which suggest themselves – that others will in due course follow.

REFERENCES

Bhagwati, J.N. (1982). ‘Shifting Comparative Advantage, Protectionist Demands,and Policy Response’, in J.N. Bhagwati (ed.), Import Competition and Response(Chicago: University of Chicago Press)

Brecher, R.A. and E. U. Choudri (1994). ‘Pareto Gains from Trade, Reconsidered’,Journal of International Economics, 36, 223–38

188 Discussion by Rudolf Winter-Ebmer

Chiswick, B. (1978). ‘Americanisation and the Earnings of Foreign-born Men’,Journal of Political Economy, 86, 897–921

Licht, G. and V. Steiner (1994). ‘Assimilation, Labour Market Experience andEarnings Profiles of Temporary and Permanent Immigrant Workers inGermany’, International Review of Applied Economics, 8, 130–56

Winter-Ebmer, R. and J. Zweimüller (1995). ‘Internal Labour Markets and Firm-specific Determination of Earnings in the Presence of Immigrant Workers’,Economics Letters, 48, 185–91

(1997). ‘Unequal Assignment and Unequal Promotion in Job Ladders’, Journalof Labour Economics, 15, 43–71


8 Globalisation and migratorypressures from developing countries:a simulation analysis

RICCARDO FAINI, JEAN-MARIE GRETHER

AND JAIME DE MELO

1 Introduction

Opposition to immigration in the North has been on the rise, notably in theEU and in the USA. This rising opposition can be traced to the regime shiftin their labour markets which have moved from tight labour markets up tothe early 1970s to rising unemployment (in the EU) and increasing wageinequalities (in the USA) since then. On the other hand, migratory pres-sures have not subsided. For many developing countries, stagnant growthin the wake of the debt crisis and explosive demographics have led to amarked deterioration in labour market conditions. The thawing ofEast–West relations has also contributed to greater migration pressure.Paradoxically, Western European governments, who used to criticise social-ist regimes in Eastern Europe for imposing draconian restrictions onpeople mobility, reacted to the 1989 events by swiftly implementing newand more effective migration control measures.

As pointed out by Schiff (1996), the change in attitude in the North towardsimmigration from the South is well exemplified by the change in attitudes inthe International Labour Organisation (ILO). Not surprisingly, during theperiod of tight labour markets in the North, the ILO recommended attract-ing suitable labour from the South (ILO, 1949). But by the mid-1980s, the ILOwas recommending that the receiving countries should

endeavour to cooperate more fully in the development of such countriesby appropriate intensified capital movements, the expansion of trade, thetransfer of technical knowledge. (ILO, 1984)

More recently, the ILO (1992) has recommended that developed countriesprovide foreign assistance to developing countries to reduce migratorypressures.

This policy shift on migration can be criticised on both ethical and eco-nomic grounds (see, for example, Bhagwati, 1992). Yet, migration policy is

190

not made in a vacuum, but reflects the values and the attitudes of voters andpressure groups reflected in what Schiff (1996) calls social capital’. The factis that the attitude toward migration has changed markedly in industrial-ised countries. According to an EC survey, 64 per cent of respondents inItaly believe that migrants are ‘too many’. In France, Germany and the UKthe percentage of those who believe migration to be excessive is over 50 percent. Since 1989, Europe’s newspapers have been filled with stories of racialintolerance. Under these conditions, Europe’s policy-makers have littlechoice but to try to stem the flow of new immigrants while at the same timepromoting the social and economic assimilation of old ones. This pressureis likely to subsist in spite of the ageing of European population. The issuethen arises of how industrial countries can discourage massive and largelyundesired population flows. That is the issue investigated in this chapter.

The policy options appear to be limited. Migration controls can work in thevery short run but are often ineffective in the medium to long run, as typifiedby the recurring use both in Europe and in the USA of migration regularisa-tion programmes, invariably coupled with the promise that future controlswill be tightened up. Foreign aid as well as trade policies have often been advo-cated as more effective tools to cope with rising migrations. Foreign aid canboost growth in sending countries, thereby reducing the incentive to emigrateto the North. Similarly, trade liberalisation can foster factor-price conver-gence and limit the incentive for factor mobility. The diffusion ofNorth–South regional trade integration agreements during the 1990s hasoften been attributed to the desire to stem migration pressure, as epitomisedby President Salinas’ quote that ‘Mexico wants to export goods, not people’.The EU’s drive toward signing Association Agreements with its close tradingpartners in the East and in the South has been similarly motivated.

Two channels of action have been identified in the policy debate. The firstare what one could call ‘direct measures’. These take the form of increasedaid to raise income in the South. Increased costs to migration through, say,more effective border controls represent an alternative direct measure. Thesecond channel includes ‘policy reform measures’ whose effects on migra-tory pressures are less direct. Paramount here are the trade reforms of theUruguay Round and beyond, leading to a reduction in barriers to trade ingoods (especially in agriculture and light manufactures such as textiles andapparel) and in services (a reduction in barriers to Foreign DirectInvestment (FDI) and eventually to trade in services).

The impact on migratory pressure of any measure will depend mostly onthe determinants of migration in sending countries in response to changesin incentives to migrate. To take an example, one would expect that migrantsfrom Sub-Saharan African (SSA) countries would be mostly low-skilllabour whereas Southern Mediterranean migrants in Northern countries

Globalisation and migratory pressures from developing countries 191

would be more predominantly medium- and high-skill labour. Migrantsfrom SSA countries would also tend to be viewed as more permanent inreceiving countries whereas migrants from Mediterranean countries wouldbe more susceptible to responding positively to an improvement in eco-nomic conditions in their home countries, and thus would exhibit ‘return’migration behaviour.1 If one considers SSA countries as typical of Low-Income (LI) economies and Southern Mediterranean countries as typical ofMiddle-Income (MI) economies, one would expect different patterns ofmigratory pressures in each group. This chapter uses simulation analysis toinvestigate if direct and indirect measures are likely to elicit different migra-tory pressures in different countries. Two questions are asked: (1) Woulddirect and indirect attempts to reduce migratory pressures yield qualitativelydifferent migratory responses for an archetypical developing country, be ita LI or a MI economy? (2) Would identical measures have different effectsin LI and MI economies?

Simulations are carried out in a single-country framework with model-ling efforts directed at the sending country. To this effect we build aRicardo–Viner (RV) economy with traded and non-traded sectors.2 Theeconomy is disaggregated on the labour and household sides with house-holds offering labour services at home and abroad. Calibration is carriedout for two ‘archetype’ LI and MI economies. The archetypes differ system-atically in the structure of production, consumption, trade and factorsupply across sectors while elasticities and the mapping of factor income tohouseholds is purposely kept the same. This allows us to investigate quitesystematically the likely effect of differences in economic structure onmigratory pressures.

One might object that modelling should be carried out in a two-countryNorth–South world. Such an approach would be desirable for an analysis ofsystemic changes in which repercussion effects are likely to be important. ANorth–South trade model would also be desirable to study factor-price-convergence effects and for a more thorough analysis of the complementar-ity–substitutability issue between trade and migration. A more elaboratemodel than the one offered here would also be desirable to investigate someof the issues which the theoretical literature has identified such as migrationin response to relative deprivation, risk and uncertainty and sunk costs inmigration. We ignore or only partially address these issues. Nonetheless, ourfocus on a single-country model enables us to give a better road-map of thechannels through which Southern countries’ migratory pressures exertthemselves. Finally, since the receiving country is not modelled, increasedmigratory pressure in this single-country framework translates directly intoan increased supply of workers abroad (in effect, we assume an infinitelyelastic demand for labour from the sending country over the relevant range).


The remainder of the chapter is organised as follows. Section 2 developsthe modelling of migration in this RV framework. It also outlines the mainfeatures of the simulation model which is given in the appendix (p. 212).Section 3 describes the calibration to the two archetype economies and thesimulations proposed to decompose the likely effects of globalisation onSouth–North migratory pressures. Section 4 reports on the simulationresults comparing the effects on migratory pressures across archetypes fora base model cloture. Section 5 examines the sensitivity of results tochanges in model cloture (elasticities and assumptions about the mappingof factor income to households). It also proposes likely orders of magni-tude for what one might call the most relevant model for each ‘archetype’.Since much of the data is constructed or invented, it is probably best tointerpret the results in relative terms – i.e. differences in results across arche-types or between measures. Section 6 draws some conclusions.

2 A Ricardo–Viner model of migration

The traditional reason for migration is earnings differential: the prospect ofa higher net (of moving and other fixed costs) expected wage in the receiv-ing country is the main reason for migration. In the traditional literaturederived from Harris and Todaro (1970), risk-neutrality on the part ofhomogeneous migrants implies that expected earning differentials are thesole determinants of migration. It has been recognised, however, thatmigration costs are also a determinant of migration. These include non-monetary costs (idiosyncratic tastes for location as in Djajic andMilbourne, 1988, or Faini and Venturini, 1994) and monetary costs (Lopezand Schiff, 1995). If agents are risk-averse, matters are somewhat morecomplicated. First, risk considerations may have an ambiguous impacteven from the point of view of an individual agent. On the one hand greateruncertainty about perspectives in the destination region should determigration. Conversely, higher income risk at home may encourage agentsto move. However, if migration is at least partially irreversible, as wouldhappen if it were subject to sunk costs, then potential migrants might preferto wait for uncertainty to dissipate before making their location decision(Burda, 1993). At the household level, on the other hand, greater uncer-tainty may boost migration provided that incomes in different location arenot perfectly correlated. Under these circumstances, the household canindeed reduce its total risk exposure by having some of its members migrateto a different location (Stark, 1991). The theoretical literature has alsoinvestigated additional factors, such as relative deprivation and asymmet-ric information, that might affect the migration decision.

While there is some evidence to support these suggestions (see Schiff,


1996, for a survey of the evidence and Daveri and Faini, 1996, for recentevidence from Southern Italy), by and large the evidence is scant. Also wehave little knowledge of the relative importance of these (and other) factorsin determining the supply of migrants. For this reason, we develop a simplemodel that accommodates the traditional motive based on earningsdifferentials and we assume that, in their location decision, households faceconcave costs to relocation. This implies that households will diversify, i.e.they will not locate entirely in one country.

2.1 Modelling migration

Consider a heterogeneous household (households are indexed over he H)that maximises a utility function that includes leisure and the consump-tion of goods as arguments. In view of the numerical application thatfollows, for simplicity, we assume strong separability in the choice oflabour supply and consumption of goods. This gives us a two-stage deci-sion process. In a first stage the household decides on the amount oflabour to supply, and in a second stage on the allocation of labourbetween the home and foreign labour markets. In the first stage, thesupply of labour by household, LSh, is determined from the maximisationof the following LES utility function:

Uh(.)5 (Ci,h2di,h)bi,h, (1)

where Sni 5 0bi,h51 and C0 is leisure. Maximisation of (1) subject to the

household’s budget constraint ((25) in table 8A.1) determines the house-hold’s supply of labour services ((12) in table 8A.1) and the consumption ofgoods ((17) in table 8A.1). Given that we choose an LES, this means thatleisure and goods are substitutes as are all goods. The allocation of labourservices LSh between the domestic (Lh) and foreign (Lh*) labour market takesplace in the second stage. It depends on the relative wage in each destina-tion with increasing costs to relocate from one labour market to the other.This gives a concave locus describing how the households’ labour servicescan be relocated. A convenient form is the familiar Constant Elasticity ofTransformation (CET) function given by:

LSh5 h[(12nh)2Vh(Lh)

11Vh1 (nh)2Vh(Lh*)11Vh]1/(11Vh)

5Ch(Lh,Lh*,Vh)5Ch(.), (2)

where h and nh are positive parameters (0nh1) and Vh is the elasticity oftransformation.

The CET function, ch(.), is concave implying increasing costs to labour

B

B

pn

i50


relocation. It has the advantage of tractability: it is easy to calibratenumerically as it requires only one parameter Vh for implementation. Inthe second stage, household h maximises labour income, WhLSh, where Whis the revenue-maximising wage under the condition of increasing reloca-tion costs ((11) in table 8.A1 gives the expression for Wh). Maximisation ofwage income is subject to the constraint that labour relocation between thehome and foreign labour markets takes place under increasing costs, i.e.

max L(Lh,Lh*,lh)5whLh1ERwh*Lh*2lhCh(.), (3)

where lh is a Lagrange multiplier, ER is the conversion factor of foreign cur-rency units into domestic currency units, and wh is the domestic wage. Inthis maximisation, the foreign wage, wh* is exogenous, reflecting an infinitelyelastic foreign demand for labour over the relevant range. The result of themaximisation of (3) is the allocation of labour services, Lh and Lh* ((13) and(14) in table 8A.1).

Figure 8.1 depicts the allocation of labour services for a given aggregatehousehold labour supply, LSh. Wage-income-maximisation leads to the solu-tion (Lh, Lh*). Since by the choice of numéraire, a price index of non-tradedgoods is kept fixed (see (28) in table 8A.1), ER is in effect the value of the realexchange rate – i.e. the relative price of tradables. Hence, we assume in (3)that the household makes its labour-allocation decision on the basis of thepurchasing power of foreign wages in terms of home (non-traded) goods.


Figure 8.1 Labour-allocation decision

Labour will then be reallocated either in response to a change in the domes-tic wage, or to a change in the value of the real exchange rate, ER.

In the above formulations, we are modelling a medium-term labour-alloca-tion decision rather than an irreversible migration decision. As argued above,this ‘guest-worker’ view of migration is perhaps more prevalent among MIthan LI developing countries. There is recent evidence in the case of Mexicanmigration to the USA that Mexico–US migration patterns are responsive tochanges in the Mexican-US bilateral exchange rate (see Markusen andZahniser, 1997, table 1). As an alternative, we assume that migration is morepermanent at least in the sense that the household does not consider thedomestic currency purchasing power of foreign wages. This amounts to fixingER to its initial value in (3). We shall call this version the ‘permanent-migra-tion’ view. It is offered here both as an alternative relevant to the migrationdecision in LI economies and as a way of isolating the real exchange rate com-ponent of migratory pressures.3 It is considered in section 5.

Whether these views of the labour allocation are realistic or not is a mootpoint: all that can be said is that they are ‘consistent’ with observations onhousehold diversification.4 As will be confirmed in the simulations below,this formulation implies that there is a resource cost to a relocation oflabour across the border because of the concavity of the CET function.5

2.2 The simulation model

In the specification of the RV simulation model, atomistic firms with con-stant-returns-to-scale (CRS) production functions maximise profits andatomistic households maximise utility. Each agent has a constraint, and theeconomy has an external constraint. Firms have a Cobb–Douglas technol-ogy for value-added and Leontief technology for intermediates andbetween intermediates and value-added. As in the traditional RV model,the economy is a price-taker in the markets for traded goods. There arehowever, non-traded activities as well, for which there is an endogenousmarket-clearing price. To accommodate the fact that much of imports indeveloping countries are non-competitive with local production (the struc-ture of imports reveals that intermediates and capital goods account for thebulk of imports), we assume that each sector also uses non-competitiveimports. Firms pay their mobile factors, the different categories of labour,their value marginal product.

As some sectors are non-traded, the real exchange rate adjusts to main-tain external and internal balance. In turn, adjustments in the real exchangerate affect both the allocation of labour as depicted in figure 8.1 as well asthe domestic-currency value of remittances. For traded goods, domestic andforeign-produced goods are perfect substitutes which would tend to lead toextreme specialisation among traded activities in response to relative price


changes. It is the presence of sector-specific factors that gives enoughconcavity to the production possibilities frontier to prevent extreme special-isation.

The mapping from factor to household income is described below insection 3 (see table 8.2, p. 200), and in the appendix, which gives the fullspecification of the model. Suffice it to note here that, to simplify, eachhousehold only supplies one type of labour, and we exclude the possibilitythat households would invest capital income abroad. The income fromcapital (the specific factor in each sector) is distributed across households(see table 8.2). This mapping of factor income to households implies thatthere is some diversification in each household’s source of income, adiversification which attenuates household income fluctuations.

It is clear that this model is very close to the standard RV model withnon-traded goods, except that it is augmented to include migration and thatthere are several mobile factors of production and several households. Toapproximate the standard model as closely as possible, the government’sonly role is to tax trade (tariffs and export taxes are the only distortions inthe model). Government revenue is redistributed lump-sum to each house-hold in proportion to that household’s income. Remittance incomeadjusted for the (exogenous) trade deficit in foreign-currency units isdistributed to households, also according to each household’s share in totalincome. Thus changes in trade policy do not have a direct effect on incomedistribution via government transfers.

In this set-up, policy changes such as a trade liberalisation that alters rel-ative prices affects both the aggregate supply of labour on the part of eachhousehold according to the resulting changes in household wage incomeand total income and the allocation of labour between the domestic andforeign labour market depending on changes in relative wages in the homeand foreign labour market. An appreciation of the real exchange rate, forexample, will make the home labour market more attractive if there is nocompensating change in the domestic wage. On the other hand, the homemarket wage rate for each labour category will change following traditionalHeckscher–Ohlin (HO) factor- intensity differentials across sectors. Whenprotection is reduced, for example, if export sectors are relatively intensivein the use of medium-skill labour, the wage rate for medium-skill labour willrise relative to other wages.

3 Decomposing the effects of globalisation on the supply of migrantsfor two archetype economies

We now examine numerically how changes in the economic environmentvia either changes in policy or in the external environment are likely toaffect the supply of migrants from the South to the North. This section


describes the two archetype economies to which the RV model is calibratedand the set of simulations to decompose the direct and indirect channelsthrough which policy changes and changes in the external environment arelikely to affect South–North migratory pressures.

3.1 The archetype economies

The RV model is calibrated to data for two archetypical developingeconomies. The data are adapted from Bourguignon et al. (1992). Asexplained in the appendix, these data are representative (in the sense ofChenery and Syrquin, 1986) of a low-income (LI) developing economy(say, a SSA economy) and a middle-income (MI) developing economy (say,a Latin American or North African economy). The details on the data areprovided in the appendix. To control for the dimension along which the MIand LI economies differ, we assume that the two archetypes are of equalsize, have the same sectoral disaggregation, the same factor-use mapping bysectors and the same household composition. This limited differencebetween archetypes makes it much easier to interpret simulation results.

The two archetypes differ in structure in the following stylised way. Theprincipal structural characteristics of the LI and the MI economies aregiven in tables 8.1a and 8.1b (also see tables 8A.3 and 8A.4 in the appen-dix). Tables 8.1a and 8.1b show the difference in production structure,trade, trade barriers and labour allocation across sectors and abroad. Thetwo economies engage in the same activities: subsistence agriculture for theLI economy which is a non-traded activity by definition and import-com-peting agriculture in the MI economy. In the LI economy, the output of theexport agriculture sector (which could also represent a mining activity) isentirely exported as there is no domestic final and no intermediateconsumption (cocoa or copper). It is taxed and is the only export activityin the calibrated scenario for the LI economy. This representation of the LIeconomy corresponds to the traditional stylised description of small LIeconomies that export only a few commodities (i.e. ‘monoculture’economies).6

Two manufacturing activities (light and heavy manufacturing) are dis-tinguished, the latter essentially a non-competing import-substitutingactivity, the former representing the typical HO labour-intensive exports.Note the structure of trade taxes in each archetype: imports and exportsare taxed, with taxation higher in the LI archetype. Otherwise, the arche-types also differ in their inter-industry linkages, these being weaker in theLI economy which also has a larger share of non-competitive imports intotal imports (see tables 8A.3a and 8A.3b in the appendix). This calibra-tion results in the LI archetype having a much larger share of activity in


agriculture. However, to control for the sources of differences in economicstructure, the same Cobb–Douglas technology for value-added is imposedfor each archetype (see table 8A.4 in the appendix). Both the Service andthe Informal sectors are non-traded activities in both economies. Thus, asubstantial share of GDP is non-traded in both calibrated archetypes.These differences in production structure are reflected in the allocation ofthe three labour categories – (low-skill5LSL), (medium-skill5MSL),(high-skill5HSL) – across sectors. (The percentage distribution across


Table 8.1. Structure of production, demand and factor allocation

a low-income (L1) archetype

Out. Net Tariff Low Medium Highstruc exp. out. exp. tax skill skill skill(1) (2) (3) (4) (5) (6)

Sub. ag. 0.20 0.0 0.0 0.65 0.0 0.0Exp. ag. 0.10 1.0 20.2 0.0 0.25 0.0Light mfg. 0.22 20.2 0.5 0.0 0.17 0.31Heavy mfg. 0.04 21.5 0.4 0.0 0.0 0.07Services 0.31 0.0 0.0 0.0 0.31 0.57Informal 0.13 0.0 0.0 0.26 0.17 0.0Lab abroada (0.2)c 0.09 0.09 0.05Lab abroadb 0.63 0.32 0.05

b middle-income (M1) archetype

Out. Net Tariff Low Medium Highstruc exp. out. exp. tax skill skill skill(1) (2) (3) (4) (5) (6)

Sub. ag. 0.09 20.3 0.2 0.46 0.0 0.0Exp. ag. 0.05 0.4 20.1 0.0 0.06 0.0Light mfg. 0.24 0.2 0.0 0.0 0.19 0.18Heavy mfg. 0.14 20.3 0.4 0.0 0.0 0.33Services 0.33 0.0 0.0 0.0 0.46 0.45Informal 0.15 0.0 0.0 0.45 0.21 0.0Lab abroada (0.1)c 0.09 0.09 0.05Lab abroadb 0.35 0.47 0.18

Notes:a Share of migrants in skill category.b Share of skill category in total migrants.c Tariff on non-competitive imports.

sectors is given in columns (3)–(6) in tables 8.1a and 8.1b.) Note in row 7of the table, the percentage of each category of labour abroad. Thesefigures are assumed to be the same for each archetype, even though thecomposition of migrants abroad from the LI and MI archetype is different(see row 8).

Finally, table 8.2 gives the mapping from factor to household income. Itis the same for both archetypes. The top part of the table gives the mappingof the mobile factor to each one of the four households (LS HH, MS HH,HS HH, CAP HH) and the bottom part of the table gives the mapping ofthe fixed factor to these households. Except for the HSL labour categorywhich is supplied by both the HS HH (75 per cent) and the CAP HH (25per cent), each labour category is supplied by only one household. LSL isthe least diversified labour category as it is employed in only two sectors,both of which are non-tradable in the LI archetype. The other labour cate-gories are employed in three or four sectors, with a split between tradableand non-tradable activities, so that the effects of changes in the relativeprice of tradables on household income is dampened by this assumeddiversification in sectoral employment.

3.2 Decomposing the effects of globalisation

To study the determinants of migratory pressures in this single-countryframework three sets of policy experiments are carried out to find out the


Table 8.2. Factor–household mapping: L1 and L2 archetypes

Low- Medium- High- Capitalistskilled skilled skilled householdLS HH MS HH HS HH CAP HH

Mobile factor:LSL 1 0.0 0.0 0.0MSL 0.0 1 0.0 0.0HSL 0.0 0.0 0.75 0.25

Specific factor:Sub. ag. 0.5 0.0 0.0 0.5Exp. ag. 0.0 0.5 0.0 0.5Light mfg. 0.0 0.25 0.25 0.5Heavy mfg. 0.0 0.0 0.25 0.75Services 0.0 0.25 0.25 0.5Informal 0.5 0.5 0.0 0.0

best way to achieve the ‘Export goods, not people’ objective. These aredescribed in table 8.3. Within each set, experiments are usually cumulative.In the first experiments (E-1 and E-2), the attempt is direct: increase inforeign aid to the South or raise migration costs by lowering the (exoge-nous) effective wage in the North. In the second set (E-3 and E-4), tradeliberalisation in the South is examined with the potential by-product of anincrease in Northern FDI in the actual (and potential) exporting sectors.The third set (E-5–E-7) attempts to capture the effects of a reduction in theprotection of agricultural and manufactured products exported by theNorth. Finally we consider the potential effects of increased globalisationof economic activity by combining increased openness to trade in theNorth and the South (E-8). In this single-country model, it is difficult tocapture the effects of a trade liberalisation in the North. It is done here byan exogenously imposed improvement in the terms of trade for the South(except in E-6 for the LI archetype which is a net importer of light manu-factures – see tables 8.1a and 8.1b). Arguably, this may not be an accuratedescription of the likely effects of trade liberalisation in the North, as it isnot clear that it would improve the terms of trade for the South, thereby


Table 8.3. Policy experiments

Direct interventionE-1 Increased transfer of resources [ITR] to the South (a relaxation of the

external constraint by 10 per cent)E-2 Increase of migration costs (E211decrease in the wage in the North by 10

per cent)

Trade liberalisation in the SouthE-3 Trade liberalisation in the South [TLS] (removal of tariffs on imports and

taxes on traditional exports)E-4 Increase in FDI in the south [FDI] (E-3110 per cent increase in capital

stock in agricultural export and light manufacturing with profit repatriationto the North)

Trade liberalisation in the NorthE-5 Agricultural trade liberalisation in the North [ATLN] (a 10 per cent

increase in the world price of agricultural exports)E-6 Manufacturing trade liberalisation in the North [MTLN] (a 10 per cent

increase in the world price of light manufactures)E-7 Across-the-board trade liberalisation in the North [TLN] (E-51E-6)

Increased globalisationE-8 Global trade liberalisation in North and South [GTL] (combination of E-7

and E-4)

presumably stemming migratory pressures. We are aware of this, but wishedto capture the hoped-for increase in the demand for exports of manufac-tures from the South that increased integration of world markets wouldpresumably bring about. In any event, it is these simulations that are carriedout for the two archetype economies mentioned above.

4 Simulation results

In this RV model, the simulations described above affect the supply ofmigrants to the North through several channels. At the macro level,changes in purchasing power (via transfers or changes in the terms of trade)affect the aggregate supply of labour as the relative price of leisure changes.At the household level, income changes result from changes in the pur-chasing power (in terms of home goods) of remittances and from changesin remunerations via adjustments in relative wages across labour categoriesand in sectoral rents. Finally there is also a redistribution of purchasingpower via changes in the cost of living as a given change in relative priceshas a differential impact across households (here mostly via Engel effects inconsumption as LS HH have both a lower income and a lower price elastic-ity of demand for agricultural products). In addition to changes in pur-chasing power, there are also changes in relative rewards for fixed andmobile factors which affect household income and hence both the supplyand the location of labour activities. We report first the aggregate effects,then the compositional effects at the household and sectoral levels.

4.1 Aggregate migratory pressures

Tables 8.4a and 8.4b give the aggregate results in terms of total labour supply,its allocation at home and abroad, along with the changes in the equilibriumvalue of the real exchange rate and real GDP. A first glance at the results intables 8.4a and 8.4b reveal both that economic structure matters and that theobjective of decreased migration is not always met for the LI archetype. Infact, because trade liberalisation in the LI economy is accompanied by astrong real depreciation, it becomes more profitable to seek employmentabroad. While FDI attenuates this effect, and trade liberalisation in the Northas well, one still gets an increase in migratory pressure with globalisation. Itwould appear that LI economies are more prone to exhibit trade–migrationcomplementarity than MI economies. Indeed, for the MI economy, under allscenarios, there is a reduction in aggregate labour supply abroad even thoughsome labour categories increase their allocation abroad (see table 8.5b, p. 206).

The contrast between the two archetypes has several causes, but theprominent one is the different effect of trade liberalisation in the South


(TLS) on the equilibrium value of the real exchange rate. Whereas for theLI economy TLS is accompanied by a strong real exchange rate deprecia-tion (18.5 per cent), for the MI economy, TLS brings about a slight realexchange appreciation. This is because of greater supply response to thechange in relative prices in the MI economy than in the LI economy whichexhibits what is called ‘structuralism’. Indeed, the elimination of the exporttax on agricultural exports generates a stronger export-supply response in


Table 8.4. Macro results: labour supply and home–foreign-labour allocation

a L1 archetypea

TLab.b TFLab.c TDLab.d RERe RGDPf

E-1 ITR 20.1 21.5 0.2 21.4 0.1E-2 IMC 20.3 29.3 1.6 20.2 0.8E-3 TLS 0.1 8.0 21.5 18.7 23.1E-4 FDI 0.0 2.6 20.5 13.4 21.3E-5 ATLN 20.4 27.5 1.1 26.0 0.5E-6 MTLN 0.1 23.8 0.9 24.2 0.3E-7 TLN 20.3 210.6 1.9 29.5 0.9E-8 GTL 20.2 27.5 1.4 2.8 20.3

b M1 archetype

TLab.b TFLab.c TDLab.d RERe RGDPf

E-1 ITR 20.1 20.8 0.0 20.9 0.0E-2 IMC 20.4 211.4 1.8 20.8 0.9E-3 TLS 20.2 22.0 0.1 20.2 21.9E-4 FDI 20.4 25.6 0.6 23.7 20.7E-5 ATLN 20.2 23.9 0.6 23.3 0.1E-6 MTLN 20.3 27.5 1.1 27.7 0.3E-7 TLN 20.5 210.4 1.5 210.1 0.6E-8 GTL 20.9 214.6 1.9 213.5 21.1

Notes:a Percentage changes from initial values.b Aggregate labour supply.c Foreign labour allocation.d Domestic labour allocation.e Real exchange rate index.f Real GDP.See table 8.3 for definition of policy experiments.

the MI economy as agriculture exports (or mining) are also consumeddomestically whereas in the LI economy agriculture exports are not. Hencea removal of the export tax induces a strong export- supply response in theMI economy which must be matched by an increase in imports to meet theexogenous current-account constraint. Hence the real exchange rateappreciation to induce the resource shift out of import-competing activities(see also the results in table 8.6, p. 207).7

On the other hand, there are similarities in the pattern of results. This isso because we have only allowed for selected differences between the twoarchetypes. Consider for instance trade liberalisation in the North (TLN).For both archetypes, there is approximately the same 10 per cent reductionin labour supply abroad. This is because this simulation results in the sameimprovement in the terms of trade for both economies which induces thesame amount of real exchange appreciation. For both economies, a declinein real GDP occurs for the simulations that involve large labour relocations.As explained in section 2, this is due to the concavity of the labour alloca-tion frontier and reflects adjustment costs.

The same pattern of results obtains when one compares ITR with IMCand TLS with FDI which are also cumulative experiments. Compare forexample increased resource transfer (ITR) with IMC (increased migrationcosts) in which the increase in foreign transfer is also accompanied byincreased migration costs. Reducing the wage received abroad (perhaps bytighter immigration controls) naturally increases the amount of returnmigration. This larger return migration induces a larger decline in the homewage which lowers the aggregate supply of labour.

Finally note the general equilibrium effects at work in the global tradeliberalisation (GTL) experiment. Though GTL5FDI1TLN in design, thereduction in migratory pressure is less than would have been predicted in apartial equilibrium calculation that would have aggregated the results fromFDI and TLN separately.

4.2 Household and sectoral effects

The aggregate results have identified the role of differential changes in thereal exchange rate in the LI and MI economies, and its implications for thepattern of migratory pressures. These aggregate changes, however, maskcomposition effects at the household level. Also they are the result ofdifferent patterns of resource reallocation across sectors. Tables 8.5a and8.5b give, for each household, the change in real income, the foreign rela-tive wage expressed in domestic-currency units, the change in labour supplyabroad and the change in the distribution of income. Table 8.6 comparesthe resource shifts in the two archetypes for the TLS and GTL simulations.


Table 8.5. Household incomes and labour-allocation decisions

a L1 archetypea

Gini LSK LSK LSK MSK MSK MSK HSK HSK HSK CAP CAP CAPcoef.b INCc RWd FLSe INC RW FLS INC RW FLS INC RW FLS

E-1 ITR 20.1 1.4 21.5 21.6 1.0 20.8 21.1 1.3 21.2 21.9 1.3 21.2 21.9E-2 IMC 2.9 21.0 4.1 26.6 2.4 20.1 213.4 4.4 22.2 217.2 3.9 22.2 217.1E-3 TLS 25.3 0.1 9.7 10.0 10.6 25.9 27.8 222.3 50.0 79.9 23.7 50.0 75.4E-4 FDI 25.7 2.5 5.1 3.2 12.4 210.1 213.4 219.8 42.4 67.2 22.7 42.3 63.4E-5 ATLN 0.4 2.6 25.5 26.0 4.9 27.9 210.7 1.8 24.0 25.9 4.3 24.0 26.2E-6 MTLN 2.7 23.3 22.1 21.9 21.4 24.6 25.9 3.1 29.1 213.1 0.6 29.1 212.8E-7 TLN 2.8 20.7 27.3 27.7 3.2 211.6 215.3 4.6 212.1 217.3 4.5 212.1 217.3E-8 GTL 22.9 1.7 22.0 21.9 15.6 219.8 226.0 215.5 24.3 38.1 1.9 24.3 35.1

b M1 archetypea

Gini LSK LSK LSK MSK MSK MSK HSK HSK HSK CAP CAP CAPcoef.f INCc RWd FLSe INC RW FLS INC RW FLS INC RW FLS

E-1 ITR 0.0 0.8 20.3 20.4 1.2 20.8 21.1 1.0 20.5 20.8 0.9 20.5 20.8E-2 IMC 2.9 0.3 2.0 29.2 1.4 0.6 211.8 2.7 20.5 214.6 3.5 20.5 214.7E-3 TLS 26.5 25.2 11.2 12.3 12.9 215.8 220.0 24.5 11.9 17.1 23.6 11.9 16.9E-4 FDI 26.9 24.3 10.0 10.9 15.2 219.3 224.4 22.7 8.0 11.3 23.2 8.0 11.3E-5 ATLN 20.5 20.0 20.5 20.6 3.5 25.9 27.4 0.0 20.8 21.2 1.2 20.8 21.4E-6 MTLN 3.0 22.3 20.4 20.3 3.9 29.6 212.1 3.5 27.1 210.1 2.4 27.1 29.9E-7 TLN 2.4 22.4 20.8 20.7 6.8 213.7 217.4 3.3 27.6 210.8 3.2 27.6 210.8E-8 GTL 23.0 27.8 8.9 10.1 20.3 229.4 237.1 0.9 23.0 24.3 0.8 23.0 24.4

Notes:a Percentage changes from initial values.b Initial value G50.41.c Real income.d wh* ER/wh.e Foreign labour supply.f Initial value G50.26.See table 8.3 for definition of policy experiments.

Start with the effects on households. As noted above, the largestdifferences between the two archetypes are for the TLS and GTL simula-tions. Comparing the effects of trade liberalisation, there a strikingdifference in the effects across households even though TLS reduces incomeinequality by approximately the same amount for both archetypes. In theLI economy the purchasing power of the low-skill household (LS HH) doesnot fall whereas it declines by 5 per cent for the MI economy. This is becausethe real depreciation lowers the cost of living following the fall in the priceof non-traded subsistence agriculture. However, in the absence of HOeffects, one would have also expected that the relative wage (RW) for the LSHH would have fallen more in the LI economy (because of the deprecia-tion). But TLS increases the demand for low-skill labour (LSL) in the LIeconomy, thereby attenuating the decline induced by the real exchange ratedepreciation. In fact, the fall in the relative wage is approximately the samefor both economies. The HS HH is the great loser of TLS in the LI economyas the fall in the domestic wage exacerbates the effect of the real deprecia-tion. Not surprisingly, TLS (and to a lesser GTL) brings an exodus of HSand CAP HH. However, because both households are only a small fractionof migrants abroad, the net effect on migratory pressure is small.

Except for the TLS and FDI results for the LI economy, we have seen thatall other measures generate a net decrease in migratory pressure. Inspectionof tables 8.5a and 8.5b indicates that return migration occurs for all HHgroups. There are however some differences in the pattern of changes acrossHH. For example, even though the CAP and HS HH face the same relativewage, their incomes deviate because of their different pattern of specific-factor


Table 8.6. Micro results: net price and sectoral output shiftsa

TLS TLS TLS TLS GTL GTL GTL GTLLI LI MI MI LI LI MI MI

Experiment Net Sec. Net Sec. Net Sec. Net Sec.archetype price out. price out. price out. price out.

Sub. ag. 4.3 23.6 221.8 212.9 3.7 21.1 236.6 220.2Exp. ag. 70.6 35.3 76.9 49.3 63.7 40.6 77.3 59.3Light mfg. 232.7 232.7 20.1 16.8 233.6 229.1 19.0 25.3Heavy mfg. 235.7 218.7 242.5 235.6 253.6 243.8 253.5 247.9Services 21.2 21.1 3.6 0.8 5.7 2.7 8.5 3.9Informal 13.9 22.5 3.6 0.4 15.1 20.7 1.0 2.5

Notes:a Percentage change from initial values.See table 8.3 for definition of policy experiments.

ownership. Again TLS in the LI economy barely affects the income of CAPHH while decreasing the income of HS HH by 20 per cent. Also in the agri-cultural trade liberalisation in the North (ATLN), return migration is morepronounced for the LI economy. This is because the terms of trade improve-ment is not dampened in the LI economy as there households do not consumethis good.

Other patterns could be drawn from a closer inspection of the results intables 8.5a and 8.5b. Two will be mentioned here. First, insofar as migra-tory pressure could respond positively to increases in relative deprivation(see Stark, 1991), then TLS in the South, especially if accompanied by FDIin exporting activities, could lead to a decrease in migratory pressure as thedistribution of income becomes less unequal. Second, migratory pressureaffects the skill composition of the labour force. Here TLS in the LIeconomy has a particularly strong adverse effect on the skill compositionof the labour force because of the Stolper–Samuelson (SS) effects leadingto an exodous of HSL. Insofar as there are positive externalities associatedwith a higher skill content of the labour force in LI economies, trade liber-alisation in the South could have adverse effects on future growth.

Table 8.6 compares the resource-pull effects of TLS and GTL across thetwo archetypes. Inspection of the results confirms the contraction in non-traded sectors in the LI economy (subsistence agriculture and informalsectors). A comparison of net price and output shifts does not give a clearpattern. This is mostly because, although both economies have the sameCobb–Douglas technology for value-added, the MI economy is moreintensive in the use of intermediates and is better endowed in MS and HSlabour.

5 Sensitivity analysis

It is obvious that the simulation results described above are likely to be sen-sitive to the choice of elasticities and, more importantly, to the modellingassumptions. In this section, we report briefly on the sensitivity of theresults to changes in elasticities and in assumptions about migration. Thisis far from systematic sensitivity analysis as only a few parameter valuesare varied and only one change in model cloture is contemplated.8 Theresults of these simulations are reported in table 8.7 using the followingnotation: ES-1(LI) indicates the results of sensitivity analysis on E-1 forthe LI economy. Multiple variations are indicated by letters (a for fixedlabour supply, b for financial constraints, etc.). In all the results in table8.7, the initial calibrated solution is the same as in tables 8.4a and 8.4b sothe results in table 8.7 are directly comparable with those in tables 8.4a and8.4b.


a Fixed labour supply

We consider first the sensitivity of the results to changes in the householdelasticity of labour supply by assuming a zero income elasticity of laboursupply for each HH. This change is done for the ITR and GTL simulations.In the aggregate for the ITR simulation for the MI economy, there is nochange in labour supply. This is because the fall in the returns to workingabroad are compensated by an increase in the domestic wage, leaving theaggregate HH wage W unchanged.

b Financial constraints to migration

The theoretical literature has pointed out that outmigration may be con-strained by financial costs, especially for low-income potential migrants(see, for example, Schiff, 1995; Lopez and Schiff, 1995). There is also evi-dence that financing constraints and migration costs can be significant in themigration decision.9 To test the sensitivity of our results to a potential finan-cial constraint, we suppose that low-skill households face a financing con-straint. We assume that only the LS HH faces a migration constraint – i.e.Vh50 for this HH group. This means that this household cannot emigrate(or return home) in response to a change in relative earnings. Introducing afinancial constraint on the migration of LS HH in the TLS for the LIeconomy has drastic effects as the aggregate migration pressure falls from 8per cent to 1 per cent. This is because migratory pressure is from the LS HH.


Table 8.7. Sensitivity analysis: labour supply and home–foreign-labourallocationa

TLab. TFLab. TDLab. Realsup. sup. sup. ER RGDP

ES-1a(M1)b ITR 20.0 20.6 0.1 20.9 0.1ES-8a(LI) GTL 20.2 27.5 1.4 2.6 20.3ES-3b(LI) TLS 0.0 1.1 20.2 22.9 22.4ES-3c(LI) TLS 0.1 8.1 21.6 19.3 23.1ES-5d(MI) ATLN 20.2 25.9 0.8 23.3 0.3ES-3e(LI) TLS 20.1 27.0 1.4 21.7 21.6ES-8e(LI) GTL 20.3 29.8 1.8 3.2 20.1

Notes:a Percentage change from initial values.b Archetype in parenthesis.See table 8.3 for definition of policy experiments.

c Concentrated factor ownership

Though the mapping matrix of factor to household income is guesswork, itrepresents a view of the world in which households have some assetdiversification as they receive both income from activities in several sectorsand from several factors. While this is certainly closer to reality than thetextbook case in which there is no asset-ownership diversification, it is inter-esting to see how sensitive results are to this assumption. To this end, wemodified the factor ownership shares in table 8.2, assuming that the CAPHH receives all its income from the ownership of specific factors. Hence, itdoes not supply labour services, and therefore cannot migrate. In the aggre-gate, increased concentration of factor ownership does not affect migratorypressure. This is so because the CAP HH represents an insignificant shareof migrants. However, this modification leads to a weaker improvement inincome distribution (Gini decrease by 3.4 per cent versus 5.3 per cent).

d Reduction in migration costs

A reduction in migration costs should increase both relocation of labour torelative wage changes and lead to less loss in GDP. In ES-5d(MI) (ATLN),for all households the value of Vh is increased from 1.5 to 2.5. As expected,there is a magnification effect on return migration. Also GDP increasesmore than in the corresponding scenario with higher migration costs.

e Permanent migration

Finally we consider the effects of a household labour-allocation decisionthat does not take into account variations in the purchasing power offoreign wages in terms of domestic goods. As argued in section 2, this cor-responds more closely to a permanent view of migration and is probablymore applicable to LI economies. Hence we consider only the effects of thisoption for the LI economy for two scenarios: TLS and GTL. Now the HOwage effects of trade liberalisation dominate the results and instead of an 8per cent increase in emigration, there is a 7 per cent decrease in foreignlabour supply due the increase in the unskilled wage caused by the expan-sion of unskilled labour-intensive export industries. The same effects are atwork in the GTL experiment with a magnification effect of return migra-tion compared with the results in table 8.4a.

In the simulations and sensitivity analysis, the approach has been taxo-nomic, with variations in only one parameter at a time. If one were tochoose the model most appropriate for a LI and MI economy, one couldargue that a ‘return migration’ model is more relevant for a MI economy


while a ‘permanent migration’ model with a financial constraint forunskilled labour migration is the more relevant model for a LI economy.Since the permanent migration closure results in return migration in theTLS and GTL experiments, the financial constraint cloture would not bebinding. Thus the results from rows 6 and 7 of table 8.7 would probably bethe more representative expected results for a LI economy. One wouldtherefore conclude that direct or indirect channels to stem migration, atleast as modelled in this chapter, would likely achieve their objective ofreducing migratory pressure.

6 Conclusions

Whether direct and indirect measures aimed at reducing South–Northmigratory pressures are effective cannot be determined on a priori grounds,although for direct measures (e.g. foreign aid and tighter controls on migra-tion), in the absence of financial costs to migration, the net impact is mostlikely a reduction in migratory pressure. It is for the indirect measures thatboth theoretical and empirical results are ambiguous. For example, at thetheoretical level, the effect of a more liberal trade regime is ambiguous,since if trade and factor movements are substitutes (complements), thentrade liberalisation will (will not) discourage migration.

This chapter shows that empirically, even in the context of a simple RVmodel, the effects of trade liberalisation are quite complex and could leadto ambiguous effects on migratory pressures. If the responsiveness ofexports to the new set of incentives is strong enough, trade liberalisationwill be accompanied by a real appreciation which could lead to returnmigration, at least if migration is not permanent, a prediction that has beencorrobated by data on Mexican–US migration. However, if exports areslow to respond to the new trade regime, trade liberalisation will be accom-panied by a strong real exchange rate depreciation. This depreciation willincrease migratory pressure. It would thus appear that trade liberalisationin LI economies is more likely to lead to an increase in migratory pressurethan in MI economies.

However, differences in structure and constraints between LI and MIeconomies should also be taken into account. Migration cost constraints forlow-skill workers in LI economies will alleviate migratory pressure. Also, forLI economies, it is likely that the pattern of migratory pressure resulting fromindirect measures such as trade liberalisation will adversely affect the skillcomposition of the labour force, as it is the medium and high-skill workersthat will lose in relative terms and will find migration more attractive.

At a more general level, there are many analogies between the ‘trade andjobs’ controversy and the debate on how to cope with growing migration


pressure. Growing trade and investment links with developing countries areoften faulted for the fall in unskilled wages in the USA and the growth inunemployment in Europe. There is indeed some evidence that trade withlabour-abundant countries has been somewhat inimical to unskilled workersin industrial countries, albeit its effects are limited. Similarly, additionalmigration is typically resisted on the ground that it would lead to a markeddeterioration in the labour market conditions of the unskilled. However, bycreating jobs for the household groups most probably wishing to emigrate,FDI is likely to reduce South–North migratory pressure. The simulations inthis chapter have identified only some of the channels through which policiesand environmental changes might affect migratory pressures from the sendingcountry. At the very least, a more complete analysis should also attempt totake into account adjustments during the transition. It is easy to imagine thatchanges in the human capital stock resulting from out- (or return) migrationwould have effects on investment and hence on transitional growth.

More than the numerical results and the differences across MI and LIarchetypes and differences in institutional settings, what is to be retainedfrom this chapter is that trade and migration policies cannot be assessedseparately. A liberal trade regime is probably likely to lead, under most cir-cumstances, to less migration. Conversely, a shift toward protectionism willadd to migration pressure. Those opposing trade liberalisation with devel-oping countries should reflect on the fact that protection breeds furthermigration, both by discouraging labour-intensive exports in sending coun-tries and by boosting the demand for foreign labour in receiving countries.Moreover, in the light of the current debate, those favouring a cut in foreignaid to developing countries on the basis of its inefficiency should thinkabout its consequences on migratory pressure. Hence industrial countriesface a dilemma. They cannot reduce aid and become more protectionistwith developing countries and still hope to reduce migratory pressures:they must make a choice.

APPENDIX

The appendix specifies the full set of equations of the model sketched in themain text and describes how we assembled the data for each archetypeeconomy.

1 The simulation model

The model is a standard general equilibrium RV model. The economy isdivided into i51, …, N sectors. These include traded and non-traded


sectors with corresponding subscripts t and nt. There are v51, …, V cate-gories of labour and h51, …, H categories of households. Sectors areindexed over i,j, factors of production over v and households over h. Pricesdenoted in foreign currency units have an asterisk – e.g. P*t is the foreign-currency price of tradables. An asterisk is also used to denote labour allo-cated abroad – e.g. L*h (L*v) is the amount of labour by household h (ofcategory v) allocated abroad. Finally a bar over a variable indicates anexogenous variable.

The equations describing the simulation model used in the text are con-tained in table 8A.1. Table 8A.2 lists the endogenous variables and theexogenous variables and parameters. The first block of equations in table8A.1 indicates a multi-factor Cobb–Douglas technology for value-addedand a Leontief technology for intermediates and between intermediatesand value-added ((4), (5)). Factor demands are determined by the first-order conditions for profit-maximisation (6) with (7) determining rents forthe fixed factors. Note the presence of non-competitive imports in thedefinition of net prices (5).

The second block gives the factor-to-household mapping. In the nota-tion used in (8) and (9) lL is a (V,H) matrix indicating the mapping oflabour by category to labour ownership by household. In this applicationthis is a matrix of ones and zeroes as each household only supplies one cat-egory of labour, though it may receive income from more than one factorand some labour categories are supplied by more than one household.Equation (10) gives the mapping of capital from sectors to households.

The third block describes the allocation of household labour betweendomestic and foreign destinations according to the description in the maintext. Equation (11) is the cost function associated with (2). Maximisationof (3) gives the division of labour between home and foreign markets givenby (13) and (14). The wage rate for each category of labour is determinedby (16).

The following block determines net trade from the material balances forthe traded sectors (19) while the material balance equations for non-tradedsectors determines prices for the non-traded sectors (20). All final demandis for household consumption as in the standard trade-theoretic models.The small-country assumption is embodied in eq. (21) and (22). The exo-geneity of the current account (24) insures that there are no free luncheswhile the choice of normalisation (28) insures that the value of ER is theequilibrium value of the real exchange rate (the price of tradables in termsof non-tradables). Finally the determination of household income (25)indicates that government and remittance redistributions do not affect thedistribution of income between households.

In the simulations with FDI, the stock of capital in the agriculturalexport and light manufacturing sectors is exogenously increased with the



Table 8A.1. Model equations

Technology and factor demand

Xi5AiK iai (Ld

i,v)gi,v (4)

gi,v1ai51

PNi5Pi2 ai,jPj2 b0,iP0 (5)

Ldi,v5 (gi,vPNiXi)/wv (6)

Ri5ai(PNiXi)/Ki (7)

Factor-household income mapping

wh5 wvlLv,h (8)

wh*5 wv*l*Lv,h (9)

lKi5 HKi,h (10)

Household home–foreign-labour allocation

Wh51/Bh[(12nh)2Vh(wh)11Vh1(nh)2Vh(wh*ER)11Vh](1/(11Vh)) (11)

LSh5MHh 2(b0,h/Wh)((Yh2fh)/(12b0,h)) (12)

Lh5 (1/bh)(11Vh)((12nh)Wh/wh)2VhLSh (13)

Lh*5(1/bh)(11Vh)(nhWh/wh*ER)2VhLSh (14)

Wage determination

Lsv5 LShl

Lv,h (15)

Lsv5 Ld

i,v (16)

Demand and material balances

Ci,h5di,h1(bi,h/Pi)(Yh2fh)/(12bo,h)

fh5 Pi,hdi,h (17)on

i51

oi

oh

oh

ov

ov

oi

ov

pv


Table 8A.1. (cont.)

bh1b0,h51

IDi5 ai,jXj (18)

Xt5 Ct,h1IDt 1NEt (19)

Xnt 5 Cnt,h1IDnt (20)

Trade and external balance constraint

Pt5Pt*ER(11 Tt) (21)

P05 P0*ER(11T0) (22)

TRB5 P t*NEt 2 b0P0*Xi (23)

CA5TRB1 Lh*wh*ER (24)

Government revenue and income–expenditure constraint

Yh 5LShWh 1 RiHKi,h1sh[GR2CA.ER] (25)

GR52 TtPt*NEtER1 T0b0P0*Xi (26)

sh5 (LShWh 1 RiHKi,h / LShWh 1 RiHKi,h (27)

Price normalisation

PNORM5 (PntX0nt) / P0

ntX0nt (28)2o

nt1o

nt

24oi

3oh

1oi

oi

ot

oi

oh

oi

ot

oh

oh

oj

oi


Table 8A.2. Variables and parameters

Endogenous variables

Pnt Non-traded price ntNEt Net trade tYh Household income HER Real exchange rate lXi Sectoral gross output NPNi Net price NLd

i,v Labour demand for category v N.Vwv Wage for labour category v Vwh Wage for labour category h HRi Rental for fixed factor in sector i NWh Aggregate wage income for household h HLSh Labour supply for household h HLh Home-labour supply for household h HLh* Foreign-labour supply for household h HCi,h Consumption by household h N.Hsh Income share for household h hIDi Intermediate demand for sector i NNEt Net exports for sector t tCA Current account lYh Income for household h HGR Government revenue l

Exogenous variables and parameters

Pt* World price for competitive imports tP0* World price for non-competitive imports lTRB Trade balance lKi Specific factor NMHh Maximum number of hourts available to h Hwh* Household wage abroad HVh Elasticity of transformation hfh Non-discretionary expenditure for household h hdi,h Exogenous consumption of i N.Hbi,h Marginal budget share of i N.Hgi,v Output elasticity for factor v N.Vai Factor-specific share in value-added N

share of profits corresponding to the foreign-owned capital being repatri-ated abroad. In the simulations with a zero elasticity of labour supply to theforeign market VLSK50, L*LSK is fixed, and (14) is dropped for this house-hold.

2 The data

The data for the structure of production, for input–output coefficients andfor the structure of trade are largely inspired from Bourguignon et al.(1992). In turn, these data were assembled from Chenery and Syrquin(1987), who construct what they call ‘archetypical’ primary and manufac-turing exporters (they also consider a third archetype, the ‘large’ economywith a population over 20 million which is less open to trade). Roughlyspeaking, the two archetypes correspond to what Chenery and Syrquin call‘low-income’ and ‘middle -income’ economies. Both economies are cali-brated to be of the same size, but the LI archetype has a much larger shareof resources in agriculture, a less dense input–output structure, and a larger


Table 8A.3. Output and final demand structure

a L1 archetype

Hous. Net Intm. Value- Non-comp.Output cons. exp. sal. added imports

Sub. ag 30.0 17.9 0.0 12.1 18.3 6.0Exp. ag 15.7 20.0 15.7 20.0 11.0 1.9Light mfg. 33.0 24.9 26.6 14.7 15.1 2.3Heavy mfg. 6.4 25.4 29.6 10.6 21.8 1.9Services 47.3 36.4 0.0 10.9 27.2 1.9Informal 20.0 18.0 0.0 22.0 14.7 0.0

b M1 archetype

Hous. Net Intm. Value- Non-comp.Output cons. exp. sal. added imports

Sub. ag 15.5 27.8 24.7 12.3 27.7 0.8Exp. ag 29.0 22.7 3.6 22.7 22.3 0.3Light mfg. 40.7 11.1 8.1 21.5 13.8 2.0Heavy mfg. 24.0 11.9 27.2 19.3 12.4 2.4Services 55.4 37.4 0.0 18.0 33.8 0.6Informal 26.0 20.6 0.0 25.4 15.3 0.0

share of non-competitive imports in total imports (see table 8A.3). Thesame sectoral technology was imposed for both archetypes. Output elastic-ities (and factor shares in value-added for the Cobb–Douglas technologyspecified here) are given in table 8A.4. Except for the sensitivity analysis,the same values for Vh51.5, the elasticity of transformation for labourallocation between home and foreign markets, was set for all householdsfor both archetypes.

Household consumption patterns were calibrated in the usual way start-ing from income elasticities of demand for each household along with esti-mates of the Frisch parameter, Fh (Fh 52Yh/(Yh 2fh) ranging from 23.0for the LS HH to 21.5 for the CAP HH.

NOTES

We thank Sanoussi Bilal, Marcelo Olarreaga, André Sapir, Alessandra Venturiniand Klaus F. Zimmermann for helpful comments on a draft and the FNRS forsupport under grant no. 12-42011.94.1 For evidence of return migration in Southern European countries (Southern

Italy, Spain and Turkey over the period 1962–88, see Faini and Venturini (1995).2 For an empirical justification of the RV model, see Kohli (1993), who showed that

it performs better than the HO framework in explaining the US experience.3 Although this is a static model, the ‘temporary-migration’ view could also be

interpreted as the inclusion of perfect foresight expectations about the exchangerate in the migration decision.

4 It is hard to predict to what extent households would diversify across countriesin the absence of barriers to migration. It is likely, though, that heterogeneitywithin the household would be important enough to insure some diversification,even though one observes agglomerations of ethnic groups (e.g. Chinatowns andLittle Italys) which suggests convex relocation costs.

5 It could be shown that risk-aversion by risk-neutral households receiving sto-chastic incomes from different sources would lead to concave iso-risk loci. Thesecould be viewed as consistent with this approach as households would spreadtheir labour services across destinations to reduce risk.


Table 8A.4. Factor shares in value-added

Low Medium High Capitalskill skill skill stock

Sub. ag 0.5 0.0 0.0 0.5Exp. ag 0.0 0.5 0.0 0.5Light mfg. 0.0 0.25 0.25 0.5Heavy mfg. 0.0 0.0 0.5 0.5Services 0.0 0.25 0.25 0.5Informal 0.25 0.25 0.0 0.5

6 A large LI economy like Bangladesh would have a more diversified economicstructure that would call for a different archetype representation. The choice ofagriculture as import-competing in the MI economy is, of course, arbitrary. Itturns out, however, to be convenient in the simulations capturing the effects oftrade liberalisation as it gives the same number of sectors with net imports forboth archetypes.

7 Neary (1988), discusses the conditions under which a transfer and a change in theterms of trade will lead to an appreciation of the real exchange rate and Edwardsand van Wijnbergen (1987) give conditions under which trade liberalisation in aRV model will lead to a depreciation of the real exchange rate. Neither studyincludes migration, but the reasoning is broadly applicable to the discussion here.

8 A systematic analysis of alternative cloture rules would have included the role ofeconomies of scale in the possibility of generating a complementarity relation-ship between trade and migration. See Venables (1997) for a summary of theresults in the trade-theoretic literature on the substitutability–complementarityrelationship between trade and migration.

9 See, for example, Adams (1996); Freeman (1993); Funkhouse (1992).

REFERENCES

Adams, R. (1996). ‘Remittances, Inequality and Asset Accumulation: The Case ofRural Pakistan’, in D. O’Connor and L. Farsakh (eds.), Development Strategy,Employment and Migration: Country Experiences (Paris: OECD DevelopmentCentre), 149–70

Bhagwati, J. (1992). ‘Free Traders and Free Immigrationists: Strangers or Friends?’,Working Paper, 20 (New York: Russell Sage Foundation)

Bourguignon, F., J. de Melo and A. Suwa (1992). ‘Distributional Effects ofAdjustment Policies: Simulations for Archetype Economies in Africa andLatin America’, World Bank Economic Review, 5, 339–66

Burda, M. (1993). ‘The Determinants of East–West German Migration’, EuropeanEconomic Review, 37, 452–61

Chenery, H. and M. Syrquin (1987). ‘Typical Patterns of Transformation’, in H.Chenery, S. Robinson and M. Syrquin (eds.), Industrialization and Growth(Oxford: Oxford University Press)

Daveri, F. and R. Faini (1996). ‘Where Do Migrants Go? Risk-Aversion, MobilityCosts and the Locational Choice of Migrants’, University of Brescia, mimeo

Djajic, S. and R. Milbourne (1988). ‘A General Equilibrium Model of Guest-Worker Migration’, Journal of International Economics, 25, 335–51

Edwards, S. and S. van Wijnbergen (1987). ‘Tariffs, the Real Exchange Rate and theTerms of Trade: On Two Popular Propositions in International Economics’,Oxford Economic Papers, 39, 458–64

Faini, R. and J.M. Grether (1994), ‘ L’ouverture au commerce, peut–elle réduire lamigration Nord–Sud?’, University of Neuchâtel, mimeo

Faini, R. and J. de Melo (1995). ‘Trade Policy, Employment and Migration: SomeSimulation Results from Morocco’, CEPR, Discussion Paper, 1198

(1994). ‘Migration and Growth: The Experience of Southern Europe’, CEPR,Discussion Paper, 964


Faini, R. and A.Venturini (1995), ‘Trade, Aid and Migration: Some Basic PolicyIssues’, European Economic Review, 37, 435–42

Freeman, R. (1993). ‘Immigration from Poor to Wealthy Countries: Experiencesfrom the US’, European Economic Review, 37, 443–51

Funkhouse, E. (1992). ‘Mass Migration, Remittances and Economic Adjustment:The Case of El Salvador’, in G. Borjas and R. Freeman (eds.), Immigration andthe Work Force: Economic Consequences for the United States and the SourceAreas (Chicago: University of Chicago Press for the NBER)

Harris, J. and M. Todaro (1970), ‘Migration, Unemployment, and Development: ATwo–Sector Analysis’, American Economic Review, 60, 126–42

ILO (1949). Migration for Employment Recommendation (Geneva: ILO)(1984). Employment Policy Recommendation (supplementary provision)

(Geneva: ILO)(1992). ‘ODA as a Means to Reduce Economic and Social Emigration Pressure’,

paper for joint ILO–UNHCR Meeting in International Aid as a Means toreduce the Need for Emigration: Informal Summary Record, Geneva

Kohli, U. (1993). ‘US Technology and the Specific–factors Model’, Journal ofInternational Economics, 34, 115–36

Lopez, R. and M. Schiff (1995). ‘Migration and the Skill Composition of theLabour Force: The Impact of Trade Liberalisation in LDCs’, PRWP, 1493(Washington, DC: World Bank)

Markusen, J. (1983). ‘Factor Trade and Commodity Trade as Complements’,Journal of International Economics, 13, 341–56

Markusen, J. and S. Zahniser (1997). ‘Liberalisation and Incentives for LabourMigration: Theory with Applications to NAFTA’, Boulder, University ofColorado, mimeo

Massey, D.S. (1989). ‘Theories of International Migration: A Survey’, Populationand Development Review, 19, 431–66

Melo, J. de and Tarr, D. (1992). A General Equilibrium Analysis of US Foreign TradePolicy (Cambridge, MA: MIT Press)

Neary, P. (1988). ‘Determinants of the Equilibrium Real Exchange Rate’, AmericanEconomic Review, 78, 211–15

Schiff, M. (1995). ‘Trade Policy and International Migration in the Short and theLong Run’, Revue d’économie du Development, 4, 3–25

(1996). ‘South–North Migration and Trade: A Survey’, PRWP, 1696(Washington, DC: World Bank)

Stark, O. (1991). The Migration of People (Oxford: Basil Blackwell)Venables, A. (1997). ‘Trade Liberalisation and Factor Mobility:An Overview’, LSE,

mimeo; see also chapter 2 in this volumeWong, K.–Y. (1995). International Trade in Goods and Factor Mobility (Cambridge,

MA: MIT Press)


Discussion

ALESSANDRA VENTURINI

Chapter 8 by Faini et al. is very interesting. It adds to the ongoing debateon whether it is possible to reduce migration through trade liberalisation.It begins with the premise that the countries of destination would like, ifpossible, to find a way of reducing immigration pressures. Before I considerthe specific case described in the chapter, it is necessary to point out that ifthis assumption is generally true, in the future it will be less tenable in thecase of Europe, where the population is ageing rapidly. In fact, generationalfactors in Europe will create a short-term demand for immigrants. InNorthern Italy, the area with the highest proportion of aged people, therewill be one-and-a-half jobs available for every young person who enters thelabour market in 2006, because of the number of workers reaching retire-ment age at that date. Thus, if the participation rate does not changedramatically and technological change does not cause the number of avail-able jobs to fall by more than 30 per cent, then there will be a relativelystrong demand for immigrant labour.

Let us return to the Faini et al. model. It provides a useful framework butsuffers some weaknesses.:

• First, as with all simulation models,the results are strictly determinedby the parameters of the specification.

• Second, it explains migratory flows only in terms of supply-side deci-sions, in fact the destination model is very static. For instance, the skillcomposition of emigrants’pull will be determined only by changes in thepattern of production induced by ‘trade policies’adopted in the countryof origin and not by changes in the demand in the country of arrival.

• Last, the time-horizon of the model is not very clear. In principal ageneral equilibrium model is not actually a short-run model; here,however, capital accumulation and intersectoral capital mobility arenot considered, as they should be in a medium-run model.

Still, the model has many interesting features – for instance, the fact thatthe wage differential is expressed in term of the consumption basket in thecountry of origin gives a role to the real exchange rate both affectingemigration and return migration.

A crucial issue from the policy-maker’s point of view is if the model pro-vides an adequate description ofreal cases, for instance, the Mediterraneancase – the main source of immigration to Europe.


In the Faini et al. simulation, any trade and foreign direct investment(FDI) liberalisation in the Mediterranean countries, which are middle-income (MI) countries (figure 8.3b), is found to have a positive effect on thedomestic balance of payment deficit. This means the exchange rate appre-ciates and emigration is discouraged. Such a result is closely linked to thekind of tax system adopted, and if it is mainly concentrated on exports, areduction of taxes will favour exports more than imports, inducing the pos-itive effects already mentioned.

However, at least in the short run, imports will react more swiftly tochanges in incentives and thus may lead to a short-run current accountdeterioration and this is why the researches of Cogneau and Tapinos (1995)for Morocco and Kebabjian (1994) for Tunisia shows after a trade liberal-isation in the South a devaluation of the exchange rate, with the effect ofmaking emigration more attractive, and a rise in the interest rate to com-pensate the balance of payment deficit which reduces investment (see table8.1a, 8.1b). A further example comes from the US–Mexico case where tradeliberalisation introduced by the NAFTA will sweep away traditionalMexican agriculture, increasing urbanisation and emigration out of agri-culture (Martin, 1994).

In Faini et al., when trade flows are less price-elastic, trade liberalisationleads to more migration but only in the low-income (LI) countries’ case.

Income distribution considerations play a crucial role. In the MI coun-tries’ case, medium-skill households are concentrated in exportable agri-culture and light industries (see table 8.2), both sectors which are favouredby the different intervention policies suggested in the model, and the incen-tive to emigrate is clearly reduced. The opposite is true for the other threegroups, which ends in the final case – global trade liberalisation (GTL) –with a higher propensity to emigrate both in LI and MI countries.

It is unlikely that this will occur, because migration is not usually fre-quent in either the lowest- or the highest-skill household, but in themedium-skill one and it is the result of the only positive effect of TradeLiberalisation (TL) and increase of foreign direct investment (FDI) in theSouth where there is a push to emigrate except for the medium-skill house-hold.

One way of getting a more reliable result would be to assume a differentsectoral distribution of households or to allow for the low-skilled con-straints to emigration. The authors are somewhat aware of this fact and inone simulation they introduce the liquidity constraint for low-skill house-hold.

As expected, the best results for migration pressure are obtained in thecase of trade liberalisation in the North (TLN), and no ‘policy’ changesbeing introduced in the South, a result which suggests that any Global free

222 Discussion by Alessandra Venturini

trade agreement (GTL) imposes high restructuring costs on the South andin the short run could also make things worse for the North.

FDI provides us with surprisingly poor results, and it should be pointedout that reductions in the cost of adopting technological advance and thefalls in the risk premium of capital movement, both of which have positivemultiplier effects – as the work of Cogneau et al. (1996) shows for theMediterranean case – have not been considered, while the overall scenariois less optimistic than it might be.

Even with these limitations, the results of the research are importantbecause they combine emigration and trade-development policies within asingle interactive framework. Although migration decisions are morecomplex and do not respond only to wage differentials, the trend of poten-tial migrants is down. Migration will be less the more the Northern coun-tries open their markets to goods produced in the developing countries. Theresult will probably not be reached in the short run, as the model shows, butin the medium term; and potential increases in emigration during thetransition period can be reduced by specific intervention in favour of low-skill workers in the South. Restrictions should also be introduced to reducethe immigration-chain mechanism which fosters mobility even when theincentives to emigrate are disappearing.

REFERENCES

Cogneau, D. and G. Tapinos (1995). ‘Libre-échange, répartition du revenu et migra-tions au Maroc’, Revue d’Economie du Dévelopment,1

Cogneau, D., J.C. Dumant and P. Izzo (1996). ‘Regional Integration, DirectInvestment and Migration in the European Mediterranean Area: ResultsObtained from a Computable General Equilibrium Model’, in Migration, FreeTrade andRegional Integration in the Mediterranean Basin, proceedings from aseminar jointly organised on 31 October and 1 November 1996 by the OECDand the Greek authorities

Kebabdjian, G. (1994). ‘Les Pays du Maghreb ont-ils intéret à une zone de libre-échange avec l’Union Européenne?’, in Vers une zone de libre-échange Europe-Maghreb, cahier du GEMDEV, 22, Paris, 59–77

Martin, P. (1994). ‘Immigration to the United States: Journey to an UncertainDestination’, Population Bulletin, 49


Part ThreeHistorical and contemporaryevidence

9 Were trade and factor mobilitysubstitutes in history?

WILLIAM J. COLLINS, KEVIN O’ROURKE

AND JEFFREY G. WILLIAMSON

Did the growth of protection in the late nineteenth century in NorthAmerica stimulate the large labour and capital inflows of that period(assuming land to have been the abundant factor)? Did the increased pro-tection in Britain in this century stimulate capital export? Did the break-down in international factor movements in the interwar period stimulatetrade? (Mundell, 1957, p. 335)

1 Introduction

Robert Mundell’s seminal article on ‘International Trade and FactorMobility’ was published 40 years ago, and its conclusion threw down achallenge: Can history tell us whether (and when) trade and factor mobil-ity are substitutes? 40 years seems long enough to wait, so this chapterfinally rises to Mundell’s challenge by interpreting trade theory in the lightof history and by investigating empirically the connection between tradeand factor flows in the Atlantic economy between 1870 and 1940. Theepoch prior to the First World War is especially interesting in this regardsince it is a liberal world environment providing a good natural experiment.

In the 40 years since Mundell wrote, trade theorists have come to under-stand that theory is ambiguous on this issue. Whether trade and factormobility are complements or substitutes depends on the assumptions madein the theory. If ever there was an open invitation for empirical research,this surely is it. Yet, few efforts have been made to identify econometricallythe complementarity or substitutability between factor flows and interna-tional trade (see Wong, 1988). Perhaps history may prove to be a valuableguide to choosing among these assumptions. But if theory is ambiguous,history may be, too: the appropriate model for one historical period maynot be the appropriate model for another. After all, there may be regimeswitches in history which correspond to alternative assumptions whichtrade theorists invoke to yield complementarity or substitutability.Nonetheless, even if history cannot identify the ‘right’ model for thinking

227

about international trade and factor mobility in the 1990s, it certainly canprovide valuable insights into the nature of the connection and, by exten-sion, illuminate important aspects of the contemporary phenomenon.

The first section of this chapter briefly discusses the theoretical ambigu-ity concerning trade and factor mobility within relatively simple models: itoverlaps somewhat with Anthony Venables’ chapter 2 in this volume, butwe need to cover the same ground independently so as to motivate the eco-nomic history. The remainder of the chapter will use that theory to explorethe experience of three New World countries and seven Old World coun-tries between 1870 and 1940. We approach the data from two angles. First,we focus on business cycles and what the literature of the 1950s and 1960scalled ‘long swings’, the latter exhibiting booms and busts over 15–20-yearperiods. Our interest is whether there was strict substitutability over thelong swing between trade on the one hand and capital and migration flowson the other and, if not, whether there is any evidence of complementarity;whether the resource-abundant New World and resource-scarce Old Worldobeyed the same laws of complementarity or substitutability; and whetherthere was a regime switch around the First World War, after which frontiersare said to have ‘disappeared’ and the world economy de-globalised.Second, we investigate the trade versus factor-flow relationship usingdecade-averaged panel data, thus shifting the focus from shorter-run rela-tionships generated by macro-instability in the time series to longer-runrelationships manifested over decades and across trading partners of widelydifferent endowments. Finally, the chapter turns to policy experience, andasks: Did New World policy-makers act as if they viewed migration andtrade as substitutes?

2 Theory

In order to guide the economic history in this chapter, this section willdiscuss whether and when trade and factor mobility should be substitutesor complements within the context of standard trade theory. Venables’contribution to this volume (chapter 2) tackles the question by deriving theeffects of (possibly small) reductions in the costs of trade or factor mobil-ity. We are interested in the same issues but, for the sake of simplicity, con-sider more dramatic changes in transport costs. Throughout, we areinterested in two questions: First, when countries move from autarchy tofree trade, does the incentive for factors to migrate increase or diminish?Second, when countries move from factor immobility to active participa-tion in global factor markets, does the volume of trade increase or dimin-ish? Unlike chapter 2, we only consider constant-returns-to-scale (CRS)models, so that trade is due solely to endowment or technology differences

228 William J. Collins et al.

between countries. This is more likely to have been true for our period thanfor the late twentieth century; Estevadeordal (1993) shows that endow-ments do a very good job of explaining international trade on the eve of theFirst World War.

2.1 232 models

The relationship between trade and factor mobility in the standard 2x2model is well known. (We refer to the 2x2 model rather than theHeckscher–Ohlin (HO) model, since Heckscher and Ohlin had in mind afar richer structure than that developed by Lerner, Samuelson, Rybczynskiand the rest of their followers.) Let Britain and Ireland produce machinesand textiles, using labour and capital. Textile production is relativelylabour-intensive, and Ireland is relatively labour-abundant. Wages are ini-tially low in Ireland and high in Britain, while profits are higher in Ireland.When trade opens up, Ireland exports textiles and imports machines.Wages rise in Ireland and fall in Britain, while profits fall in Ireland and risein Britain, implying that trade induces factor-price convergence (Ohlin,1924, reprinted in Flam and Flanders, 1991). Under restrictive circum-stances, there will be complete factor-price equalisation (FPE) (Heckscher,1919; Lerner, 1952; Samuelson, 1949). Trade reduces or completely elimi-nates the incentive for labour to move to Britain and for capital to flow toIreland; trade and factor mobility are therefore substitutes.

Alternatively, it is clear that factor mobility tends to equalise endow-ments across countries, raising Irish capital–labour ratios and loweringBritish capital–labour ratios. Factor mobility thus erodes the basis for trade(Mundell, 1957). Once again, the conclusion is that trade and factor mobil-ity are substitutes.

As Markusen (1983) has shown, these classic results are derived from theassumption that factor endowments alone are the basis for trade: countriesare assumed to be otherwise identical. If the basis for trade is some otherdifference between countries, trade and factor flows can be complements,even in the context of 232 models. Markusen’s most intuitive demonstra-tion of this principle concerns differences in technology between countries.Now let Ireland and Britain have identical factor endowments, but also lettechnology in the British machine industry be (Hicks-neutral) superior tothat in the Irish machine industry. Britain will have a comparative advan-tage in machine production. When trade opens up, Britain will exportmachines: an increase in capital-intensive machine production will increasethe demand for capital in Britain, hence raising British profits and lower-ing British wages. When Ireland specialises in textiles, Irish wages rise andIrish profits fall. Markusen shows that in trade equilibrium, wages are

Were trade and factor mobility substitutes in history? 229

higher in Ireland, while profits are higher in Britain.1 Now allow factors toflow in response to factor-price differentials. Capital will flow from Irelandto Britain, and labour from Britain to Ireland. In line with the Rybczynskitheorem (1955), machine production in Britain will expand, and Britishtextile production will contract, while the opposite production responseswill occur in Ireland. This will further increase British exports of machinesand imports of textiles. Factor mobility has increased trade. Thus, tradeand factor mobility are complements.

2.2 The specific-factors and 332 models

Heckscher and Ohlin were motivated by the intercontinental exchange offood for manufactured goods that characterised the late nineteenth century.Land-scarce and labour-abundant Europe exported manufactures to theresource-abundant, labour-scarce New World, in exchange for grain, wool,raw cotton and other agricultural products. Obviously, it is difficult to ration-alise such trade in the context of simple 232 models: land, labour and capitalwere all relevant factors of production. In 1966, an economic historian, PeterTemin, wrote a chapter on ante bellum Anglo-American industry in which heassumed that manufactured goods were produced with labour and capital,while food was produced with labour and land. This model was fully devel-oped by Jones (1971), and dubbed the specific-factors model of trade. Wedraw attention to the cliometric antecedents of this model not just out ofteam spirit, but in order to highlight a theme of this chapter: the ‘correct’trade model may vary with the period being studied.

The relationship between trade and factor mobility in the specific-factorsmodel is ambiguous. Imagine two economies, Europe and America, and tokeep matters simple, assume that they differ only in their land endowments:America has more land. What does this imply for American autarchicfactor prices, relative to European factor prices?2

One way to answer this question is to distinguish between factor mobil-ity responses and price effects due to differences in land availability (fol-lowing Corden and Neary, 1982).3 With more land, the marginal productof labour in agriculture increases. Wages increase in both sectors, as labourmoves into agriculture, and profits as well as rents decline. On the otherhand, more land increases the supply of food relative to manufacturedgoods. This will lower the relative price of food, tending to lower rents butincrease profits. American rents will definitely be lower than Europeanrents, but American profits may be higher or lower than profits in Europe.

Profits in the USA will be higher if the manufacturing sector’s labour-force and output are larger than Europe. More land and higher incomesmean a greater demand for manufactures, while a lower relative price of


food implies the opposite. The net impact of an increased land endowmenton profits thus depends on the relative size of these income and substitu-tion effects.

Thus, there are two possibilities. Either land and capital are cheaper inthe USA than in Europe (while labour is more expensive); or US land ischeap, while US capital and labour are expensive (the dual-scarcity case).In the first case, US capital should flow to Europe, while European labourshould migrate to the USA; in the second (dual-scarcity) case, there is anincentive for both labour and capital to flow from Europe to America.

What does a move from autarchy to free trade imply for these migrationincentives? The USA exports food, and the relative price of food in the USAincreases. This increases US rents, and lowers US profits; meanwhile, tradeincreases European profits, but lowers European rents. In the first case(land and capital cheaper in the US) the incentive for US capital to emigratehas increased. Thus, trade and factor mobility are complements. In thesecond case (cheap US land but expensive US capital), the incentive forEuropean capital to emigrate has declined. In effect, trade and factormobility are substitutes. As for labour, the impact of relative price changeson real wages is ambiguous: trade could be a substitute or a complement tomigration.

The same ambiguity emerges where the impact of factor flows on tradeis the issue. Labour migration’s impact on trade is ambiguous because itdepends on the migration’s impact on relative outputs in Europe and theUSA. Similarly, if capital is cheaper in the USA than in Europe, then capitalflows from America to Europe will increase trade (lowering relative suppliesof food in Europe, and increasing them in the USA); but in the ‘dual-scarcity’ case, capital flows to the USA will increase US manufacturing anddiminish US agriculture, lowering the volume of trade. This ambiguity hasbeen rigorously demonstrated by Neary (1995).

Clearly, the relationship between trade and factor mobility is ambiguousin the specific-factors model. Indeed, from the above discussion it is clearthat trade could be complementary to migration, but a substitute forcapital flows, or vice versa. To take just one example from a range ofpossibilities, in the dual-scarcity case, capital flows and trade are sub-stitutes; this does not preclude the possibility that emigration wouldincrease the supply of manufactures relative to food in Europe, and thatimmigration would increase the supply of food relative to manufactures inthe USA (in which case, migration and trade are complements). Goingfrom the specific-factors model to a more general 332 model (in whichland and capital are mobile between sectors) only increases the range ofuncertainty. Thompson (1985, 1986), for example, has shown that pricemovements can have paradoxical effects on rents and profits.


2.3 Frontier arguments

The existence of a third factor like land and other natural resources is notenough per se to generate a strong presumption of complementaritybetween trade and factor mobility. Such a presumption can emerge in thecontext of models where the effective supply of the third factor is endoge-nous – a natural assumption in the context of late nineteenth-century NewWorld frontier economies or of labour-scarce Southeast Asian colonies. Inparticular, as food prices rise due to trade, there is an incentive to extendthe frontier and to increase the effective supply of land. However, extend-ing the frontier requires heavy capital inputs (e.g. financing the railroads)and even labour (e.g. building the new towns). When the USA opens totrade, and the relative price of food increases, the specific-factors modelmay conclude that US profits fall. In the dual-scarcity case, this lessens theincentive for capital to flow from Europe. However, if the increase in thedemand for capital associated with frontier extension is not completelyoffset by the decline in manufacturing’s demand, then the return to capitalin the USA, and even the US–Europe profit differential, might increase. Inthis case, the incentive for capital to flow to the USA would also increase.Similarly, if extending the frontier sufficiently increased the US demand forlabour, this might create the presumption that US–Europe wage gapswould also increase. The existence of an endogenous frontier seems toincrease the likelihood that trade and factor mobility will be complements,rather than substitutes. However, the elegant and simple models of anendogenous New World frontier which Findlay (1995) has constructedshow that even in this case, trade and factor mobility can be substitutes.

Very stylised models of North–South trade can produce unambiguouscomplementarity. If the North exports manufactures and imports naturalresources, and if northern investment in the South is required for theexploitation of the South’s resources, then complementarity is assured.Such a model was outlined by Kemp and Ohyama (1978), even though theywere interested in a different set of issues. Schmitz and Helmberger (1970)argued strongly for complementarity between capital flows and trade, onthe basis that capital inflows are required to develop the exports of extrac-tive industries in LDCs. Williams (1929) also argued that capital mobilityand trade in resources were complements rather than substitutes. Thisargument resembles the frontier argument just put forward: in both cases,there are surplus resources which require foreign investment before theycan be vented onto world markets.

The complementary frontier argument can be extended to a third-worldcontext, too. City-building is very capital-intensive, and to the extent thatglobalisation generates an export boom of labour-intensive manufactures


in the third world, it also induces city growth (Kelley and Williamson, 1984;IBRD, 1995). This, in turn, generates heavy social overhead demand, highreturns to capital and foreign investment (Lewis, 1978). Thus, trade andcapital flows are complements, not substitutes.

Of course, this argument has its reverse. With the disappearance of nine-teenth-century frontiers, town-building slowed down and railroad andother ‘population-sensitive’ investment dropped off, and this, in turn, mayhave caused international factor mobility to subside, complementarity tovanish and substitutability to emerge. Likewise, with the completion oftwentieth-century urban social overhead in the third world, factor mobil-ity may subside, complementary may vanish and substitutability mayemerge.

Three morals emerge from this survey of what standard theory has to sayabout the substitutability versus complementarity question. First, theory isambiguous, as chapter 2 confirms. The ambiguity can be resolved only withevidence, and history is certainly a good place to look for that evidence.Second, endogenous-frontier models seem to offer a clear and historicallyrelevant case where trade and factor mobility are more likely to be comple-ments. To the extent that frontier settlement is a transition process, weshould look to disequilibrium transitions for evidence of complementarity.A transition has ended when a frontier has disappeared or an economy isfully urbanised. Third, the ‘correct’ model will vary with the period beingstudied. Regime switches are likely to be common, especially at the begin-ning and the end of a transition.

With this brief survey of theory to guide us, we are ready to look at pre-Second World War history. We start with macro-instability and the longswing.

3 Time series: macro-instability and the long swing, 1870–1940

In 1968, Moses Abramovitz gave a University Lecture at the LondonSchool of Economics on ‘The Passing of the Kuznets Cycle’, that specialform of macro-instability which characterised so much of the Atlanticeconomy prior to 1914. Abramovitz preferred to call these events Kuznetscycles, after their discoverer (Kuznets, 1930, 1958). But these 15–20-yeargrowth surges and slowdowns were also called ‘long swings’ and, unlessinterrupted by war, they typically ended in severe depressions (e.g. the mid-1870s, the mid-1890s, and the 1930s). Abramovitz and other scholars(Lewis and O’Leary, 1955; Abramovitz, 1961; Easterlin, 1966, 1968) docu-mented long swings in the rate of frontier settlement, accumulation, labour-force growth, productivity advance, city-building and other underlyingfundamentals associated with long-run growth.


These booms and busts were global in scope. Given that primary productmarkets always exhibit greater instability during global booms and busts, itis not surprising that the long swings were most dramatic in the resource-abundant New World. Nonetheless, they were also apparent in much ofEurope where they generated waves of urbanisation and industrial growth,typically producing Old World long swings in domestic accumulation rateswhich moved inversely with those in the New World.

The global character of these long swings also manifested itself in factorflows, relative prices and, to some extent, trade. The pattern appeared in therhythm of the mass migrations, which recorded spectacular flows fromEurope to the New World during booms, and net return migration duringbusts (Thomas, 1941; Thomas, 1954; Easterlin, 1961; Wilkinson, 1967;Williamson, 1974; Hatton and Williamson, 1997). It also appeared as wavesin net international capital movements, which recorded massive outflowsfrom Germany, France and especially Britain during booms abroad(slumps at home) and a flight back home during slumps abroad (booms athome) (Cairncross, 1953; Williamson, 1964). It appeared in relative prices(Abramovitz, 1961; Williamson, 1964; Rostow, 1978) and, for someprimary product exports at least, in trade. Transport costs fell globally soevery trading partner could have enjoyed an improved terms of trade priorto the First World War. But around that trend, the New World terms oftrade improved during New World booms, while it slumped in Europe. Wehave always thought that the growth in primary product exports from theNew World surged and collapsed over these long swings, but in the long runcommodity market globalisation was achieved (Williamson, 1996). Figures9.1–9.4 illustrate these ‘long swings’ for the two most important economiesin the Atlantic community, the USA and the UK, where the epochs arebroken up into the pre-war globalisation phase and the post-war de-glob-alisation phase. These figures show clearly the enormous instability in thetime series, that migration (mig) and capital flows (rlCA) are highly corre-lated (capital chases after labour), but that aggregate trade levels (rltrade)are, somewhat surprisingly, not well correlated with either.

Designing an empirical framework to capture the complementarity orsubstitutability of trade and factor flows is challenging, and the dearth ofempirical investigations of this phenomenon must be due in large part tothe seeming intractability of the problems at hand: issues of endogeneity,measurement and interpretation are formidable obstacles. Undaunted, wepursued two means of measuring the association of trade volumes withfactor flows in each country’s time series. First, we simply asked whetherrelatively large (small) factor flows are associated with relatively small(large) trade flows ceteris paribus, regardless of the direction of the factorflows. The focus in this first approach was on magnitudes alone, and the


absolute values of factor flows were entered in the regressions. But with agiven endowment, should a factor inflow and a factor outflow of the samemagnitude be expected to have the same impact on trade volumes? It couldbe argued that if trade and factor flows are substitutes, then emigration (orcapital exports) is expected to lessen trade only if labour (or capital) isleaving a relatively labour- (or capital-) abundant country. Our secondapproach incorporated a notion of factor scarcity and abundance based onmeasurements of land, labour and capital for each country in the sample.This approach asked whether factor flows in the ‘right’ direction (into afactor-scarce country or out of a factor-abundant country) were associated


Figure 9.1 Long swings in trade and factor flows: USA, 1872–1913Note: All series are three-year moving averages and rescaled for presentation.

Figure 9.2 Long swings in trade and factor flows: USA, 1920–39Note: All series are three-year moving averages and rescaled for presentation.

with more or less trade. The determination of factor abundance/scarcitywas based upon the country’s share in the sample’s total endowment of agiven factor relative to the country’s share in the sample’s total GDP.4 Sinceit turned out that these two approaches yielded much the same results, wepresent only the first in what follows.

Table 9.1a reports results when real trade values are regressed onabsolute real values of factor flows (gross immigration or emigration migand net capital flows as measured by the current-account balance rlCA),plus additional variables to be discussed in a moment; table 9.1b reportsresults when first differences in trade values are regressed on first differencesin factor flows; and table 9.1c reports results when first differences in trade


Figure 9.3 Long swings in trade and factor flows: UK, 1870–1913Note: All series are three-year moving averages and rescaled for presentation.

Figure 9.4 Long swings in trade and factor flows: UK, 1920–37Note: All series are three-year moving averages and rescaled for presentation.


Table 9.1. Factor mobility correlates with trade, 1870–1913 and 1919–36

a Dependent variable: rltrade; factor mobility variables: ambig, abrlCA

Migration Curr. act Trans. cost Tariff N R2

AUS: 1870–1913 0.169 20.584 2000.021 2162.6 32 0.95(3.11) (2.49) (0.14) (1.97)

AUS: 1919–36 0.681 21.06 20.274 2271.9 15 0.78(2.24) (2.38) (0.66) (3.12)

CAN: 1870–1913 0.810 20.219 2.11 10812. 13 0.96(1.18) (0.53) (0.51) (1.52)

CAN: 1919–36 0.121 20.075 32.11 4649. 17 0.22(0.02) (0.05) (1.13) (0.25)

USA: 1870–1913 0.882 0.063 5.01 22530. 42 0.95(4.29) (0.31) (1.07) (1.53)

USA: 1919–36 22.73 3.70 68.07 226639. 17 0.65(1.18) (2.84) (1.33) (2.33)

DEN: 1870–1913 1.37 0.263 1.08 21515. 37 0.68(0.31) (1.09) (1.00) (1.60)

DEN: 1919–36 18.94 3.17 9.79 243684. 13 0.79(1.09) (2.60) (1.32) (3.66)

FRA: 1870–1913 .— 0.153 7.04 231405. 43 0.72(0.94) (0.87) (3.55)

FRA: 1919–36 .— 20.501 2150.2 268824. 16 0.88(0.49) (2.23) (5.18)

GER: 1870–1913 3.39 20.511 20.01 216617. 33 0.61(0.90) (1.56) (1.78) (0.97)

GER: 1919–36 115.14 1.40 175.62 285107. 10 0.99(8.95) (5.97) (7.09) (45.21)

ITA: 1870–1913 5.93 20.580 24.02 258523. 43 0.94(2.02) (2.29) (0.15) (7.92)

ITA: 1919–36 11.78 1.68 71.63 251277. 17 0.84(1.97) (3.19) (1.51) (2.06)

NOR: 1870–1913 0.0001 0.193 20.0377 2759.3 29 0.59(0.14) (0.71) (0.06) (2.18)

NOR: 1919–36 0.0071 21.69 4.18 24892. 17 0.67(1.10) (2.59) (3.54) (3.08)

SWE: 1870–1913 0.874 0.337 2.85 23017. 43 0.88(0.91) (1.05) (2.50) (2.41)

SWE: 1919–36 41.48 1.84 21.33 212551. 15 0.80(1.84) (2.14) (0.17) (2.68)

UK: 1870–1913 0.427 20.0001 20.262 23703. 43 0.94(4.38) (0.00) (0.44) (3.03)

UK: 1919–36 3.34 0.803 25.72 21934 17 0.90(4.12) (1.04) (2.17) (3.26)

Notes:The dependent variable (rltrade) is the deflated sum of imports and exports. Themigration variable (abmig) measures the absolute value of gross immigration for


Table 9.1. (cont.)

a (cont.)

New World countries (except Australia which reports net immigration) or grossintercontinental emigration for Old World countries. Adequate migration data arenot available for France. The current account variable (abrlCA) measures theabsolute value of the deflated overall current balance. The transport cost measureis the same for all countries. The tariff measure is customs revenue divided byimport values. Time and time-squared terms are not reported in the table. Allregressions are Cochrane–Orcutt. t-statistics are in parentheses.

b Dependent variable: Change in rltrade; factor mobility variables: change inabmig, abrlCA


AUS: 1870–1913 0.111 20.0636 0.238 2152.46 29 0.24(1.22) (0.20) (0.12) (1.52)

AUS: 1919–36 20.0877 20.383 20.222 2110.47 14 0.30(0.20) (0.93) (0.58) (1.42)

CAN: 1870–1913 0.499 20.205 20.267 4507 12 0.12(0.58) (0.38) (0.04) (0.44)

CAN: 1919–36 3.04 20.691 30.40 214819 17 0.22(0.37) (0.52) (1.39) (0.82)

USA: 1870–1913 0.894 20.0070 0.317 22226 41 0.40(4.02) (0.04) (0.07) (1.23)

USA: 1919–36 0.163 1.95 23.85 240327 17 0.71(0.07) (2.58) (0.31) (2.73)

DEN: 1870–1913 1.51 0.242 1.04 21785 35 0.44(0.34) (1.01) (0.95) (1.86)

DEN: 1919–36 12.73 4.00 12.37 252751 11 0.85(0.62) (2.63) (1.28) (3.63)

FRA: 1870–1913 .— 0.277 8.08 235070 42 0.40(1.62) (0.99) (3.91)

FRA: 1919–36 .— 0.408 211.03 29280 15 0.29(0.41) (0.16) (0.42)

GER: 1870–1913 5.53 20.459 21.17 219254 32 0.46(0.08) (1.37) (1.92) (1.10)

ITA: 1870–1913 2.12 0.0988 222.58 213737 42 0.22(0.81) (0.37) (0.89) (0.86)

ITA: 1919–36 13.60 1.35 248040 31.48 17 0.83(2.42) (2.92) (1.88) (0.73)

NOR: 1870–1913 0.0001 20.154 0.0835 2718 27 0.40(0.16) (0.54) (0.13) (1.79)


Table 9.1. (cont.)

b (cont.)


NOR: 1919–36 0.0079 0.914 1.45 24820 17 0.24(0.85) (0.96) (0.61) (1.45)

SWE: 1870–1913 0.686 0.424 1.97 22837 42 0.32(0.72) (1.39) (1.77) (2.54)

SWE: 1919–36 35.14 1.99 23.73 211017 14 0.79(1.07) (2.54) (1.41) (1.97)

UK: 1870–1913 0.436 0.129 20.373 24218 42 0.59(4.41) (0.52) (0.61) (3.10)

UK: 1919–36 1.76 0.549 1.97 22872 17 0.72(2.14) (0.65) (1.41) (2.64)

Notes:The dependent variable (change in rltrade) is the first difference of the deflatedsum of imports and exports. The migration variable (change in abmig) measuresthe first difference of the absolute value of gross immigration for New Worldcountries (except Australia which reports net immigration) or grossintercontinental emigration for Old World countries. Adequate migration data arenot available for France. The current account variable (change in abrlCA)measures the first difference of the absolute value of the deflated overall currentbalance. Data limitations preclude estimation for inter-war Germany. Thetransport cost measure is the same for all countries and is also first-differenced.The tariff measure is the first difference in customs revenue divided by importvalues. Time and time-squared terms are not reported in the table. All regressionsare Cochrane–Orcutt. t-statistics are in parentheses.

c Dependent variable: change in rltrade; factor mobility variables: abmig and abrtlCA


AUS: 1870–1913 20.0083 20.0821 0.124 2140.6 29 0.19(0.15) (0.36) (0.59) (1.37)

AUS: 1919–36 0.250 20.819 0.010 2119.16 15 0.76(1.04) (4.00) (0.04) (3.01)

CAN: 1870–1913 20.460 0.713 28.86 213040 13 0.53(0.76) (2.54) (1.81) (1.43)

CAN: 1919–36 0.126 0.728 19.29 24212 17 0.20(0.02) (0.59) (1.22) (0.39)

USA: 1870–1913 0.383 0.032 21.16 22982 41 0.18(1.60) (0.13) (0.23) (1.48)


Table 9.1. (cont.)

c (cont.)


USA: 1919–36 27.42 1.053 0.967 213678 17 0.45(2.08) (0.51) (0.02) (1.23)

DEN: 1870–1913 2.05 20.163 1.52 21853 37 0.43(0.72) (0.65) (1.57) (2.04)

DEN: 1919–36 31.07 2.09 9.92 231387 13 0.74(1.43) (1.69) (1.50) (3.41)

FRA: 1870–1913 .— 0.183 3.75 230751 42 0.28(1.48) (0.47) (3.30)

FRA: 1919–36 .— 0.709 211.16 29198 15 0.29(0.54) (0.17) (0.42)

GER: 1870–1913 2.75 0.600 14.57 246931 32 0.57(1.13) (1.62) (1.36) (2.82)

GER: 1919–36 119.2 0.704 169.7 249832 9 0.97(3.27) (1.01) (3.14) (4.84)

ITA: 1870–1913 2.24 0.182 212.68 28790 42 0.23(1.29) (1.02) (0.51) (0.57)

ITA: 1919–36 32.87 20.116 30.96 279047 17 0.58(2.22) (0.12) (0.30) (2.21)

NOR: 1870–1913 20.0003 20.356 0.0011 2632.29 27 0.52(0.56) (1.96) (0.002) (1.68)

NOR: 1919–36 0.020 21.31 24.12 22585 17 0.71(2.39) (2.09) (3.55) (1.18)

SWE: 1870–1913 0.676 20.306 2.16 23148 42 0.33(1.06) (1.14) (1.87) (2.80)

SWE: 1919–36 45.28 20.216 28.86 213153 14 0.93(3.57) (0.19) (3.72) (3.91)

UK: 1870–1913 0.215 0.383 20.372 24846 43 0.37(1.87) (1.86) (0.52) (2.94)

UK: 1919–36 20.867 1.58 0.501 24900 17 0.63(0.63) (1.18) (0.29) (3.71)

Notes:The dependent variable (change in rltrade) is the first difference of the deflated sum ofimports and exports. The migration variable (abmig) measures the first absolute valueof gross immigration for New World countries (except Australia which reports netimmigration) or gross intercontinental emigration for Old World countries. Adequatemigration data are not available for France. The current account variable (abrlCA)measures the absolute value of the deflated overall current balance. The transportcost measure is the same for all countries. The tariff measure is the customs revenuedivided by import values. Time and time-squared terms are not reported in the table.All regressions are Cochrane–Orcutt. t-statistics are in parentheses.

values are regressed on factor flows. The dependent trade variable is simplythe deflated sum of exports and imports (rltrade).5 All of the regressions usethe Cochrane–Orcutt procedure to account for serial correlation in theerror terms. The results in tables 9.1a–9.1c appear to be consistent.Furthermore, the results seem to be robust to many other specifications wetried.6

The regressions were estimated with time and time squared (coefficientsnot reported in tables 9.1a–9.1c) to isolate macro-instability (e.g., longswings) from the trends in the country time series.

Tariff rates (tariff) were introduced to reflect the fact that many of thesecountries retreated from globalisation in the late nineteenth century, andthat all of these countries participated in the race towards de-globalisationafter the First World War.7 We expected negative coefficients here, and ourexpectations are confirmed (with the odd exception of Canada). We alsointroduced a crude measure for transport costs (Isserlis, 1938);8 it almostalways has the wrong sign in these long- swing regressions probably becausetransport costs were endogenous to trade volumes over the long swing,rising during booms as space in tramp bottoms got scarce, and fallingduring slumps as space in tramp bottoms got abundant. We expect differentsigns on transport costs when these cycles and swings are removed from thedata, and thus where exogenous supply-side forces are more likely to dom-inate.

The impact of the frontier and/or policy change is explored by estimat-ing separately relationships for the pre-war years (1870–1913) and for theinter-war years (1919–39). Our expectation was that complementaritywould appear more frequently in the pre-war decades than afterwards,especially for the resource-abundant New World.

Most important for the purposes of this study are the factor-flow vari-ables. Migration measures gross intercontinental emigration from theEuropean countries and gross immigration to the New World countries(except for Australia, which reports net in-migration). The current-accountvariable is the overall current balance, deflated by a GDP deflator for thesake of intertemporal comparability. True, no respectable economic histo-rian could possibly hold the view that mass migration (Hatton andWilliamson, 1998) and international capital flows (Zevin, 1992; Taylor,1996) were exogenous, and to the extent that endogeneity exists (forexample, as a feedback from trade to the factor flows) the estimatedcoefficients on the factor-flow variables will be confounded. Yet, it is fareasier to make that assertive statement than to construct instrumental vari-ables annually to deal with endogeneity. In any case, we are not arguing thatthe regressions identify the causal mechanism running from factor mobil-ity to trade. Rather, we are only exploring correlations, and tables 9.1a–9.1care titled accordingly.



Table 9.2. Trade and factor mobility, time-series complements orsubstitutes?

a c or s is determined at the 10 per cent level of significance from Table 9.1a

1870–1913 1919–36

Labour Capital Labour Capital

AUS c s c sCAN n n n nUSA c n n cDEN n n n cFRA — n — nGER n n c cITA c s c cNOR n n n sSWE n n c cUK c n c n

Source: Taken from table 9.1a. A positive and significant (at the 10 per cent level)coefficient relating trade to the factor flow variable in table 9.1a is a designated‘complementary’. A negative and significant (at the 10 per cent level) coefficienton the factor-flow variable is designated ‘substitutable’. Insignificant coefficientsare designated ‘neutral’.

b c or s is determined at the 10 per cent level of significance from Table 9.1b

1870–1913 1919–36


AUS n n n nCAN n n n nUSA c n n cDEN n n n cFRA — n — nGER n n — —ITA n n c cNOR n n n nSWE n n n cUK c n c n

Source: Taken from table 9.1b. A positive and significant (at the 10 per cent level)coefficient relating trade to the factor flow variable in table 9.1b is a designated‘complementary’. A negative and significant (at the 10 per cent level) coefficienton the factor-flow variable is designated ‘substitutable’. Insignificant coefficientsare designated ‘neutral’.

With these caveats firmly in mind, we next explore those results most rel-evant to the central question, namely: What are the coefficients on thefactor mobility variables? Tables 9.2a–9.2c summarise the results in termsof the hypothesis being tested: complements or substitutes? Deriving anytable 9.2 from the relevant table 9.1 is straightforward: a positive and sig-nificant coefficient is interpreted as complementarity, a negative and sig-nificant coefficient is interpreted as substitutability and any coefficient notstatistically different from zero is deemed neutral.9

Neutral entries dominate all three tables (53, 78 and 68 per cent in tables9.2a, 9.2b and 9.2c, respectively, or 66 per cent overall), suggesting that thetheoretical ambiguities highlighted above find their match in empiricalambiguities. Nevertheless, complementarity is far more common than sub-stitutability, and it accounts for 26 per cent of the entries in all three tablescombined. In addition, complementarity is associated almost twice as oftenwith migration (34 per cent) than with capital flows (19 per cent).Substitutability appears in only 8 per cent of the cases, and all but one ofthose cases involves capital flows, not migration. This distinction betweencapital flows and migration suggests that some factor flows are more closelyrelated to trade than others.


Table 9.2. (cont.)

c c or s is determined at the 10 per cent level of significance from Table 9.1c

1870–1913 1919–36


AUS n n n sCAN n c n nUSA n n s nDEN n n n nFRA — n — nGER n n c nITA n c c nNOR n s c sSWE n n c nUK c s n n

Source: Taken from table 9.1c. A positive and significant (at the 10 per cent level)coefficient relating trade to the factor flow variable in table 9.1c is a designated‘complementary’. A negative and significant (at the 10 per cent level) coefficienton the factor-flow variable is designated ‘substitutable’. Insignificant coefficientsare designated ‘neutral’.

So far, history soundly rejects substitutability. However, there are twounpleasant surprises that also appear in tables 9.2a–9.2c. First, and con-trary to prediction, complementarity was not more pervasive in theresource-abundant New World (19 per cent in all three tables combined)than in the resource-scarce Old World (29 per cent in all three tables com-bined). Second, and again contrary to prediction, complementarity wasthree times more common in the ‘post-frontier’ period: complementarityoccurred 36 per cent of the time during 1919–39 but only 12 per cent of thetime during 1870–1913.

Our priors strongly suggested complementarity and regime switch awayfrom complementarity and towards substitutability. Much to our sur-prise, the evidence rejects the latter in the sense that it supports a regimeswitch towards complementarity, when the ‘disappearing frontier’ thesispredicted a regime switch from complementarity. Perhaps the comple-mentarity is actually driven by policy, the simultaneous breakdown ofinternational commodity and factor markets. And although there is someevidence to support complementarity, we expected stronger and morecomprehensive confirmation given the plausible inferences emerging fromthe newer endogenous-frontier models and the older long-swing litera-ture. Another blow to any endogenous-frontier model favouring comple-mentarity is the finding that the New World with the frontier offers noevidence of stronger complementarity than the Old World without thefrontier. In any case, the long- swing evidence certainly rejects substitut-ability.

4 Panel data from history: the Atlantic community, 1870–1940

This section takes another step towards filling the empirical void in the sub-stitutability versus complementarity debate by exploring long-run relation-ships in panel data for the Atlantic community (plus Australia) between1870 and 1940.10 Trade volume is now measured as the sum of imports andexports relative to GDP (as opposed to section 3, which measured ‘real’trade levels for each country). One important advantage of this new trade-volume measure is that it provides a simple metric for comparison acrosscountries which does not require the difficult translation of each country’strade into a common (real) currency via exchange rates and import andexport price indices. One potential disadvantage of the new measure,however, is that it departs from HOV theory which relates the volume of acountry’s trade, unscaled by GDP, to the gap between its factor endowmentand its factor consumption. Nevertheless, an effort will be made to developa measure of the distance between endowment and factor consumptionpoints which includes the necessary scaling. In particular, three endowments


are measured for each of the 10 countries: population, arable land andcapital stock proxied by coal consumption.11 The calculation of distance, D,through three-dimensional space is complicated by the different units ofmeasurement along each of the three endowment axes, and so all units areconverted to proportions of the sample total. Distance is calculated asfollows:

Di25 (Popi /Popw 2 gi)

2 1 (Landi /Landw 2 gi)21(Ki/Kw 2 gi)

2,

where i denotes a particular country, w denotes the sample total, and girepresents GDPi /GDPw. If the dependent variable in the panel were simplytotal trade (unscaled) in a constant and common currency, then we woulduse D (unscaled) to help explain the volume of trade. But because thedependent variable is trade/GDP, we scale D by g.12 For a given gapbetween the factor consumption and endowment points, a simple substi-tutable relationship between factor flows and commodity trade wouldsuggest that larger factor flows would be associated with smaller commod-ity flows.

A higher level of tariffs is expected to decrease the volume of trade, andthis expectation was confirmed over long swings in section 3. We have onlya rough guide to tariff levels: (customs revenue)/imports. But depending onthe mix of tariffs, quotas and prohibitions, this measure can still serve as auseful proxy for protection over time and across countries. It is well knownthat transport costs declined substantially over the full course of thisperiod, and the Isserlis index of these costs is included in the regressions.However, since the index number is identical for each country, it can beuseful, potentially at least, only in explaining variation in trade/GDP overtime.

The explanatory variables of greatest interest to this chapter are averageannual net migration rates and the current account relative to GDP foreach decade between 1870 and 1940. With three exceptions, the migrationrate used in this section is a net rate calculated using the followingmethod:13 first, the period’s average rate of natural increase is applied tothe beginning-of-period population, which provides an estimate of theend-of-period population if no one has emigrated or immigrated; second,the actual end-of-period population is subtracted from the counterfactualend-of-period population to provide an estimate of net migration over theperiod. This is expressed as an annual rate per thousand population in thebase year.

Table 9.3 pursues a simple question: ceteris paribus, do larger flows ofpeople and capital tend to be associated with higher or lower volumes oftrade?14 The table reports several specifications of the same basic regressionequation, where Endow is the distance between the endowment point and


consumption point in three-dimensional space calculated as describedabove. The basic regression is specified as follows:

(Trade/GDP)it 5a1b1*Emig.Rateit1b2*(CA/GDP)it 1b3*Tariffit 1b4*Transportt1b5*Endowit1eit.

A formidable challenge for such a simple empirical strategy is to accountfor the endogeneity of the factor-flow variables. Certainly, the relationshipbetween factor-flows and trade is not a one-way street, and this must bekept in mind when interpreting the coefficient estimates of b1 and b2. Aninstrumental-variables approach could be useful in meeting this challenge,although it is extremely difficult to identify truly exogenous variables thatare correlated with the factor flows but not affected in turn by the volumeof trade or by omitted variables which might affect both trade and factormobility.15

Column (1) of table 9.3 reports regression results from simply pooling allof the available decade-averaged observations (N558). The measure oftariffs enters strongly negative, as expected, and the measure of the endow-ment point’s distance from the assumed factor-consumption point enterspositively and surpasses the 10 per cent significance level. The transportcost variable has no discernable correlation with trade/GDP. Both thecurrent-account variable and the emigration rate are statistically insignifi-cant in the pooled sample. Very similar results are attained in column (2)when using fixed effects by period, thus relying on variation across coun-tries, to estimate the coefficients. This fact suggests that variation acrosscountries dominates variation over time in determining the pooled regres-sion results.

Using the panel in column (3) to carry out fixed-effects estimation bycountry, however, provides a view of the data that is more compatible withthe long-swing analysis in section 3. The tariff coefficient is still statisticallysignificant and negative, but it is less than half the size of the coefficient esti-mated in the pooled regression. The endowment distance variable enterspositively at approximately the same size as in columns (1) and (2), but it isfar from statistically significant. The current-account variable is positiveand statistically significant, whereas the migration rate is negatively (butnot significantly) associated with the level of trade/GDP. Employing a fullset of country and period dummies yields column (4)’s results, which againfind a negative correlation between trade/GDP and labour flows, but a pos-itive relationship between trade/GDP and capital flows.

Even in columns (3) and (4), where the factor-flow variables receivetheir largest coefficient estimates, the coefficients imply a weak economiclink between factor flows and trade flows. A one-standard-deviation


increase in the absolute value of the emigration rate (3.46) decreasestrade/GDP by only 0.009 according to column (3). A one-standard-deviation increase in the absolute value of the current-account/GDP(0.020) increases trade/GDP by only 0.023 according to column (3).These magnitudes appear to be small relative to the average trade/GDPshare (0.34).

Like the long-swing analysis in section 3, the panel data provide virtuallyno support for the substitutability hypothesis. The panel data do providesome limited support (once country dummies are included) for comple-mentarity between capital flows and trade. Even when significant, however,the size of all the estimated coefficients indicates a weak economic linkbetween factor and trade flows.16


Table 9.3. Trade and factor mobility correlations: decade averaged panel datafor the Atlantic Economy, 1870–1940

Dependent variable: trade/GDP

(1) (2) (3) (4)

Emig. rate 0.00169 0.000550 20.00218 20.0318(0.56) (0.14) (1.09) (1.87)

GA/GDP 20.131 20.491 1.08 0.522(0.23) (0.73) (3.35) (2.16)

Tariff 21.07 20.999 20.382 20.190(4.94) (4.57) (3.49) (1.94)

Trans. cost 20.000140 — 0.000212 —(0.24) (0.83

Endow. 0.0341 0.0355 0.0349 0.0120(1.77) (1.85) (0.85) (0.29)

Constant 0.449 0.438 0.315 0.430(6.53) (13.67) (6.04) (12.47)

Country dum. No No Yes YesTime dum. No Yes No YesN 58 58 58 58R2 0.33 0.34 0.91 0.95

Notes:The factor-flow variables are the absolute value of the net emigration rate and theabsolute value of the current account divided by GDP. The transport cost variableis the same for all countries (and therefore drops out when time dummies areincluded). The tariff variable is customs revenue divided by import values. Thecalculation of the ‘endowment distance’ variable is described in the text. Standarderrors are adjusted for heteroscedasticity. t-statistics are in parentheses.

5 Political-economy connections: tariffs and immigrationrestrictions17

International trade takes pride in emphasising the endogeneity of manyvariables. In that spirit, this section reports an effort to endogenise NewWorld immigration policy. Did New World immigration policy behave in afashion which suggests that the decision-makers of that time viewed tradeand factor flows as substitutes?

5.1 The stylised policy facts

After the 1880s, there was a gradual closing of New World doors to immi-grants. The doors did not suddenly slam shut on American immigrantswith the Congressional over-ride of President Wilson’s veto of the immi-grant literacy test in February 1917 or by the Emergency Quota Act of May1921. The USA started imposing restrictions on what had been freeimmigration not too long after the Civil War, a half-century prior to theLiteracy Act. Furthermore, the USA was hardly alone. Australia,Argentina, Brazil and Canada did the same, although the timing was oftendifferent and it often took the form of an enormous drop in or even dis-appearance of what had been large immigrant subsidies rather than of out-right exclusion. In short, there was considerable variance in immigrationpolicy across these five New World countries over the half-century; it wasnot simply one big policy switch around the First World War.

And what was true of immigration policy was also true of trade policy.Globalisation proceeded in fits and starts in the three decades after 1846when Britain repealed the Corn Laws and started a liberal trend towardsfree trade on the Continent. Add to that enormous declines in internationaltransport costs, and you have the ingredients for the first great globalisa-tion boom. As we have seen, it took the form of a trade boom, mass migra-tions and international capital flows at levels never reached before or since.The liberal trend did not last long, however, and there was a globalisationbacklash. Tariffs started to rise on the European Continent, while immigra-tion and trade restrictions started to rise in the New World. With the endof the First World War, the world economy plunged into a dark age of de-globalisation and hostile policy towards open factor and goods markets.This long inter-war period of darkness was followed by a liberal renaissanceembedded in Bretton Woods, GATT, guest worker arrangements inEurope, the 1965 repeal of national origins quotas in the USA, foreigncapital market deregulation and other policy events consistent with global-isation, especially after the early 1970s.

What explains changing immigration, tariff and external capital market


policy, and how are they related? For world globalisation to persist, richnations must play the liberal game. After all, typically they have the bigmarkets. Poor nations, in turn, are dependent on those big markets to selltheir labour-intensive exports, to secure finance for projects in their capital-scarce economies and to vent their surplus labour on rich countries willingto absorb unskilled immigrants. The problem for rich countries is thatglobalisation can reduce high real wages there, or at least produce inequal-ity.

So, what determined trade and immigration policy in the rich, labour-scarce nations in the late nineteenth century when there was still a frontier?Did these nations pursue legislation then which suggests that politiciansand voters believed that trade and immigration were substitutes? If theythought so, then they would have believed that it was no use restricting justone when workers’ living standards were under threat, but rather that theyhad to restrict both. Alternatively, if trade and factor flows were comple-ments, then restricting either trade or immigration would have done thetrick.

The USA was indeed protectionist, especially after the Civil War whenlanded and slave interests in the South had been defeated. And Canada wasprotectionist, especially after 1878. The fact that both countries graduallymoved to restrict immigration is consistent with our earlier suggestion that,as the frontier closed, trade and factor flows moved from being comple-ments to being substitutes. However, the closing of the frontier may havemattered more directly for the immigration debate.

Americans thought the Western frontier had disappeared in the 1890s,Argentinians thought the Pampas had filled up shortly thereafter,Brazilians in São Paulo must have thought the same as coffee prices col-lapsed in the 1920s and Canadians must have shared this view when declin-ing grain prices brought an end to their Prairie boom. Simple theory wouldpredict that the closing of the frontier had an impact on attitudes towardsimmigration: the wage impact of immigration should have been lower whenelastic supplies of land were available, since they could absorb new workers.The timing of New World immigration restrictions seems roughly consis-tent with this view.

Is the timing of immigration restrictions adequately explained by the lesselastic New World demand for labour implied by closing frontiers? Or dida changing relationship between trade and factor flows – first complements,then substitutes – also play a role? Were immigration restrictions positivelyassociated with tariffs, as would be true if trade and factor flows were sub-stitutes, and did the association between tariffs and migration policychange over time?

More generally, what explains immigration restriction itself ? The answer


ought to be consistent with the correlations invoked elsewhere in thischapter, but three candidates dominate all others – rising inequality, greaterawareness of that inequality by the politically powerful and greater votingpower in the hands of those hurt most by liberal policy – the working poor.

Surprisingly, there have been few attempts to introduce these long-runfundamentals into explicit models of immigration policy, although therehave been many attempts to make trade policy endogenous. Furthermore,we are aware of only three efforts to test immigration policy models withhistorical evidence (Shughart et al., 1986; Foreman-Peck, 1992; Goldin,1994). Two papers by Ashley Timmer and one of the present authors(Timmer and Williamson, 1996, 1997) add to this new literature. Theirresults will be summarised in a moment, but only to the extent that the com-plements-versus-substitutes issue can be placed in proper perspective.

5.2 Policy complements or substitutes? An empirical test

Unskilled labour at the bottom of the distribution is scarce and expensivein rich countries, and that fact by itself should make income distributionsmore egalitarian than in poor countries. Immigration, especially low-quality immigration from poor countries, should erode unskilled labourscarcity in rich countries and raise inequality. Labour-intensive commod-ity imports should do the same. How should policy respond? If, driven bythe presence of a frontier, trade and immigration were complements in thepre-1914 rich New World, protection and immigration restriction need nothave gone hand in hand. Such a correlation might have emerged as the fron-tier disappeared.

Among the many results reported by Timmer and Williamson (1996), table9.4 presents just one which captures the central issue in this chapter. Theunderlying panel data are for five countries – Argentina, Australia, Brazil,Canada and the USA, from about 1860 to 1930, and where the unit ofobservation is a half-decade. The dependent variable measures the extent towhich policy is pro-immigration (POLICY): the index varies from -5 to15,the former reflecting severe anti-immigration restriction, including dis-crimination in labour markets, the latter reflecting strong pro-immigration,including large subsidies, while 0 reflects neutrality. The econometric estima-tion uses fixed effects, with the lagged dependent variable on the right-handside (see table 9.4). Four variables matter most. First, policies persist: thelagged dependent variable is large and significant. However, politicians werenot total prisoners of the recent political past, since the coefficient on thelagged dependent variable is 0.8, not one. Second, the lagged relative positionof the unskilled at the bottom of the distribution matters. The numerator inthe wage–GDP-per-worker ratio is the unskilled wage, so the ratio measures


the economic distance between the bottom and the average. When theunskilled were relatively scarce (w/y was high), policy was more liberaltowards immigration. When the unskilled were relatively abundant (w/y waslow), New World policy was relatively restrictive. Third, a lagged rise in animmigrant threat to unskilled workers’ wages tends to generate more restric-tive policy (the coefficient on lagged THREAT is negative), where THREATis the combined influence of the immigration rate and the inverse of immi-grant quality. Fourth, the quality of the immigrants relative to natives (andprevious immigrants) matters with a lag. IMWAGE is calculated as the wagedifference between receiving and sending countries, and the bigger the‘quality difference’, the more restrictive the immigration policy.


Table 9.4. The determinants of immigration policy, c.1860–1930

Dependent variable: POLICY

CoefficientVariable (t-statistic)

Lagged dependent variables: POLICY (21) 0.809***(9.744)

Lagged ratio wage to GDP per worker: w/y (21) 0.015***(2.646)

Lagged real wage of the unskilled: WAGER (21) 0.005(0.637)

Unemployment rates: UNEMP 0.009(1.242

Trade share measure of ‘openness’: [X1M]/Y 0.005(0.800)

Lagged unskilled wage, immigrant origin: IMWAGE (21) 20.028*(1.896)

Lagged immigrant supply threat: THREAT (21) 20.847**(2.573)

R2 0.867Adjusted R2 0.839SE of regression 0.911No. of observations 56.

Notes:*** refers to 1, 5 and 10 per cent significance levels. POLICY measuresliberal (15) versus restrictive (25) policies toward immigration. Unit ofobservation is half-decade. Estimated by fixed effects, although the fixedeffects coefficients are excluded from the table. See text for furtherdiscussion.Source: Timmer and Williamson (1996).

The remaining right-hand-side variables are insignificant, so we will notdiscuss them here, except for one: the ratio of trade to GDP ([X1M]/Y).Any economist who has explored trade policy knows how difficult it is toconstruct a satisfying summary statistic of ‘protection’ and ‘openness’(Anderson and Neary, 1994; Sachs and Warner, 1995; O’Rourke andWilliamson, 1997; O’Rourke, 1997). We confess that the openness measureused here, [X1M]/Y, is a crude index of trade policy. Nonetheless, there isno evidence of a significant positive correlation between trade andimmigration policy. Between 1860 and 1930, policy did not behave as ifNew World politicians and voters thought trade and immigration were sub-stitutes.

6 History’s bottom line

The elegant model associated with Eli Heckscher and Bertil Ohlin makesthe unambiguous statement that trade and factor mobility are substitutes.In the absence of factor mobility, trade can serve as an indirect substitute.In the presence of factor mobility, trade is no longer necessary. The issueis important for the policy implications it generates. If immigration threat-ens egalitarian distributions by creating labour abundance at the bottomof the distribution, immigration quotas will simply provoke more tradeand a flood of ‘unfair competition’ from unskilled workers making labour-intensive goods in distant lands of labour surplus from which they can nolonger emigrate. If rich countries retreat behind tariff walls, there will befar more immigrants hammering at their gates since opportunities formaking labour-intensive export goods at home have dried up. Theeconomics is so plausible that it survived with little challenge over the fourdecades from the early post-First World War years, when Heckscher andOhlin were writing, to 1957 when Robert Mundell published his seminalpiece. In the four decades since, the challenges to that conventionalwisdom have been so many that nothing but theoretical ambiguityremains.

The ambiguity invites empirical analysis. It is all the more surprising,therefore, that there has been so little econometric work on the question.This chapter has started to fill the vacuum by exploring the historical expe-rience of the Atlantic economy between 1870 and 1940.

The historical bottom line is this: when we look at the long swingsembedded in the time-series data, we find that trade and capital flows wererarely substitutes and often complements. The same was true for trade andmigration. When we look at longer-run relationships in the panel data, wefind once again that trade and capital flows were rarely substitutes andoften complements. Trade and migration were never substitutes. When we


look at immigration policy in the New World, it appears that policy-makers never acted as if they viewed trade and migration as substituteseither.

The history of the Atlantic economy between 1870 and 1940 rejects thethesis that trade and factor mobility were substitutes. It is a little morecomfortable with the thesis that they were complements. Whether the sameis true of the late twentieth century is another matter entirely, but we haveattempted to show a way to discriminate between the propositions.

APPENDIX

Most of the data is available in the following volumes of historical statisticscompiled by B.R. Mitchell: International Historical Statistics: Europe1750–1988 (1992, hereafter IHSE); International Historical Statistics – TheAmericas (1993, hereafter IHSA); International Historical Statistics –Africa, Asia and Oceania (1995, hereafter IHSAAO); and InternationalHistorical Statistics – The Americas and Australasia (1983, hereafterIHSAA).

Imports and exports

Generally, statistics are for merchandise trade only, and imports are valuedc.i.f., whereas exports are valued f.o.b. European figures are from Mitchell,IHSE. Complete data are unavailable for Germany before 1880, and from1871 to 1918 Alsace Lorraine is included in Germany rather than in France.Trade between Ireland and Britain is treated as international trade begin-ning in 1923. Canadian and US figures are from Mitchell, IHSA.Australian figures are from Mitchell, IHSAAO. GDP deflators are used toconvert nominal trade figures to real trade figures in the ‘long-swing’section of the chapter.

GDP (nominal)

Current-price GDP (or GNP) figures are used to scale the current-pricetrade figures in the panel data section of the chapter. European figures arefrom Mitchell, IHSE. Germany’s statistics are for net national product.Ireland is excluded from the UK total beginning in 1921. US figures arefrom Mitchell, IHSA. Canadian figures were extracted from data providedby Michael Bordo and used in Bordo and Jonung (1987). Australian figuresare from Mitchell, IHSAAO.


Transport costs

Calculated from Isserlis (1938, table VIII). The nominal freight index(column (2)) is deflated by the Statist wholesale price index (column (6)).

GDP (real)

GDP levels measured in million 1990 Geary–Khamis dollars are takenfrom Maddison (1995, appendix C). Maddison (1995) reports GDP for theFederal Republic area only, and so overall German figures were calculatedby multiplying per capita income (Maddison, 1995, appendix D) by thepopulation figures described below.

Population

For Australia, Denmark, France, Italy, Norway and Sweden, annualpopulation figures are from Maddison (1995, appendix A). For Canada,Germany, the UK and the USA, annual population figures are fromMaddison (1991, appendix B). The UK figures include Ireland (fromMaddison, 1995) throughout the period.

Land

Agricultural land is taken from Estevadeordal (1993, p.64). The figures arebased on a League of Nations study for 1913 which defined agriculturalland as ‘arable land, pasture, meadow and prairie, trees, shrubs and bushes,but not forests’ (Estevadeordal, 1993, p.21). These quantities are assumedconstant over time.

Coal

Coal consumption is calculated as the sum of coal production plus coalimports minus coal exports. Brown coal is converted to hard coal equiva-lents using conversion factors taken from Darmstadter (1971, p.828) foreach country. European data for production and trade are from Mitchell,IHSE. Canadian and US production and trade figures are from Mitchell,IHSAA. Australian production figures are from Kalix et al. (1966) andtrade figures are from Mitchell, IHSAA.

Tariff revenue

European figures are from Mitchell, IHSE. Italian tariffs for 1870 to 1929were kindly provided by Giovanni Federico. Canadian and US data are


from Mitchell, IHSA. Tariff rates for Australia prior to 1901 are con-structed from customs and import data available in the Victorian Year-Book (various editions) for Victoria. Beginning in 1901 Australian customsrevenue figures are available in Mitchell, IHSAAO.

Current account

The overall current balance for European countries is from Mitchell, IHSE.Australian, Canadian and US figures are from Mitchell, IHSAA. TheCanadian series does not begin until 1900.

Gross migration (annual)

Gross annual emigration (Old World) or immigration (New World) figuresare used in the time-series analysis. These statistics relate to inter-conti-nental migration. In general, it is not possible to construct reliable netmigration figures on an annual basis. European statistics are from Mitchell,IHSE. Canadian and US statistics are from Mitchell, IHSA. Australianmigration figures are reported as net immigrants (that is, immigration andemigration are not reported separately) in Mitchell, IHSAAO.

Net emigration (decade)

For most countries these figures were estimated using a combination ofcensus population figures and average annual rates of natural increase asdescribed in detail in the text. However, Italian figures are taken from Bacci(1978), Australian figures are derived from the net immigration series inMitchell, IHSAAO, and the US figures are from Eldridge and Thomas(1964). The UK figures are constructed to include Ireland throughout theperiod using Mitchell (1988). The panel data section of the chapter uses netemigration rates where average annual emigration is expressed as a rate per1,000 population at the period’s beginning.

NOTES

We acknowledge with thanks the help of Don Davis, Riccardo Faini, ElhananHelpman, Arye Hillman, Jim Markusen, Giani Toniolo, Tony Venables, KlausZimmermann and participants at the Venice conference (25–26 January 1997). Thischapter has been revised extensively from the conference version.11 Textile production is higher in Ireland than in Britain. It follows that

capital–labour ratios are higher in Irish than in British textile production, whichimplies higher wages and lower profits in Ireland.

12 The treatment here follows Findlay (1995, ch. 1).


13 Corden and Neary (1982) consider the impact of a Hicks-neutral technologicalimprovement in one sector, and distinguish between the resource movement andspending effects of the associated boom.

14 As section 4 will show at greater length, this alternative approach relied onassumptions about which way factors ‘should’ flow based on Heckscher–Ohlin–Vanek (HOV) reasoning. Relative abundance or scarcity was determinedby using endowment measures of labour, land and capital for each of the 10countries in combination with the country’s share in the sample’s total GDP. Asthe text indicates, this second approach yielded roughly the same results as thoseimplied by tables 9.1a–9.1c, and thus are not reported here.

15 Data sources are discussed in the appendix (p. 253).16 The same results emerged when trade shares in GDP were regressed on factor-

mobility rates. Furthermore, the same results emerged whether we estimated theregressions separately for mig and rlCA, or together as in tables 9.1a–9.1c in thetext.

17 The tariff rate measure is crude, customs revenues relative to imports, but itshould be effective enough over a long swing in the time series.

18 It is, appropriately, a deflated freight-rate index, but it is a generic world-wideindex rather than a country-specific index.

19 Tables 9.2a–9.2c use a 10 per cent level in the hypothesis testing, but the samequalitative results emerge when a 5 per cent level is used. The only difference isthat in the latter case some complementarity results fall into the neutral cate-gory. In short, the substitutability hypothesis is still soundly defeated.

10 As in section 3’s long-swing analysis, the sample includes Australia, Canada,Denmark, France, Germany, Italy, Norway, Sweden, the UK and the USA.

11 See Estevadeordal (1993, p.20) or Landes (1969, p. 293) for examples and justi-fications of this proxy.

12 If country A0 has endowment vector V and country B has endowment vector2V, we would expect B to have a larger (unscaled) volume of trade and a largermeasure of D, but the countries should have the same amount of trade/GDP andwe would want them to have the same measure of D/g, ceteris paribus.

13 Italian emigration is taken from Bacci (1978, p. 39). The US rate is taken fromEldridge and Thomas (1964). The Australian rate is calculated using netimmigration figures from Mitchell (1983). We calculated net migration rates forthe other seven countries using the method described in the text.

14 A previous version of this chapter took account of the direction of factor flows,rather than focusing on size exclusively. Each country in each time period wasdesignated ‘abundant’ in a factor if its share in the total sample’s endowmentexceeded its share in the total sample’s GDP. Abundant countries are expectedto export that factor to close the gap between its endowment point and itsfactor-consumption point (assumed to be along the diagonal of the world factor‘box’). Observing the direction of actual factor flows, we then asked whetherfactor flows in the ‘right’ direction decreased trade flows, as a substitutable rela-tionship would suggest.

15 The previous version of this chapter attempted to use the lagged rate of naturalincrease and a New World dummy variable to instrument for the emigration rate,and used the dependency rate (proportion of population under age 15) to instru-ment for the current account balance (Taylor and Williamson, 1994). The currentframework, which employs the absolute value of factor-flow measures, is not wellsuited for these instruments. Moreover, such instruments are far from exogenous.


16 This finding is echoed in the empirical literature on trade, foreign domesticinvestment and multinational corporations (Grubert and Mutti, 1991).

17 This section draws heavily on Timmer and Williamson (1996).

REFERENCES

Abramovitz, M. (1961). ‘The Nature and Significance of Kuznets Cycles’, EconomicDevelopment and Cultural Change, 9, 225–48

Anderson, J. and J. P. Neary (1994). ‘Measuring the Restrictiveness of Trade Policy’,The World Bank Economic Review, 8, 151–69

Bacci, M. L. (1978). ‘Fattori Demographici Dello Sviluppo Economico’, in G. Fua(ed.), Lo Sviluppo Economico in Italia, vol. II (Milan: F. Angeli)

Bordo, M. and L. Jonung (1987). The Long-Run Behavior of the Velocity ofCirculation (Cambridge: Cambridge University Press)

Cairncross, A. (1953). Home and Foreign Investment (Cambridge: CambridgeUniversity Press)

Corden, W. M. and J. P. Neary (1982). ‘Booming Sector and De-Industrializationin a Small Open Economy’, Economic Journal, 92, 825–48

Darmstadter, J. (1971). Energy in the World Economy (Baltimore: Johns HopkinsUniversity Press)

Easterlin, R. A. (1966). ‘Economic–Demographic Interactions and Long Swings inEconomic Growth’, American Economic Review, 56, 1063–4

(1968). Population, Labor Force and Long Swings in Economic Growth (NewYork: NBER)

Eldridge, H. and D. Thomas (1964). Population Redistribution and EconomicGrowth, United States, 1870–1950: Demographic Analyses and Interrelations(Philadelphia: American Philosophical Society)

Estevadeordal, A. (1993). ‘Historical Essays on Comparative Advantage’, PhDthesis, Harvard University, unpublished

Findlay, R. F. (1995). Factor Proportions, Trade, and Growth (Cambridge, MA: MITPress)

Flam, H. and M. J. Flanders (1991). Heckscher–Ohlin Trade Theory (Cambridge,MA: MIT Press)

Foreman-Peck, J. (1992). ‘A Political Economy Model of International Migration,1815–1914’, The Manchester School, 60, 359–76

Goldin, C. (1994). ‘The Political Economy of Immigration Restriction in the US, 1890to 1921’, in C. Goldin and G. Libecap (eds.), The Regulated Economy: AHistorical Approach to Political Economy (Chicago: University of Chicago Press)

Grubert, H. and J. Mutti (1991). ‘Taxes, Tariffs and Transfer Pricing inMultinational Corporate Decisions Making’, Review of Economics andStatistics, 73, 285–93

Hatton, T. J. and J. G. Williamson (1998). The Age of Mass Migration: An EconomicAnalysis (New York: Oxford University Press)

Heckscher, E. (1919). ‘The Effect of Foreign Trade on the Distribution of Income’,Ekonomisk Tidskrift, 21, 497–512


International Bank for Reconstruction and Development (1995). WorldDevelopment Report 1995 (New York: Oxford University Press)

Isserlis, L. (1938). ‘Tramp Shipping Cargoes and Freights’, Journal of the RoyalStatistical Society, 101, pt. 1, 304–417

Jones, R. W. (1971). ‘A Three-Factor Model in Theory, Trade, and History’, in J. N.Bhagwati et al. (eds.), Trade, Balance of Payments, and Growth (Amsterdam:North-Holland), 3–21

Kalix, Z. (1966). Australian Mineral Industry: Production and Trade, 1842–1964(Canberra: Bureau of Mineral Resources, Geology and Geophysics)

Kelley, A. C. and J. G. Williamson (1984). What Drives Third World City Growth?(Princeton: Princeton University Press)

Kemp, M. C. and M. Ohyama (1978). ‘On the Sharing of Trade Gains by ResourcePoor and Resource-Rich Countries’, Journal of International Economics, 8,93–115

Kuznets, S. (1930). Secular Movements in Production and Prices (New York: NBER)

(1958). ‘Long Swings in the Growth of Population and in Related EconomicVariables’, Proceedings of the American Philosophical Society, 102, 25–52

Landes, D. S. (1969). The Unbound Prometheus (Cambridge: Cambridge UniversityPress)

Lerner, A. P. (1952). ‘Factor Prices and International Trade’, Economica, 19, 1–15

Lewis, W. A. (1978). The Evolution of the International Economic Order (Princeton:Princeton University Press)

Lewis, W. A. and P. J. O’Leary (1955). ‘Secular Swings in Production and Trade,1870–1913’, The Manchester School, 13, 113–52

Maddison, A. (1991). Dynamic Forces in Capitalist Development: A Long-RunComparative View (New York: Oxford University Press)

(1995). Monitoring the World Economy, 1820–1992 (Paris: OECD)Markusen, J. R. (1983). ‘Factor Movements and Commodity Trade as

Complements’, Journal of International Economics, 13, 341–56Mitchell, B. R. (1983). International Historical Statistics: The Americas and

Australasia (Detroit: Gale Research Co.)(1988). British Historical Statistics (New York: Cambridge University Press)(1992). International Historical Statistics: Europe, 1750–1988 (New York:

Stockton Press)(1993). International Historical Statistics: The Americas, 1750–1988 (New York:

Stockton Press)(1995). International Historical Statistics: Africa, Asia & Oceania, 1750–1988

(New York: Stockton Press)Mundell, R. A. (1957). ‘International Trade and Factor Mobility’, American

Economic Review, 47, 321–35Neary, J. P. (1995). ‘Factor Mobility and International Trade’, Canadian Journal of

Economics, 28, special issue, S4–S23O’Rourke, K. H. (1997). ‘Measuring Protection: A Cautionary Tale’, Journal of

Development Economics, 53, 169–83


O’Rourke, K. H. and J. G. Williamson (1997). ‘Around the European Periphery1870–1913: Globalization, Schooling and Growth’, European Review ofEconomic History, 1, 153–90

Rostow, W. W. (1978). World Economy: History and Prospect (Austin: Univers i tyof Texas Press)

Rybczynski, T. M. (1955). ‘Factor Endowments and Relative Commodity Prices’,Economica, 22, 336–41

Sachs, J. D. and A. Warner (1995a). ‘Economic Reform and the Process of GlobalIntegration’, Brookings Papers on Economic Activity, 1, 1–118

(1995b). ‘Economic Convergence and Economic Policies’, NBER, WorkingPaper, 5039

Samuelson, P. A. (1949). ‘International Factor–Price Equalization Once Again’,Economic Journal, 59, 181–97

Schmitz, A. P. and P. Helmberger (1970). ‘Factor Mobility and International Trade:The Case of Complementarity’, American Economic Review, 60, 761–7

Shughart, W., R. Tollison and M. Kimenyi (1986). ‘The Political Economy ofImmigration Restrictions’, Yale Journal on Regulation, 51

Taylor, A. M. (1996). ‘International Capital Mobility in History: TheSaving–Investment Relationship’, NBER, Working Paper, 5743

Taylor, A. M. and J. G. Williamson (1994). ‘Capital Flows to the New World as anIntergenerational Transfer’, Journal of Political Economy, 102, 348–71

Temin, P. (1966). ‘Labor Scarcity and the Problem of American IndustrialEfficiency’, Journal of Economic History, 26, 277–98

Thomas, B. (1954). Migration and Economic Growth (Cambridge: CambridgeUniversity Press)

Thomas, D. S. (1941). Social and Economic Aspects of Swedish PopulationMovements (New York: Macmillan)

Thompson, H. (1985). ‘Complementarity in a Simple General EquilibriumProduction Model’, Canadian Journal of Economics, 18, 616–21

(1986). ‘Free Trade and Factor–Price Polarization’, European Economic Review,30, 419–25

Timmer, A. and J. G. Williamson (1996). ‘Racism, Xenophobia or Markets? ThePolitical Economy of Immigration Policy Prior to the Thirties’, NBER,Working Paper, 5867

(1997). ‘Immigration Policy Prior to the Thirties: Labor Markets, PolicyInteractions and Globalization Backlash’, Department of Economics,Harvard University, March

United States Bureau of the Census (1975). Historical Statistics of the UnitedStates, Colonial Times to 1970 (Washington, DC: GPO)

Urquhart, M. C. (1965). Historical Statistics of Canada (Cambridge: CambridgeUniversity Press)

Venables, A. J. (1997). ‘Trade Liberalisation and Factor Mobility: An Overview’,chapter 2 in this volume

Wilkinson, M. (1967). ‘Evidences of Long Swings in the Growth of SwedishPopulation and Related Economic Variables 1860–1965’, Journal of EconomicHistory, 27, 17–38


Williams, J. H. (1929). ‘The Theory of International Trade Reconsidered’, EconomicJournal, 39, 195–209

Williamson, J. G. (1964). American Growth and the Balance of Payments, 1820–1913 (Chapel Hill, NC: University of North Carolina Press)

(1974). ‘Migration to the New World: Long Term Influences and Impact’,Explorations in Economic History, 11, 357–90

(1996). ‘Globalization, Convergence, and History’, Journal of Economic History,56, 277–306

Wong, K.-Y. (1988). ‘On Choosing among Trade in Goods and InternationalCapital and Labor Liability: A Theoretical Analysis’, Journal of InternationalEconomics, 14, 223–50

Zevin, R. B. (1992). ‘Are World Financial Markets More Open? If So Why and WithWhat Effects?’, in T. Banuri and J. Schor (eds.), Financial Openness andNational Autonomy (Oxford: Oxford University Press)

Discussion

GIANNI TONIOLO

Before reading chapter 9, the introduction of an endogenous frontier hadproved to be a useful tool in my economic history classes, helping under-graduates, trained in 232 trade models, make sense of the late nineteenth-century apparent paradox: the prima facie evidence that labour and capitalflowed in the same direction, while at the same time the volume of interna-tional trade rapidly increased. After reading Collins et al., I will have torevise my syllabus. Or will I?

It was over 10 years ago when Jeff Williamson embarked upon what soonbecame a major effort in describing, understanding and explaining themultifarious aspects of the so-called ‘first globalisation’ – the one, that is,that came to an abrupt end in August 1914. Since then, Williamson and hisnumerous co-authors have patiently constructed an incredibly large data-base on flows of goods, labour and capital as well as on prices of products,factors of production, transport and communication services. Armed withsuch a powerful tool, they set out systematically to tackle the many con-vergence issues underlying the very notion of globalisation. This effortresulted in an incredibly large output of papers – only few of which featureas references to this chapter – dealing in the first place with individual


markets and then, as in the present case, with the interaction amongmarkets.

Here the question is: were trade and factor mobility complement or sub-stitute during the first great wave of globalisation? Collins et al. remind usthat, as far as theory is concerned, ‘the relationship between trade andfactor mobility is ambiguous’, so that ‘trade could be complementary tomigration, but a substitute for capital flows, or vice versa’ . Neither does theintroduction of an endogenous frontier per se guarantee complementarity.The latter depends on the underlying assumptions; in the economic histo-rian’s parlance, ‘the “correct” model will vary with the period beingstudied’ (a point as obvious as it is overlooked).

So, what was the ‘correct’ model for the Atlantic economy between 1870and 1940? The authors’ prior derived from the economic historian’sreceived wisdom, the one we teach our undergraduate classes: comple-mentarity would prevail in the presence of an endogenous frontier andwould decline with its slow disappearance (that is, they argue, after 1914).The empirical findings in this chapter show that this was not the case: if any-thing, complementarity between trade and factor flows appears more oftenafter the Great War than before it. Neither did policy-makers believe in sub-stitution between trade and factor mobility.

In fact, empirical analysis seems to be consistent with theory in that theyboth remain ambiguous. Neutrality rather than complementarity or sub-stitutability prevailed throughout: most coefficients in the trade-factor-mobility regressions are not significant at the 10 per cent confidence level.The authors are themselves surprised by their own results and offer onlyscanty tentative explanations for the unexpected outcome of their efforts.

One question suggested by the results regards the most appropriate levelof aggregation in dealing with the complex questions raised by the authorsat the beginning of the chapter. On the one hand, ‘total trade’ is likely to betoo aggregate a variable in testing complementarity or substitutability withfactor mobility. Capital may be positively linked with trade if an element oftransfer of embodied technology is involved but this would be capturedonly by imports and exports of technology-intensive goods. Likewise,labour migration may indeed increase trade in idiosyncratic consumptiongoods (Irish migrating to America will probably increase the imports oftheir favourite beer brands, at least until they learn to brew them in the hostcountry, Italians will do the same with pasta and peeled tomatoes). At thesame time, there might be sectors where substitution prevails, particularlythose where large multinational companies operate. At the aggregate levelthese two effects may simply offset each other and disappear. On the otherhand, a multi-country approach may be too disaggregate meaningfully totest the frontier hypothesis. Here it might prove fruitful to lump together in


a single geographic unit the New-World countries on the one hand and theEuropean countries on the other. This would be a more accurate way oftesting a 233 model, by controlling – so to speak – for infra-frontier andinfra-European trade and factor movements.

But, of course, there are many more candidate explanations than just thefrontier to account for the apparent paradox of over-time increasing (ratherthan decreasing) complementarity between capital flows and trade. Was itnot American capital (first private then in the form of inter-allied loans)that financed those large convoys defying German U-boats? And whatabout Schuker’s ‘American Reparations to Germany’: were they not theengine of transatlantic trade throughout the 1920s? Then, in the 1930s, theflows of both goods and capitals dried up as a result of all sorts of barriersto trade in ‘foreign exchange controls’ : this pattern, too, would appear inthe regression as a (?pseudo-)complementarity.

Perhaps, for the time being, I shall just add a number of caveats to thepart of my syllabus dealing with the international economy in the forty-oddyears before the Great War rather than throwing it altogether into thedustbin. Incidentally, this would be very educational: a good teacher ismeasured by the number of problems he raises, by the food for thought hegives his students rather than by the answers–solutions he provides.

The last part of the chapter deals with the intriguing issue of the end ofthe first globalisation, raised elsewhere by Williamson, again in a problem-atic way. Here I have one minor reservation about the appropriateness ofthe tariff variable included in the regression since, after the mid-1890s, whilenominal tariffs did not change, real import duties were significantly loweredby commercial treaties and by moderate inflation.

At this stage Collins et al. raise more questions than they answer andnobody is more aware of it than the three authors themselves. This chapteris nonetheless important as a yet another building block in the hugeconstruction undertaken by Williamson et al. They are assembling factsand explanations about the zenith and nadir of the first globalisation:sooner rather than later we shall be able to see the solid and coherent build-ing already discernible in most of its constituent parts.

262 Discussion by Gianni Toniolo

10 Liberalisation and incentives forlabour migration: theory withapplications to NAFTA

JAMES R. MARKUSEN AND

STEVEN ZAHNISER

1 Introduction

Regionalism has been an important phenomenon in the world of interna-tional trade in the last decade. Groups of countries, typically geograph-ically concentrated, are banding together to liberalise trade and investmentamong themselves. The EU is surely the furthest along, with relativelyliberal provisions for labour migration added to trade and investment liber-alisation.

One interesting and relatively novel feature of some of the new regionaltrade agreements (RTAs) is that they combine partners of very differentlevels of development. Typically this was not the case during previousdecades, when such agreements tended to be among countries of similar percapita income levels. The NAFTA was pioneering in this respect, and maybe expanded to include other Latin American countries in the next fewdecades. Similarly, the EU will surely consider substantial liberalisations inthe future with countries from Eastern Europe and the former Soviet Union.

Factors which motivate and encourage these new ‘North–South’ or‘East–West’ agreements may also differ from the older agreements amonghighly developed countries. The latter were in large part motivated by theobjective of creating large internal markets in order to capture scaleeconomies and other production efficiencies. But the newer agreementshave a somewhat different focus. First, the developed partner(s) may beseeking a low-wage partner who can provide low-cost labour for labour-intensive tasks of the developed country’s firms. The less-developedpartner(s) may be seeking access to inward investment and newer tech-nologies. A somewhat more subtle motive for the less-developed country(LDC) is to obtain ‘insurance’ against capricious policy changes by thedeveloped countries.

Lastly, the developed partner(s) may be seeking to reduce the pressuresof immigration and migration from the developing partner(s).1 Some

263

observers believe that influxes of workers from LDCs cause various socialand economic problems in the developed countries, and the underlyingstrategy is to help create jobs in the LDCs in order to keep these personsand their families at home. Both the USA and the EU have experienced thistype of inward movement of labour. Between 1986 and 1995, the USAadmitted over 2.8 million legal immigrants from Mexico (see table 10.1).This number includes some 2 million individuals who entered the USA ille-gally but were granted amnesty under the terms of the 1986 ImmigrationReform and Control Act (IRCA). Barring any dramatic reform of USimmigration law and any substantial long-term improvement in theMexican economy, it is not unreasonable to expect that the USA will con-tinue to admit sizable numbers of additional legal immigrants from Mexicoeach year for some time to come, especially as persons originally admittedunder IRCA become entitled to seek the entrance of their relatives.

Undocumented migration from Mexico to the USA also continues to besubstantial. No one knows with any great precision just how manyMexicans reside illegally in the USA, but widely circulated ‘guesstimates’range between 4 and 10 million. What is certain is that undocumented

264 James R. Markusen and Steven Zahniser

Table 10.1. Legal immigration to the USA by persons born in Mexico anddeportations of Mexicans by the US border-patrol, FY 1981–95

Total Total No. of MexicansFiscal legal minus IRCA apprehended and deportedyear immigration legalisations by the US border patrol

1981 101,268 101,268 1,797,9231982 256,106 256,106 1,795,3621983 259,106 259,106 1,076,3451984 257,557 257,557 1,104,4291985 261,290 261,290 1,218,6951986 266,533 266,533 1,635,7021987 272,351 272,351 1,123,7251988 295,039 295,039 1,928,2781989 405,172 266,445 1,830,9851990 679,068 256,549 1,054,8491991 946,167 252,866 1,045,1221992 213,802 291,332 1,168,9461993 126,561 109,027 1,230,1241994 111,398 106,995 1,965,1441995 289,932 286,960 1,381,465

Sources: INS (various issues); INS (1996).

Mexican migration now extends to virtually every region of the USA andthat this phenomenon often outmatches the resources allocated by the USgovernment to control it. Of course, not everyone sees undocumentedmigration from Mexico to the USA as a problem because illegal migrantsare a source of considerable profit for many US firms and farmers.

The purpose of this chapter is to consider the links between trade andinvestment liberalisation on the one hand and the incentives for the migra-tion of less-skilled labour on the other. The analysis will be mostly theoret-ical, using models firmly rooted in the stylised facts of NAFTA. Ourdiscussion focuses largely on Mexico and the USA but not Canada, whichhopefully is a forgivable transgression since undocumented migration fromMexico is a more prominent concern in the USA than in Canada. The prin-cipal question at hand is whether we should expect that trade and invest-ment liberalisation between a developed and a less-developed partner willlead to convergence in the wages of unskilled workers between the twocountries and hence reduce the incentives of workers to migrate from theless-developed to the developed country. In other words, will trade andinvestment liberalisation substitute for migration?

Section 2 of the chapter presents a few background facts about NAFTAand the Mexican economy, especially with respect to the Mexican labourmarket and the country’s agricultural sector. Although it is far too early tojudge the effects of NAFTA, especially in the wake of a major macro-economic downturn in Mexico in late 1994 and 1995, we echo the widelyheld view that NAFTA will do little to achieve wage convergence in the nextdecade or two. We then outline the basic mechanism of theHeckscher–Ohlin (HO) model which argues in the other direction, infavour of wage convergence. The HO model suggests that the wages ofunskilled labour in Mexico should rise and that of unskilled labour in theUSA should fall following trade and/or investment liberalisation.

The next three sections of the chapter outline some of the reasons whythe gap between wages in the USA and Mexico may not fall, and indeedwhy the gap between skilled and unskilled wages may actually rise in bothcountries, as some evidence seems to suggest. Section 3 considers a modeland associated evidence presented in two chapters by Feenstra and Hanson(1995a, 1995b). Section 4 examines a mechanism proposed by Markusenand Venables (1995, 1997a, 1997b) and Markusen (1997). Both of thesemodels focus on a crucial role played by investment liberalisation, and bothsuggest the possibility of a rising skilled–unskilled wage gap in both thedeveloped countries and LDCs.

Section 5 considers three other mechanisms, all of which operate in tradi-tional competitive, constant-returns models and all of which seem empir-ically relevant to NAFTA. The first is a model based on multiple techniques

Liberalisation and incentives for labour migration 265

of production in agriculture – maize in particular – motivated by theobservation that maize production is very capital-intensive in the USA (rel-ative to other US sectors) and very labour-intensive in Mexico (relative toother Mexican sectors). The argument is similar to but not quite the sameas one suggested by Schiff (1996), which relates tangentially to Burfisher etal. (1994) as well. The second is a simple specific-factors model developedby Neary (1995) and Markusen (1983) in which trade in goods and andtrade in factors are complements. An alternative version of the specific-factors argument is presented by Schiff (1996). The final model is drawnfrom a suggestion by Martin (1996) that Mexican agriculture requirespublic infrastructure more than some other sectors, such as maquiladoraassembly plants. Hence maize is a disadvantaged sector in Mexico despitebeing intensive in the use of the abundant factor, namely unskilled labour.Section 6 draws some conclusions.

2 NAFTA

NAFTA took effect on 1 January 1994, following the Canadian–AmericanFree Trade (CAFTA) of the late 1980s and early 1990s. In the simplest terms,NAFTA lowered trade and investment barriers within North America butcontained only minor provisions regarding labour migration. Mexicoundertook substantial unilateral liberalisations during the period following1985, so the actual reforms introduced by NAFTA were not profound insome overall sense. Most estimates put US tariffs against Mexico prior toNAFTA at about 4 per cent and Mexican protection levels at somethingover 10 per cent, with a variety of non-tariff barriers (NTBs) also in place.

The leaders of the three signatories to NAFTA stated that the agreementaimed to increase the growth and income levels of all three countries. Tradeand investment liberalisation were predicted to lead to higher incomes,investment, growth, employment and all things bright and beautiful. But itis widely believed that there was a sub-agenda guiding the agreement thatwas rarely stated in public, in particular regarding the USA and Mexico.

Four things may have been important from the US point of view. First,there was the hope that NAFTA would improve wages in Mexico, therebyreducing the tendency of unskilled Mexican workers to migrate to the USA.But others wonder how confident the US government really was about this,and perhaps it was privately sceptical that NAFTA would have much of animpact over one or two decades on illegal migration. Second, US firmsargued fairly persuasively that they needed a low-wage partner for routine,less-skilled operations such as assembly in order to compete with suppliersor branches of Japanese multinationals in LDCs. They argued, perhapsgenuinely in many cases, that jobs going to Mexico would displace jobs in


Asia, not in the USA. Third, some believe that the USA was greatly frus-trated with the slow pace of traditional multilateral negotiations andwanted to create something of an example in order to get things movingagain. A fourth explanation, and one that we favour, is that a foreign-policyobjective dominated the economic ones: the USA had witnessed sub-stantial liberalisation in Mexico in both the economic and political arenasand wanted to do something to help lock-in these reforms. Under thisexplanation, it is quite possible that US policy-makers had little confidencein their own statements about growth, jobs and reduced inward migration.Those issues were irrelevant.

The Mexican government did make clear statements about its desire toreduce undocumented migration to the USA (‘we want to export goods, notpeople’), but these may have been issued largely to cultivate support forNAFTA in the USA. Many analysts believe that the Mexican governmentalso saw NAFTA as a way to lock-in reforms that it had brought into effect.NAFTA was thus ‘insurance’ against domestic backsliding. But it is at leastas likely that the Mexican government viewed NAFTA as insurance againstUS backsliding. The fact that late in the game Mexico seemed to give in onmany bargaining points (in the face of a scare that NAFTA might not passthe US Congress) in the opinion of many reinforces this view. In any case, ourpoint is that it is not necessarily true that either government really believedthat NAFTA would reduce migration in the span of one or even two decades.

Some statistics may help set the context for this chapter. Table 10.2 gives


Table 10.2. Canadian, Mexican and US civilian employment, by economicactivity, 1994 (thousand persons and per cent)

Agriculture, hunting, forestry and fishing ,545 4.1 8,361 25.8 3,586 2.9Mining and quarrying ,157 1.2 ,152 0.5 ,669 0.5Manufacturing 1,949 14.7 5,127 15.8 20,157 16.4Electricity, gas and water ,144 1.1 ,80 0.2 1,216 1.0Construction ,750 5.6 1,828 5.6 7,493 6.1Wholesale and retail trade, 3,151 23.7 6,962 21.5 27,163 22.1

restaurants and hotelsTransport, storage and communication ,835 6.3 1,467 4.5 6,750 5.5Financing, insurance, real estate 1,611 12.1 1,111 3.4 13,566 11.0

and business servicesCommunity, social and personal services 4,149 31.2 7,337 22.6 42,460 34.5Activities not adequately defined — ,14 0.0 —

Total 13,292 32,439 123,060

Source: OECD (1996b),

the distribution of civilian employment within the Canadian, Mexican andUS economies. Mexico’s labour force is far more concentrated in agriculturethan that of either Canada or the USA. In 1994, 25.8 per cent of the Mexicanwork force was occupied in agriculture, as compared to 4.1 and 2.9 per centfor Canada and the USA, respectively. In fact, the number of personsworking in agriculture in Mexico is more than double that of Canada and theUSA combined (8,361,000 versus 4,131,000). Moreover, agriculture’s shareof the Mexican work force has held steady between 1991 and 1994.2 In con-trast, the US and Canadian labour forces are distinguished by relatively highproportions employed in two categories: (1) finance, insurance, real estateand business services; and (2) community, social and professional services.Using 1988 data, Burfisher et al. (1994) further identify that Mexico’s urbanlabour force is much less skilled than that of the USA (see table 10.3). Whenone considers both urban and rural workers, it is clear that Mexico is abun-dant in unskilled labour relative to the USA and Canada.

Although agriculture employs roughly one-quarter of Mexico’s workforce, it accounts for only about 6 per cent of the country’s gross domesticproduct (GDP) (INEGI, 1997). Depending on the region, much of ruralproduction is concentrated in one crop: maize (referred to as corn in theUSA). Most Mexican farmers grow maize in some amount, perhaps fortheir private consumption if nothing else. But production is inefficient andyields are low, averaging about 2.4 metric tons per hectare (see table 10.4).Mexican yields begin to approach US and Canadian levels of roughly 7tons per hectare for producers during the Spring–Summer growing seasonand when farmers have access to irrigation.

Maize is also a highly protected sector in Mexico. Mexico’s bilateralimport barrier on US food corn is 45.0 per cent, while the correspondingUS barrier on Mexican food corn is nil (Burfisher et al., 1994). Thus wehave an odd situation where the crop that uses Mexico’s abundant factor


Table 10.3. Comparison of US and Mexican employmentstructures, 1988, per cent

Sector of work force Mexico USA

Rural labour 223.8 221.1Urban unskilled labour 214.1 217.7Urban skilled labour 237.1 248.5White-collar workers 225.0 232.7Total 100.0 100.0

Source: Burfisher et al. (1994).

intensively is an import-competing sector. The USA lobbied hard forreform to the various forms of maize protection and basically got its way,phased in over a number of years (initially 15, but Mexico has apparentlyspeeded this up). Mexico also has high protection levels on what are some-times called ‘program crops’ in the USA because they are subject to specialagricultural programmes. These include feed corn, food grains, soybeansand cotton. Mexico’s bilateral barrier on such crops from the USA is 12.9per cent, while the corresponding US barrier is again zero.

This characterisation of a major part of Mexican agriculture as highlyprotected yet intensive in unskilled labour is both tremendously importantfor discussing the possible effects of NAFTA on migration to the USA andalso somewhat of a puzzle for economic theory. Most trade economists’quick intuition about trade policy is derived from the HO model. In thatsimple, two-sector, two-good, two-country model, each country exports thegood whose production intensively uses its abundant factor. This appar-ently contradicts the Mexican situation if we think of two goods, maize and


Table 10.4. Canadian, Mexican and US maize production, 1992–6

Area Production Yieldharvested (000 metric (metric tons

Country Year/Growing season (000 hectares) tons) per hectare)

Canada 1994 ,955 7,043 7.41993 ,985 6,501 6.61992 ,858 4,883 5.7

Mexico 1996 5,921 14,000 2.41996 Spring/Summer 5,059 10,840 2.11995/1996 Fall/Winter ,862 3,160 3.7

1995 5,855 13,421 2.31995 Spring/Summer 4,748 9,678 2.01994/1995 Fall/Winter 1,107 3,744 3.4

1994 7,853 19,193 2.41993 7,536 18,648 2.51992 7,249 16,929 2.3

USA 1994 29,508 256,629 8.71993 25,464 160,954 6.31992 29,169 240,719 8.3

Note:Mexican data for 1955 and 1996 may not be strictly comparable with 1992–4 data.Sources: For 1992–4, Food and Agriculture Organization (1995); for 1995–6,Secretaria de Agricultura, Ganaderia, y Desarrollo Rural (1997).

a ‘composite’ good, and two factors, unskilled labour and a compositefactor. Trade liberalisation should raise the return to each country’s abun-dant factor, and hence unskilled labour in Mexico should benefit.

Few analysts seem to believe that the HO story is the right one. There isthe problem with maize and several other crops, as we just noted. We willattempt to model this more in section 5. But there are also doubts regard-ing the prospects for workers in Mexican manufacturing. Some believe thatliberalisation will tend to favour more-skilled Mexican workers rather thanthe less-skilled in manufacturing. These ideas will be discussed in sections3 and 4.

Three additional aspects of the Mexican economy further complicatethe evaluation of NAFTA. First, the Mexican population is muchyounger that that of its two NAFTA partners and is growing at a faster,albeit decelerating, rate (see table 10.5). In 1995, the median age inMexico was 21.7 years, in contrast to 34.7 in Canada and 34.2 in theUSA. Mexico implemented family-planning policies a decade or twolater than other middle-income countries such as Taiwan and SouthKorea, and this delay is reflected in the high growth rate of its popula-tion. Mexico has made tremendous strides in this area, and recent figuresindicate that the country’s population growth rate has slowed to 1.8 percent (Associated Press, 1996). Nevertheless, the large number of youngMexicans who enter the work force each year both expands the pool ofprospective migrants and is likely to exert a downward pressure onMexican wages.

Second, high rates of unemployment and under-employment are endur-ing features of the Mexican economy. During the last 10 years, Mexico’scombined rate of unemployment and under-employment has never dippedbelow 20 per cent, and the rate often exceeded 25 per cent (see table 10.6).In the face of such phenomena, along with the demographic characteristicsdescribed above, less-skilled Mexican workers may not experience dramaticincreases in their wages or employment levels, even if post-NAFTA changesin the relative demand for their labour conform to those found in the HOmodel.

Third, Mexico suffered a devastating macroeconomic disruption inlate 1994 and 1995, less than a year after NAFTA’s implementation.3 Wedo not believe that NAFTA was the primary cause of this crisis, but wedo view the agreement as one of a multitude of factors that entered intoplay. It is also important to note that the crisis abruptly terminated a six-year period during which real per capita income in Mexico rose mod-estly. Between 1988 and 1994, real income climbed from 17,327 to18,491 new pesos per capita, as expressed in 1995 prices (see table 10.6).This period of rising incomes falls neatly within the broader period of


Table 10.5. Population, median age and population growth in the three NAFTA countries, estimates and projections, 1970–2025

Total population (000) Median age (years) Population growth rate (per cent)

Year Canada Mexico USA Canada Mexico USA Period Canada Mexico USA

Estimates1970 24,324 250,328 205,051 26.0 16.6 27.9 1965–70 1.61 3.21 1.081975 22,727 258,876 215,972 27.5 16.6 28.7 1970–5 1.27 3.14 1.041980 24,070 267,046 227,757 29.3 17.5 30.0 1975–80 1.15 2.60 1.061985 25,181 275,594 238,466 31.4 18.6 31.5 1980–5 0.90 2.40 0.921990 26,639 284,486 249,975 33.1 20.0 32.9 1985–90 1.13 2.22 0.94

Projectionsa

1995 28,537 293,670 261,138 34.7 21.7 34.2 1990–5 1.38 2.06 1.032000 30,425 102,555 275,324 36.2 23.2 35.5 1995–2000 1.28 1.81 0.902005 32,293 110,810 285,931 37.2 24.8 36.6 2000–5 1.19 1.55 0.762010 34,070 118,455 296,089 37.9 26.5 37.4 2005–10 1.07 1.33 0.702015 35,674 125,484 305,622 38.6 28.3 38.1 2010–15 0.92 1.15 0.632020 37,125 131,885 314,541 39.2 30.1 38.8 2015–20 0.80 0.99 0.572025 38,356 137,783 322,007 39.9 32.0 39.5 2020–5 0.65 0.83 0.47

Note:aProjections are the medium-variant projections.Source: United Nations (1993).

Mexican liberalisation that we outlined above. However, even at the endof this six-year period, real per capita income still had not returned toits 1980 level.

We end this section on a note of caution. Unfortunately, it is much tooearly to assess the effects of NAFTA. After all, even given the variouscomplexities of the Mexican economy, the agreement is still only three yearsold. However, there is some hope for prompt empirical work on the subjectof Mexican liberalisation, insofar as many Mexican reforms predatedNAFTA by as much as 10 years. In any case, we expect that most expertswould concur that NAFTA is unlikely to have much of an effect in reduc-ing Mexico–US migration over at least the next decade.


Table 10.6. Selected economic statistics for Mexico, 1980–95

GDP Population Real GDP Under-employment(millions (mid-year per capita and unemploymentof new pesos, estimate, (new pesos, rate

Year 1995 prices) millions) 1995 prices) (per cent)

1980 1,323,964 69.66 19.006 —1981 1,429,169 71.35 20.030 —1982 1,420,066 73.02 19.448 —1983 1,360,750 74.67 18.224 —1984 1,408,615 76.31 18.459 —1985 1,444,146 77.94 18.529 —1986 1,391,585 79.57 17.489 —1987 1,416,837 81.20 17.449 25.61988 1,435,337 82.84 17.327 24.11989 1,482,613 84.27 17.594 21.81990 1,549,565 86.15 17.987 22.11991 1,605,943 87.84 18.283 21.21992 1,650.285 89.54 18.431 23.61993 1,661,285 91.21 18.219 23.51994 1,719,879 93.01 18.491 27.81995 1,600,953 94.78 16.891 27.2

Note:Under-employment and unemployment rate indicates proportion of economicallyactive population that was either unemployed or employed for less than 35 hoursper week in sampled urban areas. The figures displayed are for the second quarterof the given calendar year.Sources: For GDP and population figures, International Monetary Fund (1997); for under-employment and unemployment rate, INEGI, EncuestaNacional de Empleo, as cited in INEGI (1997).

3 Investment liberalisation and income distribution: I – theFeenstra–Hanson model

In this section and the next, we focus largely on manufacturing. We con-sider some evidence and associated theory that suggests that NAFTA maydo little to help less-skilled workers in Mexico and hence have little effecton the tendency for these workers to seek jobs in the USA over the short tomedium term. Evidence is present in a number of studies, includingFeenstra and Hanson (1995a, 1995b), Aitken et al. (1994), Aitken et al.(1995), Hanson and Harrison (1995) and OECD (1996a), that foreign firmspay higher wages, with a higher share of wages going to more-skilledworkers, and that wages gaps in manufacturing between skilled andunskilled workers have increased since Mexican liberalisation began in themid-1980s.4

Table 10.7 reproduces data from Feenstra and Hanson (1995b). They useproduction and non-production workers as proxies for skilled andunskilled workers which is subject to limitations of which they are surelyaware. The table shows the ratio of non-production to production wagesand the share of non-production wages in total wages. The years 1970–85were years of high protection in Mexico, with substantial liberalisationsbeginning about 1984. The data indicate that the ‘wage gap’ (the ratio ofnon-production to production wages) fell in all regions between 1975 and1985, and that the share of non-production wages rose modestly. The bigchange occurred after 1985, with the wage gap rising significantly in allregions and the non-production share rising in all regions except the North.The largest increases in both measures is in the Mexico–US border region,where liberalisation in both trade and investment barriers had by far thelargest impact.

Tables 10. 8 and 10.9 present more recent data on changing compensa-tion levels in the Mexican economy. Between 1985 and 1993, the overallpattern is one of shrinking levels of real compensation per employee from1985 to 1988 and recovering levels thereafter (see table 10.8). The netincrease in the average annual compensation per salaried employee overthis eight-year period was a mere 1.8 per cent. Workers in agriculture faredextremely poorly. The average level of real anuual compensation in thesector dropped by 35.8 per cent between 1985 and 1993, to a meagre 2,776new pesos per salaried employee. Workers in construction and commercealso experienced a net decline in their average compensation, while manu-facturing employees received a net increase.

More detailed evidence pertaining to the manufacturing sector is foundin table 10.9. These data indicate that the average real compensation levelin Mexican manufacturing climbed 45.0 per cent between 1987 and 1994,


from 3,180 to 4,610 new pesos per month, as measured in June 1995 prices.The three manufacturing sectors that provided the highest levels ofcompensation in 1994 are also ones that require workers with particulartechnical skills: chemicals, petroleum derivatives and rubber and plasticproducts (V); non-metallic mineral products (VI); and basic metal indus-tries (VII).

Table 10.9 also contains separate measures of compensation for produc-tion and non-production workers. For manufacturing as a whole, thecompensation gap between the two groups widened between 1987 and1994, from 1,830 to 3,820 new pesos per month. In the three manufactur-ing sectors mentioned above, as well as in metal products, machinery andequipment (VIII), the difference in compensation between production and


Table 10.7. Relative wage and wage shares, by Mexican region, 1975–88

Non-production wage/ Non-productionProduction wage share of total wages

Change ChangeRegion Year Level (per cent) Level (per cent)

Border 1975 2,104 0,3421980 2,048 20.537 0,365 0.4771985 2,073 0.245 0,373 0.1541988 2,517 6.464 0,415 1.398

North 1975 1,963 0,3041980 1,964 0.011 0,335 0.6321985 1,813 21.599 0,358 0.4531988 2,085 4.659 0,353 20.178

Centre 1975 1,838 0,3131980 1,824 20.156 0,330 0.3291985 1,719 21.181 0,341 0.2301988 2,085 3.048 0,363 0.733

Mexico City 1975 2,145 0,4161980 2,022 21.185 0,410 20.1171985 1,772 22.634 0,435 0.4821988 2,137 6.237 0,466 1.055

South 1975 2,090 0,2881980 1,518 26.400 0,292 0.0751985 1,530 0.159 0,313 0.4251988 1,699 3.490 0,330 0.568

Source: Feenstra and Hanson (1995b, table 2).

Table 10.8. Average annual compensation, by sector, 1980–93 (new pesos per salaried employee, 1995 prices)

FinancialTransportation services, Community

Agriculture, Electricity Commerce, warehouses, insurance social andpoultry Manu- gas and restaurants and commu- and personal

Year Total and fishing Mining facturing Construction water and hotels nication real estate services

1980 22,290 4,658 40,562 37,332 26,912 282,998 23,031 31,023 51,635 26,6621981 29,892 6,376 52,917 49,076 34,744 108,897 29,836 41,712 68,672 36,2461982 25,753 5,391 43,118 43,192 27,929 101,189 24,225 34,634 58,987 31,8501983 20,783 4,563 34,775 35,055 23,705 279,547 20,826 30,409 47,145 25,1491984 18,010 3,957 29,579 30,012 20,249 261,940 17,711 26,233 41,097 22,0631985 19,571 4,321 35,545 32,782 21,958 264,072 19,040 28,756 45,385 23,8321986 17,633 4,315 30,383 30,028 19,508 255,706 17,093 26,290 42,301 20,7151987 17,352 4,030 30,074 29,883 18,911 261,516 16,197 25,921 44,097 20,5751988 15,810 3,433 28,248 29,709 16,587 251,473 14,323 25,629 40,066 18,4381989 16,606 3,281 27,897 31,681 15,838 251,435 14,756 25,592 42,518 19,9951990 17,076 3,143 27,056 32,820 15,743 251,401 14,696 24,496 46,280 20,5231991 17,643 2,911 27,091 34,562 16,036 253,582 15,017 24,965 50,685 21,7001992 (P) 19,009 2,831 29,129 37,502 16,709 259,532 15,306 26,171 56,879 24,0971993 (P) 19,922 2,776 27,044 38,952 17,219 264,894 16,057 26,531 63,396 26,030

Note:P5Provisional.Source: INEGI, Dirección General de Contalbilidad Nacional, Estudios Socioeconómicos y Precios; and Dirección de ContabilidadNacional, Sistema de Cuentas Nacionales de Mexico (various years); both as cited INEDI (1996), Cuadro 4.5.

Table 10.9. Monthly compensation per worker in Mexican manufacturing, 129 classes of activity, 1987–95

Food products, beverages Textiles, clothing and Wood industriesAll manufacturing and tobacco (I) leather industry (II) and wood products (III)

Non- Non- Non- Non-Production production Production production Production production Production production

Year Total workers workers Total workers workers Total workers workers Total workers workers

1987 3.18 1.71 3.54 2.76 1.46 2.74 2.14 1.49 2.48 2.05 1.35 2.531988 3.26 1.70 3.71 2.47 1.51 2.76 2.21 1.49 2.51 1.97 1.21 2.411989 3.53 1.76 4.10 2.80 1.59 3.21 2.41 1.57 2.78 2.05 1.22 2.771990 3.64 1.79 4.41 2.85 1.60 3.45 2.35 1.45 2.92 2.17 1.31 2.891991 3.96 1.79 4.87 3.08 1.62 3.71 2.49 1.43 3.18 2.27 1.33 3.201992 4.21 1.91 5.32 3.34 1.68 4.04 2.60 1.49 3.55 2.35 1.34 3.261993 4.43 1.97 5.64 3.75 1.80 4.42 2.66 1.49 3.63 2.49 1.34 3.451994 4.61 2.05 5.87 3.90 1.83 4.66 2.82 1.56 3.79 2.47 1.24 3.741995 (P) 3.99 1.66 5.08 3.26 1.52 3.72 2.16 1.18 2.88 1.89 0.95 2.77

Paper, paper products, Chemicals, petroleum Non-metallic mineral productsprinted matter and editorial derivatives, and rubber and except those derived from

material (IV) plastic products (V) petroleum or carbon (VI)

Non- Non- Non-Production production Production production Production production

Year Total workers workers Total workers workers Total workers workers

1987 3.27 1.94 3.51 4.03 2.01 4.11 3.93 2.01 4.631988 3.51 1.81 3.97 4.21 2.05 4.36 4.14 2.00 4.921989 3.66 1.93 4.10 4.53 2.16 4.84 4.21 1.91 5.241990 3.61 1.81 4.39 4.68 2.11 5.02 4.52 1.93 5.851991 5.21 1.85 4.76 4.97 2.19 5.45 4.92 1.96 6.431992 4.18 1.98 5.41 5.60 2.39 6.08 5.36 2.12 7.281993 4.15 1.97 5.42 5.65 2.30 6.55 5.59 2.24 7.871994 4.16 1.90 5.53 6.05 2.44 6.85 5.78 2.36 7.791995 (P) 3.45 1.60 4.65 5.47 2.08 6.13 5.16 2.02 7.35

Metal products, machinery, Other manufacturingBasic metal industries (VII) and equipment (VIII) industries (IX)

Non- Non- Non-Production production Production production Production production

Year Total workers workers Total workers workers Total workers workers

1987 4.19 2.11 3.99 3.22 1.63 3.63 2.47 1.26 3.441988 4.14 2.14 4.49 3.23 1.57 3.71 2.40 1.22 3.401989 4.55 2.08 4.57 3.52 1.63 4.17 2.64 1.29 3.781990 4.74 2.04 5.02 3.69 1.84 4.60 2.58 1.27 3.961991 5.27 2.01 5.99 3.91 1.80 5.19 2.81 1.32 4.271992 5.78 2.23 6.31 4.18 1.93 5.59 3.09 1.53 4.721993 6.26 2.30 6.67 4.46 2.05 5.94 3.41 1.68 5.061994 6.05 2.41 6.78 4.58 2.13 6.24 3.56 1.65 5.201995 (P) 5.22 2.08 5.66 3.97 1.62 5.53 3.22 1.29 4.39

Notes:Observations are for the month of June of each calendar year. Figures for total compensation per worker include benefits, while those forproduction and non-production workers do not.P5Provisional.Source: INEGI, Encuesta Industrial Mensual, as cited in INEGI (1997).

non-production workers exceeded 4,000 new pesos per month in 1994, witha gap of 5,430 new pesos existing in the non-metallic mineral productssector.

Two broad patterns in wage movements are thus apparent during thefirst decade of Mexican liberalisation (1985–94). First, wages divergedacross different sectors of the Mexican economy. Compensation levels inagriculture, construction and commerce actually shrank during the1985–93 period, and these sectors have high concentrations of less-skilledworkers. A similar pattern of divergence existed among various sectorswithin manufacturing. Second, the compensation gap for production andnon-production workers within manufacturing widened, particularly insectors where overall compensation levels increased the most. Finally, we


Table 10.10. Foreign investment in Mexico, 1980–95 (millions of USdollars)

Direct foreign investment (flows) Net foreign investment (stock)

Year Total USA Canada Total USA Canada

1980 1,622.8 1,078.6 17.5 8,458.8 5,836.6 126.91981 1,701.1 1,072.1 5.2 10,159.9 6,908.7 132.11982 ,626.5 ,426.1 8.1 10,786.4 7,334.8 140.21983 ,683.7 ,266.6 22.1 11,470.1 7,601.4 162.31984 1,429.8 ,912.0 32.5 12,899.9 8,513.4 194.81985 1,729.0 1,326.8 34.9 14,628.9 9,840.2 229.71986 2,424.2 1,206.4 40.6 17,053.1 11,046.6 270.31987 3,877.2 2,669.6 19.3 20,930.3 13,716.2 289.61988 3,157.1 1,241.6 33.9 24,087.4 14,957.8 323.51989 2,499.7 1,813.9 37.4 26,587.1 16,771.7 360.91990 3,722.4 2,308.0 56.1 30,309.5 19,079.7 417.01991 3,565.0 2,386.1 74.2 33,874.5 21,465.8 491.21992 3,599.6 1,651.7 88.4 37,474.1 23,117.5 579.61993 4,900.1 3,503.6 74.2 42,374.2 26,621.1 653.81994 8,026.2 4,004.5 163.5 50,400.4 30,625.6 817.31995 6,534.4 4,176.3 77.9 — — —

Notes:Net foreign investment does not include investments in the stock market norforeign capital derived from authorisations granted by the Comisión Nacional deInversiones to firms that participate in the stock market.1995 figures for direct investment include certain maquiladora imports.Source: SECOFI, Dirección General de Inversión Extranjera, as cited in INEGI(1997).

note that there was a large jump in foreign investment following liberal-isation after 1985 (table 10.10), coincident with the wage movements justdiscussed.

Econometric analysis by Feenstra and Hanson provides strong supportfor the hypothesis that foreign investment leads to an increase in the wagegap and the non-production wage share in Mexican manufacturing. Butthis is again apparently at odds with the HO model, insofar as less-skilledlabour would seem to be Mexico’s abundant factor. Furthermore, it hasbeen widely observed that the wage gap is rising in the USA (reviews of theliterature and evidence are found in Freeman, 1995; Richardson, 1995;Wood, 1994, 1995). According to HO theory, the factor–price ratios shouldmove in opposite directions in the two countries, as the return to the abun-dant factor rises in each country.

Feenstra and Hanson provide an intriguing model, and indeed a modelthat is very much in the tradition of HO, to explain this phenomenon. Themodel has three factors: capital, skilled labour and unskilled labour. Thereis a single composite consumption commodity that is ‘assembled’ from acontinuum of intermediate inputs. All these intermediate inputs have thesame capital intensity but may be ranked according to skilled-labour inten-sity. The authors assume that initially Mexico has a higher return to capitalthan the USA. There is a dividing point in the continuum of intermediates,with less-skilled-labour-intensive goods produced in Mexico and more-skilled-labour-intensive goods produced in the USA.

The effect of investment liberalisation is to move capital from the USAto Mexico. This lowers the cost of producing all intermediates in Mexicoand raises it in the USA. In equilibrium, it shifts the dividing line in the con-tinuum toward Mexico. More goods are produced in Mexico. The situationis illustrated in figure 10.1, where the horizontal axis indexes the commod-ities, with the least-skilled-labour-intensive goods on the left. There is someinitial (pre-liberalisation) dividing line in the continuum as shown.Liberalisation shifts this dividing line to the right. But now we have thecrucial insight. The goods shifted are relatively skilled-labour-intensivefrom Mexico’s point of view (they are to the right of the goods that Mexicowas producing), but they are unskilled-labour-intensive from the US pointof view (they are to the left of the goods that the USA continues toproduce). The effect is thus to raise the relative demand for skilled labourin both Mexico and the USA. The relative wage of skilled labour will risein both countries in equilibrium.

Empirical results in the chapter provide support for the simple, ingeni-ous model. However, they do not provide support for the notion thatNAFTA will mitigate the problem of unskilled-labour migration to theUSA.


4 Investment liberalisation and income distribution: II – theMarkusen–Venables model

A second approach to the wage-gap issue explicitly considers the role andstructure of multinationals (Markusen and Venables, 1995, 1997a, 1997b;Markusen, 1997). Some features of the basic model, expanded in the latterpaper to include vertical multinationals, are as follows:

1. Two homogeneous goods, X and YTwo countries, h and tTwo factors, unskilled labour: L skilled labour: S

2. Y – competitive, constant returns to scale, L-intensive3. X – imperfectly competitive, increasing returns to scale, S-intensive

overall ‘Headquarters’ and ‘plant’ may be geographically separated; Afirms may have plants in one or both countries

4. There are six firm types, with free entry and exit into and out of firmtypes; regime denotes a set of firm types active in equilibrium:

Type mh – horizontal multinationals which maintain plants in both coun-tries, headquarters is located in country hType mf – horizontal multinationals which maintain plants in both coun-tries, headquarters is located in country fType nh – national firms that maintain a single plant and headquarters incountry h; type-nh firms may or may not export to country f


Figure 10.1 Investment liberalisation and wage gaps: the Feenstra–Hanson model

Type nf – national firms that maintain a single plant and headquarters incountry f; type-nf firms may or may not export to country hType vh – vertical multinationals that maintain a single plant in countryf, headquarters in country h; type-vh firms may or may not export tocountry hType vf – vertical multinationals that maintain a single plant in countryh, headquarters in country h; type-vf firms may or may not export tocountry f

Crucial to the story are assumptions concerning the factor intensities ofvarious activities. In this we draw indirect inferences from a number ofempirical sources, including Feenstra and Hanson (1995a, 1995b),Blomstrom and Wolff (1994); Slaughter (1995). First, assume that X-sectorproduction is skilled-labour-intensive overall relative to the Y-sector. Thesecond crucial assumption is that branch plants (e.g. plants of US firms inMexico) are more skilled-labour-intensive than the Y-sector but lessskilled-labour-intensive than local integrated X-producers. The branchplants need local managers, engineers, technicians and so forth, and theserequirements make them more skilled-labour-intensive than composite Y-production. But much of the branch plants’ skilled-labour requirements arenevertheless supplied from the home firm (e.g. the US parent). There is atransfer of ‘producer services’ from the home firm to the subsidiary in theform of research and development and other assets. The third assumptionis that two-plant multinational firms require more skilled labour in theirheadquarters than the one-plant firms. These represent the ‘technologytransfer’ costs of doing business abroad. Fourth, headquarters’ activitiesare more skilled-labour-intensive than a production plant (indeed, head-quarters use only skilled labour in the model). The full set of factor-inten-sity assumptions are as follows:

Factor-intensity assumptions: ranked from most skilled-labour-intensive toleast skilled-labour-intensive

Activities[headquarters]. [integrated X]. [branch plant]. [Y]

Firm Types[type-m firms]. [type-v and type-n firms]

Operations within a country[local type-v firm]. [local type-m firm]. [local type-n firm].

[plant of foreign type-m or type-v firm]. [Y]

The consequences of trade and investment liberalisation in this modeldepend very much on the initial parameters. If trade costs are high, invest-ment liberalisation involves the creation of type-m firm if the two countries


are not extremely different, with headquarters concentrated in theskilled-labour-abundant country. If trade costs are low, investment liber-alisation results in the entry of type-v firms if countries differ significantlyin relative endowments, with their headquarters in the skilled-labour-abundant country and their single plant in the unskilled-labour-abun-dant country.

Consider a parameterisation of the model which resembles theUS–Mexico situation, with the countries referred to as the north and thesouth. The north is skilled-labour-abundant and large, and the south isunskilled-labour-abundant and small. Figure 10.2 presents a generaloutline of the effects of investment liberalisation given such an initial situa-tion. These results are qualitatively independent of whether or not trade isalso liberalised or restricted. The horizontal axis of Figure 10.2 ranks threeactivities according to their skilled-labour intensity, Y-production beingthe least skilled-labour-intensive, then final X-production and then head-quarters’ services.

In the initial protected situation in which multinationals cannot exist, thesouth produces little or no X, since it is severely short of skilled labour forheadquarters’ activities. Investment liberalisation leads to branch plants oftype-m (higher trade costs) or type-v firms (lower trade costs) headquar-tered in the north. But since the south was not producing much or any Xinitially, resources for the branch plants are drawn from the Y-sector. But


Figure 10.2 Investment liberalisation and wage gaps: the Markusen–Venablesmodel (activity shifts)

the branch plants are more skilled-labour-intensive than Y, since the plantsrequire managers, engineers, technicians and so forth, as we noted above.Thus investment liberalisation increases the demand for skilled labour inthe south. Under the assumptions noted, local skilled labour is a comple-ment, not a substitute, for the imported producer services produced withnorthern skilled labour. The relative wage of skilled labour can rise in thesouth in equilibrium.

The situation in the north is perhaps more intuitive. It is the skilled-labour-abundant region, and initially it has a relatively low price for skilledlabour. Investment liberalisation leads to a shift of headquarters toward thenorth and a shift of X-production toward the south. Even if all headquar-ters are initially in the north (no southern X-production), liberalisation isanalogous to a cost reduction or positive technical change, so the sectorexpands with more headquarters’ activities in the north in the new equilib-rium. This, of course, increases the relative demand for skilled labour in thenorth, with some of the unskilled labour released from the X-productionbeing ‘soaked up’ by the Y-sector.

Figure 10.3 presents for the sake of completeness the ‘world’ factor box,with country h measured from the southwest corner and the country f fromthe northeast corner. This is a rough composite of the results of investmentliberalisation for different levels of trade costs. We see substantial regions


Figure 10.3 Investment liberalisation and wage gaps: the Markusen–Venablesmodel (factor-price effects)

in which the wage of skilled labour rises in both countries. This is particu-larly likely to occur when one country is both larger and skilled-labour-abundant. If we are near the top and to the right of centre of the box, forexample, country h is the north (e.g. the USA) and country f is the south(e.g. Mexico).

One final point should be noted. In this type of model, it is quite possi-ble for the efficiency gains associated with liberalisation to lead to anincrease in the prices of both factors. Still, the overall message of this modelis not to assume that the HO mechanism will operate in favour of unskilledlabour in the south and decreased incentives for migration.

5 Technology and the maize sector

In this section, we focus on some aspects of the maize sector in Mexico,since the reforms in that sector plus trade liberalisation through NAFTAare possibly the greatest force operating to send more, not fewer, unskilledMexicans northward. The first feature about this sector is that it is verylabour-intensive in Mexico and very capital-intensive in the USA. While wedo not have precise statistics on this point, we believe that is true not just inthe cross-country comparison, but also within in each country. That is,cereal grain production in the USA is relatively capital-intensive relative tomany other US tradable sectors, and maize production is labour-intensivein Mexico relative to other tradable sectors there.

This suggests a model based on multiple techniques of production inthe maize sector. A very capital-intensive technique exists (and is clearlyavailable to Mexican farmers) and a labour-intensive technique exists.The technique adopted in a country depends on factor prices, which inturn depend on relative endowments. The situation is shown in figure10.4, for an initial equilibrium with positive protection. There is theagricultural sector and a composite Y-sector. The unit value isoquantfor Y lies between the unit-value isoquants for the more capital-inten-sive technique and the labour-intensive technique. There exist two‘cones of diversification’ in the diagram, one producing Y and agricul-ture with the capital-intensive technique and one producing Y and agri-culture with the labour-intensive technique. We assume in the initialequilibrium that the north’s factor endowment lies in the former (upper)cone and that the south’s factor endowment lies in the latter (lower)cone.

The initial factor–price ratios are as shown in figure 10.4, with the northhaving a relatively high wage–rental ratio and the south having a lowwage–rental ratio. Note that from each country’s point of view, the unusedtechnique is not profitable at equilibrium factor prices. We assume that in


the initial protected equilibrium, the north exports agriculture and importsthe composite (in fact, Mexican protection was so high that there was vir-tually no trade in maize).

Now consider tariff reduction for one or both countries, using Y as thenumeraire for expositional convenience. The results are shown in figure10.5. The price of agriculture (maize) rises in the north, shifting its unit-value isoquant inward. The price of maize falls in the south, shifting its unitvalue isoquant outward. The effect of these changes on the relative pricesof factors moves in the same direction in the two countries. The wage–rentalratio falls in both the north and the south because, from each of their pointsof view, the price of the capital-intensive good has risen. Mexican labourmay of course still gain somewhat relative to US unskilled labour, but theeffect is not going to be dramatic and could go the other way.

In order to verify that such an outcome is indeed possible, we constructeda simple numerical example using Rutherford’s (1994, 1995) non-linearcomplementarity solver. Complementarity is necessary for the problem,because we need to verify that each country does not choose to use theother available technique and more generally that the proposed solution isindeed the equilibrium. The model was benchmarked initially with 20 percent import tariffs in both directions.


Figure 10.4 Competitive model with multiple techniques of agriculturalproduction

Results for the simulations are shown in table 10.11, where the first rowfor tariff 0.20 is the benchmark replication. Factor prices are reported inreal terms, the nominal price divided by the consumer price index (the unitexpenditure function). At all levels of protection, neither country shifts tothe alternative maize technique. Results of the simulations are as suggestedin figure 10.5. Tariff reduction reduces the real wage of labour in both coun-tries, but raises the real return to capital. The real wage is reduced some-what proportionately less in the south, but there is little wage convergence.Specialisation is reached at a tariff rate of about 0.08. Note that for furtherreductions in the tariff, the relative factor prices in the two countries remainunchanged, but the real prices of all factors rise somewhat, reflecting thecapture of further gains from trade (eliminating the consumption distor-tions even though production ceases to change). The countries aresufficiently different in this example that there is no factor-price equalisa-tion (FPE) at free trade.

This example is suggestive, but obviously needs some empirical work.But as in the case of the Feenstra–Hanson model, it is firmly rooted in tradi-tional HO theory, if differing from the very narrow HO model. The latterhas been used by several authors to argue that factor-proportions tradetheory is false, because it is inconsistent with the stylised facts.


Figure 10.5 Competitive model with multiple techniques of agriculturalproduction: trade liberalisation

A second way of thinking about the maize sector uses the specific-factorsmodel, drawing on ideas in Markusen (1983) and Neary (1995). Supposethat we once again view the Mexican economy as having a maize sector anda composite sector. Maize uses a specific factor, R, and the composite sectoruses a specific factor, K. Labour is homogeneous, and can be used in maize(M) or the composite sector (C) or it can migrate to the USA. Assume thatinitially, trade protection raises the price of maize and that investmentrestrictions limit foreign investment in C.

Figure 10.6 gives the familiar cross-diagram, with the value-of-marginal-product curves for maize and the composite, with the labour input to maizemeasured from the left-hand axis and the labour input to C measured fromthe right-hand axis. The heavy lines give the initial marginal product curvesin the protected equilibrium. The twist on the standard story is that thereis an ‘outside option’, indicated by the US wage, drawn as a horizontal linein figure 10.6. The initial equilibrium allocation of Mexican labour betweenM, C and migration is given by the intersection of the M and C value-of-marginal-product curves with the US wage line.

There are some complicated issues here about price indices in determin-ing real wages, such as whether or not migrant workers use US prices orMexican prices to evaluate their wages. Indeed, one should probably usedifferent urban and rural price indices within Mexico. In order to get theidea across, we will ignore this important difficulty here and assume that


Table 10.11. Simulation results for the alternative-technologies model

Tariff wh wf rh rf wh/rh wf /rf

0.20 2.00 1.00 1.00 2.00 2.00 0.500.18 1.91 0.98 1.03 2.08 1.86 0.470.16 1.83 0.95 1.06 2.17 1.73 0.440.14 1.74 0.93 1.09 2.25 1.61 0.410.12 1.66 0.91 1.12 2.35 1.47 0.390.10 1.58 0.89 1.15 2.44 1.37 0.360.08 1.56 0.89 1.17 2.49 1.33 0.360.06 1.57 0.90 1.18 2.52 1.33 0.360.04 1.59 0.91 1.19 2.54 1.33 0.360.02 1.60 0.92 1.20 2.56 1.33 0.360.00 1.62 0.93 1.21 2.59 1.33 0.36

Notes:a Factor prices are in real units: price divided by the consumer price index.b Specialisation is reached at a tariff near 0.08. After that point, tradeliberalisation does not affect relative factor prices in the two countries.

the same price index is used regardless of whether one is a C-worker, an M-worker or a migrant. This assumption in turn allows us to ignore the priceindex altogether: the price index affects real income, but not the allocationof labour among the three activities.

Adopt the price of C, as the numeraire. Trade liberalisation drops theprice of the protected good, in this case maize. This is shown as a down-ward shift in the value-of-marginal-product curve (the price of maize timesthe physical marginal product) for maize in figure 10.6. Under the assump-tion that the US wage does not change, the new equilibrium must involve ashift in Mexican labour from the rural maize sector to the USA, with nochange in the labour allocated to C. Trade liberalisation worsens the migra-tion problem.

Investment liberalisation, on the other hand, shifts up the value-of-mar-ginal-product curve for C in figure 10.6 by bringing sector-specific capitalinto the C-sector. The C-sector expands and all of the expansion is in theform of reduced labour migration, rural employment is held constant.Investment liberalisation can in this way relieve some of the migration pres-sure.

Many variations on this theme can be presented. Obviously, manyfactors are ignored, including the price-index issue and the issue of themultiple types of labour which formed the focus of section 4. Our purposehere is not to suggest what the exact effects of NAFTA might be, butrather to suggest some empirically relevant possibilities, as we notedearlier. Figure 10.6 also serves the function of emphasising that trade and


Figure 10.6 Specific-factors model with an outside opportunity

investment liberalisation might have quite different effects, contrary to thesymmetry between the two in the HO model.

A final suggestion about the agriculture–migration dilemma is drawnfrom Martin (1993, 1996). He notes the importance of public-sector infra-structure in agriculture, particularly roads and other transportation inputs.It is somewhat unclear to us what the general proposition is, but it seems tobe that these are more important in agriculture than in maquiladora-typemanufacturing plants in the US–Mexico border region, and that is clearlyplausible. The second point is that this infrastructure is far better developedin the USA than in Mexico. The lack of infrastructure in Mexico is analo-gous to having a poorer technology in that sector relative to the USA.Figure 10.7 illustrates the ideal with two production frontiers, one with ahigh level of public transport-sector capital and one with a low level. Maize(or agriculture more generally) is assumed to be more sensitive to thiscapital stock, and that turns into a source of comparative advantage anddisadvantage.

Markusen (1983) notes that in such a situation in which one countryhas a superior technology in one sector with all other things being equal,the real return to the factor used intensively in that sector will be higherin the advanced country. In the present situation, the real wage tounskilled rural Mexican workers will be less than the corresponding real


Figure 10.7 Model with a public intermediate good: public infrastructure

wage in the USA. This difference may not be eliminated by free trade, evenif Mexico is abundant in unskilled labour (specialisation is now a neces-sary condition for FPE). It is only by moving Mexican workers to theUSA to work with US ‘technology’ (infrastructure) that wages can beginto equalise. Martin presents evidence that this productivity difference isrelevant, with Mexican workers being significantly more productive in theUSA on the same crops.

6 Summary and conclusions

The purpose of this chapter is to consider the migration of unskilled, ruralMexican workers to the USA and how migration incentives may be alteredby NAFTA. We have not yet done any formal empirical work, and it is fartoo early to infer such effects from an agreement that has just turned fouryears of age, even though Mexican liberalisation began in the mid-1980s, aswe noted. This task is further compounded by a major macroeconomic dis-ruption in Mexico in late 1994 and 1995. Yet what evidence does exist plusthe opinions of many experts suggests that NAFTA is not likely to havemuch of an effect on the incentives to migrate.

The chapter therefore more or less accepts this conclusion and inquiresinto the reasons why it might be true. After all, it apparently contradictsour most cherished trade model, the HO model, despite the considerablebroadening of the theory. Several models are presented, all of whichimply that NAFTA may not raise the wages of unskilled Mexicanworkers very much relative to their potential wages as legal or illegalimmigrants to the USA. While there are potentially an unlimited numberof such models, we believe that the ones we present are all empiricallyplausible and relevant.

The first one is due to Feenstra and Hanson and involves a continuumof goods ranked by their intensity in skilled labour. The effect of invest-ment liberalisation is to move to the south production of goods whichare skilled-labour-intensive from the south’s point of view, butunskilled-labour-intensive from the north’s point of view, thereforeraising the relative demand for skilled labour in both countries. Thesecond one is by Markusen and Venables, where the ‘unbundling’ ofactivities permitted by investment liberalisation raises the relativedemand for skilled labour in both countries (but may increase the realincomes of all factors as well). The final three models focus on the maizesector and consider the role of production technologies, specific factorsand public infrastructure.

We might note again that, with the exception of the Markusen–Venablesmodel, all of these explanations are solidly within the tradition of HO, if


differing from the very specific model of that name. This is of some impor-tance insofar as some economists have used the inconsistency of certainstylised facts within the HO model to dismiss all factor-proportions tradetheory. More exotic explanations are then sought in terms of industrial-organisation features and/or ‘technical change’. The latter is almost neverestimated, an unexplained residual is just defined to be technical change. Atthis point, all models remain candidates for explaining the wage-gap phe-nomenon, and of course all of them could plausibly be contributing to theobserved data.

NOTES

Markusen acknowledges support from a National Science Foundation grant.Zahniser acknowledges a grant from the Social Science Research Council’sInternational Migration Program, funded by the Andrew W. Mellon Foundation.The authors also thank Rose Ellen Flandes for her assistance in collecting statisticsused in this chapter.1 We distinguish between legal immigration, in which persons obtain legal residency

and even citizenship from their new country of residence, and undocumented orillegal migration, in which persons enter a country without its government’sexpressed permission and reside there for an indefinite period of time.

2 The available statistics measuring civilian employment in Mexico do not allow aneasy comparison with years prior to 1991. Specifically, some sort of methodolog-ical change appears to have occurred between the generation of the 1990 and the1991 statistics, perhaps in conjunction with the 1991 Mexican census. In fact, thedata for years prior to 1991 probably under-estimate agricultural employment inMexico, as OECD (1996b) indicates that civilian employment in agriculture was5,300,000 in 1990 and 7,532,000 in 1991. The Mexican data for five of the othereconomic activities listed in table 10.2 also experience a profound shift between1990 and 1991.

3 Although the crisis manifested itself in full force in December 1994, when theMexican government sharply devalued the peso, various statistics presented inthis chapter indicate that the Mexican economy was slowing down during thecourse of 1994.

4 For related labour market issues, see Hinojosa-Ojeda and McCleery (1992);Hinojosa-Ojeda and Robinson (1992); Leamer (1993); Tan and Batra (1995);Brainard and Riker (1995); and Riker and Brainard (1996).

REFERENCES

Aitken, B., A. Harrison and R. E. Lipsey (1995). ‘Wages and Foreign Ownership:A Comparative Study of Mexico, Venezuela and the USA’, Journal ofInternational Economics, 40, 345–76

Aitken, B., G. Hanson and A. Harrison (1994). ‘Spillovers, Foreign Investment, andExport Behavior’, NBER, Working Paper, 4967

Associated Press (1996). ‘Population Rises to Estimated 93 million’, 16 December


Blomstrom, M. and E. Wolff (1994). ‘Multinational Corporations and ProductivityConvergence in Mexico’, in W. Baumol, R. Nelson and E. Wolff (eds.),Convergence of Productivity: Cross–National Studies and Historical Evidence(Oxford: Oxford University Press)

Brainard, S. L. and D. Riker (1995). ‘Are US Multinationals Exporting US Jobs?’,MIT, Working Paper

Burfisher, M. E., S. Robinson and K. E. Thierfelder (1994). ‘Wage Changes in aUS–Mexico Free Trade Area: Migration Versus Stolper–Samuelson Effects’, inC. Shields and J. Francois (eds.), Modelling Trade Policy: AGE Models of NorthAmerican Free Trade (Cambridge: Cambridge University Press), 195–209

Feenstra, R. C. and G. H. Hanson (1995a). ‘Foreign Investment, Outsourcing, andRelative Wages’, in R. C. Feenstra, G. M. Grossman and D. A. Irwin (eds.),Political Economy of Trade Policy: Essays in Honor of Jagdish Bhagwati(Cambridge, MA: MIT Press)

(1995b). ‘Foreign Direct Investment and Relative Wages: Evidence fromMexico’s Maquiladoras’, NBER, Working Paper, 5122

Food and Agriculture Organization (FAO) (1995). FAO Yearbook, Production,1994, vol. 48, FAO Statistics Series, 125 (Rome: FAO)

Freeman, R. B. (1995). ‘Are Your Wages Set in Beijing?’, Journal of EconomicPerspectives, 9, 15–32

Hanson, G. H. and A. Harrison (1995). ‘Trade, Technology, and Wage Inequality’,NBER, Working Paper, 5110

Hinojosa–Ojeda, R. and R. McCleery (1992). ‘US–Mexican Interdependence,Social Pacts, and Policy Perspectives: A Computable General-EquilibriumApproach’, in J. Bustamante, C.W. Reynolds and R. Hinojosa–Ojeda (eds.),US–Mexico Relations: Labour Market Interdependence (Stanford: StanfordUniversity Press)

Hinojosa–Ojeda, R. and S. Robinson (1992). ‘Labour Issues in a North American FreeTrade Area’, in N. Lustig, B. Bosworth and R. Lawrence (eds.), North AmericanFree Trade: Assessing the Impact (Washington, DC: Brookings Institution)

Immigration and Naturalization Service (INS) (various issues). StatisticalYearbook of the Immigration and Naturalization Service (Washington, DC:INS)

(1996). INS Home Page, Statistical Information, web site, last modified 17October, http://www.usdoj.gov/ins/public/stats

Instituto Nacional de Estadística, Geografía, e Informática (INEGI)(1997). Banco de Información Económica INEGI, web sitehttp://dgcnesyp.inegi.gob.mx/bie.html–ssi

(1995). Anuario Estadístico de los Estados Unidos Mexicanos, 1995International Monetary Fund (IMF) (1996). International Financial Statistics

Yearbook 1996 (Washington, DC: IMF)Leamer, E. E. (1993). ‘Wage Effects of a US–Mexican Free Trade Agreement’, in P.

M. Garber (ed.), The Mexico–US Free Trade Agreement (Cambridge, MA:MIT Press)

Markusen, J. R. (1983). ‘Factor Movements and Commodity Trade asComplements’, Journal of International Economics, 13, 341–56


(1997). ‘Asymmetric Effects of Trade and Investment Liberalisation’, WorkingPaper

Markusen, J. R. and A. J. Venables (1995). ‘Multinational Firms and the New TradeTheory’, NBER, Working Paper, 5036

(1997a). ‘Multinational Production, Skilled Labour, and Real Wages’, in R.Baldwin and J. Francois (eds.), Dynamic Issues in Applied Commercial PolicyAnalysis (Cambridge: Cambridge University Press)

(1997b). ‘The Role of Multinational Firms in the Wage-Gap Debate’, Review ofInternational Economics, 5, 435–51

Martin, P. L. (1993). Trade and Migration: NAFTA and Agriculture (Washington,DC: Institute for International Economics)

(1996). ‘Trade and Migration Linkages: The Case of NAFTA’, University ofCalifornia, Davis, Working Paper,

Neary, J. P. (1995). ‘Factor Mobility and International Trade’, Canadian Journal ofEconomics, special issue, 28, S4–S23

Organisation for Economic Cooperation and Development (OECD) (1996a).‘Evidence on Trade and Wages in the Developing World’, Development CentreTechnical Paper, 119

(1996b). Labour Force Statistics, 1974–1994 (Paris: OECD)Richardson, J. D. (1995). ‘Income Inequality and Trade: What to Think, What to

Conclude’, Journal of Economic Perspectives, 9, 33–56Riker, D. and S. L. Brainard (1996). ‘US Multinationals and Competition from

Low Wage Countries’, MIT, Working PaperRutherford, T. F. (1994). ‘Applied General-Equilibrium Modelling with MPS/GE

as a GAMS subsystem’, Working Paper(1995). ‘Extensions of GAMS for Complementarity Problems Arising in

Applied Economics’, Journal of Economic Dynamics and Control, 19,1299–1324

Secretaria de Agricultura, Ganadería, y Desarrollo Rural, Centro de EstadísticaAgropecuaria (1997) Centro de Estadística Agropecuaria web site,http://www.sagar.gob.mx/cea.htm

Schiff, M. (1996) ‘Trade Policy and International Migration: Substitutes orComplements’, in J. E. Taylor (ed.), Development Strategy, Employment andMigration: Insights from Models (Paris: OECD Development Centre)

Slaughter, M. J. (1995). ‘Multinational Corporations, Outsourcing, and AmericanRelative Wage Divergence’, NBER, Working Paper, 5253

Tan, H. and G. Batra (1995). ‘Technology and Industry Wage Differentials:Evidence from Three Developing Countries’ (Washington, DC: WorldBank)

United Nations (1993). World Population Prospects, The 1992 Revision (New York:UN)

Wood, A. (1994). North–South Trade, Employment and Inequality (Oxford:Clarendon Press)

(1995). ‘How Trade Hurt Unskilled Workers’, Journal of Economic Perspectives,9, 57–80


Discussion

PASQUALE M. SGRO

Chapter 10 by Markusen and Zahniser argues that NAFTA and morerecent regional trade agreements (RTAs) have a different focus to the olderagreements, which were essentially concerned with trying to achieve largeinternal markets to capture economies of scale. Among the reasons for suchagreements, which also now involve countries with different levels ofdevelopment, might be the motive to reduce immigration pressures sincemigrants from LDCs may create various kinds of social and economicproblems in the developed country. If this is indeed a motive for NAFTA,it is rather ironic that in all these regional agreements one of the importantclauses is the freeing up of cross-border migration and trade. This is yetanother example of the gap between rhetoric and practice. The chapter goeson to consider the theoretical link between trade and investment liberalisa-tion on the one hand and the incentives for the migration of less-skilledlabour on the other. Will there be a convergence of wages between Mexicoand the USA, and hence less migration? Putting this proposition anotherway, will trade and investment substitute for migration?

The chapter postulates four particular reasons that the USA had forforming NAFTA, one of which is to improve the Mexican wages ofunskilled workers to reduce pressures on migration to the USA. The otherthree reasons, which appear more substantial and achievable, are: providea low-wage partner for US firms to compete with third-world suppliers andJapanese multinationals, set an example of speed in multinational negotia-tions and help Mexico lock in its economic reforms. The authors argue thatthe wage-convergence motive was the main driving force in the agreement.What appears clear about Mexico is that it is abundant in unskilled labourrelative to the USA, the rural areas are poor and the agricultural sector ishighly protected.

A number of theoretical models, in the Heckscher–Ohlin (HO) tradition,are considered that all imply a rising skilled–unskilled wage gap in both thedeveloped and the developing country. The models discussed are theFeenstra–Hanson model, the Markusen–Venables model and three morespecific models to do with technology and public infrastructure. TheFeenstra-Hanson model concludes that foreign investment leads to anincrease in the wage gap and in the unskilled wage share of Mexican man-ufacturing. This leads to an increase in demand (and the relative wage) for

294 Discussion by Pasquale M. Sgro

skilled labour in both the USA and Mexico. The Markusen–Venablesmodel deals with the role of multinationals and plant location. They find,inter alia, that the price of skilled labour would increase in both Mexico andthe USA. The three specific models are concerned with (1) multiple tech-niques of production in agriculture, (2) the specific-factor model, withcapital and land as specific factors and (3) the public- infrastructure (roadsand other transport inputs) model. In all of these models, the tendency istowards non-wage convergence. Although NAFTA has been in existencefor only four years and extensive hard data are still not available, there issome evidence that the wage gap has widened rather than narrowed. All themodels discussed by Markusen and Zahniser provide theoretical reasonswhy one might expect the wage gap and migration to persist despiteNAFTA. What would be of interest in this case would be to examine if therelocation by US firms of their plant and production processes across theborder to take advantage of the cheaper labour has had an impact. One pre-sumes that US firms would prefer to employ the cheaper Mexican labourrather than pay the same labour a higher wage when it came to the USA.Having a common border, the problem of illegal migration and the cross-flows of such migrants along with repatriation of income presents somedefinitional and data-collection problems. For example, it is not unusual forMexicans to cross and re-cross the border to return home for various festi-vals like weddings, anniversaries and carnivals.

Another class of HO-type trade models concerned with distortions ineither product or factor markets (the ‘old’ trade theory) also have featuresthat lead to non-wage convergence. As more hard data become available,these models also merit attention if one is looking for a theoretical explana-tion for non-convergence. For example, models that emphasise the role ofexchange rate movements between Mexico and the USA ought to play arole.

All of the models tend to concentrate on wage differences as the mainmotivation for migration. In general, other factors do play a role – forexample, public-good provision, educational and social infrastructure. Amore complete explanation would need to incorporate these and otherfactors. What is clear is that there are a large number of HO-type factor-proportions’ models quite capable of providing a reasonable explanationfor the observed non-wage convergence.

It is difficult to come to any firm conclusion on whether NAFTA hasachieved or will in fact achieve wage convergence (and hence reduce migra-tion pressures) until more hard data is available. Markusen and Zahniserdo present some models and argue (based on what little data is available)that convergence will be unlikely in the medium term (10 years). However,such a conclusion is premature, it is still too early for any firm conclusions.


11 East–West trade and migration: theAustro–German case

RUDOLF WINTER-EBMER AND KLAUS F.

ZIMMERMANN

1 Introduction

The demise of the political system in Central and Eastern Europe (CEE)and the Former Soviet Union (FSU) has created a new challenge for theEU – economic integration and enlargement. Eastern enlargement con-cerns the association of 10 CEE countries (CEECs), the Visegrád-4 (CzechRepublic, Hungary, Slovak Republic and Poland), the Balkan-3 (Bulgaria,Rumania and Slovenia) and the Baltic-3 (Estonia, Latvia and Lithuania).The integration issue, however, also covers the other states from the FSU.While enlargement is seen both as a political necessity and a historicalopportunity, the economic consequences are not yet well understood. Arising number of EU Member States seem more reluctant to take in newmembers, often because they fear that such a move will be expensive. Onemajor concern stems from the potential labour market effects of integrationand enlargement. Opening markets will encourage factor flows and trade,and hence very likely cause adjustments in wages and employment oppor-tunities in EU economies. Given the ever-rising unemployment rates, andthe relative decline of unskilled wages in Western Europe, the Easternenlargement is seen as a threat to native labour markets. As a result of geog-raphy and historical ties, Austria and Germany have already received dis-proportionately more immigrants and stronger increases of trade flowsthan other countries. It has to be expected that this trend will continue inany process of economic integration in the East. Hence, the objective of thischapter is to assess present and potential future developments against thebackground of the Austro-German case.

Two largely separated strands of literature have investigated the effects oftrade and migration on native labour markets. Labour economists haveexamined the hypothesis that immigration is causing a decline in wages andan increase in unemployment among natives, especially among low-skilledworkers. However, they did not find much support for these concerns. Trade

296

economists and labour economists have investigated the issue whethertrade, in particular with the developing world, has caused the measurednegative trends in the labour market, especially for its manual and unskilledsegments. However, mostly trade economists seem to believe that this is nottrue but can be attributed to skill-biased technical progress. So far, only fewstudies have dealt with these issues in the context of Eastern integration andEU enlargement. Section 2 of the chapter will survey the literature on thelabour market effects of trade and migration, with an emphasis on existingEuropean – especially Austrian and German – studies. Section 3 documentsmajor trends in East–West migration and trade flows, largely for Austriaand Germany. Section 4 will present econometric sectoral panel studies forboth countries to measure the employment and wages effects in the recentdecade. Section 5 draws some conclusions.

2 Trade, migration and the labour market consequences

The empirical literature on the labour market impacts of migration is con-centrated on the US experience as a major immigration country.Furthermore, NAFTA and the high US trade deficit in recent years hastriggered further research on trade. In many cases, both lines of research areunconnected. We offer a selective review of empirical studies with a heavyconcentration on European cases (see also Zimmermann, 1995a and 1995bfor a review and evaluation of European studies). This differentiation isvery important, because of the different labour market organisation.Krugman (1995) paradigmatically distinguishes an ‘American model’ froma ‘European model’ in the way globalisation interacts with the labourmarket. (Zimmermann, 1994, makes the case for a specific Europeanmigration problem.) In the more flexible USA, wage effects should be moreprominent; in the European case unemployment effects should be moreimportant. US immigration studies generally use regional variation inimmigration as the major explanatory variable – i.e. they compare locallabour markets with differing immigration rates. (Freeman and Katz, 1991,find a positive, though insignificant, association between the share of immi-grants and the change in annual hours worked in an estimation for a panelof 428 US industries.) LaLonde and Topel (1991) and Altonji and Card(1991) use US Census data in a regional context, and both find very smallunemployment effects. Borjas et al. (1997) use California as a primeexample of a high-immigration state and compare the development ofCalifornia with other states. These studies may suffer from reverse causal-ity: immmigrants tend to concentrate in states or cities with favourablelabour market conditions. Furthermore, regional out-migration (orreduced in-migration) of native workers from regions hit by migration

East–West trade and migration: the Austro–German case 297

streams might counteract the immediate impact of migration (Filer, 1992).Borjas et al. (1997, p. 25) are themselves critical about the spatial concept:

spatial correlation[s] between changes in native outcomes and immigra-tion do not, in fact, measure what we want them to measure. The incon-sistency in the signs of the correlations over time provides littleinformation about the structural impact of immigration on the nativelabour market.

They suggest instead using occupations as units, but cannot apply this ideabecause of data problems.

This possible simultaneity bias is circumvented in historical case studiesof an exogenous influx of immigrants such as the ‘Mariel boatlift’ ofCubans to Miami in 1980 (Card, 1990) or the repatriation of French citi-zens from Algeria to southern France in 1962 (Hunt, 1992). Both studiesfind only minor transitory adaptation problems in these labour markets. Inthe French case, the labour force rose by 1.6 per cent in the regions con-cerned. In the early years significantly lower wage growth resulted, as wellas higher unemployment. The quantitative effect was modest, though.More severe negative impacts are found by Carrington and de Lima (1996)for Portuguese ‘Retornados’ from Africa. Here, labour supply rose byalmost 10 per cent, and the new workers were on average highly qualified.Using comparisons with France and Spain, the authors conclude thatunemployment in Portugal rose significantly between 1974 and 1980.

The situation in Austria and Germany after the fall of the Iron Curtaincould also be considered a sort of historic experiment because of themassive inflow of immigrants and the sudden change in the trade regime.Labour market reactions to immigration and trade are especially inter-esting, therefore. In contrast to studies for the USA, most studies use indus-tries instead of regions as the unit of measurement. One reason is – ofcourse – the smallness of the countries, another is a more rigid employmentstructure across industries.

Table 11.1 summarises recent econometric studies for Austria andGermany. Winkelmann and Zimmermann (1993) use micro panel data forGermany and the count data methodology to find that immigration in the1970s increases the frequency of unemployment spells. However, no nega-tive impact for the 1980s could be detected in a later study using a panelprobit approach (Mühleisen and Zimmermann, 1994). This is explained bya sufficient wage flexibility that is also obtained in the study by De New andZimmermann (1994). Pischke and Velling (1993) use aggregate data forGerman counties. Simple regression analyses reveal high employmenteffects, but the effect vanishes once the mean reversion process ofunemployment rates is accounted for. Hatzius (1994) uses a two-stageapproach to study immigration effects in Germany. In a first stage he

298 Klaus F. Zimmermann

Table 11.1 Empirical studies on the effects of trade and migration on the labour market in Austria and Germany, 1990–7

Authors Data, methods anddependent variables Time period Findings

A AustriaAiginger et al. (1996) Social security records, manu- 1988–91 u: exports (2), imports (~)

facturing workers, individual Dw: exports (1), imports (~)unemployment rate, Dw

Pichelmann and Walterskirchen (1995) Macro model u: migrant share (1)

Pollan (1990) Aggregate time series, 1965–89 s(w): migrant share (1)wage dispersion (s(w))

Winter-Ebmer and Zweimüller (1996) Social security records, young 1981–91 w: migrant share (1)workers, w, Dw Dw: migrant share (~)

Winter-Ebmer and Zweimüller (1998) Social security records, manu- 1988–91 u risk: migrant share (~),facturing workers, unemployment imports(~), exports (~)risk, unemployment duration u duration: migrant share (1),

imports (~), export share (~)

B GermanyBauer (1997) German labour-force survey, 1990 Small elasticities of

estimation of a Translog complementarity. Most of the production function; wages immigrant groups are

complements to natives.

De New and Zimmermann (1994) German socio-economic panel 1984–9 w: migrant-labour share(SOEP), micro panel with sectoral- total (2),migrant share; wages (w) blue-collar (2), white-collar (1)

Haisken-De New and Zimmermann (1999) SOEP, panel with migrant share 1984–92 w: migrant-labour share (1)differentiated according to trade-deficit ratio (2)industry and state (Bundesländer); m: migrant-labour share (~)wages (w) and mobility (m) trade-deficit ratio (2)

Hatzius (1994) Regional-migrant share w: migrant share (2)u: migrant share (~)

Lücke (1996) 32 industries 1970–92 w: relative price of unskilled-w for unskilled labour labour-intensive goods (~)

Mühleisen and Zimmermann (1994) SOEP, individual unemployment, 1984–9 u risk: migrant-labour share (~)sectoral-migrant shares

Pischke and Velling (1994) County data, unemployment rate 1985–9 u: Dmigrant-population share (~)

Winkelmann and Zimmermann (1993) SOEP, unemployment risk, 1974–84 u risk: migrant-labour share (1)sectoral-migrant share

Notes:(1) The coefficient is positive and significant in most reported specifications.(2) The coefficient is negative and significant in most reported specifications.(~) The coefficient has mixed sign or is not significant in most specifications.

regresses individual unemployment on a set of region-by-period dummiesand individual characteristics. He then uses the estimated coefficients in asecond regression: differentiating between foreigners, East Germans andethnic Germans he finds no significant effect of the presence of any of theseimmigrant groups on unemployment probabilities of natives. He therebyconfirms the findings of Mühleisen and Zimmermann (1994). Gang andRivera-Batiz (1993) offer an interesting exercise in political economy: inaccordance with the above-mentioned studies they find that unemploymentrisk is not furthered by the presence of foreigners, but the individual assess-ment by workers is different. Unemployed workers believe that the presenceof foreigners is responsible for their dismal situation, a belief that is notshared by employed workers.

Wage impacts are studied by De New and Zimmermann (1994) who usedata from the German socio-economic panel together with industryconcentration of foreigners. They find large negative impacts on hourlywages in a random-effects panel model for blue-collar workers, but positiveeffects for white-collar workers.They hence conclude that foreigners aresubstitutes to native blue-collar workers, but complements to native white-collar workers. These results are corroborated by Hatzius (1994), who usesregional variation in foreign shares.

More recent studies investigating wage effects include Bauer (1997) andHaisken-De New and Zimmermann (1999). The former contribution esti-mates a Translog production function to calculate elasticities of comple-mentarity, while the latter chapter is an update of De New andZimmermann (1994) using a longer time series of a panel of individuals,more detailed data on immigrants and different measures of high-skilled/unskilled work. Both contributions suggest that the degree ofcomplementarity is much more strongly present than was found in previ-ous exercises. Hence, immigration hardly affects native wages, at least notnegatively, and mostly positively.

The effect of trade on the German labour market was investigated byLücke (1996) and Haisken-De New and Zimmermann (1999). Lücke (1996)cannot detect a relevant effect of the relative price of unskilled-labour-inten-sive goods on wages. Haisken-De New and Zimmermann (1999) study wageand mobility effects of trade and migration, and find that trade mattersmore than migration. Wages are affected negatively by a relative increase inimports (relative to exports). Trade seems to depress occupational mobilityand movements within firms, but stimulates inter-firm changes.

For Austria, several studies with very different methodologies exist.Pichelmann and Walterskirchen (1995) simulate increased migration usinga standard Keynesian macro-model. The increased labour-supply result issplit up approximately equally between increased employment and


unemployment. Brandel et al. (1994) study turnover processes in firms andconclude that the recent surge of new immigrants into Austria led to a sig-nificant displacement of guest-workers of earlier generations, but also ofnatives: 60 per cent of all firms in their sample with shrinking employmentof natives extended the engagement of foreigners in the period 1989–91.However, the latter study uses only descriptive techniques rather thanregression analysis. Moreover, their measure of shrinking firms does notcorrespond exactly to the notion of displacement, because firm size canchange for a variety of reasons: retirement, voluntary quits, etc. In aneconometric study using individual data for workers, Winter-Ebmer andZweimüller (1998) conclude that increased immigration did not result inhigher unemployment entry of Austrian manufacturing workers, althoughit increased the duration of unemployment: an increase in the immigrantshare by one percentage point increased unemployment duration byapproximately 5 per cent – i.e. five days.

Wage impacts in Austria are studied by Pollan (1990), who found that thedispersion of industry wages in the 1965–89 period was positively associ-ated with the share of foreigners in the economy at large. Winter-Ebmer andZweimüller’s (1996) analysis concentrates on young blue-collar workers. Incross-section, wages correlate positively with foreigner shares at a regional,industry or even firm level. The pattern is more mixed in a wage-growthformulation: Mobile workers – those who changed industry or region – canprofit from higher foreign shares, whereas immobile workers lose. Theauthors explain these results by a bargaining model: the presence of foreignworkers in a firm will reduce the bargaining power of insiders because of‘threat effects’; on the other hand, insiders may benefit from a high share ofimmigrants if they are able to exploit them in a two-tier wage system.

Assessing the impact of international trade on the labour market – in linewith the equilibrium character of trade theory – would ideally call for theuse of a computable general equilibrium (CGE) model. (See, for example,Kohler, 1991, for an application to trade liberalisation in Austria; seeBrown, 1992, for a survey of CGE models applied to the consequences ofNAFTA.) These studies typically analyse the effects ex ante rather than expost. Some recent studies in the USA use simulation techniques to inferwage effects from trade (see Borjas et al., 1992 and Murphy and Welch,1991; Baldwin, 1995, provides a survey of trade effects on employment andwages). In a recent update Borjas et al. (1997) attribute a significant pro-portion of the wage decline of US high-school dropouts to increasedimmigration, but less to trade impacts.

Factor-content studies have been heavily criticised by several authors(e.g. Wood, 1995; Bhagwati and Deheijia, 1994; Leamer, 1996). The mainproblems are: (1) Are trade flows the right measures or only prices? (2) Is itpossible to measure the right competing groups of workers? (3) What unit-


labour coefficients should be used? Furthermore, a significant element oftrade competition is a simple threat of re-location of industries abroad,which can change wages significantly without any trade flows at all (andalso without any foreign direct investment (FDI) flows). Rodrik (1997)argues that these effects make labour and product demand functions moreelastic in the home country and labour more vulnerable to demand shifts.

For Europe, only a few studies on employment effects of trade withEastern Europe exist. Cadot and de Melo (1994) provide simulation resultsfor the regional distribution of possible job creation and destruction causedby CEEC trade with France. Looking only at emerging trade patterns withthe CEECs, no general problems for EU markets as a whole as well as forspecific industries, like metals or textiles, are found by a study edited byFaini and Portes (1995) – mainly because the level of EU–CEEC trade isstill very low. The Austrian Institute of Economic Research (WIFO)(Aiginger, 1993), based on industry studies, calculates a positive employ-ment balance of Austrian trade with the CEECs. This is mainly due tohigher exports as well as cheaper inputs for manufacturing firms. Aigingeret al. (1996) look at a panel of Austrian workers in manufacturing and con-clude that individual unemployment rates over a period of three years reactsignificantly negatively to increased export volumes and (only insignifi-cantly) positively to import volumes. However, significant positive importimpacts are found for sub-groups of blue-collar workers, the elderly andlow-income earners. These results are based on the period from 1988 to1991 – i.e. the immediate aftermath of the fall of the Iron Curtain. As thetrade structure was heavily distorted before that time, these results have tobe considered as transitional adaptation processes, far away from equilib-rium. In any case, they have to be complemented by more recent data. Bycalculating the labour content of trade flows using industry-specific pro-ductivity data, Altzinger (1995) finds that increased net exports to CEECshave lead to positive employment effects on the Austrian labour market.For Germany, Lücke (1996) analysed 32 industries in the period 1970–92and found that wages for non-skilled labour did not react to relative pricechanges of unskilled labour-intensive goods.

Apart from the relation between trade and migration, capital flows haveto be considered as well. Growing FDI can influence export performancesignificantly. The traditional Heckscher–Ohlin (HO) trade theory viewsFDI as export of capital motivated by higher returns abroad, and it sug-gests that FDI substitutes for exports, thus lowering the production ofcapital-intensive goods in the home country. Re-allocation of capital in theEast can be viewed in this sense as a further impediment to the employmentin capital-intensive industries. This conclusion is questioned by new modelswith monopolistic competition and horizontally differentiated products(e.g. Helpman, 1984). If endogenous growth is based on an expanding stock


of knowledge, one can show that FDI, the number of multinational firmsand exports, increase complementarily because of increasing productvariety.

Pfaffermayr (1996) tests for complementarity or substitutability ofexports and FDI using a panel of Austrian industries. The results indicatea significant – although numerically small – complementarity between FDIand exports in the 1980s and early 1990s.

3 East–West flows of goods and people

This section investigates the flow of factors and trade between Austria andGermany and the East, and documents the performance of both countries.A summary of the economic conditions in the major countries in CEE isgiven in table 11.2, where we provide data on real growth, inflation andunemployment. The situation has improved considerably in many coun-tries, although unemployment remains high. It seems that there is a largepotential for out-migration and trade. According to gravity models(Holzmann and Zukowska-Gagelmann, 1996) trade with Eastern Europeis still much below equilibrium projections. These expectations about tradecreation have yet to be augmented by the effects of EU East enlargement.

Figure 11.1 studies the evolution of the foreigner share in the labour forceand the import ratio (total imports divided by GNP) for West Germany andAustria between 1986 and 1994. In total, there is a trend in favour of immigra-tion, while the evolution of trade is stagnating. Both countries are similar inthe sense that immigration and trade was expanding until 1990, but after-wards only the immigration level increased while the import share declined.Austria always had a higher import share, but its foreigner share was signifi-cantly below the foreigner share in Germany at the end of the 1980s, before itincreased to the German level in the mid-1990s. Since in the German case,immigration of ethnic Germans is not covered by the data, figure 11.1 under-reports this fact for Germany. Figure 11.2 investigates the issue using data onmigrants (the share of East European workers as a percentage of total foreignworkers) and trade from Eastern Europe (the share of imports from the Eastin total imports). Both shares have increased for Germany, while for Austria,the increase is only in trade. In general, figure 11.2 demonstrates that Austriawas and still is much more involved with the East than Germany, althoughthis is true only in relative terms and the differences are shrinking.

3.1 Austria

Table 11.3a summarises the main developments in Austria’s labour marketperformance with respect to internationalisation between 1986 and 1995.


Table 11.2. Economic development in Central and Eastern Europe,a 1991–7

CzechPoland Republic Slovakia Hungary Slovenia Rumania Bulgaria Russia

Real Growth1991 127.0 214.2 211.2 211.9 1129.3 1212.9 11211.7 11125.01992 212.6 123.3 126.5 123.1 1125.5 1128.8 11127.3 11214.51993 213.8 120.6 123.7 120.6 2112.8 2111.5 11121.5 11128.71994 215.2 212.6 214.9 212.9 2115.3 2113.9 21111.8 11212.61995 217.0 214.8 216.8 211.5 2113.9 2117.1 21112.1 11124.01996 216.1 214.4 217.0 211.1 2113.5 2114.1 11210.9 11122.81997 215.5 211.0 215.0 213.0 2113.5 1122.0 11127.0 21110.5

Inflation1991 276.5 256.7 261.1 234.8 2115.2 2174.4 12338.5 21193.01992 243.0 211.1 210.0 223.0 2201.3 2210.4 21191.2 21526.61993 235.3 220.8 223.2 222.5 2132.3 2256.1 21172.8 21873.51994 233.2 210.0 213.4 218.9 2119.8 2136.8 21196.0 21307.61995 227.8 219.0 219.9 228.2 2112.6 2132.3 21162.1 21197.51996 219.9 218.8 215.8 223.5 2119.7 2138.8 21123.0 21147.61997 216.0 219.0 216.5 218.0 2119.0 2150.0 21100.0 21115.0

Unemployment Rate1991 219.2 212.8 217.1 215.4 2118.2 — 21111.1 —1992 212.9 213.0 211.3 210.7 2111.5 2116.2 21113.2 21110.41993 214.9 213.0 212.9 212.8 2114.4 2119.2 21116.3 21111.01994 214.4 213.2 214.8 210.8 2119.1 2110.9 21112.8 21116.01995 213.3 213.1 213.1 210.3 2117.4 2118.9 21110.8 21118.01996 212.4 213.5 212.5 210.0 2117.3 2118.5 21112.5 21119.01997 211.5 214.5 212.5 219.0 2117.0 2110.0 21114.0 21110.0

Note:a Real growth: GDP; inflation: consumer price index. All numbers in per cent. 1997: Estimates of the German Council of EconomicExpertsSource: Sachverständigenrat (1996/7, p. 33) and Sachverständigenrat (1997/8, p. 30).

Employment rose nearly constantly, albeit at a very slow pace. Three-quar-ters of the rise in employment was driven by increasing employment offoreign workers. The share of foreign workers almost doubled in thisperiod, its increase being particularly high in the years 1990 and 1991.However, heavy restrictions for immigrants have damped its furtherdevelopment. Although the rise in foreign employment coincided with thefall of the Iron Curtain, most immigrants came from Austria’s mainsending countries, namely Former Yugoslavia and Turkey; foreign workersfrom Eastern Europe have only a modest share in the foreign population atlarge. Unemployment increased slightly over the period with constantlyhigher rates for foreigners than for natives.

Trade with Eastern Europe was almost unchanged over the period, with


Figure 11.1 Immigration and trade, 1986–96Note: Foreigner share: total share of foreigners of employment. Imports/GNP: total importsdivided by GNP.Sources: See tables 11.3a and 11.3b.

Im

an export and import share of between 2 and 3 per cent of domestic output.This general picture conceals important structural changes in terms ofindustries, as well as in terms of countries. Whereas trade with the FSUalmost collapsed, trade with the CEECs increased considerably. Moreover,the import structure changed from raw materials mainly from the FSU, tomore refined products.

The development of FDI is shown in table 11.4. Between 1989 and 1994Austria’s outward FDI into Eastern Europe rose by more than 30-fold. Themain part of these investments went into the neighbouring countries – theCzech and Slovak Republics and Hungary: in 1992, Austrian firms partic-ipated in more than 20 per cent of all joint ventures in these countries. Thehigh international capital inflow into Eastern Europe should lead – in prin-ciple – to enhanced competition for foreign capital. This competition hasnot affected Austria’s inward FDI’s very greatly.

The opening of the Eastern borders had regionally different impacts.


Figure 11.2 East–West migration and trade: Germany, 1986–94; Austria, 1990–4Sources: See tables 11.3a and 11.3b.

Table 11.3. Labour, migration and trade, Austria and Germany, 1986–95

a Austria

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

Employment (000) 12,780 12,785 12,765 12,815 12,881 12,940 12,964 12,956 12,972 12,970

Unemployment rate, natives (per cent) 1,1115.2 1,1115.6 1,1115.3 1,1115.0 1,1115.4 1,1115.8 1,1115.9 1,1116.8 1,1116.5 1,1116.6

Unemployment rate, foreigners (per cent) 1,1116.4 1,1115.7 1,1116.2 1,1115.9 1,1117.8 1,1117.1 1,1118.6 1,1119.1 1,1118.3 1,1117.9

Foreign share in employment (per cent) 1,1115.2 1,1115.3 1,1115.4 1,1115.8 1,1117.2 1,1118.6 1,1119.0 1,1119.0 1,1119.5 1,1119.8

Foreign workers from Former Yugoslavia as per cent of all foreigners 1,1157.3 1,1155.9 1,1155.0 1,1154.2 1,1150.7 1,1148.5 1,1149.6 1,1150.6 1,1148.9 1,1150.7

Foreigners from Eastern Europea as per cent of all foreigners 1,1113.0b — — — 1,1114.5 1,1114.7 1,1115.6 1,1115.1 1,1114.7 1,1114.1

Eastern Europe exports as per cent of domestic output 1,1112.69 1,1112.411,1112.571,1112.66 1,1112.511,1112.571,1112.64 1,1112.65 1,1112.86 —

Eastern Europe imports as per cent of domestic output 1,1112.77 1,1112.191,1112.111,1112.14 1,1112.121,1112.131,1112.15 1,1112.07 1,1112.41 —

Table 11.3. (cont.)

a Austria

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

Eastern Europe exports as per cent of total exports 1,1119.63 1,1119.021,1119.141,1119.04 1,1118.841,1118.981,1119.68 1,1110.48 1,1111.00 —

Eastern Europe imports as per cent of total imports 1,1118.34 1,1116.801,1116.371,1116.06 1,1116.001,1116.041,1116.47 1,1116.76 1,1117.54 —

Notes:a Excluding workers from Former Yugoslavia, which is the major sending country for Austria.b 1981.

b Germany

Employed workers (000) 23,910 24,141 24,365 24,750 25,460 25,920 26,066 25,611 25,242 25,022

Unemployment rate, natives (per cent) 1,1119.0 1,1118.9 1,1118.7 1,1117.9 1,1117.2 1,1116.3 1,1116.6 1,1118.2 1,1119.2 1,1119.3

Unemployment rate, foreigners (per cent) 1,1113.7 1,1114.3 1,1114.4 1,1112.2 1,1110.9 1,1110.7 1,1112.2 1,1115.1 1,1116.2 1,1116.6

Foreign share in employment (per cent) 1,1117.7 1,1117.6 1,1117.6 1,1117.8 1,1118.0 1,1118.2 1,1118.7 1,1119.5 1,1119.4 1,1119.4

Foreign workers from Eastern Europe as per cent of all foreign workers 1,1112.9 1,1113.1 1,1113.4 1,1113.8 1,1114.2 1,1115.5 1,1117.3 1,1118.1 1,1117.7 1,1117.8

Inflow of ethnic Germans (000) 1,1142.7 1,1178.5 1,1202.6 1,1377.0 1,1397.1 1,1122.0 1,1230.5 1,1218.9 1,1222.6 1,1217.9

Inflow of ethnic Germans from Russia as per cent of total inflow 1,1111.76 1,1118.5 1,1123.5 1,1126.0 1,1137.0 1,1166.4 1,1184.9 1,1194.7 1,1195.8 1,1196.1

per cent of domestic output 1,1111.04 1,1110.901,1111.001,1111.17 1,1111.031,1111.361,1110.81 1,1111.40 1,1111.51 1,1111.62

Eastern Europe imports as per cent of domestic output 1,1110.99 1,1110.801,1110.811,1110.90 1,1110.911,1111.151,1110.78 1,1111.14 1,1111.36 1,1111.62

Eastern Europe exports as per cent of total exports 1,1113.45 1,1113.131,1113.391111,3.71 1,1113.211,1115.361,1113.42 1,1116.30 1,1116.60 1,1116.85

Eastern Europe imports as per cent of total imports 1,1113.96 1,1113.361,1113.331,1113.44 1,1113.461,1114.531,1113.27 1,1115.29 1,1116.11 1,1117.11

Source: Statistisches Handbuch der Republik Österreich (various years); WIFO Data base, Vienna. Employment according toHauptverband der Sozialversicherungstäger, excluding those below the minimum social security contributions level.

Table 11.4. Foreign direct investment (stocks in ECU million), 1989–95

1989 1990 1991 1992 1993 1994 1995

AustriaFDI into Eastern Europe 11,147.6 11,240.7 11,516.9 11,733.9 111,043.2 111,558.0 —Austrian inward FDI — — 12,137 12,028 113,230 114,408 —

GermanyFDI into Eastern Europe — — 11,076 11,857 113,312 114,864 116,599German inward FDI 71,493 83,089 93,986 96,812 105,517 113,432 122,284

Source: Mitteilungen des Direktoriums der Österreichischen Nationalbank (various years); Statistisches Jahrbuch für dieBundersrepublik Deutchland (various years); Sachverständigenrat (1996/7). Own calculations.

Austria’s North-eastern parts had suffered in the past from the ‘deadborder’ and the severe trade and traffic restrictions. Now, employment inthe Eastern parts, especially the counties in the Czech, Slovak andHungarian borderland, increased disproportionately (Geldner, 1993),whereas employment growth between 1983 and 1989 had been slow. Thisemployment growth is completely due to new immigration: between 1989and 1992 foreign employment in Austria increased by 68 per cent, in theborderland by more than 200 per cent, dragging the foreign share there upto the national average. Consequently, native employment stayed almostconstant.

3.2 Germany

Data on migration, trade and FDI in Germany are contained in tables11.3b and 11.4. Table 11.3b indicates that employment in Germany wasstagnating, with between 24 and 26 million workers. However, the size offoreign employment in general has increased, as well as the inflow of EastEuropeans and ethnic Germans. While the share of foreign employmentwas less than 8 per cent in 1986, it reached a level of more than 9 per centin 1995. The share of foreign workers from Eastern Europe as a percentageof all foreign workers has increased from about 3 per cent to about 8 percent. There was a dramatic increase (by a factor of 10) in the inflow ofethnic Germans per year from the mid-1980s to 1990. Since then, quotasset by the German government caused a decline, to a level of about 200,000per year. Among ethnic Germans, the inflow has more and more concen-trated on migrants from Russia; while there were fewer than 2 per centRussians among ethnic German migrants in 1986, their share had increasedto nearly 100 per cent in 1995.

Native unemployment was comparatively low (less that 7 per cent) in theearly 1990s, but returned to higher levels in the mid-1990s (9 per cent, as in1986–7). The unemployment rates of foreigners were rising even more: theywere about 14 per cent in 1986 and about 17 per cent in 1995; they were thennearly twice as high as those of the natives. This increase was primarilycaused by differences in job composition (and the related unemploymentrisk) between natives and foreigners, and not by a higher unemploymentrisk of foreigners per se.

Table 11.3b also suggests that trade with Eastern Europe (exports andimports) was expanding significantly, relative to both domestic output andtotal imports/exports. Exports and imports to and from Eastern Europe inGermany rose from about 1 per cent of domestic output in 1986 to 1.6 percent in 1995; in terms of total imports or exports, the share rose in the sameperiod from less than 4 per cent to about 7 per cent.


Table 11.4 indicates that FDI in Eastern Europe has increased sub-stantially in the last few years; it was been six times higher in 1995 than itwas in 1991. In contrast, German inward FDI grew only by about 30 percent over the same period. However, the size of German FDI into EasternEurope is still somewhat modest in comparison with the Austrian involve-ment (only about 2–3 times larger), and in relationship to FDI in Germany.German FDI into Eastern Europe in 1994 was about 4 per cent of Germaninward FDI, whereas the figure for Austria is 11 per cent.

4 Effects on the labour markets

In this section, we investigate the effects of immigration change and exportand import flows on employment and wage growth. Employment growth isalso studied for natives only. The time period chosen is 1986–94 to cover aperiod before and after the demise of the socialist system. We use all avail-able sectors at the two-digit ISCO level, excluding those that have no inter-national trade. The data sets have a panel structure. The Austrian sample issomewhat larger, since we obtained a more complete set of information inthis country. Owing to data problems and different structural issues, themodel specification differs somewhat between both countries. We employ areduced-form approach where growth rates of (total and native) employ-ment and wages depend on a time trend and various measures of changesof foreigner shares and trade shares. The basic method is weighted regres-sion, where we use the sectoral employment shares as weights. The migra-tion flows are instrumented to capture endogeneity problems. Theinstruments were valid only for migration flows, and hence were not appliedfor the other regressors.

4.1 Austria

The data for Austria capture 30 industries for the years 1985–94.Employment and wage statistics were obtained from the Ministry ofLabour. Contrary to Germany, the number of immigrant workers is notdifferentiated between those from Eastern Europe and elsewhere. As mostimmigrants to Austria are from the Former Yugoslavia and from Turkey,the human capital as well as the language knowledge of these immigrantscan be assumed to be similar to that of the more recent immigrants fromEastern Europe. As an indicator for wage growth, we take the medianmonthly gross wage in the industry. Trade shares are nominal exports(imports) divided by nominal output in the industry, which in fact assumesequal price indices for imports and domestically produced goods. As thebulk of manufactured-goods trade with Eastern Europe took place with


Czecho-Slovakia, Hungary and Poland, the Eastern trade measures in theAustrian regressions use only these three (four) countries. As trade data usethe SITC nomenclature, while the labour market data apply the ISIC code,we had to use a concordance which was developed by WIFO; this concor-dance defines which ISIC code is nearest to an SITC code at the three-digitlevel.

Results for the reduced-form regressions for employment and wagegrowth are in tables 11.5–11.7 for Austria. We present separate results for


Table 11.5 Employment growth, Austria,a 1987–94

High- All Low-wage High-import immigrationindustries industries industries industries

Dln (Foreign share) 20.001 20.031 20.042 20.030(0.04) (0.87) (1.12) (0.88)

Dln (East import share) 20.036 20.045 20.019 20.076(2.59) (2.12) (1.06) (3.05)

Dln (ROW import share) 20.055 20.150 0.049 20.172(1.25) (1.91) (0.89) (1.93)

Dln (East export share) 20.011 20.009 0.016 0.001(0.98) (0.40) (1.27) (0.02)

Dln (ROW export share) 20.038 20.001 20.039 20.038(1.02) (0.02) (0.85) (0.54)

Dln (Output) 0.268 0.328 0.394 0.276(5.90) (4.73) (7.51) (3.76)

Time trend 20.001 0.001 20.002 0.001(0.40) (1.08) (1.66) (0.64)

R2 0.232 0.255 0.279 0.282

F 11.33 7.22 8.05 8.13

N 240 128 128 128

Notes:a All regressions include a constant. Absolute t-values in parentheses. Regressionsweighted by employment share. Time period 1987–94. Dln (Foreign share) isinstrumented. Instruments are lagged levels and changes in minimum wages,shares of blue-collar workers and EU output as well as lagged levels of immigrantshares. Low-wage industries refers to those with a mean wage level 1987–94 up tothe industry median; respectively for high-import and high-immigrationindustries.

employment growth at large and for employment of domestic workers. Asthe pooled time-series cross-section approach assumes constant coefficientsfor all industries, we study also some sub-samples, who might be moreendangered by increasing internationalisation. Immigrant shares areinstrumented by industry minimum wages, shares of blue-collar workersand EU output growth.

Immigration is found to have no impact on industry employment atlarge. This is equally so for the three sub-samples we study. In contrast to


Table 11.6 Native employment growth, Austria,a 1987–94







Dln (Output) 0.313 0.402 0.381 0.373(6.95) (5.90) (7.13) (5.39)

Time trend 0.001 0.002 20.003 0.001(0.21) (1.44) (1.90) (0.96)

R2 0.324 0.373 0.297 0.466

F 17.39 11.28 8.66 16.84

N 240 128 128 128

Notes: a All regressions include a constant. Absolute t-values in parentheses. Regressionsweighted by employment share. Time period 1987–94. Dln (Foreign share) isinstrumented. Instruments are lagged levels and changes in minimum wages,shares of blue-collar workers and EU output as well as lagged levels of immigrantshares. Low-wage industries refers to those with a mean wage level 1987–94 up tothe industry median; respectively for high-import and high-immigrationindustries.

that, rising import penetration has employment costs. Interestingly, thisapplies mainly to imports from Eastern Europe. The quantitative impact isrelatively low: a 1 per cent increase in the import share leads to reducedemployment by 0.03 per cent. It has to be noticed that this elasticity appliesto a very low import share of Eastern firms. The elasticity is somewhathigher in low-wage industries and especially in industries characterised bya high share of foreign workers. For these sub-groups, rising imports fromROW countries also have detrimental employment effects. Exports are


Table 11.7 Wage growth, Austria,a 1987–94







Dln (Output) 0.307 0.227 0.250 0.178(3.50) (2.06) (2.06) (1.42)

Time trend 20.015 20.013 20.017 20.013(7.41) (5.38) (5.51) (5.38)

R2 0.316 0.259 0.328 0.267

F 16.75 7.35 9.86 7.60

N 240 128 128 128

Notes: a All regressions include a constant. Absolute t-values in parentheses. Regressionsweighted by employment share. Time period 1987–94. Dln (Foreign share) isinstrumented. Instruments are lagged levels and changes in minimum wages,shares of blue-collar workers and EU output as well as lagged levels of immigrantshares. Low-wage industries are those with a mean wage level 1987–94 up to theindustry median; the same principle is applied for high-import and high-immigration industries.

never significant. Domestic industry output growth always has the expectedpositive employment impact, although with a rather small elasticity.

This pattern is very similar for domestic employment growth, but we cansee that a rising foreigner share with practically constant total employmentwill reduce domestic employment. A 1 per cent increase in the foreignershare in industry employment reduces domestic employment by 0.13 percent. This means that increasing immigration led to an almost complete dis-placement of native workers in the industries concerned, with the highesteffects in high-immigration industries. Again imports from Eastern Europehave a minor negative employment effect and ROW imports are significantonly in low-wage industries.

Wages react negatively to increased immigration with an elasticity of 0.16.The wage-depressing effects are lower in already low-wage industries andinsignificant in industries with high-import as well as high-immigrationshares. This might be explained by a higher tendency for binding minimumwages in low-wage industries; other explanations would refer to a higherpotential for rent reductions in high-wage and often more unionised indus-tries. Freeman and Katz (1991), for instance, find for the USA that wageresponsiveness was higher in high-wage and highly unionised sectors.Industry-wage growth is not affected by rising imports. This corroboratesresults by Gaston and Trefler (1994), who study the Canadian employmentand wage reaction to increased trade with the USA in a very similar frame-work to ours. Contrary to employment reactions, exports into ROW countriessignificantly further wage growth, with an elasticity of 0.24 and an even higherresponse rate in high-immigrant industries. As we observe only median wagesfor Austria, this differing response could be the result of a structural employ-ment effect: the creation of relatively high-paying jobs in export industries.These issues can be resolved only by the use of more detailed data on incomedistribution within industries and is beyond the scope of the present study.Again, output growth has a significant positive wage impact.

4.2 Germany

Results for Germany are given in tables 11.8–11.10 for 12 industries in1987–94. Wherever possible, the data refer to Western Germany, and theexceptions will be noted below. The regressions contain a time trend,output growth (measured as real gross value-added in per cent), andvarious growth rates of foreigner shares in the labour force and growth ratesof export and import ratios. Wage, output, employment and trade data arefrom various sources of the German Statistical Office. Numbers on workersfrom Eastern Europe are from a report of the Council of EconomicExperts, and the figures on ethnic Germans are from a government report.



Table 11.8 Employment growth, Germany,a 1986–94

High- Low-wage High-import immigration

All industries industries industriesindustries only only only

Constant 0.010 0.023 0.006 0.035(0.62) (1.18) (0.35) (1.20)


Dln (Eastern Europe 0.966 1.763 0.377 1.014foreigner share) (2.82) (3.43) (1.12) (1.67)

Dln (Share of ethnic 0.018 20.051 0.045 20.049Germans) (0.62) (1.44) (1.29) (0.84)

Russian share3Dln 20.047 0.121 20.147 20.020(Share of ethnic Germans) (0.89) (2.11) (2.31) (0.22)

Dln (Import share) 0.074 0.039 0.110 0.106(2.99) (1.58) (1.89) (2.90)

Dln (Export share) 20.056 0.015 20.058 20.069(2.18) (0.67) (1.34) (1.78)



Dln (Output) 0.238 0.321 0.235 0.294(3.95) (2.78) (3.80) (2.30)

Time trend 20.006 0.001 20.005 20.008(1.72) (0.30) (1.31) (1.56)

R2 0.600 0.665 0.867 0.550

F 14.56 8.52 28.02 5.25

N 108 254 254 254

Notes: a All regressions include a constant. Absolute t-values in parentheses. Regressionsweighted by employment share. Time period 1987–94. All foreigner shares areinstrumented. Instruments are lagged levels and changes in union wages, shares ofblue-collar workers and EU output as well as lagged levels of immigrant shares.Low-wage industries are those with a mean wage level 1987–94 up to the industrymedian; the same principle is applied for high-import and high-immigrationindustries.


Table 11.9 Native employment growth, Germany,a 1986–94



Constant 0.008 0.015 0.011 0.022(0.49) (0.53) (0.62) (0.70)





Dln (Import share) 0.049 0.032 0.119 0.084(1.82) (0.89) (2.01) (2.12)

Dln (Export share) 20.041 20.009 20.058 20.055(1.45) (0.28) (1.32) (1.30)



Dln (Output) 0.250 0.394 0.243 0.366(3.80) (2.33) (3.87) (2.64)

Time trend 20.006 0.000 20.006 20.007(1.78) (0.02) (1.57) (1.24)

R2 0.574 0.484 0.867 0.558

F 13.05 4.03 27.92 5.42

N 108 54 54 54

Notes: a All regressions include a constant. Absolute t-values in parentheses. Regressionsweighted by employment share. Time period 1987–94. Al foreigner shares areinstrumented. Instruments are lagged levels and changes in union wages, shares ofblue-collar workers and EU output as well as lagged levels of immigrant shares.Low-wage industries are those with a mean wage level 1987–94 up to the industrymedian; the same principle is applied for high-import and high-immigrationindustries.


Table 11.10 Wage growth, Germany,a 1986–94



Constant 0.023 0.021 0.019 0.022(2.39) (1.30) (1.29) (1.37)





Dln (Import share) 20.019 20.022 0.033 20.039(1.21) (1.14) (0.67) (1.93)

Dln (Export share) 0.003 20.014 0.040 0.011(0.17) (0.79) (1.10) (0.51)



Dln (Output) 0.140 0.186 0.216 0.061(3.73) (2.01) (4.20) (0.86)

Time trend 0.000 0.001 20.000 0.002(0.10) (0.40) (0.11) (0.77)

R2 0.375 0.420 0.545 0.307

F 5.81 3.11 5.14 1.90

N 108 54 54 54

Notes: a All regressions include a constant. Absolute t-values in parentheses. Regressionsweighted by employment share. Time period 1987–94. All foreigner sharers areinstrumented. Instruments are lagged levels and changes in union wages, shares ofblue-collar workers and EU output as well as lagged levels of immigrant shares.Low-wage industries refers to those with a mean wage level 1987–94 up to theindustry median; the same principle is applied for high-import and high-immigration industries.

The growth rate of the foreigner share in the labour force is the changeof foreign workers compared to the total change in the labour force. Aseparate term for the Eastern Europe workers in the labour force allows foran examination of the hypothesis whether this group of workers has effectsdifferent from the general group of foreigners. If they behave equally, theestimated coefficient for the East Europeans should not be statisticallydifferent from zero. While the total foreigner share in the labour force isdifferentiated according to industry, the employment of East Europeans isnot available at the industry level. Hence, we have distributed the EastEuropeans using the sectoral-employment shares of the workers from theFormer Yugoslavia, assuming that the employment structure is similar.

The inflow of ethnic Germans into the labour market is not properlyrecorded. This is because the only statistic available is the total inflow ofethnic Germans according to sending regions. Since they immediatelyobtain German citizenship, they disappear into German statistics; there areno records about their labour-force participation and their sectoral involve-ment. Hence, we approximate the percentage change of the share of ethnicGermans in the labour force by the difference between (1) the inflow ofethnic Germans divided by lagged total employment and (2) the growthrate of employment.

Import and export data are collected on a somewhat different classifica-tion scheme, but can be merged with sufficient precision. Imports andexports are measured in unit values, but are available only either on anindustry breakdown (and not differentiated by regions) or differentiatedfor sending regions (but not disaggregated at the industry level). Hence,we decided to include variables that capture both levels of measurement.The import and export data at the sectoral levels were divided by the sec-toral outputs (real gross value-added), respectively. We also included(real) total imports and exports from East Europe divided by real GDP.A further problem is that imports and exports after unification are notdifferentiated according to Western and Eastern Germany; this is not toomuch of an issue, since most of the trade seems to take place with WesternGermany.

The regressions for employment growth, native employment growth andwages are contained in Tables 11.8–11.10. All regressions are weighted withthe industry employment shares, and migration flows are instrumented asdetailed in the footnotes to the tables. As for the Austrian case, each tablecontains results for all industries and for those industries with low wages,high imports and many foreigners.

Table 11.8 refers to total employment growth that is largely driven byoutput growth in all samples, where all the estimated parameters have larget-ratios, while the trend variable is hardly significant at conventional levels.


There are no clear indications that immigration has affected total employ-ment negatively. However, a clear negative effect (with significant t-valuesfor low-wage and high-import industries only) was obtained for Russianethnic German immigrants, while an increase in the foreigner share fromEastern Europe has exhibited a positive (and mostly significant) effect ontotal employment. The total foreigner share is statistically significant onlyfor high-import industries with a positive parameter estimate. All importshares affect employment positively, and exports negatively, although notalways with statistically significant t-values.

Table 11.9 contains the same set of regressors, but using native employ-ment growth as the dependent variable. Again, output growth is drivingemployment, and the time trend plays no effective role. However, thegeneral foreigner share now exhibits a negative coefficient in three out offour cases (with the exception of high-import industries), and the negativeparameters are even significant for low-wage and high-immigration indus-tries. East European immigration does not exhibit much of a relevant addi-tional contribution. Global trade shows up with the signs of the parametersas before; they are all positive for imports and negative for exports, but onlyfew are statistically significant. In addition, all East European trade vari-ables have insignificant parameter estimates.

Table 11.10 studies wage growth, which again depends largely on outputgrowth and is unaffected by the time trend. Here, foreigners (especially EastEuropeans) seem to have a positive effect on general wages, but the esti-mates are significant in only a few cases: for high-immigration industriesand the growth rate of the general foreigner share; for high-import indus-tries (and the total sample) and the growth rate of the East European for-eigner share. Finally, none of the trade variables has any significant impacton wage growth.

The simple labour market model would suggest that more foreignersshould depress wages (if migrants are substitutes for natives), and increasetotal employment while native employment should remain constant (ifmarkets are flexible). In the findings reported here, wages are not declining,at best they are increasing (especially if related to the inflow of EastEuropeans). Total employment is increasing, although mainly due to EastEuropeans. Native employment is somewhat (and negatively) affected byforeigners in general, but positively by East Europeans. This all suggeststhat East Europeans and German workers are complements, and that theoverall effect of migrants on the German labour market is unproblematic.

The results for the trade variables are also controversial. Traditionaltheory would expect a positive impact of exports on employment andwages, and the reverse for imports. Trade does not affect wages at all,perhaps because one should differentiate wages for qualification levels and


trade for regions to obtain more sensible findings, but we cannot do this dueto lack of data. In the employment equations, imports (exports) have a neg-ative (positive) impact on employment, but the estimates are in most casesstatistically insignificant. This can be explained by the relationship of tradeto technical progress, which is insufficiently measured by the time trend. Itis well known that technical progress is more intense in exporting indus-tries, while importing industries are less innovative; hence, we may beunable to separate these effects. In sum, we cannot find harmful effects oftrade on the German labour market.

5 Conclusions

Trade and migration have become more important in recent years forAustria and Germany. The transition in CEE has played an important rolein this development. The derived labour market consequences are not fullyclear so far. The Austrian findings suggest that immigration exhibits nega-tive effects on native employment and wages, and has no effects on totalemployment. Imports affect employment negatively and exports have apositive effect on wages. The German results indicate that immigration andtrade is not harming employment and wages. Natives seem to be comple-ments to migrants, at least to those from East Europe. Trade does not affectwages at all, and hardly affects employment. These results are in line withrecent findings for both Austria and Germany. Hence the conclusions are:while the Austrian labour market might be somewhat negatively affected bythe Eastern enlargement of the EU, the German labour market is not.Further research and better data are needed to understand these findingsproperly.

NOTE

We wish to thank Thomas Bauer, Ralph Rotte, Andreas Million, Anja Thalmaierand Michael Vogler for able research assistance and conference participants, espe-cially Riccardo Faini, for helpful comments.

REFERENCES

Aiginger, K. (ed.) (1993). Chancen und Gefährdungspotentiale der Ostöffnung(Vienna: Austrian Institute for Economic Research)

Aiginger, K., R. Winter–Ebmer and J. Zweimüller (1996). ‘Eastern European Tradeand the Austrian Labour Market’, Weltwirtschaftliches Archiv, 132, 476–500

Altonji, J. G. and D. Card (1991). ‘The Effects of Immigration on the LabourMarket Output of Less-skilled Natives’, in J. M. Abowd and R. B. Freeman


(eds.), Immigration, Trade, and the Labour Market (Chicago: University ofChicago Press for NBER), 201–34

Altzinger, W. (1995). ‘Beschäftigungseffekte des österreichischen Osthandels’,Working Paper, 34, Department of Economics, Vienna University ofEconomics and Business Administration

Baldwin, R. E. (1995). ‘The Effect of Trade and Foreign Direct Investment onEmployment and Relative Wages’, NBER, Working Paper, 5037

Bauer, T. (1997). ‘Native Wage Impacts of Foreign Labour: Further Evidence forGermany’, Mitteilungen zur Arbeitsmarkt- und Berufsforschung, 30, 652–6

Bhagwati, J. and V. H. Dehejia (1994). ‘Freer Trade and Wages of the Unskilled –Is Marx Striking Again?’, in J. Bhagwati and M. H. Kosters (eds.), Trade andWages: Leveling Wages Down? (Washington, DC: American EnterpriseInstitute)

Borjas, G. J., R.B. Freeman and L. H. Katz (1992). ‘On the Labour Market Effectsof Immigration and Trade’, in G.J. Borjas and R.B. Freeman (eds.),Immigration and the Work Force (Chicago: University of Chicago Press),213–44

(1997). ‘How much do Immigration and Trade Affect Labour MarketOutcomes?’, Brookings Papers on Economic Activity, 1, 1–90

Brandel, F., H. Hofer and K. Pichelmann (1994). ‘Verdrängungsprozesse amArbeitsmarkt’, Research Memorandum, 345 (Vienna: Institute for AdvancedStudies)

Brown, D. K. (1992). ‘The Impact of the North American Free Trade Agreement:Applied General Equilibrium Models’, in N. Lustig et al. (eds.), NorthAmerican Free Trade: Assessing the Impact (Washington, DC: BrookingsInstitution)

Cadot, O. and J. de Melo (1994). ‘France and the CEECs: Adjusting to AnotherEnlargement’, INSEAD, paper presented to a CEPR Conference, Brussels,mimeo

Card, D. (1990). ‘The Impact of the Mariel Boatlift on the Miami Labour Market’,Industrial and Labour Relations Review, 43, 245–57

Carrington, W. I. and P. de Lima (1996). ‘The Impact of 1970s’ Repatriates fromAfrica on the Portuguese Labour Market’, Industrial and Labour RelationsReview, 49, 330–47

De New, J. P. and K. F. Zimmermann (1994). ‘Native Wage Impacts of ForeignLabor: A Random Effects Panel Analysis’, Journal of Population Economics,7, 177–92

Faini, R. and R. Portes (eds.) (1995). ‘European Union Trade with Eastern Europe:Adjustment and Opportunities’ (London: CEPR)

Filer, R. K. (1992). ‘The Effect of Immigrant Arrivals on Migratory Patterns ofNative Workers’, in G.J. Borjas and R. B. Freeman (eds.), Immigration and theWork Force (Chicago: University of Chicago Press), 245–70

Freeman, R. B. and L. F. Katz (1991). ‘Industrial Wage and EmploymentDetermination in an Open Economy’, in J.M. Abowd and R. B. Freeman(eds.), Immigration, Trade and the Labour Market (Chicago: University ofChicago Press), 235–60


Gang, I. N. and F. L. Rivera–Batiz (1993). ‘Unemployment and Attitudes againstForeigners in Germany’, Rutgers University, mimeo

Gaston, and Trefler (1994). ‘Protection, Trade and Wages: Evidence from USManufacturing’, Industrial and Labor Relations Review, 47, 574–93

Geldner, N. (1993). ‘Beschäftigungseffekte der Ostöffnung’, in K. Aiginger (ed.),Chancen und Gefährdungspotential der Ostöffnung (Vienna: Austrian Instituteof Economic Research), 41–6

Haisken–De New, J. P. and K. F. Zimmermann (1999). ‘Wage and Mobility Effectsof Trade and Migration’, forthcoming in M. Dewatripont and A. Sapir (eds.),International Trade and Employment: The European Experience (Oxford:Oxford University Press)

Hatzius, J. (1994). ‘The Unemployment and Earnings Effects of GermanImmigration’, Merton College, Oxford, mimeo

Helpman, E. (1984). ‘A Simple Theory of International Trade with MultinationalCorporations’, Journal of Political Economy, 92, 451–71

Holzmann, R. and K. Zukowska–Gagelmann (1996). ‘Exportchancen undImportschocks: Eine Untersuchung zur Asymmetrie der Handelsanpassungder OECD-Länder im Gefolge der Marktöffnung in Mittel- und Osteuropa’,in R. Holzmann and R. Neck (eds.), Ostöffnung: Wirtschaftliche Folgen fürÖsterreich (Vienna: Manz), 193–236

Hunt, J. (1992). ‘The Impact of the 1962 Repatriates from Algeria on the FrenchLabour Market’, Industrial and Labour Relations Review, 45, 556–72

Kohler, W. (1991). ‘Multilateral Trade Liberalization: Some General EquilibriumSimulation Results for Austria’, Empirica, 18, 167–200

Krugman, P. (1995). ‘Growing World Trade: Causes and Consequences’, BrookingsPapers on Economic Activity, at proof. , 327–77

LaLonde, R. J. and R. H. Topel (1991). ‘Labour Market Adjustments to IncreasedImmigration’, in J.M. Abowd and R. B. Freeman (eds.), Immigration, Trade, andthe Labour Market (Chicago: University of Chicago Press for NBER), 167–200

Leamer, E. E. (1996). ‘What’s the Use of Factor Contents?’, NBER, Working Paper,5448

Lücke, M. (1996). ‘Has Trade with Low-Wage Countries Hurt Unskilled Labour inWest Germany?’, Kiel, Institute of World Economics, mimeo

Mühleisen, M. and K. F. Zimmermann (1994). ‘A Panel Analysis of Job Changesand Unemployment’, European Economic Review, 38, 793–801

Murphy, K. M. and F. Welch (1991). ‘The Role of International Trade in WageDifferentials’, in M.H. Kosters (ed.), Workers and their Wages (Washington,DC: American Enterprise Institute)

Pfaffermayr, M. (1996). ‘Foreign Outward Direct Investment and Exports inAustrian Manufacturing: Substitutes or Complements?’, WeltwirtschaftlichesArchiv, 132, 501–22

Pichelmann, K. and E. Walterskirchen (1995). ‘East–West Integration and itsImpact on Workers: The Austrian Example’, Vienna, Institute for AdvancedStudies, Reprint 14

Pischke, J.-S. and J. Velling (1993). ‘Ausländische Werkvertragsarbeitnehmer amBau: Bauboom oder Verdrängungswettbewerb’, ZEW, Discussion Paper, 9315,Mannheim


(1994). ‘Labour Market Effects of Foreign Employment in Germany’, CEPR,Working Paper, 935

Pollan, W. (1990). ‘Lohnunterschiede in der Industrie’, Monatsberichte des Österre-ichischen Instituts für Wirtschaftsforschung, 63, 616–22

Rodrik, D. (1997). ‘Has Globalisation Gone Too Far?’ (Washington, DC: Institutefor International Economics)

Sachverständigenrat (1996). Jahrgutachten 1996/7 des Sachsverständigenrates zurBegutachtung des gesamtwirtschaftlichen Entwicklung, Bundesdrucksache13/6200, Bonn

(1999). Jahrsgutachten 1997/8 des sachsverständigenrates zur Begutachtung desgesamtwirtschaftlichen Entwicklung, Bundestagsdrucksache 13/9090, Bonn

Winkelmann, R. and K. F. Zimmermann (1993). ‘Ageing, Migration and LabourMobility’, in P. Johnson and K. F. Zimmermann (eds.), Labour Markets in anAgeing Europe (Cambridge: Cambridge University Press), 255–83

Winter–Ebmer, R. and J. Zweimüller (1996). ‘Immigration and the Earnings ofYoung Native Workers’, Oxford Economic Papers, 48, 473–91

(1998). ‘Immigration, Trade and Austrian Unemployment’, in M. Landesmannand E. Streissler (eds.), Unemployment in Europe (London: Macmillan)

Wood, A. (1995). ‘How Trade Hurts Unskilled Workers’, Journal of EconomicPerspectives, 9, 57–80

Zimmermann, K. F. (1994). ‘Some General Lessons for Europe´s MigrationProblem’, in H. Giersch (ed.), Economic Aspects of International Migration(Heidelberg: Springer-Verlag), 249–73

(1995a). ‘European Migration: Push and Pull’, Proceedings volume of the WorldBank Annual Conference on Development Economics, 1994, Supplement toThe World Bank Economic Review and The World Bank Research Observer,313–42; reprinted in International Regional Science Review, 19 (1996), 95–128

(1995b). ‘Tackling the European Migration Problem’, Journal of EconomicPerspectives, 9, 45–62; reprinted in M. N. Jovanovic (ed.), InternationalEconomic Integration. Critical Perspectives of the World Economy (London:Routledge, 1998)

Discussion

MARINA SCHENKEL

Chapter 11 is a good piece of empirical work, since it helps to understandthe peculiarities of Austria and Germany, not only with regard to interna-tional migrations (i.e. their impact on employment and wages) but also as


regards the broader picture of investments abroad and trade flows withEastern Europe.

In my opinion, however, some problems arise with regard to the inter-pretation of the econometric estimates. Even if the reduced-form approachis correctly carried out, the results are sometimes not easy to understand.

Since an articulated model of the functioning of the labour market is notpresented, one cannot appreciate if the specification which has been chosenis apt to assure that the coefficients capture just the additional effect ofmigrations (and segments of them), and of trade on the main labour marketvariables.

In the case of Germany, we read in the chapter that between 1992 and1995 almost 1 million jobs were lost, between 1986 and 1993 about 600,000more foreign workers were employed, but in the following two years foreignworkers decreased again by more than 480,000. During the same period,workers from Eastern Europe increased by more than 130,000 until 1993,and decreased by 27,000 thereafter.

Since in the chapter nothing is said about the internal migration flowsfollowing reunification, it is possible to think that migration flows of abigger entity and other origin can have contributed to produce those effectsin the labour market.

As far as various types of heterogeneity are concerned, the commoditycomposition of the trade flows is controlled for, since the study is performedon an inter-industry base. Also, it can be assumed that the temporaryversus the permanent character of migration, which could not have beentaken into account, makes little difference to labour market effects, at leastin the short run.

We can therefore agree with the chapter’s two main conclusions: (1) themigration consequences for the labour market are not yet fully clear and (2)the Austrian labour market is more significantly affected by the Easternenlargment than the German one.

328 Discussion by Marina Schenkel

Abramovitz, M. 233, 234agglomeration 11

country size effect 39from integration 51, 60–6see also location decisions

aggregationchoice of 121–2level 261

agriculture 15, 268, 273public infrastructure 266, 289tariff levels 285–6technology 284–90

Aiginger, K. 303Aitken, B. 273Allen–Uzawa elasticities of substitution

118, 123–4, 135–6, 140Altonji, J. G. 297Altzinger, W. 303Anderson, J. 252Antille, G. 176Argentinia 248–9, 250Atkinson, A. B. 152, 154, 163, 176, 185Atlantic economy

(1870–1940) 227–57panel data 244–7

Australia 82, 250Austria 17, 80, 93, 188

links with the east 296–324, 327–8

Bacci, M. L. 255Bairoch, P. 23Baldwin, R. E. 302Banerjee, B. 10Basevi, G. 68–75Bauer, T. 301Begg, D. 23Ben-David, D. 2Benhabib, J. 86, 155Berndt, E. R. 133Bernholz, P. 82

Berry, R. A. 153Bhagwati, J. 84, 93, 95, 151, 165, 166, 186,

190, 302Blomstrom, M. 281Bond, E. W. 84Borjas, G. J. 84, 85, 155, 297, 298, 302Bourguignon, F. 198, 217brain drain 84, 94–5Brandel, F. 302Brazil 248–9, 250Brecher, R. A. 95, 187Brown, D. K. 302Buckley, F. H. 87, 88Bulow, J. I. 165Burda, M. 193Burfisher, M. E. 266, 268

Cadot, O. 303Cairncross, A. 234Canada 82, 87–8, 248–9, 250, 268

see also NAFTAcapital

market (1870–1940) 248–50mobility 122, 148movements 7, 234, 303–4: level of 245–7;

measures of 241; policies on 77–80see also foreign direct investment;

investment liberalisationCard, D. 297, 298Carrington, W. I. 298chauvinism 80, 93Chen, T. J. 84Chenery, H. 198, 217Chiswick, B. 185Choudri, E. U. 187Christensen, L. R. 130Clark, R. L. 85Cochrane–Orcutt procedure 241Cogneau, D. 222–3Collins, W. J. 15, 148, 227–57, 260–2

329

Index

Collinson, S. 7competitive models 23–34Constant Elasticity of Transformation

(CET) 194, 196Corden, W. M. 230core–periphery 24, 39–45, 51–2cost of living 37, 56

effects 39–45index 24

country size 37–9, 74see also market size

culture 76, 92–3preferences 80–3as public good 81

cumulative causation 34–45

Dallas, H. 74Darmstadter, J. 254Daveri, F. 91–3, 194Davies, J. B. 78de Coulon, J. P. 152, 180de Lima, P. 298de Melo, J. 1–19, 190–219, 221–3, 303De New, J. P. 298, 301De Vries, C. G. 74Deheijia, V. H. 302demand-linkages 51–3demography 6, 221, 254, 270–1Dickens, W. T. 180Diewart, W. E. 120, 131–2, 143direct-democracy model 86, 155discrimination 152–3, 155–8, 160–1, 185Dixit, A. K. 35, 95Djajic, S. 193Docquier, F. 84dual labour market 158, 165, 167–8, 186

see also discrimination; labour marketsegmentation

Dustmann, C. 152, 165, 182

earnings differentials 193–4see also wage gap

Easterlin, R. A. 233, 234efficiency 85, 152, 162elasticities

Allen–Uzawa 118, 123–4, 135–6, 140CET 194, 196Hicksian 118, 123–4, 135–6, 140import demand 137–9price and quantity 124–7, 129resident labour demand 137–9

Eldridge, H. 255employment

equilibrium pattern 59–62fluctuations 153

Epstein, G. 87, 88

Estevadeordal, A. 229, 254Ethier, W. J. 84, 153ethnicity 149–50European Union (EU) 16–17, 263

CEEC trade 303enlargement 66, 296four-pillar freedoms 94inward migration 264liberalising sequence 74politicians 48–51stylised facts 62

European–US migration 28–9, 30, 227–57

factorflows 234–44intensity assumptions 281prices 26–7scarcity and abundance 235–6

factor mobility 28, 261falling barriers 51–2incentives for 25–6, 36–7trade complement or substitute 8–9, 94,

99–100, 140, 227–57, 261see also immigration; migration

factor-price equalisation (FPE) 29, 78, 94, 286Faini, R. 1–19, 190–219, 221–3, 303Favarger, P. 132Feenstra, R. C. 265, 273, 281Feenstra–Hanson model 273–9, 286, 290, 294Ferenczi, I. 4Filer, R. K. 298financial constraints model 9–10Findlay, R. F. 232Finger, M. 3Flam, H. 229Flanders, M. J. 229Flückiger, Y. 152, 163, 180foreign aid 7, 191foreign direct investment (FDI) 3, 49

Austria 308and export performance 303–4Germany 313–14versus migration 191, 201, 204, 207, 212,

213, 222–3see also investment liberalisation

Foreman-Peck, J. 86, 250Former Yugoslavia 314, 322Fox, K. J. 131, 143France 80, 93, 191, 234, 298, 303Freeman, R. B. 279, 297, 318Friedberg, R. 6functional forms, flexibility of 120, 130–2,

143, 149

Gaillard, S. 176Galeotti, M. 147–50

330 Index

Gang, I. N. 149, 301Gaston 318Gatsios, K. 94–111GATT 248

General Agreement on Trade in Services(GATS) 94

Uruguay Round 3, 94, 191Geldner, N. 313Germany 17, 80, 93, 152, 191, 234

links with the east 296–324, 327–8globalisation

effects decomposed 200–2effects on migrant supply 197–202see also trade liberalisation

Goldin, C. 250Greenwood, M. J. 147Grether, J.-M. 14, 190–219, 221–3Grossman, B. 117Grubel, H. B. 160guest-worker system 14, 88, 151–83, 180,

196

Haisken-De New, J. P. 301Hamada, K. 95Hamermesh, T. 182Hammar, T. 152Hanson, G. H. 265, 273, 281Haque, N. U. 84Harley, K. 23Harris, J. 193Harrison, A. 273Hatton, T. J. 234, 241Hatzipanayotou, P. 12, 94–111Hatzius, J. 298, 301Heckscher, E. 229, 230, 252Heckscher–Ohlin (HO) model 24–34, 48–9,

77–9, 94, 186, 197, 265, 269–70, 284,286, 290–1, 294, 303

3x2 229–31Helmberger, P. 232Helpman, E. 24, 37, 39–41, 45, 48, 303Helpman–Krugman model 39–45, 48Hicksian elasticities of complementarity

118, 123–4, 135–6, 140Hillman, A. L. 6, 11, 12, 76–89, 91–3Hoffmann-Nowotny, H.-J. 170Holzmann, R. 304Huddle, D. 85Hungary 308, 315Hunt, J. 6, 298

immigrantsconsumption of public goods 95illegal 86–7, 92, 264, 266as inputs 121permanent 170–1, 174, 186, 196, 210–11

skill-mix 173–4, 191–2, 211trade substitute or complement 117see also migration

immigrationincome distribution 78–9quotas 128unemployment 139–40, 151–2versus protection 151, 166–8

immigration policy 5–6(1870–1940) 248–52direct and indirect 190–1endogenous 76, 77–83, 85–8

importsdemand for 121–4as inputs 118restriction 99–106see also trade protection

income distribution 151–2, 172–3, 222after investment liberalisation 273–9,

280–4from immigration 78–9native 163unskilled immigrants 154, 168–9welfare effects 188

industrialisation, three-stage 52inputs

foreign labour 117immigrants as 121imports as 118

International Labour Organisation (ILO)190

investment liberalisation 265, 288and income distribution 273–9, 280–4see also foreign direct investment

(FDI)Isserlis, L. 241, 254

Jahn, A. 87Jones, R. 79Jones, R. W. 230

Kalix, Z. 254Kanbur, S. M. 10Katz, L. H. 297, 318Kebabjian, G. 222Kelley, A. C. 233Kemp, M. C. 232Kim, S. J. 84King, M. 97Kohler, W. 302Kohli, U. 13, 117–45, 147–50Krugman, P. 4, 11, 24, 35, 37, 39–41,

44–5, 48, 51–3, 59–60, 68, 69, 75, 297

Kuznets, S. 233Kwok, V. 84

Index 331

labourdisplacement effects 128–9extensive margin of 148non-resident 138–9see also guest-worker system

labour demandforeign 121–4low-skilled 266–7model of 53–7non-linear 69–74

labour marketeffects 158–62segmentation 165, 167–8, 173see also dual labour market

labour mobilityand agglomeration 65–6imperfect 52–3job displacement 117redistribution effects 117secondary to primary sector 157–8

labour supply 100–3elasticity 60–2fixed 209international 51model of 57–9non-linear 69–74

LaLonde, R. J. 297land 15–16, 230–1, 254

endogenous frontier 232–3, 260Lang, K. 180Lau, L. J. 148Le Châtelier principle 138Leamer, E. E. 302Leland, H. 84Lerner, A. P. 229Lewis, W. A. 233Licht, G. 185location decisions

efficiency 81–2and factor prices 26–7firms 37, 40workers 56, 57, 75, 195see also agglomeration

long swings (1870–1940) 233–44Lopez, R. 10, 193, 209Lucas, R. E. 7Lücke, M. 301Ludema, R. D. 11, 51–68, 68–75

Maddison, A. 254maize 266, 284–90market size 24, 39–45

see also country sizeMarkusen, J. R. 9, 16, 17, 25, 196, 229,

263–91, 294–5Markusen–Venables model 280–4, 290–1, 294

Martin, P. L. 222, 266, 289Mayer, W. 79Mazza, I. 86Mexico 50, 191, 263–91, 268

incomes 270, 272maize 266, 284–90migration to US 16, 18, 196, 211, 222,

264–91population 270–1

Michael, M. S. 12, 94–111migration 12–15

controls 7, 191, 211costs 193, 209, 210, 211countries of origin 87globalisation effect 197–202and import restriction 99–103incentives 265and income differentials 49–50labour market consequences 297–304mass 234measures of 241non-pecuniary benefits 57rates 245–7, 255return 85, 192and trade liberalisation 17–18, 112–14trends 3–5and unemployment 211–12welfare effects 93see also factor mobility; immigrants;

immigration; migrantsmigratory pressure 7, 15, 190–219, 290

aggregate 202–4data 217–18factors affecting 202household and sectoral effects 204–8and investment liberalisation 288low-income and middle-income 192,

198–212sensitivity analysis 208–11simulation model 196–7, 212–17, 221and trade liberalisation 211

Milbourne, R. 193Mitchell, B. R. 253–5Molle, W. 5Mühleisen, M. 298, 301Müller, T. 13–14, 151–83, 185–8multinationals 280–4Mundell, R. A. 2, 25, 48, 78, 94, 100, 227,

229Murphy, K. M. 302Musu, I. 112–14

NAFTA 7, 10, 16, 18, 74, 222, 263–91,266–7, 294–5, 297

Neary, J. P. 155, 171, 230–1, 252, 266, 287

332 Index

nested CES function 162, 187Norman, V. D. 25, 95

Ohlin, B. 229, 230, 252Ohyama, M. 232O’Leary, P. J. 233O’Rourke, K. H. 15, 227–57, 252, 260–2ownership 79, 82–3, 210

Passel, J. S. 85Pfaffermayr, M. 304Pichelmann, K. 301Piore, M. J. 152Pischke, J.-S. 298policy asymmetries 11–12, 77–80, 92Pollan, W. 302Portes, R. 303price indices 287–8production cost 36–7production-theoretic approach 117–45,

147–50aggregation choice 148cost-function setting 124–5, 129data 149empirical results 133–40estimation techniques 132–3GNP-function setting 13, 126–7, 129,

137, 148–9input prices 150production-function setting 125–6

property rights 82–3public choice model 85–6public good, culture as 81public goods 12

consumed by immigrants 95and import restriction 103–6model 96–9and trade liberalisation 112–14

push–pull model 83–5

Rapoport, H. 84regional integration 7, 51–68, 66, 263restricted profit function 148–9returns to scale

constant 121, 196, 228increasing 24, 34–45variable 121

Ricardian models 9, 48Ricardo–Viner model 14, 192, 193–7, 196,

198, 202, 212Richardson, J. D. 78, 279Rivera-Batiz, F. L. 149, 301Rodrik, D. 303Rodriquez, C. 93Rostow, W. W. 234Ruffin, R. 79

Rutherford, T. F. 285Rybczynski, T. M. 229, 230Rybczynski effects 9, 127, 138Rybczynski theorem 78–9

Sachs, J. D. 4, 252Salinas, President 7, 191Samuelson, P. A. 229Sapir, A. 17, 48–50Schenkel, M. 327–8Schiff, M. 6, 10, 190–1, 193, 209, 266Schmidt, C. M. 153, 162Schmitz, A. P. 232Schultze, G. 77Scott, A. D. 160Sgro, P. M. 294–5Shapiro, C. 156Shephard, R. W. 123Shughart, W. 250Silber, J. 163Simon, J. L. 88Slaughter, M. J. 281Slaughter, M. L. 78Slovak Republic 308social welfare

data and calibration 176–82effects 153–5function 14, 154–5, 185–6maximisation 169–70policy experiments 164–6sensitivity analysis 172simulation: model 162–3, 175–6; results

166–74see also welfare

Soligo, R. 153specific-factors model 9, 24, 29–34, 48, 101–3,

105–6, 108–9, 161, 186, 230–1, 266, 287Stalker, P. 6Stark, O. 84, 85, 193, 208Steiner, V. 185Stiglitz, J. E. 35, 156Stilz, A. 84Stolper–Samuelson effects 78, 127, 138, 139Straubhaar, T. 87Summers, L. H. 165Swagel, P. 78Swan, P. L. 82Switzerland 13

data 132immigration policy 170–1, 175non-resident labour trends 119–20production sector 117–45social welfare 152, 160–2, 164, 166, 175–7total factor productivity (TFP) 120, 131,

140–3Syrquin, M. 198, 217

Index 333

Tapinos, G. 222tariffs 262

(1870–1940) 241, 245, 248–50and agriculture 285–6revenue 254rising 171–2see also trade protection

tax revenue 104–6Taylor, A. M. 241Taylor, J. E. 84technology

and agriculture 284–90differences 229progress 140–3

Temin, P. 230theft 82–3Thomas, B. 234Thomas, D. 255Thomas, D. S. 234Thompson, H. 231Tiebout, C. 80–3Timmer, A. 250–1Todaro, M. 193Toniolo, G. 260–2Topel, R. H. 297Tornqvist index 132trade

costs 29, 33, 42, 53, 60–2and factor mobility 8–9, 94, 99–100,

117–18, 227–57, 261labour market consequences 297–304policies on 77–80trends 2–3volumes 234–44, 245–7

trade liberalisation 7–8, 191and agglomeration 11, 63–5effects of 30–2, 33–4, 38–9, 45and factor mobility 28and migration 12–15, 17–18, 112–14, 211,

288Pareto gains from 187and public goods 95, 112–14sequence 66, 74see also globalisation

trade protection 185–6social welfare effect 155, 161–2versus immigration 151, 166–8see also imports; tariffs; trade

liberalisationtransport costs 23, 27, 51–2, 55, 56, 228,

234, 241Trefler 318

unemployment 139–40, 151–2, 211–12, 270United Kingdom 191, 234United States 48–50, 212, 234, 250, 268

Amish 81beneficial immigration 85European migrants 28–30immigration policy 87–8Immigration Reform and Control Act

(IRCA) 264labour market 297–8maize 266protectionist 248–9role of NAFTA 263–91, 294–5stylised facts 62see also Atlantic economy; Mexico;

NAFTA

van Winden, F. 86Velling, J. 298Venables, A. J. 9, 10, 11, 17, 23–47, 48–50,

52, 63, 103, 228, 265, 280Venturini, A. 10, 193, 221–3voter, median 79, 86

wage bargaining, insider-outsider model 153wage gap 221

and foreign investment 279sectoral 180–2skilled–unskilled 265, 273, 280–4see also earnings differentials

wages, relative 57Wales, T. J. 120, 131–2, 143Walterskirchen, E. 301Warner, A. 4, 252Weiss, A. 6, 11, 12, 76–89, 91–3Welch, F. 302welfare

analysis 151effects 93, 188utilitarian 6, 154–5, 158–62, 187see also social welfare

Wilkinson, M. 234Willcox, W. 4Williamson, J. G. 15, 227–57, 260–2Wilson, President 248Winkelmann, R. 298Winter-Ebmer, R. 16, 153, 185–8, 296–324Wolff, E. 281Wong, K. 76, 94Wong, K.-Y. 227Wood, A. 78, 279, 302Wooton, I. 11, 51–68, 68–75, 78

Zahniser, S. 9, 16, 17, 196, 263–91, 294–5Zellner, A. 133Zevin, R. B. 241Zimmermann, K. F. 1–19, 83, 152, 296–324Zukowska-Gagelmann, K. 304Zweimüller, J. 153, 188, 302

334 Index

migration the controversies and the evidence

Documents