You are on page 1of 315

Routledge studies in development economics

1 Economic Development in the 6 Financial Development and

Middle East Economic Growth
Rodney Wilson Theory and experiences from
developing countries
2 Monetary and Financial Policies Edited by Niels Hermes and
in Developing Countries Robert Lensink
Growth and stabilization
Akhtar Hossain and 7 The South African Economy
Anis Chowdhury Macroeconomic prospects for the
medium term
3 New Directions in Development Finn Tarp and Peter Brixen
Growth, environmental concerns 8 Public Sector Pay and
and government in the 1990s Adjustment
Edited by Mats Lundahl and Lessons from five countries
Benno J. Ndulu Edited by Christopher Colclough

9 Europe and Economic Reform

4 Financial Liberalization and
in Africa
Structural adjustment and
Kanhaya L. Gupta and
economic diplomacy
Robert Lensink
Obed O. Mailafia
5 Liberalization in the Developing 10 Post-apartheid Southern Africa
World Economic challenges and policies
Institutional and economic for the future
changes in Latin America, Edited by Lennart Petersson
Africa and Asia
Edited by 11 Financial Integration and
Alex E. Fernández Jilberto and Development
André Mommen Liberalization and reform in
sub-Saharan Africa
Ernest Aryeetey and
Machiko Nissanke
12 Regionalization and 20 Contemporary Issues in
Globalization in the Modern Development Economics
World Economy Edited by B.N. Ghosh
Perspectives on the Third World
and transitional economies 21 Mexico Beyond NAFTA
Edited by Edited by Martín Puchet Anyul
Alex E. Fernández Jilberto and and Lionello F. Punzo
André Mommen
22 Economies in Transition
13 The African Economy A guide to China, Cuba,
Policy, institutions and the future Mongolia, North Korea and
Steve Kayizzi-Mugerwa Vietnam at the turn of the
twenty­first century
14 Recovery from Armed Conflict Ian Jeffries
in Developing Countries
Edited by Geoff Harris 23 Population, Economic Growth
and Agriculture in Less
15 Small Enterprises and Economic
Developed Countries
Nadia Cuffaro
The dynamics of micro and small
24 From Crisis to Growth in
Carl Liedholm and
Donald C. Mead
Edited by Mats Lundal
16 The World Bank
New agendas in a changing world 25 The Macroeconomics of
Michelle Miller-Adams Monetary Union
An analysis of the CFA franc zone
17 Development Policy in the David Fielding
Twenty-First Century
Beyond the post-Washington 26 Endogenous Development
consensus Networking, innovation,
Edited by Ben Fine, institutions and cities
Costas Lapavitsas and Antonio Vasquez-Barquero
Jonathan Pincus
27 Labour Relations in
18 State-Owned Enterprises in the Development
Middle East and North Africa Edited by
Privatization, performance and Alex E. Fernández Jilberto and
reform Marieke Riethof
Edited by Merih Celasun
28 Globalization, Marginalization
19 Finance and Competitiveness in and Development
Developing Countries Edited by S. Mansoob Murshed
Edited by José María Fanelli and
Rohinton Medhora
29 Programme Aid and 37 Essays on Balance of Payments
Development Constrained Growth
Beyond conditionality Theory and evidence
Howard White and Geske Dijkstra Edited by J.S.L. McCombie and
A.P. Thirlwall
30 Competitiveness Strategy in
Developing Countries 38 The Private Sector after
A manual for policy analysis Communism
Edited by Ganeshan Wignaraja New entrepreneurial firms in
transition economies
31 The African Manufacturing Jan Winiecki, Vladimir Benacek
Firm and Mihaly Laki
An analysis based on firm surveys
in sub-Saharan Africa 39 Information Technology and
Dipak Mazumdar and Development
Ata Mazaheri A new paradigm for delivering
the internet to rural areas in
32 Trade Policy, Growth and developing countries
Poverty in Asian Developing Jeffrey James
Edited by Kishor Sharma 40 The Economics of Palestine
Economic policy and institutional
33 International Competitiveness, reform for a viable Palestine state
Investment and Finance Edited by David Cobham and
A case study of India Nu’man Kanafani
Edited by A. Ganesh Kumar,
Kunal Sen and 41 Development Dilemmas
Rajendra R. Vaidya The methods and political ethics
of growth policy
34 The Pattern of Aid Giving Melvin Ayogu and Don Ross
The impact of good governance on
development assistance 42 Rural Livelihoods and Poverty
Eric Neumayer Reduction Policies
Edited by Frank Ellis and
35 New International Poverty H. Ade Freeman
Reduction Strategies
Edited by Jean-Pierre Cling, 43 Beyond Market-Driven
Mireille Razafindrakoto and Development
François Roubaud Drawing on the experience of
Asia and Latin America
36 Targeting Development Edited by Makoto Noguchi and
Critical perspectives on the Costas Lapavitsas
millennium development goals
Edited by Richard Black and
Howard White
44 The Political Economy of 52 The New Famines
Reform Failure Why famines exist in an era of
Edited by Mats Lundahl and globalization
Michael L. Wyzan Edited by Stephen Devereux

45 Overcoming Inequality in Latin 53 Development Ethics at work

America Explorations – 1960–2002
Issues and challenges for the Denis Goulet
twenty­first century
Edited by Ricardo Gottschalk and 54 Law Reform in Developing and
Patricia Justino Transitional States
Edited by Tim Lindsey
46 Trade, Growth and Inequality
in the Era of Globalization 55 The Asymmetries of
Edited by Kishor Sharma and Globalization
Oliver Morrissey Edited by Pan A. Yotopoulos and
Donato Romano
47 Microfinance
Perils and prospects 56 Ideas, Policies and Economic
Edited by Jude L. Fernando Development in the Americas
Edited by
48 The IMF, World Bank and Esteban Pérez-Caldentey and
Policy Reform Matias Vernengo
Edited by Alberto Paloni and
Maurizio Zanardi 57 European Union Trade Politics
and Development
49 Managing Development Everything but arms unravelled
Globalization, economic Edited by Gerrit Faber and
restructuring and social policy Jan Orbie
Edited by Junji Nakagawa
58 Membership Based
50 Who Gains from Free Trade? Organizations of the Poor
Export­led growth, inequality and Edited by Martha Chen,
poverty in Latin America Renana Jhabvala, Ravi Kanbur
Edited by Rob Vos, and Carol Richards
Enrique Ganuza, Samuel Morley
and Sherman Robinson 59 The Politics of Aid Selectivity
Good governance criteria in World
51 Evolution of Markets and Bank, U.S. and Dutch
Institutions development assistance
A study of an emerging economy Wil Hout
Murali Patibandla
60 Economic Development, 68 Rural Poverty and Income
Education and Transnational Dynamics in Asia and Africa
Corporations Edited by Keijiro Otsuka,
Mark Hanson Jonna P. Estudillo and
Yasuyuki Sawada
61 Achieving Economic
Development in the Era of 69 Microfinance
Globalization A reader
Shalendra Sharma David Hulme and Thankom Arun

62 Sustainable Development and 70 Aid and International NGOs

Free Trade Dirk-Jan Koch
Shawkat Alam
71 Development Macroeconomics
63 The Impact of International Essays in memory of Anita Ghatak
Debt Relief Edited by Subrata Ghatak and
Geske Dijkstra Paul Levine

64 Europe’s Troubled Region 72 Taxation in a Low Income

Economic development, Economy
institutional reform and social The case of Mozambique
welfare in the Western Balkans Channing Arndt and Finn Tarp
William Bartlett
73 Labour Markets and Economic
65 Work, Female Empowerment Development
and Economic Development Edited by Ravi Kanbur and
Sara Horrell, Hazel Johnson and Jan Svejnar
Paul Mosley
74 Economic Transitions to
66 The Chronically Poor in Rural Neoliberalism in Middle-Income
Bangladesh Countries
Livelihood constraints and Policy dilemmas, crises, mass
capabilities resistance
Pk. Md. Motiur Rahman, Edited by Alfedo Saad-Filho and
Noriatsu Matsui and Galip L. Yalman
Yukio Ikemoto
75 Latecomer Development
67 Public–Private Partnerships in Innovation and knowledge for
Health Care in India economic growth
Lessons for developing countries Banji Oyelaran-Oyeyinka and
A. Venkat Raman and Padmashree Gehl Sampath
James Warner Björkman
76 Trade Relations between the EU 83 Towards New Developmentalism
and Africa Market as means rather than
Development, challenges and master
options beyond the Cotonou Edited by Shahrukh Rafi Khan and
Agreement Jens Christiansen
Edited by
Yenkong Ngangjoh-Hodu and 84 Culture, Institutions, and
Francis A.S.T. Matambalya Development
New insights into an old debate
77 The Comparative Political Edited by Jean-Philippe Platteau
Economy of Development and Robert Peccoud
Africa and South Asia
Edited by Barbara Harriss-White 85 Assessing Prospective Trade
and Judith Heyer Policy
Methods applied to EU–ACP
78 Credit Cooperatives in India economic partnership agreements
Past, present and future Edited by Oliver Morrissey
Biswa Swarup Misra
86 Social Protection for Africa’s
79 Development Economics in Children
Action (2nd edition) Edited by Sudhanshu Handa,
A study of economic policies in Stephen Devereux and
Ghana Douglas Webb
Tony Killick
87 Energy, Bio Fuels and
80 The Multinational Enterprise in Development
Developing Countries Comparing Brazil and the United
Local versus global logic States
Edited by Rick Molz, Cătălin Ratiu Edited by Edmund Amann,
and Ali Taleb Werner Baer and Don Coes

81 Monetary and Financial 88 Value Chains, Social Inclusion

Integration in West Africa and Economic Development
Temitope W. Oshikoya Contrasting theories and realities
Edited by A.H.J. (Bert) Helmsing
82 Reform and Development in and Sietze Vellema
What can China offer the 89 Market Liberalism, Growth,
developing world and Economic Development in
Edited by Ho-Mou Wu and Latin America
Yang L. Yao Edited by
Gerardo Angeles-Castro,
Ignacio Perrotini-Hernández, and
Humberto Ríos-Bolívar
Market Liberalism, Growth,
and Economic Development in
Latin America

Edited by Gerardo Angeles-Castro,

Ignacio Perrotini-Hernández, and
Humberto Ríos-Bolívar
First published 2011
by Routledge
2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
Simultaneously published in the USA and Canada
by Routledge
711 Third Avenue, New York, NY 10017

This edition published in the Taylor & Francis e-Library, 2011.

To purchase your own copy of this or any of Taylor & Francis or Routledge’s
collection of thousands of eBooks please go to

Routledge is an imprint of the Taylor & Francis Group, an informa business

© 2011 Selection and editorial matter, Gerardo Angeles­Castro, Ignacio
Perrotini­Hernández and Humberto Ríos­Bolívar; individual chapters, the
The right of Gerardo Angeles­Castro, Ignacio Perrotini­Hernández and
Humberto Ríos­Bolívar to be identified as authors of the editorial material
and of the authors for their individual chapters has been asserted by them
in accordance with the Copyright, Designs and Patent Act 1988.
All rights reserved. No part of this book may be reprinted or reproduced or
utilized in any form or by any electronic, mechanical, or other means, now
known or hereafter invented, including photocopying and recording, or in
any information storage or retrieval system, without permission in writing
from the publishers.
Trademark notice: Product or corporate names may be trademarks or
registered trademarks, and are used only for identification and explanation
without intent to infringe.
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
Market liberalism, growth, and economic development in Latin America /
edited by Gerardo Angeles­Castro, Ignacio Perrotini­Hernández and
Humberto Ríos-Bolivar.
p. cm.
Includes bibliographical references and index.
1. Latin America–Economic policy. 2. Latin America–Commercial
policy. 3. Free trade–Latin America. 4. Economic development­­Latin
America. I. Angeles­Castro, Gerardo. II. Perrotini­Hernández, Ignacio.
III. Ríos­Bolivar, Humberto.
HC125.M325185 2010

ISBN 0-203-81612-9 Master e-book ISBN

ISBN: 978­0­415­57374­0 (hbk)

ISBN: 978­0­203­81612­7 (ebk)

List of figures xiv

List of tables xvii
List of contributors xx

Foreword xxii

Acknowledgements xxv
List of acronyms xxvi

Editors’ introduction 1

Trade liberalization, development and regional integration 5

1 Has trade liberalisation in poor countries delivered the

promises expected? 7

2 Beyond the Washington Consensus: the quest for an alternative

development paradigm for Latin America 26
xii Contents
3 Foreign trade and per capita income: new evidence for
Latin America and the Caribbean 59

4 Regional integration and its effects on inward FDI in

developing countries: a comparison between North–South
(Mexico) and South–South (Brazil) integration 81

5 Trade blocs as determinants of trade flows in South

American countries: an augmented gravity approach 110

Trade reforms and development experience: case studies
in Latin America 131

6 Downhill or the long agony of Argentinian development 133


7 The determinants of FDI in Chile: a gravity model approach 149


8 Assessment of the distributive impact of trade reforms in

Uruguay 168

Economic liberalization, development and growth in Mexico 193

9 Economic liberalisation and income distribution: theory

and evidence in Mexico 195

10 How risk factors affect growth in Mexico: a free-market

liberalism approach 220
Contents xiii
11 Anti- inflationary policy and financial fragility:
a microeconomic analysis case study of Mexico, 1990–2004 233

12 Technological innovation and sectoral productivity in the

Mexican economy: regional evidence 250

13 The robustness of Okun’s law – evidence from Mexico:

a quarterly validation, 1985.1–2006.4 264

Index 277

1.1 The trade-off between growth and the balance of payments 18

2.1a Argentina’s GDP (annual growth rate), 1961–2009 32
2.1b Brazil’s GDP (annual growth rate), 1961–2009 32
2.1c Chile’s GDP (annual growth rate), 1961–2009 33
2.1d Mexico’s GDP (annual growth rate), 1961–2009 33
2.1e China’s GDP (annual growth rate), 1961–2009 34
2.2a Argentina’s annual rate of inflation, 1961–2009 35
2.2b Brazil’s annual rate of inflation, 1970–2009 35
2.2c Chile’s annual rate of inflation, 1961–2009 36
2.2d Mexico’s annual rate of inflation, 1961–2009 36
2.2e China’s annual rate of inflation, 1987–2009 37
2.3a Argentina: trade balance–GDP ratio, 1960–2009 39
2.3b Brazil: trade balance–GDP ratio, 1960–2009 39
2.3c Chile: trade balance–GDP ratio, 1960–2009 40
2.3d Mexico: trade balance–GDP ratio, 1960–2009 40
2.3e China: trade balance–GDP ratio, 1978–2008 41
2.4a Argentina’s development gap, 1960–2008 42
2.4b Brazil’s development gap, 1960–2008 42
2.4c Chile’s development gap, 1960–2008 43
2.4d Mexico’s development gap, 1960–2008 43
2.4e China’s development gap, 1960–2008 44
2.5a Argentina’s gross investment and gross savings, 1960–2009 45
2.5b Brazil’s gross investment and gross savings, 1960–2009 46
2.5c Chile’s gross investment and gross savings, 1960–2009 46
2.5d Mexico’s gross investment and gross savings, 1960–2009 47
2.5e China’s gross investment and gross savings, 1960–2008 47
2.6a Argentina’s income elasticity of imports, 1961–2007 48
2.6b Brazil’s income elasticity of imports, 1961–2007 48
2.6c Chile’s income elasticity of imports, 1961–2007 49
2.6d Mexico’s income elasticity of imports, 1961–2007 49
2.6e China’s income elasticity of imports, 1979–2007 49
4.1 Foreign direct investment, net inflows, comparison with
ROW 90
Figures xv
4.2 Foreign direct investment, net inflows, comparison with other
middle-income countries 91
4.3 Foreign direct investment as percentage of GDP 92
4.4 Mexico inward FDI stock, 1990–2000 95
4.5 Mexico inward FDI flows, 2001–2006 96
4.6 Brazil inward FDI stock, 1990–2000 97
4.7 Brazil inward FDI flows, 2001–2006 97
4.8 Mexico inward FDI stock by regions, 1990–2000 98
4.9 Mexico inward FDI flows by region, 2001–2006 98
4.10 Brazil inward FDI stock by regions, 1990–2000 99
4.11 Brazil inward FDI inflows by regions, 2001–2006 100
4.12 Dow Jones Index composite average and FDI inflows in Brazil
and Mexico, 1990–2006 103
4.13 Brazil FDI inflows and exchange rate 104
4.14 Mexico FDI inflows and exchange rate 104
5.1 Graphic evolution of exports by country member of Mercosur 114–15
5.2 Graphic evolution of exports by country member of the CAN 117–19
5.3 Graphic evolution of gravity equation fixed effects for Mercosur
and CAN 125
6.1 Income per capita 133
6.2 Business cycle 136
6.3 Output decomposition 139
6.4 Real exchange rate 140
6.5 Gini coefficients 144
6.6 Wage share 145
7.1 FDI flows into Chile 1985–2006 151
7.2 FDI inflows into Chile by geographical origin 1974–2006 154
8.1 Consumption effect: compensating variation as per cent of
income by income distribution 177
8.2 Compensating variation as per cent of income by income
distribution: traded good 180
8.3 Compensating variation as per cent of income by income
distribution: non-tradable goods effect 182
8.4 Compensating variation as per cent of income by income
distribution: labour income effect 183
8.5 Compensating variation as per cent of income by income
distribution: total effect 183
10.1 Optimal w* as a function m and h 225
10.2 Consumption as a function of x0 and t0 229
10.3 Observed growth rate from 1930 to 2002 230
11.1 Growth, returns and interest rate: aggregate levels,
1990:02–2004:04 238
11.2 Growth and firms’ indebtness: aggregate levels,
1990:02–2004:04 240
11.3 Net wealth, growth and debt: aggregate levels, 1990:02–2004:04 240
xvi Figures
11.4 Composition of firms according to financial structure,
1990:02–2004:04 241
11.5 Composition of firms’ assets according to financial structure,
1990:02–2004:04 242
11.6 Probability of hedge, speculative and Ponzi finance among
firms, 1990:02–1994:04 247
11.7 Probability of hedge, speculative and Ponzi finance among
firms, 1995:01–2004:04 248
13.1 Mexico: macroeconomic rate of unemployment, 1970–2004 266
13.2 GDP, unemployment, output gap and employment rate,
1985.1–2006.4 267
13.a.1 Model 1 first differences 270
13.a.2 Model 1: first differences 271
13.a.3 Diagnostic tests: correlogram, density, qqplot, cusum residual 271
13.a.4 Model 2: output gap 272
13.a.5 Model 2: output gap 272
13.a.6 Diagnostic tests: correlogram, density, qqplot, cusum residual 273
13.a.7 Model 3: fitted trend and elasticity 273
13.a.8 Model 3: fitted trend and elasticity 273
13.a.9 Diagnostic tests: correlogram, density, qqplot, cusum residual 274

1.1 A comparison of Gini ratios 14

3.1 Latin America and the Caribbean countries considered in the
sample 68
3.2 Panel data regressions of per capita income 71
3.3 Panel data regressions of per capita income 76–7
4.1 Inward FDI stock as a percentage of GDP, 1990–2006 94
4.2 Transaction values of cross­border M&As of privatised firms 101
4.3 Share of cross­border M&A sales in total FDI inflows in
Mexico and Brazil 102
5.1 Mercosur and the CAN 112
5.2 Gravity equation for the panels of Mercosur and CAN:
random effects 123
5.3 Gravity equation for the panels of Mercosur and CAN:
fixed effects 124
5.4 Exogeneity Hausman Test for distance in the gravity equation 126
5.5 Distance as the only regressor in the gravity equation 127
6.1 Growth and productivity 135
6.2 Fiscal policy 140
6.3 Investment composition 142
6.4 Debt sustainability indicators 143
6.5 Manufacturing exports 143
7.1 FDI inflows into Chile by sector, 1974–2006 153
7.2 FDI inflows into Chile by geographical origin, 1974–2006 155
7.3 Variable description 160
7.4 Summary statistics 161
7.5 Random effects estimation of the baseline equation 162
7.6 Random effects estimation of the augmented equation 164
8.1 Trade openness coefficient 170
8.2 Intra and extra Mercosur trade flows: simple average 170
8.3 Tariff structure: Uruguay 178
8.4 Price changes from Mercosur 179
8.5 Unit­root test: tradable and non­tradable prices 181
8.6 Prices cointegration 182
xviii Tables
8.7a Poverty: before and after trade reform 184
8.7b Poverty: before and after trade reform 185
8.8 Change in income inequality 185
8.9 Unit-root test: ADF 186
8.10 Engle-Granger: cointegration test 187
8.11 Poverty and inequality effects of liberalisation in the beef
international trade 188
8.12 Change in the probability of being employed after a Free Trade
Agreement with USA 188
9.1 Average real hourly income per level of education 201
9.2 Changes in average hourly income and share of educational
levels 202
9.3 Average real hourly income and educational attainment per
sector 203
9.4 Income bill share per sector and level of education 204
9.5 Performance of sectors (labour income) 206–7
9.6 Returns to education (labour income) 209
9.7 Average real monthly income, Gini, and composition of
income receivers 211
9.8 Decomposition of household Gini by income source 213
9.9 The effect of market openness on income (traded sector) 214
10.1 Optimal consumption shares, parameters, and estimates 231
11.1 Typology of the growth rate of firms 236
11.2 Average values for the sample of firms as a whole 239
11.3 Average percentage composition of the companies according
to their financial structure 242
11.4 Dependent variable: bit(t subscript directly beneath
superscript i) 243
11.5 Dependent variable: wit(t subscript directly beneath
superscript i) 245
11.6 Dependent variable: Fit(t subscript directly beneath
superscript i) 246
11.7 Average of the estímate probability of financial structure of
firms 248
12.1 Variables 255
12.2 Estimate for municipalities with high density of population
(urban) and weighted by the product 258
12.3 Estimate for municipalities with low density of population
(rural) and weighted by the product 259
12.4 Estimate for municipalities with high density of population
(urban) and weighted by the number of registered units 260
12.5 Estimate for municipalities with low density of population
(urban) and weighted by the number of registered units 260
13.1 Okun models 265
13.2 Mexico: Okun estimations 268
Tables xix
13.a.1 Basic statistics and unit roots, 1985.1–2006.4 269
13.a.2 Granger Causality Test, 1985.1–2006.4 for an unrestricted
VAR(5) 269
13.a.3 Mexico: Okun’s law, 1985.1–2006.4 270
13.a.4 Model 2 271
13.a.5 Model 3 273

Gerardo Angeles-Castro is Research Economist and Head of Research and

Graduate Studies at the School of Economics in the Instituto Politécnico
Nacional, Mexico.
Blanca L. Avendaño-Vargas is Research Economist at the Faculty of Eco-
nomics in the Benemérita Universidad Autónoma de Puebla, Mexico.
Fernando Borraz is Senior Researcher at the Central Bank of Uruguay and also
Head of the Department of Economics and full-time Professor at Montevideo
University, Uruguay.
Leobardo de Jesús is Research Economist at the Universidad Nacional
Autónoma de México (UNAM), Mexico.
Alcino Ferreira Câmara-Neto is Dean of the Centre for Economic and Legal
Sciences (CCJE) and Chairman of the Development Council in the Federal
University of Rio de Janeiro (UFRJ); he is also Chief Editor of the Journal of
Applied Social Science of the CCJE­UFRJ, Brazil.
Daniel Ferrés is Assistant Professor at Montevideo University, Uruguay.
Thomas Goda is PhD student in Economics at the London Metropolitan Uni-
versity, UK.
Matteo Grazzi is Economist at the Science and Technology Division of the
Inter­American Development Bank Washington, DC, USA. He is also an
external researcher at the Centre for Research on Latin American Studies and
Transition Economies Studies (ISLA) of the Bocconi University, Milan, Italy.
Clemente Hernández-Rodríguez is Coordinator of Research and Consultancy
Centre in Administration and Management, and Associate Director of
Research in the Asia Pacific Institute at Tecnológico de Monterrey, Campus
Guadalajara, Mexico.
Eduardo Loría is Research Economist at the Universidad Nacional Autónoma
de México (UNAM), Mexico.
Omar Neme-Castillo is Research Economist at the School of Economics in the
Instituto Politécnico Nacional, Mexico.
Contributors xxi
Penélope Pacheco-López is consultant for UNIDO­Mexico’s Office in the area
of the competitiveness of the Mexican manufacturing sector.
Ignacio Perrotini-Hernández is Research Economist and Head of the Graduate
Faculty of Economics at the Universidad Nacional Autónoma de México
(UNAM), Mexico; he is also the Editor of Investigación Económica (Eco-
nomic Research), a bilingual leading journal in Latin America.
Humberto Ríos-Bolívar is Research Economist at the School of Economics in
the Instituto Politécnico Nacional, Mexico; he is also the Editor of the journal
Panorama Económico (Economic Panorama).
Jaime Ros is currently Professor of Economics at the Universidad Nacional
Autónoma de México (UNAM). He also taught Economics at the University
of Notre Dame and is a Faculty Fellow at the Helen Kellogg Institute of Inter-
national Studies.
Máximo Rossi is Professor of Economics at De la República University, Uru-
A.P. Thirlwall is Professor of Applied Economics at the University of Kent,
José Carlos Trejo-García is PhD student in Economics at the School of Eco-
nomics in the Instituto Politécnico Nacional, Mexico.
Ana Lilia Valderrama-Santibáñez is Research Economist at the School of
Economics in the Instituto Politécnico Nacional, Mexico.
Juan Alberto Vázquez-Muñoz is Research Economist at the Faculty of Eco-
nomics in the Benemérita Universidad Autónoma de Puebla, Mexico.
Francisco Venegas-Martínez is Research Economist at the School of Eco-
nomics in the Instituto Politécnico Nacional, Mexico.
Matías Vernengo is Associate Professor at the Department of Economics, Uni-
versity of Utah, USA.
Jaime Ros

The impact of economic liberalization on growth and development in developing

countries is by now the subject of a vast literature that includes a variety of per-
spectives. This book takes up this subject and makes a substantial contribution to
this literature, concentrating largely, although not exclusively, on Latin America.
Have the processes of trade liberalization since the 1980s lived up to their prom-
ises? What went wrong with the policy package known as the Washington Con-
sensus? What is the case for a developmental industrial policy in order to address
the shortcomings of economic liberalization? What are the links between inter-
national trade and the level of economic development? What have been the
effects of regional trade agreements on trade flows and foreign direct investment
(FDI)? Do North–South trade agreements really attract more FDI than South–
South agreements? What about the role in FDI of bilateral investment treaties?
These are some of the key questions addressed by this book through an array of
statistical and econometric methods applied to samples of developing and, in
particular, Latin American countries, before and after and cross country compar-
isons, as well as case studies of Argentina, Chile, Mexico and Uruguay.
The first part of the book focuses on the effects of trade liberalization and the
processes of regional integration in Latin America. Trade liberalization has not
lived up to expectations regarding growth performance, poverty reduction and
income distribution as argued by Pacheco-López and Thirlwall in their review of
the evidence for developing countries (a conclusion shared by country studies in
the second and third parts of this volume). This, at least in part, is due to the fact
that trade liberalization, especially in poor countries, has had deleterious effects
on the balance of payments constraint on growth and also because, contrary to
the orthodox predictions of the Stolper–Samuelson theorem, income inequalities
have tended to increase in many developing countries as unskilled labour has not
benefited from the expansion of trade. More generally, the policy package
known as the Washington Consensus has not improved economic performance
and closed the development gap in the large Latin American economies. This is
attributed by Perrotini­Hernández, Vázquez­Muñoz, and Avendaño­Vargas to
the fall in investment rates (with the exception of Chile among the large Latin
American economies) and the increase in the income elasticity of imports which
contrasts sharply with the Chinese experience. The policy implications derived
Foreword xxiii
by the authors include an alternative development strategy centred on the imple-
mentation of an industrial policy for development (Perrotini­Hernández,
Vázquez­Muñoz, and Avendaño­Vargas) as well as a trade strategy for develop-
ment centred on acquiring dynamic comparative advantages and a call for a new
world trade order that works for development in poor countries (Pacheco­López
and Thirlwall).
All this doesn’t mean that trade liberalization and regional trade agreements
such as Mercosur and the Andean Community have not led to an increase in
trade flows among trading partners (without reducing extra­regional trade flows).
The increase in trade flows is shown by Hernández­Rodríguez using a gravity
model which relates trade to income, distance between trading partners and
membership in a trading bloc. Trade agreements have also been accompanied by
increasing flows of FDI as shown by Goda in his interesting comparison of the
experiences of Brazil under Mercosur and Mexico under NAFTA. Also, these
trends in trade and FDI have generally had positive static effects on per capita
income according to Ríos-Bolívar and Neme-Castillo in a study of a set of 21
Latin American and Caribbean countries.
The second part of the book deals with trade reforms and development experi-
ence presenting three case studies of Latin American countries (Argentina,
Chile, and Uruguay). Taking a long view of Argentina’s development, Ferreira
and Vernengo argue that the recent growth and equity performance (except for
the period since 2003) under economic liberalization compare rather unfavoura-
bly to that of the import substitution industrialization period. Grazzi examines
the links between trade and investment liberalization (following the establish-
ment of bilateral investment treaties and free trade agreements) and the flows of
FDI in Chile, the third largest recipient of foreign investment in Latin America
(after Brazil and Mexico) in absolute terms and the second in per capita terms
after Trinidad and Tobago. Using an augmented gravity model, the author finds
that bilateral investment treaties have a positive and significant impact on FDI
inflows, while there is little or no evidence of a significant effect of double taxa-
tion treaties and free trade agreements. The experience of Uruguay under Merco-
sur is the subject of the chapter by Borraz, Ferrés and Rossi. They conclude that
while specific groups of the population (those with higher and lower incomes)
reaped more of the mild gains from trade than the middle income groups, they
could not find any evidence about absolute losers resulting from Mercosur.
The third part focuses on Mexico and contains a heterogeneous collection of
chapterss dealing with different aspects of its economic performance. A first
trend that characterizes the recent past is the rise and fall of inequality during the
economic liberalization period. This is the subject of the chapter by Angeles-
Castro who finds that income inequality worsens after liberalization, mainly due
to an increase in the skilled labour premium, a reduction of agricultural incomes
relative to services and a negative impact of market openness on wages in the
traded sector. This worsening of inequality is followed after 1998 by a decrease
in inequality associated to the decrease in the returns to skilled labour and a
weaker effect of trade on wages in the traded sector. Venegas-Martínez focuses
xxiv Foreword
on growth performance from 1930 to 2002. The author brings currency, market,
debt and fiscal risk factors in an intertemporal optimizing model of endogenous
growth showing that risk factors may lead to qualitative changes in the determi-
nants of growth in contrast with the deterministic setting. Perrotini­Hernández,
Avendaño­Vargas, and Vázquez­Muñoz use a Minskyan framework to evaluate
Mexico’s inflation targeting monetary policy, concluding that the management
of the interest rate under this regime may encourage Ponzi financing. Using a
production function approach, Ríos­Bolívar and Valderrama­Santibáñez study
the relationship between the labour productivity growth and investment in
research and development (R&D) for the manufacturing sector, trade and serv-
ices, finding that investment in R&D has a positive effect on productivity growth
in the three sectors. Finally, Loría and de Jesús focus on the relationship between
unemployment and output growth, confirming for Mexico the validity of Okun’s
law that relates the change in unemployment to the rate of GDP growth.
Overall, this is a very valuable volume of essays that contributes significantly
to the literature on its subject. It must be read by all those interested in the effects
of trade and investment liberalization in developing countries, who will find in it
not only new evidence but also fresh, intelligent and generally heterodox per-
spectives on the subject matter.

The editors wish to thank Gabriela Cruz­González, Estefania Molerés­Regalado,

Jonathan Ortiz­Galindo, and Paulina Salazar­Rivera, for assisting in the editing
of this book. Usual disclaimers apply.

ACP African, Caribbean and Pacific Countries

ADF Augmented Dickey-Fuller
AFTz Andean Free Trade zone
ALADI Asociación Latinoamericana de Integración (Latin American
Association of Integration)
APEC Asian­Pacific Economic Cooperation
BADECEL Banco de Datos Estadísticos de Comercio Exterior (Statical Data
Bank for Foreing Trade)
BADEINSO Social Statistics and Indicators
BCIA Banco de Crédito Industrial Argentino (Argentinian Industrial
Bank of Credit)
BCU Central Bank of Uruguay
BIS Bank for International Settlements
BIT Bilateral Investment Treaties
BM Banco de México
CAN Andean Community
CB Central Bank
CEMPE Center of Economic Modeling and Forecasting
CEPII Centre d’Etudes Prospectives et d’Informations Internationales
(Centre for Prospective Studies and International Information)
CES Constant Elasticity of Substitution
CET Common External Tariff
CETES Certificados de Tesorería (Mexico’s Treasury bonds)
CISEA Centro de Investigaciones Sociales sobre el Estado y la
Administración (Centre for Social Research on the State and
CONAPO Consejo Nacional de Población (National Council of Population)
CPI Consumer Price Index
DTTs Double Taxation Treaties
ECH Encuesta Continua de Hogares
ECLAC Economic Commission for Latin America and the Caribbean
ENIGH Encuesta Nacional de Ingresos y Gastos de los Hogares (National
Survey of Households’ Expenditures an Incomes)
Acronyms xxvii
EPAs Economic Partnership Agreements
ESE Escuela Superior de Economía (Superior School of Economics)
EU European Union
FDI Foreign Direct Investment
FE Fixed-Effects
FEM Fixed Effects Model
FGLS Feasible Generalised Least Squares
FL Financial Liberalisation
FTAs Free Trade Agreements
FTAA Free Trade Area of the Americas
GDP Gross Domestic Product
GLS Generalised Least Squares
GNP Gross National Product
IADB Inter American Development Bank
IFAD International Fund for Agricultural Development
IHS Inverse Hyperbolic Sine
IMF International Monetary Fund
INE Instituto Nacional de Estadística (National Institute of Statistics)
INEGI Instituto Nacional de Estadística, Geografía e Informática
(National Institute for Statistics, Geography and Informatics)
IPN Instituto Politécnico Nacional (National Polytechnic Institute)
ISI Import Substituting Industrialization
IT Inflation Targeting
IV Instrumental Variables
LM Lagrange Multiplayer Test
Mercosur Mercado Común del Sur (South American Common Market)
MF Ministry of Finance
MNCs Multinational Companies
MSM Mexican Stock Market
MUR Macroeconomic Unemployment Rate
M&As Mergers and Acquisitions
NAFTA North American Free Trade Agreement
OECD Organisation for Economic Co-operation and Development
OLS Ordinary Least Squares
PCSE Panel Corrected Standard Errors Method
PCY Per Capita Income
PPP Purchasing Power Parity
RE Random-Effects
REM Random Effects Model
RIAs Regional Integration Agreements
RoO Rules of Origin
ROW Rest of the World
R&D Research and Development
SBTC Skill-Biased Technological Change
SETH Skill Enhancing Trade Hypothesis
xxviii Acronyms
SIMBAD Municipal Information System Database
SITC Standard International Trade Classification
SM Secretaría del Mercosur (Mercosur’s Secretary)
SST Stolper-Samuelson theorem
TL Trade Liberalization
TNCs Transnational Companies
TOBIT Tobit Model (it was first proposed by James Tobin, and involves
aspects of Probit analysis)
UK United Kingdom
UN United Nations
UNCTAD United Nations Conference on Trade and Development
US United States
USA United States of America
WC Washington Consensus
WDI World Development Indicators
WIDER World Institute for Development Economics Research
WTO World Trade Organization
YPF Yacimientos Petrolíferos Fiscales (Fiscal Oilfields)
Editors’ introduction
Gerardo Angeles-Castro, Ignacio Perrotini-Hernández
and Humberto Ríos-Bolívar

The principal themes pursued in this book emerge from the great transformation
that the Latin American and the Caribbean economies experienced in the aftermath
both of the foreign debt crisis of 1982 and of the macroeconomic stabilisation pol-
icies that vividly and painfully produced the so-called ‘lost decade’ of the 1980s.
Latin America implemented an economic liberalisation process during the late
1980s and the 1990s. The main policy reforms involved in that course can be
summarised as privatisation of state owned firms, trade openness, deregulation of
the foreign direct investment (FDI) regime and fiscal discipline. Latin American
countries have also embarked in regional trade agreements, the most important
ones being Mercosur and the North American Free Trade Agreement (NAFTA).
The book compares results from the experience of North–South and South–South
moulds of integration. Thus, the impacts of these policies on growth, development,
technological progress, poverty and inequality are analysed. Orthodox and hetero-
dox economic policies and theories are discussed along with relevant empirical
evidence with a view to assess the relative merits of the various policy reforms
applied by different countries in the region, on the one hand, and the experience of
integration into the global economy, on the other.
There are 13 chapters in this collection linked in varying ways to the series of
economic reforms introduced in the region in the last decades. In this introduc-
tion we do not intend to provide a comprehensive résumé of the chapters of the
present book or to exhaust the great richness of the analyses and debates con-
tained therein. We just want to give a general appraisal of the editorial structure
of the volume. There are many authors in this volume, and undoubtedly they
disagree with each other on many important economic issues and policy ques-
tions. Yet, we have chosen to bring these chapters together because they indis-
putably share a common concern, namely the search for an alternative economic
policy model that best suits Latin America’s future economic development. The
book is organised in three parts.

Part I
The first part comprises five chapters with the theme of trade liberalisation,
development and regional integration. It is very appropriate to begin the book
2 G. Angeles-Castro et al.
with a contribution written by Pacheco-López and Thirlwall, who examine the
impact of trade liberalisation on the trade-off between development and the
balance of payments. The authors conclude that a new world trade order is
needed where poor countries are allowed the acquisition of dynamic compara-
tive advantage. Perrotini-Hernández, Vázquez-Muñoz and Avendaño-Vargas
take up the issue of the impressive development gap that separates the world’s
most advanced economy, the United States, and Latin America. They discuss the
relative merits of the Washington Consensus strategy within a context where
‘classic conditions for market failure’ are ubiquitous. They make the case for a
developmental industrial policy which calls for a fiscal policy regime aimed at
stabilising investment, growth and employment. Ríos-Bolívar and Neme-Castillo
study the relationship between international trade and per capita income in 21
countries of Latin America and the Caribbean in the period 1977–2007. They
estimate the relationship between imports of capital goods, physical capital, per
capita income and human capital. Among other things, they find the existence of
a ‛literacy trap’. Thomas Goda wrestles with the relationship between regional
integration and FDI in the Mercosur and the NAFTA areas. He compares the
impacts of FDI in North–South and South–South integration environments and
finds empirical results which may be seen as conflicting with received theory.
Therefore, he states, regional integration as an effective means of attracting FDI
should not be taken for granted. Hernández-Rodríguez, in turn, builds a gravity
model to explore the factors that influence trade flows within trade blocs and
outside such blocs. He concludes that geographic factors are most important, fol-
lowed by income size.

Part II
In the second part of the volume a set of three case studies is presented. Câmara-
Neto and Vernengo offer a critical assessment of the process of economic devel-
opment of Argentina from the mid-1940s to the present time. Their historical
approach leads them to identify three stages of economic growth (the state-led
growth model from 1946 to 1976, the liberal stage from 1976 to 2001 and the
commodity boom period from 2002 to 2008) which have been unsustainable
altogether. In addition, Grazzi examines in some detail the determinants of FDI
inflows into Chile through the estimation of a gravity model from 1990 to 2005.
His econometric results lead him to conclude that investments in Chile are
mostly of the resource-seeking type; that FDI is negatively affected by distance
and positively affected by the source country’s GDP and GDP per capita. This
second part ends with an analysis of the impact of trade liberalisation on income
distribution in Uruguay. Borraz, Ferrés and Rossi study the long-term effects on
poverty and inequality resulting from Uruguay integrating into Mercosur. Their
empirical results show that even if gains from trade liberalisation can be
observed, those benefits are unevenly distributed. Therefore, trade liberalisation
is not necessarily and always a pro-poor growth strategy. The ultimate effect of
economic liberalisation and market integration depends on the impact of price
Editors’ introduction 3
changes of both tradable and non-tradable goods on the income share of the
various social classes participating in the production process.

Part III
The book ends with Part III which contains five chapters, though mainly theoreti-
cally heterogeneous in their inspiration, solidly anchored in a thorough empirical
analysis of various features of the Mexican economy. This closing part should dis-
courage any scrutiny of Mexico’s recent development experience through the
narrow prism of a particularly raw version of economic religiosity. Some of the
chapters here view the economic reform with the greatest of suspicion, whereas
others entertain a more congenial approach. Angeles-Castro deals with factors
driving changes in income distribution in post-reform Mexico. He argues that the
main factors that contributed to increase inequality following economic liberalisa-
tion are the relative expansion of the average income in the service sector in rela-
tion to the agricultural and manufacturing sectors, a negative relationship between
wages and market openness, and the increase in skill premium. On the other hand,
between 1998 and 2006, income dispersion fell gradually and some of the factors
driving this trend are the decrease in returns to skill and a weaker effect of market
openness in wages. Venegas-Martínez works out a stochastic endogenous growth
model that accounts for how risk factors (currency, market, debt and fiscal risk
factors) affect economic growth. His model, in which a Brownian motion and a
Poisson process drive exchange-rate expectations and a geometric Brownian
motion guides a tax rate on wealth, sheds new lights to carry out simulation experi-
ments and empirical research. He argues that his model explains the average and
variance of Mexico’s economic growth over the period of time under analysis. In
addition, Perrotini-Hernández, Avendaño-Vargas and Vázquez-Muñoz work with
data for 47 non-financial firms quoted in the Mexican stock exchange market from
1990 to 2004, and conduct a microeconomic analysis of financial instability in
Mexico. They conclude that financial fragility can occur when the central bank
targets low inflation and appreciates the exchange rate due to the presence of high
pass-through effects from currency volatility on to the price level. Trejo-García,
Ríos-Bolívar and Valderrama-Santibáñez examine in some detail the relationship
between human capital productivity and economic growth under market openness
conditions and, focusing on the behaviour of three economic sectors – manufactur-
ing, commerce and services – they show that investment on research and develop-
ment has a positive effect and in most cases the coefficients are statistically
significant. Lastly, Loría and de Jesús assess the validity of Okun’s Law in
Mexico. They estimate Okun’s Law for the Mexican economy from 1985 to 2006
and found robust evidence in favour of bidirectional causality between output and
Finally, the analyses comprising this volume provide rich diagnoses and
empirical analysis which, we hope, may contribute to the debate for the enhance-
ment of economic development of Latin America. May this book inspire further
research on the topics dealt with.
Part I

Trade liberalization,
development and regional
1 Has trade liberalisation in poor
countries delivered the promises
Penélope Pacheco-López and A.P. Thirlwall

Trade liberalisation has not lived up to its promises. But the basic logic of trade –
its potential to make most, if not all, better off – remains. Trade is not a zero-sum
game in which those who win do so at the cost of others; it is, or at least can be, a
positive-sum game, in which everybody is a winner. If that potential is to be real-
ised, first we must reject two of the long-standing premises of trade liberalisation:
that trade liberalisation automatically leads to more trade and growth, and that
growth will automatically ‘trickle down’ to benefit all. Neither is consistent with
economic theory or historical experience.
(Stiglitz, 2006)

1.1 Introduction
The last decades have witnessed tremendous pressure on poor developing coun-
tries to liberalise their trade. The free trade mantra preached by developed coun-
tries and major international development organisations has become like a
religion, holding out the promise that if poor countries adopt the faith, they will
somehow be ‘saved’. The broad purpose of this chapter is to challenge this sim-
plistic view. The chapter is based on a review of the vast literature of theory and
case studies (including research of our own) on the relation between trade liber-
alisation and economic performance across the world (see Thirlwall and
Pacheco-López, 2008), which leads us to four general, but important, conclu-
sions. The first is that while there can be static gains from trade (if certain crucial
assumptions are met) there is nothing in the theory of trade per se which demon-
strates conclusively that trade liberalisation will launch a country on a higher
sustainable growth path. Even Jagdish Bhagwati (2001), the high priest of free
trade, is honest about that (see below). Second, the impact of trade liberalisation
on reducing world poverty has been minimal, and may have increased it. Third,
trade liberalisation has almost certainly worsened the distribution of income
between rich and poor countries, and between unskilled wage-earners and other
workers within countries, contrary to the predictions of orthodox theory. Finally,
the evidence is fragile that the economic growth performance of countries that
have liberalised extensively is in any way superior to countries that have not.
The timing, sequencing and context of liberalisation are of prime importance in
8 P. Pacheco-López and A.P. Thirlwall
determining the impact of liberalisation. What really matters for growth per-
formance is domestic economic policy and growth-supportive institutions. This
will lead us at the end of the chapter to a brief discussion of trade strategy for

1.2 What is wrong with orthodox trade theory?

Orthodox trade theory is based on Ricardo’s (1817) law of comparative advant-
age, and the Heckscher–Ohlin theorem which argues that countries will gain by
specialising in the production of goods which use their most abundant factor of
production (Heckscher, 1919; Ohlin, 1933). Paul Samuelson (1962) cites Ricar-
do’s theory of comparative advantage as one of the few laws in economics ‘that
is both true and non-trivial’. There are, indeed, static welfare gains to be had by
countries specialising in goods in which they have the greatest comparative
advantage (or lowest opportunity cost), but two crucial, often-forgotten, assump-
tions need to be met. The first is that in the process of resources reallocation, full
employment is preserved, but this is not guaranteed. If unemployment arises, the
welfare gains from greater specialisation may be offset by the welfare losses of
unemployment. As Keynes (1930) rightly says ‘free trade assumes that if you
throw men out of work in one direction you re-employ them in another. As soon
as this link in the chain is broken the whole of the free trade argument breaks
down.’ The second crucial assumption is that in the process of freeing trade,
balance of payments equilibrium is preserved, which is also not guaranteed. In
orthodox theory, the balance of payments is assumed to look after itself without
affecting output and employment. This was the implicit assumption of the gold
standard adjustment mechanism, and is also implicit in the theory of flexible
exchange rates. But if trade liberalisation leads to a faster growth of imports than
exports and the nominal exchange rate is not an efficient balance of payments
adjustment weapon, then output will need to contract to reduce imports, leading
to welfare losses. As we shall see later, this has been the experience of many
developing countries forced to liberalise prematurely.
In fact, the existence of unemployment provides one of the major economic
arguments for protection, as outlined in Johnson’s (1964) classic paper on tariffs
and economic development. Unemployment means that the social cost of labour
is less than the private cost so that a welfare gain is possible by encouraging
more domestic employment until the social cost of production is equal to the
world price of goods. A subsidy to labour, however, is the first best policy
because an equivalent tariff would reduce consumer surplus. Johnson also out-
lines some of the other classic economic arguments for protection such as the
infant industry argument, the externalities argument and the optimal tariff argu-
ment. But Rodrik (1988) is correct that despite the body of trade theory which
legitimises protection, the arguments have still not penetrated the vast literature
on trade policy in developing countries even though the market imperfections
that the arguments reflect are more serious in developing countries than in
developed countries.
Trade liberalisation in poor countries 9
As well as potential static gains from trade (although not guaranteed, and in
any case small (see Dowrick, 1997)) there are also possible dynamic gains which
arise through the greater flow of ideas, new knowledge, investment and econo-
mies of scale if the domestic market for output is small. The dynamic effects of
trade, however, depend primarily on what countries specialise in; whether
natural resource activities or manufacturing. John Stuart Mill (1848) pointed this
out in the nineteenth century, and Stiglitz (2006) today makes the same enduring

Without protection, a country whose static comparative advantage lies in, say
agriculture, risks stagnation; its comparative advantage will remain in agri-
culture, with limited growth prospects. Broad-based industrial protection can
lead to an increase in the size of the industrial sector which is, almost every-
where, the source of innovation; many of these advances spill over into the
rest of the economy as do the benefits from the development of institutions,
like financial markets, that accompany the growth of an industrial sector.
Moreover, a large and growing industrial sector (and the tariffs on manufac-
tured goods) provides revenues with which the government can fund educa-
tion, infrastructure, and other ingredients for broad-based growth.

In other words, if trade is to be an engine of growth, poor countries need to

acquire new comparative advantage in goods that have favourable production
and demand characteristics. Structure matters for economic growth. This is rec-
ognised in ‘new’ trade theory pioneered by Krugman (1984, 1986) in the 1980s,
who shows there is a case for protecting industries with spillovers and externali-
ties, and for using import substitution for export promotion. In most standard
growth models, however, the effect of trade on growth is ambiguous. For
example, in the canonical neoclassical Solow model (1956), trade cannot affect
the steady-state growth rate, because it is treated as an exogenous constant. Only
in the ‘new’ growth theories of, for example, Grossman and Helpman (1991a,
1991b) does trade have the potential to raise the growth rate permanently
through learning and spillover effects, but they have to be continuous. Bhagwati
(2001), the most ardent advocate of free trade, even for poor developing coun-
tries, frankly admits:

Those who assert that free trade will lead necessarily to greater growth
either are ignorant of the fine nuances of the theory and the vast quantity of
literature to the contrary on the subject at hand or are nonetheless basing
their argument on a different premise; that is, that the preponderant evidence
on the issue (in the post-war period) suggests that free trade tends to lead to
greater growth after all. In fact, where theory includes several models that
can lead in different directions, the policy economist is challenged to choose
the model that is most appropriate to the reality she confronts. And I would
argue that, in the present instance, we must choose the approach that gener-
ates favourable outcomes for growth when trade is liberalised.
10 P. Pacheco-López and A.P. Thirlwall
The issue is empirical, but certainly history is not on the side of the free-traders.
None of the now-developed countries transformed their economies on the basis
of laissez-faire, laissez-passer. Great Britain started to protect and foster indus-
tries as early as the late fifteenth century under Henry VII, and did not start dis-
mantling the structure of protection until the repeal of the Corn Laws in 1848.
From then on Great Britain preached free trade, but it had already attained
technological superiority in the world economy, and such preaching, as List
(1885) remarked, was like ‘kicking away the ladder’. The United States followed
Great Britain’s protectionist route at the end of the eighteenth century under the
influence of the Treasury Secretary, Alexander Hamilton, who, in 1791, first
coined the term ‘infant industry’. Adam Smith’s advice to the United States in
his Wealth of Nations (1776) was to pursue free trade:

Were the Americans, either by combination or by any other sort of violence,

to stop the importation to European manufactures, and, by thus giving a
monopoly to such of their own countrymen as could manufacture the like
goods, divert any considerable part of their capital into this employment,
they would retard instead of accelerating the further increase in the value of
their annual produce, and would obstruct instead of promoting the progress
of their country towards real wealth and greatness.

If the United States had followed Smith’s advice, it would have remained an
economic backwater instead of becoming the richest country in the world based
on high productivity in industry. The same can be said of modern-day economic
giants, such as Japan and South Korea, whose comparative advantage once lay
in rice, but who, through selective protection, import substitution, export promo-
tion and directed credit, transformed themselves into industrial power-houses
(see Chang, 2005). The newly industrialising countries of South-East Asia, and
particularly China, are pursuing the same route to development – transforming
their industrial structure through deliberate policy intervention – and are growing
fast as a consequence. Stiglitz (2006) is right when he says that ‘economists who
promise that trade liberalisation will make everybody better off are being disin-
genuous. Economic theory (and historical experience) suggests the contrary.’ All
we know is that as countries get richer they dismantle trade restrictions, not that
they get richer because they liberalise trade. The issue for poor developing coun-
tries today is not whether to protect, but how to protect in order to ensure the
dynamic efficiency of their nascent industrial activities.

1.3 Poverty and income inequality within countries

At this moment in time, nearly one billion of the world’s population live on less
than $1 a day, and 2.7 billion live on less than $2 a day (Chen and Ravallion,
2004). In other words, over one-third of the world’s population lives in absolute
poverty. Advocates of trade liberalisation promise that the freeing of trade will
lift people out of poverty. The former European Trade Commissioner, Peter
Trade liberalisation in poor countries 11
Mandelson, wrote in the Guardian newspaper (3 October 2008) that globalisa-
tion is the greatest engine of poverty reduction the world has ever seen. If he had
looked at the facts, however, he would see that since 1980 the absolute number
of people in poverty has not decreased. The number on less than $2 a day has
increased from 2.4 billion to 2.7 billion, and the number on less than $1 a day
(excluding China) has increased from 848 million to 870 million. The reduction
in the total number living on less than $1 a day is because of a fall in China in
the early 1980s, but this was due to agricultural reforms not to trade liberalisa-
tion. Winters et al. (2004), in their survey of trade liberalisation and poverty,
claim that ‘theory provides a strong presumption that trade liberalisation will be
poverty alleviating in the long run and on average’. This is simply not true,
because, as we have seen, the theory of trade liberalisation says nothing definite
about economic growth. The impact of trade liberalisation on poverty depends
on its effects on employment and prices. Trade liberalisation can easily cause
poverty by throwing people out of work. For example, since the NAFTA agree-
ment was signed between the US, Canada and Mexico in 1994, two million
Mexican maize farmers have lost their jobs because they cannot compete with
subsidised maize from the US. Trade liberalisation can provide new opportun-
ities in the export sector, but only if the sector is prepared.
The statistical research on the relation between trade liberalisation and
poverty is very inconclusive. The most comprehensive study is by Ravallion
(2006) who takes 75 countries where there have been at least two household
surveys on poverty, and runs a simple regression of the percentage change in the
poverty rate on the percentage change in the ratio of trade to GDP (as a proxy
for liberalisation). There is a statistically significant negative coefficient of 0.84,
but the correlation is very fragile. For example, controlling for initial conditions
makes the relation insignificant, and adding other control variables makes no dif-
ference. Ravallion concludes ‘it remains clear that there is considerable variation
in the rates of poverty reduction at a given rate of expansion of trade volume’.
Equally, however, ‘based on the data available from cross-country comparisons,
it is hard to maintain the view that expanding trade, in general, is a powerful
force for poverty reduction in developing countries’.
At the same time as the absolute numbers in poverty have been increasing,
the distribution of income within poor countries has also been widening, con-
trary to the orthodox predictions of the Hecksher–Ohlin (and Stolper–Samuel-
son, 1941) theorems. Goldberg and Pavcnik (2007), in their survey of the
distributional effects of globalisation in developing countries, say: ‘while ine-
quality has many different dimensions, all existing measures for inequality in
developing countries seem to point to an increase in inequality which in some
cases is severe’. The major cause of income inequality is wage inequality
between skilled and unskilled workers. Orthodox trade theory predicts a narrow-
ing of wage inequality in poor countries because their comparative advantage
should lie in the production and export of goods using abundant unskilled labour.
This narrowing has not happened for four main reasons: first, trade-related, skill-
biased technical change; second, competition between poor countries; third,
12 P. Pacheco-López and A.P. Thirlwall
flows of FDI adding to the demand for skilled labour; and finally, in some cases,
depressed demand for unskilled labour where trade liberalisation has caused
balance of payments difficulties (see Arbache et al. 2004 for a case study of
By far the most detailed study of the impact of trade liberalisation on the dis-
tribution of income is that by Milanovic (2005a). First, in his introductory survey
of the existing literature, he remarks:

The conclusions run nearly the full gamut, from openness reducing the real
income of the poor to openness raising the income of the poor proportion-
ately less than the income of the rich to raising both the same in relative
terms. Note, however, that there are no results that show openness reducing
inequality; that is, raising the income of the poor more than the income of
the rich – let alone raising the absolute income of the poor by more.

Milanovic’s own research takes 321 household surveys from 95 countries in

1988, and 113 countries in 1993 and 1998 covering 90 per cent of the world’s
population. Income inequality is measured, not by a summary statistic such as
the Gini ratio or Theil index, but by the income of the ith decile of the popula-
tion relative to the mean level of income of the whole population. For each
decile, income inequality is then related to trade openness measured by the ratio
of total trade to GDP, and also to openness interacted with the level of income to
test whether the effect of openness on inequality varies with the level of income.
A regression is run for each of the ten deciles using the same independent varia-
bles. Two striking results emerge. First, increased openness reduces the income
share of the bottom six deciles. Second, the adverse effect of openness on ine-
quality is less the higher a country’s per capita income. The turning point for the
poor to benefit from increased trade is approximately US$7,500 at 1990 prices.
Barro (2000) and Spilimbergo et al. (1999) also find openness worsens income
inequality up to a certain point and then the effect diminishes. Milanovic con-
cludes: ‘openness would therefore seem to have a particularly negative impact
on poor and middle income groups in poor countries – directly opposite to what
would be expected from the standard Hecksher–Ohlin–Samuelson framework’.
The contrary conclusion to the above studies of Dollar and Kraay (2002,
2004) that ‘growth is good for the poor’ arises from their unusual procedure of
measuring trade in nominal US dollars, but measuring GDP at purchasing power
parity (PPP). Since GDP at PPP is much higher than in nominal dollars, this con-
siderably understates the ratio of trade to GDP in poor countries. For example,
China’s exports as a share of GDP measured at PPP are only 7 per cent com-
pared to 26 per cent if both trade and GDP are measured in nominal dollars. It is
this latter ratio that affects the income distribution, and which should be used in
studies of trade and income distribution.
Trade liberalisation in poor countries 13
1.4 International and global inequality
Not only has the distribution of income within poor countries been increasing
over time, but also the distribution of income between poor and rich countries.
Again, this contradicts the prediction of orthodox neoclassical theory (Solow,
1956) which argues that because the productivity of capital should be higher in
poor capital-scarce countries than in rich capital-saturated countries, poor coun-
tries should grow faster than the rich leading to a convergence of living stand-
ards across the world. There are many non-orthodox models to explain
divergence, associated with the names of Myrdal (1957), Hirschman (1958),
Kaldor (1970) and various Marxist writers, but nonetheless the orthodoxy pre-
vails despite the evidence.
The measurement of international inequality takes each country’s average per
capita income as a single unit, regardless of the distribution of income within
countries, and a Gini ratio can be calculated either unweighted or weighted by
population. Global inequality, by contrast, not only measures inequality between
countries but also within countries as well, giving a higher figure. The most
recent calculations of the Gini ratio of international inequality by Milanovic
(2005b), and of global inequality by Milanovic (2005b), Bourguignon and Mor-
risson (2002) and Sala-i-Martin (2002) are shown in Table 1.1. The unweighted
Gini ratio for international inequality shows a steady historical rise from 1820,
and also in the post-war period of trade liberalisation from 1952. The population-
weighted Gini ratio of international inequality shows a slight decline in recent
years due to the fast growth of populous countries such as China and India. If
China is taken from the sample, the population-weighted Gini ratio is also shown
to rise. The Gini ratio for global inequality has increased over time but has been
relatively static in recent years because while between-country inequality (popu-
lation weighted) has fallen slightly, income inequality within countries has
increased, particularly in China between the rural and urban sectors.
The question is: how much of this rising and persistent inequality across the
world is due to trade liberalisation? It is not easy to answer this question, but
attempts can be made. One methodological approach is to interact a measure of
trade openness with the level of per capita income (PCY) to test whether the
impact of openness varies with the level of development. This is what Dowrick
and Golley (2004) do, taking over 100 countries for two separate time periods,
1960–80 and 1980–2000, and regressing the growth of PCY on (1) trade as a per
cent of GDP, (2) an interaction term of trade openness and a country’s level of
PCY, (3) a dummy variable for specialisation in primary products, measured as
more than 50 per cent of exports and (4) a number of control variables. Separate
regressions are also run for developed and less developed countries. For the first
period 1960–80, a higher trade share of one percentage point is associated with
0.11 per cent faster growth, and the poorer the country, the slightly greater the
benefit from openness, meaning that trade was a force for convergence. But for
the second period, 1980–2000, this result is reversed. The impact of the trade
share on the growth of PCY is now negative (–0.072) and the interaction term
Table 1.1 A comparison of Gini ratios

Year International inequality1 Global (or world) inequality

Unweighted Population weighted Milanovic (2005b) Bourginon and Morrisson (2002) Sala-i-Martin (2002)

1820 0.2 0.12 – 0.5 –

1870 0.29 0.26 – 0.56 –
1890 0.31 0.3 – 0.59 –
1913 0.37 0.37 – 0.61 –
1929 0.35 0.4 – 0.62 –
1938 0.35 0.4 – – –
1952 0.45 0.57 – 0.64 –
1960 0.46 0.55 – 0.64 –
1978 0.47 0.54 – 0.66 0.66 (1970)
1988 0.5 0.53 0.62 – 0.65
1993 0.53 0.52 0.65 0.66 (1992) 0.64
1998 – – 0.64 – 0.63
2000 0.54 0.5 – – 0.63

1 Adapted from Milanovic (2005b), Table 11.1.
Trade liberalisation in poor countries 15
with the level of PCY is positive (+0.009) indicating that poor countries suffered
from trade openness more than rich countries, leading to divergence. Dividing
the 1980–2000 sample of countries into 33 poorest countries and the rest shows
no significant effect of the trade share on growth in the poorest countries, but the
richer countries gained about 0.012 per cent growth for a one percentage point
increase in the trade share. Specialisation in primary products had a strong negat-
ive effect on growth in the 1980–2000 period, reducing it on average by nearly 1
per cent; and the impact was even stronger in the poor country group – a differ-
ence of –1.76 per cent. Dowrick and Golley’s conclusion is that ‘trade has pro-
moted strong divergence in productivity [between countries] since 1980’.

1.5 Trade liberalisation and trade performance

The main purpose of trade liberalisation is to promote, or allow, the most effi-
cient allocation of a country’s resources to maximise its welfare. We have
already criticised the static nature of orthodox trade theory, and outlined some of
its limiting assumptions, but what has been the effect of trade liberalisation in
practice on countries’ trade performance, and ultimately on the growth of living
standards? This requires detailed statistical analysis. Research on export per-
formance before and after liberalisation gives mixed results depending on the
context in which trade liberalisation takes place, particularly the domestic eco-
nomic policy being pursued and world economic conditions. Also, in economet-
ric studies, results differ according to the methodology used and how
liberalisation is measured. The most comprehensive recent study is that of
Santos-Paulino and Thirlwall (2004) who take a panel of 22 developing coun-
tries from the four ‘regions’ of Africa, Latin America, East Asia and South Asia
that undertook significant trade liberalisation during the period 1972–97. Trade
liberalisation is measured by two indicators: first by duties on exports, and
second by a dummy variable taking the value of one in the year when trade lib-
eralisation took place (and continued) and zero otherwise. Panel data and time
series/cross-section estimation techniques are then applied to the determination
of export growth using a conventional export growth equation of the form:

xt = ao + a1(rert) + a2(zt) + a3(dxt) = a4(libt) (1.1)

where x is the growth of export volume, rer is the rate of change of the real
exchange rate, z is the growth of world income, dx is the duty on exports, lib is the
liberalisation dummy and t is time. Depending on the estimation technique used,
the central estimate is that trade liberalisation has raised export growth by approxi-
mately two percentage points, or by one-quarter compared to the pre-liberalisation
export growth rate. The estimated coefficient on the export duty variable is negat-
ive, but small (roughly –0.2). The coefficient on the liberalisation dummy variable
is consistently in the range one to two taking the full sample of 22 countries, but
the quantitative effect (shown in parentheses) differs between the four regions:
Africa (3.58), South Asia (2.54), East Asia (2.42) and Latin America (1.66).
16 P. Pacheco-López and A.P. Thirlwall
For a country’s overall economic performance to improve, however, it is not
enough for export growth to accelerate. Export growth must be shown to outpace
import growth, otherwise balance of payments difficulties will arise. In the liter-
ature on trade liberalisation, very little attention has been paid to import growth,
or to the balance between export growth and import growth. This is a serious
weakness of trade liberalisation studies, but is a reflection of the fact that in
orthodox trade and growth theory the balance of payments is either assumed to
look after itself, or deficits as regarded as a form of consumption smoothing and
have no long run effect on real variables. Country studies by Melo and Vogt
(1984) for Venezuela, Mah (1999) for Thailand, and Bertola and Faini (1991)
for Morocco all show a significant impact of trade liberalisation on import
growth and on the sensitivity of imports to domestic income growth, but the
most detailed study is that by Santos-Paulino and Thirlwall (2004) who take the
same 22 countries as for export growth and test three hypotheses: (1) trade liber-
alisation, measured by a shift dummy variable (lib), significantly increases
import growth; (2) reductions in tariffs (dm) raise import growth and (3) a more
liberal trade regime increases the income and price elasticities of demand of
imports (measured by interacting the liberalisation dummy with the growth of
income and real exchange rate variables, liby and librer, respectively). The
import growth equation specified to capture these effects is:

mt = bo + b1(rert) + b2(yt) + b3(dmt) + b4(libt) + b5(libyt) + b6(librert) (1.2)

The results may be summarised as follows. Trade liberalisation itself, control-

ling for all other factors, has increased the growth of imports by between five
and six percentage points, which represents a near doubling of the pre-
liberalisation import growth. The independent effect of tariff cuts has been to
raise the growth of imports by between 0.2 and 0.5 percentage points for a one
percentage point cut in tariff rates. Liberalisation has increased the elasticity of
imports to both domestic income and exchange rate changes by between 0.2 and
0.5 percentage points.
We have examined ourselves (Pacheco-López and Thirlwall, 2006) the direct
effect of trade liberalisation on the income elasticity of demand for imports in 17
Latin American countries over the period 1977 to 2002 using a simplified
version of equation (1.2):

mt = co + c1(rert) + π1(yt) + π2(libyt) (1.3)

where p1 is the income elasticity of demand for imports in the pre-liberalisation

period and (p1 + p2) is the income elasticity in the post-liberalisation period. We
find an increase of 0.55 from 2.08 to 2.63, which more or less offsets the increase
in export growth post-liberalisation, leaving the GDP growth rate of countries con-
sistent with balance of payments equilibrium virtually unchanged. This increase in
the income elasticity of demand for imports in Latin America as a result of trade
liberalisation is confirmed using the technique of rolling regressions taking 13
Trade liberalisation in poor countries 17
overlapping periods starting from 1977–90 and ending in 1989–2002. The esti-
mated income-elasticity starts at 2.04 and ends at 2.82, giving an annual trend rate
of increase of approximately 0.04 percentage points.
If trade liberalisation raises the growth of imports by more than exports, or
raises the income elasticity of demand for imports by more than in proportion to
the growth of exports, the balance of trade (or payments) will worsen at a given
growth of output, unless the currency can be manipulated to raise the value of
exports relative to imports. Surprisingly, very little research has been done on
the balance of payments effect of trade liberalisation. The first major studies
were by Parikh for UNCTAD (1999) and for WIDER (Parikh, 2002). The first
study examined 16 countries over the period 1970–95, with the main result that
trade liberalisation seems to have worsened the trade balance by 2.7 per cent of
GDP (which is substantial). The second study extends the analysis to 64 coun-
tries, with the general conclusion:

The exports of most of the liberalising countries have not grown fast enough
after trade liberalisation to compensate for the rapid growth of imports
during the years immediately following trade liberalisation. The evidence
suggests that trade liberalisation in developing countries has tended to lead
to a deterioration in the trade account.

Santos-Paulino and Thirlwall (2004) take the same sample of 22 developing

countries as for the impact of trade liberalisation on export and import growth
previously discussed, and estimate the following equation:


where TB/GDP is the trade balance to GDP ratio, BP/GDP is the balance of pay-
ments to GDP ratio, tt is the terms of trade and the other variables are as defined
in equations (1.1) and (1.2). The equations are estimated using panel data tech-
niques over the period 1976–98. The most important result is that the switch to a
more liberal trading regime has worsened, on average, the trade balance by 2 per
cent of GDP (which is similar to the Parikh estimate), and the balance of pay-
ments by 1 per cent of GDP. For a group of 17 Least Developed Countries over
the period 1970–2001, Santos-Paulino (2007) finds a deterioration in the trade
balance ratio of 4 per cent of GDP.
In our own research (Pacheco-López and Thirlwall, 2007) on 17 Latin Amer-
ican countries over the period 1977–2002 we find a deterioration in the trade
balance of between 1.3 and 2.3 per cent of GDP depending on the method of
estimation used (whether a panel, or time/series/cross-section, estimator, using
as control variables the first three variables in equation 1.4). All these results
show that trade liberalisation has impacted unfavourably on the trade balance
18 P. Pacheco-López and A.P. Thirlwall
and current account balance of liberalising countries. Such a deterioration, if it
cannot be financed by sustainable capital inflows, may either trigger a currency
crisis or necessitate a severe deflation of domestic demand (and therefore
growth) to control imports. As UNCTAD (2004) says in its Least Developed
Countries Report 2004: ‘this critical [balance of payments] constraint on devel-
opment and sustained poverty reduction is conspicuously absent in the current
debate on trade and poverty’; and also, it may be added, in the debate on the
wisdom of excessive trade liberalisation in poor vulnerable countries.
Indeed, the ultimate test of successful trade liberalisation, at least at the
macro-level without regard to distributional effects, is whether it lifts a country
on to a higher growth path consistent with a sustainable balance of payments; or,
in other words, whether it improves the trade-off between growth and the
balance of payments, as illustrated in Figure 1.1.
On the vertical axis of Figure 1.1 is measured the ratio of the balance of pay-
ments to GDP, and on the horizontal axis, the growth of GDP. The solid-line
curve gives the negative trade-off curve showing how the balance of payments
deteriorates as growth accelerates. The curve is drawn to represent a serious situ-
ation where the balance of payments is in deficit (point a) even at zero growth.
The objective of trade policy should be to shift the curve upwards to, say, point
b on the horizontal axis so that some positive GDP growth is possible without
running into balance of payments difficulties.


0 b
� � GDP growth (y)

Figure 1.1 The trade-off between growth and the balance of payments (source: author’s
Trade liberalisation in poor countries 19
We have estimated this trade-off curve (using the trade balance/GDP ratio as
the dependent variable) for 17 Latin American countries over the period 1977 to
2002 using pooled data (giving 425 observations) to see whether trade liberalisa-
tion has resulted in a positive shift (Pacheco-López and Thirlwall, 2007). Fitting
a linear (for simplicity) regression line, without controlling for liberalisation,
gives the result (t-statistics in parentheses):

TB/GDP = – (1.5)

The curve cuts the vertical axis in the negative quadrant, which is serious. The
average GDP growth for the sample as a whole is 2.76 per cent per annum, with
an average trade deficit of –4.69 per cent of GDP. When a liberalisation dummy
is included in the equation, the result shows a negative, not positive, effect, i.e.

TB/GDP = – – (1.6)

The pre-liberalisation deficit at zero growth is –1.387, and the post-liberalisation

deficit is (–1.387) + (–3.610) = –4.997. Liberalisation has worsened the trade-off
by 3.6 percentage points. Controlling for changes in the real exchange rate and
world income growth reduces the unfavourable impact to –2.0 percentage points,
but this is still significant.

1.6 Trade liberalisation and growth performance

While it is true that trade liberalisation has improved export performance, liber-
alisation and export growth are not the same, and should not be confused. As
Stiglitz (2006) notes:

Advocates of liberalisation cite statistical studies claiming that liberalisation

enhances growth. But a careful look at the evidence shows something quite
different. . . . It is exports – not the removal of trade barriers – that is the
driving force of growth. Studies that focus directly on the removal of trade
barriers show little relationship between liberalisation and growth. The
advocates of quick liberalisation tried an intellectual sleight of hand, hoping
that the broad-brush discussion of the benefits of globalisation would suffice
to make their case.

Our study of Latin America discussed above is the only one we know that
examines the impact of liberalisation on the trade-off between growth and the
balance of payments, but there are several time series and cross-section studies
of the relation between liberalisation and GDP growth on the one hand and trade
openness and GDP growth on the other (although trade openness is not the same
as liberalisation). The studies give mixed results, but it can definitely be said that
20 P. Pacheco-López and A.P. Thirlwall
the extravagant claims of the pro-trade liberalisers look hollow when compared
with the evidence. Early work by Edwards (1992, 1998) and Dollar (1992)
showing a positive relation between openness (or the outward orientation of
countries) and growth performance was heavily criticised by Rodriguez and
Rodrik (2000) on methodological grounds and for lack of robustness. Similar
work by Dollar and Kraay (2004) on ‘globalisation’ and economic growth was
likewise criticised by Dowrick and Golley (2004) who show that the faster
growth of Dollar and Kraay’s sample of ‘globalising’ countries is entirely due to
the fast growth of China and India. Even more telling, the so-called ‘globalising’
countries identified were not the most open or liberalised. Another major study
of trade orientation and growth is that by Sachs and Warner (1995) who found
that more open economies over the period 1979–89 grew 2.44 percentage points
faster than economies identified as closed. Wacziard and Welch (2008) extend
the Sachs–Warner study into the 1990s with more countries, and find that their
result is not robust; there appears to be no significant effect of openness on
growth performance. Greenaway, Morgan and Wright (1998, 2002) examine the
relationship between trade liberalisation and GDP growth using an impact
dummy variable for the year of liberalisation in a sample of up to 73 countries
over the period 1975–93, and find a J-curve effect with growth first deteriorating
and then improving. There is no indication, however, of how long the lagged-
growth effect lasts. Rodriguez and Rodrik (2000) conclude their evaluation of
studies of trade orientation and economic growth by saying that indicators of
openness used are either poor measures of trade barriers or are highly correlated
with other determinants of domestic performance. All studies should therefore
be treated with great caution. They are particularly concerned that the priority
given to trade policy reform has generated expectations that are unlikely to be
met, and may preclude other, institutional reforms which would have a greater
impact on economic performance. Trade liberalisation, in other words, cannot be
regarded as a panacea, or as a substitute for a comprehensive trade and develop-
ment strategy. To quote Rodrik (2001): ‘Deep trade liberalisation cannot be
relied upon to deliver high rates of economic growth and therefore does not
deserve the high priority it typically receives in the development strategies
pushed by leading organisations.’

1.7 Trade strategy for development

So what trade strategy should poor countries pursue? The overriding objective
must be to acquire dynamic comparative advantage. For this, the private sector of
an economy needs the support of the government in the form of incentives and
various types of ‘protection’ to mitigate investment risks. It is one thing to argue
against anti-export bias; it is another to argue that poor countries should abandon
all forms of protection of domestic industry. Improved market access across
developed countries for poor country exports merely perpetuates static compara-
tive advantage. As Rodrik (2001) argued in the lead-up to the recent (failed) Doha
round of trade negotiations ‘the exchange of reduced policy autonomy in the
Trade liberalisation in poor countries 21
South for improved market access in the North is a bad bargain where develop-
ment is concerned’. Poor countries need time and policy space to nurture new
(infant) industrial activities as developed countries did historically, and as many
newly industrialising economies still do today. As Hausmann and Rodrik (2003)
say in their important work on the concept of ‘self discovery’: ‘The fact that the
world’s most successful economies during the last few decades prospered doing
things that are most commonly associated with failure (e.g. protection) is some-
thing that cannot easily be dismissed.’
Hausmann and Rodrik’s argument is that there is much randomness in the
process of a country discovering what it is best at producing, and a lack of pro-
tection reduces the incentive to invest in discovering which goods and services
they are. Poor, labour abundant economies have thousands of things they could
produce and trade, but in practice their exports are highly concentrated. Some-
times, over 50 per cent of exports are accounted for by fewer than ten products.
Bangladesh and Pakistan are countries at similar levels of development, but
Bangladesh specialises in hats and Pakistan in bed sheets. This specialisation is
not the result of resource endowment; it is the result of chance choice by enter-
prising entrepreneurs who ‘discovered’ (ex post) where relative costs were low.
Other ‘chance’ investments include cut flowers in Colombia for export to North
America, camel cheese in Mauritania for export to the European Union, high-
yield maize in Malawi and squash in Tonga. The policy implications of the
Haussmann and Rodrik observation and model are that governments need to
encourage entrepreneurship and invest in new activities ex ante, but push out
unproductive firms and sectors ex post. Intervention needs to discriminate as far
as possible between innovators and imitators. Normal forms of trade protection
turn out not to be the ideal policy instruments because they do not discriminate,
and earn profits only for those selling in the domestic market. Export subsidies
avoid anti-export bias, but still do not discriminate between the innovators and
the copycats, and in any case are illegal under the rules of the World Trade
Organization (WTO). The first-best policy is public sector credit or guarantees
which can discriminate in favour of the innovator, and be used as a ‘stick’ if
firms do not perform well.
There is much that the international community can also do to promote trade
for development, as opposed to pursuing trade liberalisation for its own sake. The
whole world trade system works against the majority of poor developing coun-
tries, first because of their dependence on primary commodities (the ‘curse’ of
natural resources) and low value-added manufactures, second because the ‘rules
of the game’ governing trade between rich and poor countries are rigged and
biased in favour of the rich, and third because the agenda for trade reform is
largely set by the rich developed countries. The only permanent solution to
primary-commodity dependence is structural change which requires the establish-
ment of new, non-traditional industries; but this is what the rich developed nations
are hostile to. They want free access to poor countries’ markets, while continuing
to protect their own. The most recent example of this is the ongoing debate
between the European Union (EU) and the African, Caribbean and Pacific (ACP)
22 P. Pacheco-López and A.P. Thirlwall
countries over Economic Partnership Agreements (EPAs) to replace the trade
preferences that the ACP countries used to enjoy under the Lomé Convention.
The EU is insisting that poor developing countries reduce restrictions on imports
of manufactured goods and service activities in return for continued access to the
EU market for their agricultural products. The EU is refusing to look at altern-
atives to free trade EPAs, but by its own admission it concedes that EPAs could
lead to the collapse of the manufacturing sector in many poor countries. As
Stiglitz (2006) remarks in his powerful book Making Globalisation Work, ‘the
US and Europe have perfected the art of arguing for free trade while simultan-
eously working for trade agreements that protect themselves against imports from
developing countries’. If developed countries really wanted to help poor develop-
ing countries they could reduce and eliminate tariffs and barriers against all their
goods. Oxfam (2002) estimates that trade barriers against developing country’s
goods cost about $100 billion a year; or twice the level of official development
assistance. In addition, developing countries might be allowed ‘infant country
protection’, which would be equivalent to a currency devaluation, but have the
advantage of raising revenue for spending on public goods. One of the severe
drawbacks of tariff reductions in poor countries is a loss of tax revenue.
If trade is to promote development, the WTO, that now governs world trade,
needs radical reform and rethinking. The Agreement establishing the WTO
(1995) lists as one of its purposes:

Raising standards of living, ensuring full employment and a large and steady
growing volume of real income and effective demand, and expanding the
production of, and trade in, goods and services, while allowing for the
optimal use of the world’s resources in accordance with the objective of sus-
tainable development, seeking both to protect and preserve the environment
and to enhance the means of doing so in a manner consistent with their
respective needs and concerns at different levels of development

The aim is laudable, but unfortunately there is a divorce between rhetoric and
reality because the WTO treats trade liberalisation and economic development as
synonymous, and yet as we have seen the historical and contemporary evidence is
that domestic economic policy, institution building and the promotion of investment
opportunities are far more important than trade liberalisation and trade openness in
determining economic success in the early stages of economic development. Rodrik
(2001) reminds us (like Chang, 2002, 2005 and Reinert, 2007) that:

No country has [ever] developed simply by opening itself up to foreign trade

and investment. The trick had been to combine the opportunities offered by
world markets with a domestic investment and institution-building strategy
to stimulate the animal spirits of domestic entrepreneurs.

But now, under WTO rules, all the things that, for example, South Korea,
Taiwan and other East Asian countries did to promote economic development in
Trade liberalisation in poor countries 23
the 1960s, 1970s and 1980s are severely restricted. Some countries that break
the rules are succeeding spectacularly. China is one obvious example, but
another would be Vietnam which, while promoting FDI and exports, also pro-
tects its domestic market, maintains import monopolies and engages in state
trading. The WTO should shift away from trying to maximise the flow of trade,
to understanding and evaluating what trade regime will maximise the possibility
of development for individual poor countries. A new world trade order is
required which acts on behalf of poor countries; and poor developing countries
need a louder voice in any reformed structure.

Arbache, J., Dickerson, A. and Green, F. (2004), Trade Liberalisation and Wages in
Developing Countries, Economic Journal, February.
Barro, R. (2000), Inequality and Growth in a Panel of Countries, Journal of Economic
Growth, March.
Bertola, G. and Faini, R. (1991), Import Demand and Non-Tariff Barriers: The Impact of
Trade Liberalisation, Journal of Development Economics, November.
Bhagwati, J. (2001), Free Trade Today (Princeton, NJ: Princeton University Press).
Bourguignon, F. and Morrisson, C. (2002), Inequalities Among World Citizens:
1820–1992, American Economic Review, September.
Chang, Ha-Joon (2002), Kicking Away the Ladder: Development Strategy in Historical
Perspective (London: Anthem Press).
Chang, Ha-Joon (2005), Why Developing Countries Need Tariffs (Geneva: South Centre).
Chen, S. and Ravallion, M. (2004), How Have the World’s Poorest Fared Since the Early
1980s? World Bank Research Observer, Fall.
Dollar, D. (1992), Outward Oriented Developing Countries Really do Grow More
Rapidly: Evidence from 95 LDCs 1976–85, Economic Development and Cultural
Change, April.
Dollar, D. and Kraay, A. (2002), Growth Is Good for the Poor, Journal of Economic
Growth, September.
Dollar, D. and Kraay, A. (2004), Trade, Growth and Poverty, Economic Journal, Febru-
Dowrick, S. (1997), ‘Trade and Growth: A Survey’ in J. Fagerberg, P. Hansson, L. Lund-
berg and A. Melchior (eds), Trade, Technology, and Changes in Employment of Skilled
Labour in Swedish Manufacturing (Cheltenham: Edward Elgar).
Dowrick, S. and Golley, J. (2004), Trade Openness and Growth: Who Benefits? Oxford
Review of Economic Policy, Spring.
Edwards, S. (1992), Trade Orientation, Distortions and Growth in Developing Countries,
Journal of Development Economics, July.
Edwards, S. (1998), Openness, Productivity and Growth: What Do We Really Know?
Economic Journal, March.
Goldberg, P. and Pavcnik, N. (2007), Distributional Effects of Globalisation in Develop-
ing Countries, Journal of Economic Literature, March.
Greenaway, D., Morgan, W. and Wright, P. (1998), Trade Reform, Adjustment and
Growth: What Does the Evidence Tell Us? Economic Journal, September.
Greenaway, D., Morgan, W. and Wright, P. (2002), Trade Liberalisation and Growth in
Developing Countries, Journal of Development Economics, February.
24 P. Pacheco-López and A.P. Thirlwall
Grossman, G. and Helpman, E. (1991a), Innovation and Growth in the Global Economy
(Cambridge, MA: MIT Press).
Grossman, G. and Helpman, E. (1991b), Trade, Knowledge Spillovers and Growth, Euro-
pean Economic Review, April.
Hausmann, R. and Rodrik, D. (2003), Economic Development as Self Discovery, Journal
of Development Economics, December.
Heckscher, E. (1919), The Effect of Foreign Trade on the Distribution of Income, Ekono-
misk Tidskrift, Vol. 21.
Hirschman, A. (1958), Strategy of Economic Development (New Haven, CT: Yale Uni-
versity Press).
Johnson, H. (1964), Tariffs and Economic Development: Some Theoretical Issues,
Journal of Development Studies, October.
Kaldor, N. (1970), The Case for Regional Policies, Scottish Journal of Political Economy,
Keynes, J.M. (1930), ‘Evidence to the Macmillan Committee on Finance and Industry’ in
D. Moggridge (1981), The Collected Writings of J.M. Keynes Vol. 20: Activities
1929–1931: Rethinking Employment and Unemployment Policies (London: Macmillan).
Krugman, P. (1984), ‘Import Protection as Export Promotion: International Competition
in the Presence of Oligopoly and Economies of Scale’ in H. Kierzkowski (ed.), Monop-
olistic Competition in International Trade (Oxford: Clarendon Press).
Krugman, P. (1986), Strategic Trade Policy and the New International Economics (Cam-
bridge, MA: MIT Press).
List, F. (1885), The National System of Political Economy, translated from the original
German edition published in 1841 by Sampson Lloyd (London: Longmann, Green and
Mah, J.S. (1999), Import Demand, Liberalisation and Economic Development, Journal of
Policy Modelling, July.
Melo, O. and Vogt, M.G. (1984), Determinants of the Demand for Imports of Venezuela,
Journal of Development Economics, April.
Milanovic, B. (2005a), Can we Discern the Effect of Globalisation on Income Distribu-
tion? World Bank Economic Review, January.
Milanovic, B. (2005b), Worlds Apart: Measuring International and Global Inequality
(Princeton, NJ: Princeton University Press).
Mill, J.S. (1848), Principles of Political Economy (London: Longmann, Green and Co.).
Myrdal, G. (1957), Economic Theory and Underdeveloped Regions (London: Duck-
Ohlin, B. (1933), Interregional and International Trade (Cambridge, MA: Harvard Uni-
versity Press).
Oxfam (2002), Rigged Rules and Double Standards: Trade, Globalisation and the Fight
Against Poverty (Oxford: Oxfam).
Pacheco-López, P. and Thirlwall, A.P. (2006), Trade Liberalisation, the Income Elasticity
of Demand for Imports and Growth in Latin America, Journal of Post Keynesian Eco-
nomics, Fall.
Pacheco-López, P. and Thirlwall, A.P. (2007), Trade Liberalisation and the Trade-Off
Between Growth and the Balance of Payments in Latin America, International Review
of Applied Economics, September.
Parikh, A. (2002), Impact of Liberalization, Economic Growth and Trade Policies on
Current Accounts of Developing Countries: An Econometric Study, WDP 2002/63
(Helsinki: WIDER).
Trade liberalisation in poor countries 25
Ravallion, M. (2006), Looking Beyond Averages in the Trade and Poverty Debate, World
Development, August.
Reinert, E. (2007), How Rich Countries Got Rich and Why Poor Countries Stay Poor
(London: Constable and Robinson).
Ricardo, D. (1817), On the Principles of Political Economy and Taxation (P. Sraffa, ed.)
(Cambridge: Cambridge University Press, 1951).
Rodriguez, F. and Rodrik, D. (2000), ‘Trade Policy and Economic Growth: A Skeptic’s
Guide to the Cross-National Evidence’ in B. Bernanke and K. Rogoff (eds), Macro-
economics Annual 2000 (Cambridge, MA: MIT Press).
Rodrik, D. (1988), ‘Imperfect Competition, Scale Economies and Trade Policy in Devel-
oping Countries’ in R. Baldwin (ed.), Trade Policy Issues and Empirical Analysis
(Chicago, IL: Chicago University Press for the NBER).
Rodrik, D. (2001), The Global Governance of Trade: As if Development really Mattered
(New York: UNDP).
Sachs, J. and Warner, A. (1995), Economic Reform and the Process of Global Integra-
tion, Brookings Papers on Economic Activity No. 1.
Sala-i-Martin, X. (2002), The Disturbing ‘Rise’ of Global Income Inequality, NBER
Working Paper, 8904.
Samuelson, P. (1962), Economists and the History of Ideas, American Economic Review,
Santos-Paulino, A. (2007), Aid and Trade Sustainability under Liberalisation in Least
Developed Countries, World Economy, June.
Santos-Paulino, A. and Thirlwall, A.P. (2004), The Impact of Trade Liberalisation on
Exports, Imports, and the Balance of Payments of Developing Countries, Economic
Journal, February.
Smith, A. (1776), An Inquiry into the Nature and Causes of the Wealth of Nations
(London: Straham and Caddell).
Solow, R. (1956), A Contribution to the Theory of Economic Growth, Quarterly Journal
of Economics, February.
Spilimbergo, A., Londoño, J.L. and Székely, M. (1999), Income Distribution, Factor
Endowments and Trade Openness, Journal of Development Economics, June.
Stiglitz, J. (2006), Making Globalisation Work (New York: W.W. Norton and Co.).
Stolper, W. and Samuelson, P. (1941), Protection and Real Wages, Review of Economic
Studies, November.
Thirlwall, A.P. and Pacheco-López, P. (2008), Trade Liberalisation and the Poverty of
Nations (Cheltenham: Edward Elgar).
UNCTAD (1999), Trade and Development Report (Geneva: UNCTAD).
UNCTAD (2004), Least Developed Countries Report 2004: Linking International Trade
with Poverty Reduction (Geneva: UNCTAD).
Wacziard, R. and Welch, K. (2008), Trade Liberalisation and Growth: New Evidence,
World Bank Economic Review, Vol. 22.
Winters, A., McCullock, N. and McKay, A. (2004), Trade Liberalisation and Poverty:
The Evidence so Far, Journal of Economic Literature, March.
World Trade Organization (WTO) (1995), ‘Agreement Establishing the World Trade
Organisation’ (Geneva: WTO Information and Media Relations Divisions).
2 Beyond the Washington
The quest for an alternative
development paradigm for Latin
Ignacio Perrotini-Hernández,
Juan Alberto Vázquez-Muñoz and
Blanca L. Avendaño-Vargas

To find a growth performance in Latin America that is even close to the failure of
the last 25 years, one has to go back more than a century, and choose a 25-year
period that includes both World War I and the Great Depression.
(Mark Weisbrot, 2006, p. 3)

2.1 Introduction
The trade–development connection has been emphasised in the literature since
the inception of economics as a discipline. Adam Smith (1776 [1976: Bk.III,
ch.1, p. 405]), for example, regards the evolution from agriculture to manufac-
turing and ‘last of all to foreign commerce’ as ‘the natural course’ of develop-
ment. This pattern of development goes on through an increasingly intricate set
of transactions between advanced (‘town’) and non-advanced (‘country’) regions
and between nations. In the latter case, foreign trade plays the role of a vent for
surplus home production, thereby promoting further division of labour and
greater accumulation of capital. Smith’s trade–development nexus highlights the
principle of absolute advantage in production. Technological progress (enhanced
division of labour) changes the pattern of absolute advantage, thereby enhancing
the wealth of nations.1 The ‘extent of the market’ limits technological progress,
hence the relevance of foreign trade for bringing increasing returns to fruition.
The economy, Smith argued, possessed unlimited upward potential.
David Ricardo (1821), another classical political economist, held a more
sceptical view. He maintained that, since land rent grows as population increases,
in the long run the economy follows a path towards a standstill. Ricardo’s model
of international trade is composed of two countries and two commodities; it laid
out the theory of comparative advantage, which argued that all countries could
benefit from free trade, even if a country lacked absolute advantage at producing
all kinds of goods. Thus, even if, say, country A is more proficient in producing
both goods relative to country B, international trade can profitably continue on
Beyond the Washington Consensus 27
the basis of comparative advantage. According to Ricardo’s doctrine, countries
reap gains from specialising in what they are best at producing and trading with
each other; hence foreign trade is beneficial at any rate, albeit in the long run
increasing differential land rent will ultimately bring about economic stagnation.
Building on Ricardo’s theory of comparative advantage, Eli Heckscher and
Bertil Ohlin elaborated a theorem maintaining that relative endowments of
inputs determines comparative advantage and trade specialisation; goods and
factor flows tend to be substitutes (cf. Ohlin, 1933). The theorem essentially says
that countries will export goods that utilise their abundant and cheap factor of
production and import products that utilise the countries’ scarce factor. The
policy lesson that follows from the above is fairly straightforward: in order to
attain higher stages of economic development, under the assumption of interna-
tional immobility of labour and capital, a nation must specialise in the produc-
tion of those goods that use intensively the input that is relatively more abundant
in the domestic market. The Heckscher–Ohlin theorem became a basic constitu-
ent of the modern neoclassical theory of international trade, which states that all
countries benefit equally from (free) foreign trade, regardless of the nature and
quality of the goods being produced and exchanged internationally.
After a long process of import-substitution industrialisation – running roughly
between 1940 and 1980 – Latin America (and to a lesser extent Sub-Saharan
Africa) adopted a model of economic liberalisation in the late 1980s. The neo-
classical theory of international trade, in spite of its logical inconsistency (for
example, capital as primary factor has no method of measuring it before the
determination of profit rate), became the building block of that ‘new’ trade
policy reform, encapsulated by the so-called Washington Consensus (henceforth
WC). The WC was said to represent a framework of ‘good policies’ of trade and
financial liberalisation (Williamson, 1990). While the Heckscher–Ohlin model
gave the justification for trade liberalisation in the region in the aftermath of the
1982 foreign debt crisis, McKinnon’s theory (1973 and 1991) provided the strat-
egy for ‘the order’ of financial liberalisation. The new orthodox approach to eco-
nomic affairs was adopted under the presumption that it would restore growth
and development on a sustainable basis, thus bringing about macroeconomic
stability, convergence to optimum growth and narrowing the outstanding devel-
opment gap of the Latin American economies vis-à-vis the leading economy in
the world, the US economy.
The essential questions to be tackled in this chapter are to enquire: whether
the WC strategy – original or ‘augmented’ – contributed to bridging the develop-
ment gap; in what sense can it be argued that the decision – undertaken in the
1980s – of giving markets free rein improved the performance of the economies
under scrutiny if at all; does the development process call for ancillary institu-
tions? If the answer to the latter question is in the positive, as experience appears
to suggest (Cimoli et al., 2009a; Chang, 2003 and 2008; Ffrench-Davis, 2005;
Rodrik, 2004), one may ask whether industrial policy can play a sensible role.
We also expect to contribute to the current debate on an alternative develop-
mental paradigm for the Latin American emerging economies, an urgent need
28 I. Perrotini-Hernández et al.
dramatically prompted by the ongoing economic crisis. Our analysis is focused
on Argentina, Brazil, Chile and Mexico, the core economies of Latin America,
which altogether account for more than two-thirds of the region’s GDP. China,
today’s fastest growing economy, is also included for the sake of comparison;
the US economy appears as a benchmark with a view of measuring the evolution
of the development gap of the involved countries before and after the inception
of the WC model.
The chapter is organised in five sections in addition to this introduction. The
first section revisits the WC framework and its policy implications for economic
development. The second section highlights the performance of some key macr-
oeconomic variables before and after the reform. The next section discusses the
evolution of the development gap and the main driving forces behind its trend,
the fourth makes the case for industrial policy from a developmental point of
view and the last section concludes.

2.2 The Washington Consensus framework revisited

The widespread foreign debt crisis of 1982 was a major catalyst for change in
Latin America; the depth of the protracted crisis (and the stagflation that ensued)
was interpreted as the need for a new development strategy.
The Washington Consensus agenda was said to represent such an alternative
approach and, therefore, was adopted in the late 1980s under the presumption
that it would reignite fast and sustainable growth. The basic thrust of the original
WC recipe provided a fairly simple catalogue of policy changes that ‘were
needed more or less everywhere in Latin America’ (Williamson, 2004a:1),
namely fiscal discipline and reorientation of government expenditure, tax reform,
privatisation of state owned assets, deregulation, protected property rights, trade
liberalisation, financial liberalisation (indeed, liberalising interest rates), open-
ness to inward foreign direct investment and unified and competitive exchange
rates (Williamson, 1990). Getting a minimalist state and prices right was the
mantra of the new market-friendly approach to renewed economic development.
Even if it went unnoticed at the outset, the advent of the WC implied the crisis
of the developmental state that had been responsible, to a great extent, for Latin
America’s industrialisation in 1940–1980 (Câmara and Vernengo, 2004).
Indulgent fiscal policies leading to excessively large fiscal deficits were con-
sidered as the root of high inflation, balance of payments disequilibria and
exchange rate instability. Moreover, the failure of import-substituting industriali-
sation across the region was interpreted as associated to the deleterious effect of
government intervention through growing fiscal deficits. Many Latin American
governments, seeking macroeconomic stability and sustainable growth in the
late 1980s, coalesced at their own volition – combined with external pressure –
on conducting market friendly reforms, with the core countries Argentina,
Brazil, Chile and Mexico successively excelling as a poster child for the WC
strategy. Ocampo (2004: 67) gives an account of the economic reform, worth
quoting in full:
Beyond the Washington Consensus 29
Structural economic reforms varied in intensity across sectors and countries.
All countries in Latin America significantly liberalized international trade,
external capital flows and the domestic financial sector. Policy decisions in
these areas included reducing tariffs and their dispersion; dismantling non-
tariff barriers; eliminating most restrictions on foreign direct investment;
phasing out many or most foreign exchange regulations; granting greater or
total autonomy to central banks; dismantling regulations regarding interest
rates and credit allocation; reducing reserve requirements on domestic
deposits; and privatizing several state banks.
In the fiscal area, reforms strengthened the value added tax, reduced
income tax rates and strengthened tax administration, though with only a
limited effect on tax evasion. Social security systems were overhauled in
several countries to allow for the participation of private agents and a more
clear balance between benefits and (employers’ and workers’) contributions.

The original agenda put forth by the WC strategy failed to live up to its goals;
it did not produce macroeconomic stability with sustainable growth as had been
predicated by its proponents, although Williamson (2004a:8) warned, after 15
years of experience of free-market policies, ‘I have to admit that I too am
uncomfortable if it [the WC framework] is interpreted as a comprehensive
agenda for economic reform.’ Indeed, while per capita GDP had increased by
2.7 per cent per year during 1950–1980, the ‘lost decade’ (the 1980s) saw a
decline of 0.9 per cent annually and the WC era produced a dull recovery of less
than 1 per cent per year from 1990 to 2003 (ECLAC, 2002 and 2003). What
went wrong? Williamson (2003:2) maintains that ‘the liberalizing reforms of the
past decade and a half, or globalisation, can [not] be held responsible for the
region’s renewed travails in recent years’. The reasons accounting for the failure
are threefold, according to Williamson and Kuczynski (2003). First, while pur-
suing macroeconomic stabilisation and microeconomic reforms practitioners of
the WC incurred misguided macroeconomic policies, such as procyclical fiscal
policies and exchange rate overvaluation, which left them exposed to capital
reversals. Second, the economic reform was incomplete; and third, the WC’s
concern for accelerating growth, alas, did not include equity.
Williamson and Kuczynski (2003) addressed these topics and proceeded to
propose a set of four second-generation reforms with the aim of restarting and
ensuring ‘future growth’. First, with regards the business cycle, countries must
build up a countercyclical policy intended for isolating their economies from
adverse exogenous shocks and reducing exposure to international financial
markets volatility. In this view, it is perfectly appropriate for the government to
endorse the operation of the automatic stabilisers and reject deflationary fiscal
policies (like the one the current Mexican government has followed, we may add)
as a policy reaction to the present financial crisis. The main elements of the pro-
posed crisis-proofing or stabilisation policy are: sufficient ex ante accumulation
of budget surpluses, of foreign exchange reserves and stabilisation funds in times
of cyclical booms; adoption of a flexible exchange rate regime and a monetary
30 I. Perrotini-Hernández et al.
policy framework of (low) inflation targeting as a cushion against perverse effects
resultant from excessive capital inflows; forestalling liability dollarisation (the so-
called original sin) and currency mismatch; reinforcement of prudential regula-
tion of the banking sector; and, last but not least, improvement of domestic
savings rates. Interestingly, the volume edited by Williamson and Kuczynski
(2003:8, 77, 81, passim) endorses Maastricht-type debt and fiscal deficit rules in
order to enforce fiscal discipline in Latin America.
Second, liberalising reforms should be extended in order to make the labour
market more flexible. Excessive rigidity of the labour market ‘constitutes a
major obstacle to an expansion of employment in the formal economy’ (ibid.:
10). Inflexibility of the formal labour market, it is argued, impairs acceleration
of growth and prevents a reduction of the informal sector, which employs ‘[a]
round 50 percent of the labor force in many Latin American countries’ (William-
son, 2004a:11). The net effect of labour market flexibilisation is higher employ-
ment rates because it reduces inequality of opportunity in the labour market
(ibid.; Saavedra, 2003:237–41, 252–53).
Third, the need of building institutions. The original WC was oblivious of the
crucial role of institutions in the making of economic development, which is
thoroughly acknowledged in the augmented version. The type and quality of
institutions suitable for the enhancement of output growth and political stability,
for the internalisation of externalities, for efficiently supplying public goods and
correcting polarising income distribution, varies from one country to another. In
this respect, the government is allowed to play both the ‘old-fashioned’ role of
building a good productive infrastructure and the modern role of ‘building a
national innovation system’ and promoting R&D (Williamson and Kuczynski,
2003:12). The government can also advocate institutional reforms aimed at
enhancing property rights and prudential regulation of the financial sector. While
these reforms are not an easy task, if properly designed they tend to reduce trans-
actions costs and the risk premium.
Finally, governments should also have income distribution targets through
market-incentive mechanisms, i.e. through creation of new assets by providing
poor people with more human capital (education), secure property rights through
land reforms (in rural areas), microcredit lines (in rural and urban areas alike)
and similar instruments that encourage civil society. The crucial tenet here is that
these assets, when combined within a sensible programme, may represent a pro-
poor growth strategy that enables outcasts to work their way out of poverty (Wil-
liamson and Kuczynski, 2003:14–18; Williamson, 2004a:12).

2.3 Macroeconomic performance before and after the

structural reform
The WC policy reforms prompted a radical change in the development strategy
based on perennial fiscal austerity2 and the laissez-faire paradigm (Davidson,
2003). As pointed out, the WC famously promised that the replacement of the
import-substituting closed economy by a neoliberal open economy model in
Beyond the Washington Consensus 31
Latin America would lead to renewed and accelerating economic growth on a
more sustainable basis.

2.3.1 Economic growth

The observed pattern of output growth varies from one country to another
between 1960 and 2009. Yet, some common trends can be highlighted. Argen-
tina, Brazil and Mexico experienced episodes of growth accelerations from 1960
to 1980, albeit growth became highly volatile in Argentina towards the late
1970s, perhaps owing to domestic political instability and adverse oil shocks.
Chile’s economic growth rate shows a downward trend from 1961 to 1975 –
with some good times in between – and did not return to growth rate levels
observed in the early 1960s until 1985. The Chilean economy appears to be an
outlier in many ways: import-substituting industrialisation was not as relevant
vis-à-vis other Latin American core economies; it adopted the laissez-faire
approach to economic development in the mid-1970s, i.e. during the first years
of Pinochet’s dictatorship. Chile’s per capita GDP increased over 3 per cent
annually in 1980–2002, while Argentina, Brazil and Mexico experienced negli-
gible, even negative, growth rates of per capita GDP. A noteworthy feature about
Chile is that her approach to economic reform has been way more pragmatic
than other countries’ approach, taming liberalisation with capital and exchange
controls and non-tariff restrictions for a certain period of time.
Furthermore, beginning in the late 1980s the growth rate of the four core
Latin American economies exhibited a short-term upward trend that could not be
sustained beyond 1994. Thus, long-term economic activity became slower and
more volatile after the economic reforms. Contrary to Latin American countries,
the Chinese economy, like many other East Asian economies whose fate has not
been subject to the WC agenda, has been more vigorous and stable. If one looks
at the more dynamic pattern of behaviour shown by the Chinese economy in the
last decades, one may ask, then, what is the advantage of laissez-faire policies
for latecomers? Summing up, both GDP and per capita income slowed down in
Latin America from 1960–1980 to 1981–2006, whereas China, which kept away
from the WC agenda, experienced long-term sustainable growth accelerations
(see Figure 2.1a to 2.1e).3 While Latin America’s long-term output performance
has been stagnant, China has accumulated three decades of fast growth.

2.3.2  Inflation
Price stability through a flexible exchange rate regime and targeting a low rate of
inflation is another key condition for restarting growth,4 according to Williamson
(2003). Inflation has been conquered in the region but, since monetary anchors
had failed to cope with rampant inflation, price stabilisation necessitated a signi-
ficant role of managed pegs and exchange rate overvaluation. Actually, price
stability predates the adoption of an inflation targeting regime by central banks;
in nearly all inflation targeters inflation was already on a downward trend prior
Figure 2.1a Argentina’s GDP (annual growth rate), 1961–2009 (source: authors’ calcula-
tions using data from the World Bank and the INDEC).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.1b Brazil’s GDP (annual growth rate), 1961–2009 (source: authors’ calculations
using data from the World Bank and the IBGE).
Hodrick–Prescott Filter: smoothing parameter = 100.
Figure 2.1c Chile’s GDP (annual growth rate), 1961–2009 (source: authors’ calculations
using data from the World Bank and the Banco Central de Chile).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.1d Mexico’s GDP (annual growth rate), 1961–2009 (source: authors’ calcula-
tions using data from the World Bank and the Banco de México).
Hodrick–Prescott Filter: smoothing parameter = 100.
34 I. Perrotini-Hernández et al.

Figure 2.1e China’s GDP (annual growth rate), 1961–2009 (source: authors’ calculations
using data from the World Bank and the National Bureau of Statistics of
Hodrick–Prescott Filter: smoothing parameter = 100.

to the introduction of the new monetary policy consensus (Rochon and Rossi,
2006). The median inflation rate in Latin America went down from 32 per cent
to 14 per cent between 1990 and 1994 (see Figure 2.2a to 2.2e).
However, exchange rate-based stabilisation becomes unsustainable in the
long-term because of the Impossible Trinity proposition: governments cannot
simultaneously keep a fixed exchange rate regime, set the interest rate and have
free capital mobility. At the end of the day, when exchange rate crises erupt a
nominal exchange rate anchor must be abandoned. The Impossible Trinity helps
to explain why Brazil, Chile and Mexico moved from a variety of pegs to a flex-
ible exchange rate regime cum inflation targeting in the 1990s, while Argentina
adopted a managed floating exchange rate regime after the financial crisis of
2001 when its currency board was abandoned. Insofar as the monetary-based
anchor regime had gone astray in the 1980s and exchange rate-based anchors
had collapsed in the mid-to-late 1990s, the only monetary/exchange rate regime
left was a model that Knut Wicksell (1898) had originally put forth in which the
interest rate regulates the price level. This is the well-known inflation targeting
monetary policy framework advocated by the new monetary consensus (Bern-
anke et al., 1999; Lavoie and Seccareccia, 2005).
The transition from the exchange rate-based anchor to the new monetary con-
sensus model was also facilitated by the recent developments in the international
capital markets (international securitisation of investment instruments, financial
innovation, derivatives, financialisation of the economic activity) which, in the
Figure 2.2a Argentina’s annual rate of inflation, 1961–2009 (source: authors’ calcula-
tions using data from the World Bank and the INDEC).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.2b Brazil’s annual rate of inflation, 1970–2009 (source: authors’ calculations
using data from the World Bank and the Banco Central do Brasil).
Hodrick–Prescott Filter: smoothing parameter = 100.
Figure 2.2c Chile’s annual rate of inflation, 1961–2009 (source: authors’ calculations
using data from the World Bank and the Banco Central de Chile).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.2d Mexico’s annual rate of inflation, 1961–2009 (source: authors’ calculations
using data from the World Bank and the Banco de México).
Hodrick–Prescott Filter: smoothing parameter = 100.
Beyond the Washington Consensus 37

Figure 2.2e China’s annual rate of inflation, 1987–2009 (source: authors’ calculations
using data from the World Bank and the National Bureau of Statistics of
Hodrick–Prescott Filter: smoothing parameter = 100.

last analysis, led to the ongoing international financial crisis. Nonetheless, given
the influence of the volatility of perfect capital mobility on the dynamics of both
the domestic interest rate and the exchange rate, the very same transmission
mechanism of the international financial markets may powerfully influence the
effectiveness of developing countries’ monetary policy. As Rojas-Suárez (2003)
put it, financial liberalisation has not guaranteed constant access to international
capital markets. Therefore, it appears that a framework composed of inflation
targeting cum a flexible exchange rate regime will not necessarily circumvent
the constraints caused by the Impossible Trinity nor will it automatically estab-
lish an independent monetary policy, though a floating exchange rate will be a
better trade shock-absorber than a peg. Moreover, if inflation is not elastic to
interest rate changes as a result of exogenously determined distributive shares
(real wages and profit margins), in this case setting a low inflation rate as the
unique goal of monetary policy may impart negative effects on capital accumu-
lation, investment expenditure, output growth, the rate of employment and
income distribution (Palley, 1996; Oreiro et al., 2008).
In line with the discussion above, the high pass-through effect of exchange
rate fluctuations to the price level is a major reason that makes an arrangement
of managed exchange rate and flexible inflation targeting a superior regime than
the one proposed by the augmented WC (cf. Galindo and Ros, 2006; Ito and
Sato, 2007). Again, as shown in Figure 2.2, China achieved price stability in her
own way, as opposed to Latin America, without surrendering herself to the
38 I. Perrotini-Hernández et al.
constraints imposed on domestic monetary policy by international capital
markets and the Impossible Trinity.

2.3.3  Trade liberalisation and trade balance
Hoping to spur the level of economic activity, the core economies of Latin
America moved from import substitution to trade liberalisation in a short period
of time. The trade-to-GDP ratio almost doubled between 1986 and 2006 as a
consequence of rapid expansion of exports and imports. Free trade in most cases
meant severe import competition for many industries that still were at the
infancy stage and unable to cope with trade liberalisation.
Rodrik (2004) studied 83 growth accelerations events5 in the world economy
between 1957 and 1992; he found that the bulk of those accelerations (82 per
cent) were not brought about by economic liberalisation. Remarkably, Latin
America’s growth accelerations of the post-war period were certainly unrelated
to trade liberalisation. Some commentators have argued that, by and large, trade
liberalisation in Latin America favoured rapid expansion of exports but at the
same time the industrial sector moved to static comparative advantages, while in
China and East Asia manufacturing upgrading allowed for dynamic advantages
(Shaffaeddin, 2005). Most importantly, trade liberalisation in Latin America led
to an export boom in activities characterised by heavy imported inputs content.
The export-led growth involved an import-led growth model; the balance of pay-
ments constraint to long-term output expansion became even more binding after
trade liberalisation as a result of the deterioration of the trade balance (Moreno-
Brid, 1999; Pacheco and Thirlwall, 2006), although conjuncture improvements
in the terms of trade, owing to booming commodity prices such as copper, soja
and oil, produced temporary surplus positions which vanished over time. The
Chinese economy, instead, improved its current account position while opening
to foreign trade (cf. Figure 2.3a to 2.3e).

2.4 Evolution of the development gap

Structuralist economic theory emphasises that economic development is basi-
cally about structural change, therefore about growth and income distribution
(Ros, 2000; Taylor, 1991). There are many factors that can hamper structural
change and capital accumulation in a developing economy. These binding con-
straints often lead to a gap between actual and optimum or desired growth rates
of capital accumulation.
The structuralist literature has pointed out that ‘a gap signals macroeconomic
disequilibrium’ (ibid.:159). The foreign debt crisis of 1982 and the ensuing lost
decade of the 1980s signalled a gap in Latin America. Economic liberalisation
and market-oriented reforms were expected to remove such a gap (Rodrik,
2007). We now turn to enquire whether economic liberalisation has contributed
to narrow the development gap of a core set of Latin American economies and
the most developed economy in the world, namely the United States of America
Figure 2.3a Argentina: trade balance–GDP ratio (%), 1960–2009 (source: authors’ cal-
culations using data from the World Bank and the INDEC).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.3b Brazil: trade balance–GDP ratio (%), 1960–2009 (source: authors’ calcula-
tions using data from the World Bank and the IBGE).
Hodrick–Prescott Filter: smoothing parameter = 100.
Figure 2.3c Chile: trade balance–GDP ratio (%), 1960–2009 (source: authors’ calcula-
tions using data from the World Bank and the Banco Central de Chile).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.3d Mexico: trade balance–GDP ratio (%), 1960–2009 (source: authors’ calcula-
tions using data from the World Bank and the Banco de México).
Hodrick–Prescott Filter: smoothing parameter = 100.
Beyond the Washington Consensus 41

Figure 2.3e China: trade balance–GDP ratio (%), 1978–2008 (source: authors’ calcula-
tions using data from the World Bank).
Hodrick–Prescott Filter: smoothing parameter = 100.

(USA). The development gap for Latin America’s core economies is calculated
in terms of the per capita income of the country in question as a percentage of
the USA’s per capita income. We assess the performance of the development
gap over the whole period 1960–2008, and highlight the evolution of the gap
before and after economic liberalisation. Also, the evolution of the gap of Latin
America’s core economies is contrasted with that of China, undoubtedly today’s
fastest growing economy in the world.
A perusal analysis of Figure 2.4a to 2.4e shows that Chile is the only Latin
American economy in our sample that ends the period on a clear better off posi-
tion compared with its starting position. Chile had recovered all that it had lost
in terms of the development gap by 1993 and somewhat narrowed the gap later
on. The worse case is Argentina, followed by Mexico. Brazil lost most of the
ground it had achieved in the 1970s thanks to an aggressive industrial policy. By
and large, it can be said that import-substituting industrialisation (henceforth ISI)
made Brazil and Mexico better off. Yet, Argentina and Chile do not appear to
have improved their development gap in the ISI period.
The economic record of most Latin American countries after economic liber-
alisation is weak. Apart from Chile, none of them made great achievements
during 1986–2008. However, the case of Chile should be looked at with caution
because her big-bang economic liberalisation took place in the 1970s, in particu-
lar during 1973–1982, clearly, an era when the development gap was on a
Figure 2.4a Argentina’s development gap, 1960–2008 (source: authors’ calculations
using data from the World Bank).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.4b Brazil’s development gap, 1960–2008 (source: authors’ calculations using
data from the World Bank).
Hodrick–Prescott Filter: smoothing parameter = 100.
Figure 2.4c Chile’s development gap, 1960–2008 (source: authors’ calculations using
data from the World Bank).
Hodrick–Prescott Filter: smoothing parameter = 100.

Figure 2.4d Mexico’s development gap, 1960–2008 (source: authors’ calculations using
data from the World Bank).
Hodrick–Prescott Filter: smoothing parameter = 100.
44 I. Perrotini-Hernández et al.

Figure 2.4e China’s development gap, 1960–2008 (source: authors’ calculations using
data from the World Bank).
Hodrick–Prescott Filter: smoothing parameter = 100.

widening trend. As a reaction to the foreign debt crisis of 1982, the Chilean gov-
ernment reformed the liberal reforms (Ffrench-Davis, 2005:149), tilting its trade
policy towards a more pragmatic and heterodox view (with exchange rate deval-
uations, antidumping measures, price bands for some agricultural goods, the
drawback system, the uniform tariff rate went up from 10 per cent to 35 per cent
in 1984, capital controls and multiple exchange rates). So the latter sub-period
was not a laissez-faire environment altogether in this country. According to
Ffrench-Davis (2005), there was one radical trade reform in 1974–1979 and
another moderate, more pragmatic reform in 1983–1991. The first one produced
a widening development gap and the second one narrowed it.
As for Argentina, Brazil and Mexico, their development gap has widened, in
some cases very drastically with a fairly slight improvement towards the end of
the period under analysis. In contradistinction, China has continuously been
catching up vis-à-vis the US economy; China’s development gap has narrowed
rapidly, paradoxically, since it opened up to international trade.
What are the driving forces behind Latin America’s poor performance? Need-
less to say, a number of factors call for the phenomenon. We argue that struc-
tural change in Latin America has failed chiefly because the process of economic
liberalisation has involved a low gross investment to GDP ratio and a drastic
increase of the income elasticity of imports. Actually, in most Latin American
countries the gross investment ratio declined vis-à-vis the investment coefficient
of the golden age of the ISI model. In Argentina, seemingly the worst case, the
gross investment coefficient peaked in 1977 (at 31 per cent of GDP) and reached
Beyond the Washington Consensus 45
a minimum level in 2002 (12 per cent). Brazil’s gross investment ratio peaked in
1989 and has been on a downward trend ever since, whereas the Chilean ratio
exhibits an opposite (upward) drift from 1983 (9.8 per cent) to 2009 (25 per
cent). Mexico’s gross investment ratio has been low, on average around 20 per
cent, during the whole economic reform period. In contrast, China’s fast long-
term economic growth has relied on a steadfastly upward drift of its gross invest-
ment ratio from the early 1960s (10.8 per cent in 1962) to the present time (close
to 40 per cent).
Liberalisation of the capital market, an important aspect of the market-
oriented economic reform, tends to produce higher interest rates. From a free-
market perspective, higher rates of interest for both lenders and borrowers are
viewed as the automatic balancing mechanism that equalises desired saving and
desired net investment. The rate of interest sets free the dynamism that turns
investment away from inferior projects so as to promote technological progress
and economic development (McKinnon, 1973 and 1991). Nonetheless, as
Shackle (1983:154) points out, ‘the rate of interest arises from uncertainty about
the future prices of the bonds given by borrowers in exchange for loans’. In such
a case, the automatic balancing mechanism can be self-destructive. It appears
that under deregulated financial markets uncertainty increases the asymmetry
between the distribution of costs and benefits of capital account liberalisation, on
one hand, and the level of economic and institutional development, on the other,
becomes a binding constraint: while the benefits of financial liberalisation have a
tendency to increase beyond certain point, the costs associated to free capital
flows tend to augment below that dividing line (cf. Prasad and Rajan, 2008). The

Figure 2.5a Argentina’s gross investment and gross savings (% of GDP), 1960–2009
(source: authors’ calculations using data from the World Bank and the
Figure 2.5b Brazil’s gross investment and gross savings (% of GDP), 1960–2009
(source: authors’ calculations using data from the World Bank and the

Figure 2.5c Chile’s gross investment and gross savings (% of GDP), 1960–2009 (source:
authors’ calculations using data from the World Bank and the Banco Central
de Chile).
Figure 2.5d Mexico’s gross investment and gross savings (% of GDP), 1960–2009
(source: authors’ calculations using data from the World Bank and the

Figure 2.5e China’s gross investment and gross savings (% of GDP), 1960–2008 (source:
authors’ calculations using data from the World Bank).
48 I. Perrotini-Hernández et al.
policy implication of this is that the government ought to adopt a pragmatic view
on capital account liberalisation rather than an overly laissez-faire approach. The
asymmetric experience with capital account liberalisation across emerging
market economies reveals that, while financial liberalisation in Latin America
has given rise to cycles of ephemeral recovery and prosperity, exchange rate
instability, protracted stagnation, recession and sometimes complete growth
meltdown, China and other Asian economies tend to exhibit growth acceleration
A second driving force behind the enlargement of the development gap during
the period of the market-oriented reforms, we contend, is the high-income elas-
ticity of the demand for imports (henceforth π). The slow growth pattern fol-
lowed by most Latin American economies6 in the last decades is associated with
the distortion of import demand patterns generated by economic liberalisation. A
higher π means a tighter balance of payments constraint to economic expansion,
therefore a lower growth rate of output consistent with balance of payments
equilibrium (Thirlwall, 1979). Following Pacheco-López and Thirlwall (2006),
we applied the technique of rolling regressions to assess the performance of p in

Figure 2.6a Argentina’s income elasticity of imports, 1961–2007.

Figure 2.6b Brazil’s income elasticity of imports, 1961–2007.

Figure 2.6c Chile’s income elasticity of imports, 1961–2007.

Figure 2.6d Mexico’s income elasticity of imports, 1961–2007.

Figure 2.6e China’s income elasticity of imports, 1979–2007.

50 I. Perrotini-Hernández et al.
our sample of four Latin American core economies from 1961 to 2007.7 The
behaviour of China’s π from 1979 to 2007 is also calculated for the sake of
Rising trends of π for the Latin American countries can be detected during the
whole period of trade liberalisation. Opposite to that, China has experienced a
drastic reduction in its π. The question arises as to why China has exhibited a
declining income elasticity of imports and an ever-diminishing development gap,
while Latin America has experienced exactly the opposed trends in the period of
the market-friendly reforms.
Before the WC agenda, the Latin American economies (Argentina aside) kept
a downward drift of π during 1960–1980; several growth acceleration episodes
took place along this period, despite the ISI suffering from an anti-export bias.
The WC economic reform promoted an export-led growth model with outright
import liberalisation. Unfortunately, the propensity to import has increased dras-
tically since the 1980s, thus discouraging the positive effect of booming exports.
Hence slow growth. In contradistinction, China (and other East Asian econo-
mies) adopted an export-led growth strategy that has triggered growth accelera-
tion events largely because her government prevented π from rising.
The expectations of the policy of free markets and sound money, augmented
with ‘good governance’ amendments, have not been fulfilled. The growth experi-
ence of Latin America under the mainstream consensus, particularly if contrasted
with that of China or the successful East Asian economies, rejects the idea that
the ‘Victorian’ free markets and sound money approach (Krugman, 1995) is the
only and best recipe available in town to economic development. Clearly, an
alternative developmental strategy for Latin America should be put in place.
This ‘new’ development programme needs not imply an a priori rejection of the
export-led model. It must, nonetheless, target a sensible reduction of p. The
equation of the alternative developmental strategy can be a strong pro-growth
state plus strong market plus diminution of the balance of payments constraint.
There is plentiful historical evidence supporting the hypothesis that nowadays’
advanced economies followed a similar path in which industrial policy impelled
growth and development (cf. Chang, 2003, 2008; Di Maio, 2009; List, 1885).

2.5 Beyond the Washington Consensus: the case for

industrial policy
Williamson (2004a:14) admits that ‘of course we need to go beyond it [the
Washington Consensus]’ because it ‘did not contain all the answers to the ques-
tions of 1989, let alone that it addresses all the new issues that have arisen since
then’. Williamson’s idea of going beyond the WC advocates the need of institu-
tional reforms, but he is too willing to ignore the role industrial policy has played
in the history of economic development. He goes on to emphatically reject the
adoption of industrial policy whereby governments ‘pick winners’. Instead, a
‘cousin of industrial policy’ is welcome, namely a ‘Schumpeterian innovation . . .
a national system of innovation’ (Williamson, 2004a:12).8
Beyond the Washington Consensus 51
One may agree that a national system of technological innovation would con-
tribute to substantially improve Latin America’s economic performance, but it is
hard to see how Schumpeter’s approach to economic development, with its
emphasis on the role of the state and institutions as driving forces of the consti-
tution of markets and industrialisation, can be made compatible with the
microeconomic-behavior-based, Washington Consensus laissez-faire approach
to growth. The latter assumes a minimal government. The ‘new agenda’ put
forth in Kuczynski and Williamson (2003) and Williamson (2004a:5) acknowl-
edges the importance of institutions, albeit it emphasises laissez-faire institu-
tional reforms such as liberalisation of the labour market, privatisation, asset
titling and the stipulation of property rights. Interestingly, the new agenda
endorses countercyclical policies (allegedly) ‘à la Keynes’ and improvements in
the region’s inequality in income distribution. The former requires targeting a
budget balance over the business cycle; the latter calls for policies aimed at
increasing the poor’s accumulation of human capital.
Unlike John Williamson, the recent Schumpeterian literature considers that
industrial policy does not necessarily thwart the market mechanism; an industrial
policy,9 undertaken as a strategic collaboration or cooperative game between the
government and the market, is still viable and reasonable, in particular for the
enhancement of latecomers (cf. Aghion and Howitt, 2005; Cimoli et al., 2009a;
Dosi et al., 1990; Rodrik, 2007). Industrial policy has been a built-in essential
element of every successful development experience (Cimoli et al., 2009a). Wil-
liamson’s dismissal of industrial policy runs counter to historical evidence;
developmental industrial policies have played a fundamental role in the process
of structuring the market and have been part and parcel of institutional building
in every historically observed catching-up process of economic development.
Importantly, macroeconomic policies must be consistent with developmental
industrial policy. As shown above, the free-market reform has induced a delete-
rious effect on investment, the income elasticity of imports and the level of eco-
nomic activity. The macroeconomic environment as represented by the WC
agenda involves supply-side incentives for entrepreneurship, but, paradoxically
enough, it has destroyed industrial production capacity and the domestic econo-
mies’ ability to absorb, adapt and apply technological capabilities (Castaldi et al.
2009). The destruction of the domestic production chain is encapsulated in the
recent sharp increase in the income elasticity of imports, an unconstructive de-
industrialisation effect not seen in the Chinese case.10
The indisputable fact that sometimes industrial policy has been prone to
waste, inefficiency and political corruption should not lead to the conclusion that
industrial policy never works. The market mechanism may also fail. Privatisa-
tion in Latin America and elsewhere has often given rise to picking wrong
winners and crony capitalism. As Rodrik (2007:151) correctly points out:

It is not true that there is a shortage of evidence on the benefits of industrial

policy . . . it is difficult to come up with real winners in the developing world
that are not a product of industrial policies of some sort.
52 I. Perrotini-Hernández et al.
Industrial policy from this standpoint, challenges the principle of comparative
advantage as conceived by mainstream economics. Inasmuch as industrial policy
interacts with adequate institutions, it supplies public goods for improving
technological capabilities and knowledge accumulation. Historical experience
confirms that in most cases successful industrialisation did not follow a path of
productive sectoral specialisation according to the logic of comparative advant-
age (Imbs and Wacziarg, 2003). Instead, it revolved around industrial policies
targeting the acquisition of technological knowledge, highly skilled factors and
capabilities that diversified the composition of aggregate output away from given
input endowments and from ‘ricardian’ comparative advantages (Cimoli et al.,
2009a; Wood, 1995).
A Schumpeterian ‘cousin’ of industrial policy, that is, a national system of
innovation presupposes that production occurs through a variety of institutions,
namely the firm, the market and the government. The existence of market imper-
fections, externalities, public goods and increasing returns to scale furnish an
economic rationale for government interventions through industrial policies and
regulation. An adequate industrial policy can help reduce and bridge the techno-
logical and development gaps, while regulation frameworks minimise the
harmful effects of externalities that the market mechanism either generates or is
unable to internalise. Furthermore, it is hard to think of such a national system of
innovation without the role of a Developmentist State acting to set up the clus-
tering of diffusion processes of technological innovations resultant from micro-
economic behaviour. The view that often the economic impact of a technological
innovation is particular and specific in nature is a central feature of Schumpeter’s
theory. Technical change remains within the realm of singular firms and/or
industries for a certain amount of time before it is spread over a wider socio-
economic environment. If, owing to oligopolistic structures, asymmetric
information, differential access to new technologies or effective demand con-
straints, the market mechanism itself fails to disseminate technical innovations
over a broader dominion and at a faster speed, then public institution interven-
tions can enhance technological diffusion, boost the potential economic impact
of technical change and induce the multiplier effects of technical innovations,
thus leading to higher growth rates. Clearly, only a combination of both techni-
cal innovation and institutional innovation can engender constellations of suc-
cessful industrialisation and economic development. This is the manner in which
Schumpeter’s dynamic approach (1934, 1943) envisaged the relevance of
technological innovation in a theory of long-term growth, which is far away, we
may contend, from the augmented WC interpretation.
On the other hand, technological innovation increases productivity. Hence
changes the demand for labour per unit of output and the requirement of capital
and intermediate inputs in production. Therefore, aggregate supply and aggreg-
ate demand change accordingly. The newly augmented productive capacity and
the effective demand must increase at the same rate, uniformly, for the economy
to grow at full employment and full utilisation of actual productive capacity.
However, by and large, demand and productivity grow at a different pace across
Beyond the Washington Consensus 53
industrial sectors and in aggregate terms, in which case economic growth will
not necessarily involve full employment. Technological innovation, therefore, in
this scenario, produces structural unemployment. The WC’s new agenda recom-
mends labour market flexibility to cope with unemployment and slow growth,
under the presumption that flexible wages will provide a full compensation
mechanism of technologically induced unemployment. This hypothesis relies on
the strong assumption of perfect and complete substitutability of labour and
capital. Nonetheless, a number of stylised facts associated with the introduction
of technological innovation rule out the prediction of endogenous compensation
induced by labour market flexibility as assumed by the new agenda: differential
labour skills between displaced labour force and the new input requirements; dif-
ferential factor intensities between old and new products and industries; hetero-
geneous capital; changes in the composition of GDP after the introduction of
technical innovation; price stickiness in goods markets; and the presence of
asymmetric information in goods, labour and credit markets. All these stylised
facts are pervasive economy-wide. Therefore, labour market flexibility is no sen-
sible solution to overcome slow growth, structural unemployment, skewed dis-
tribution of income and social inequality in Latin America.

2.6 Final remarks

The Washington Consensus recommended microeconomic flexibility, macroeco-
nomic discipline and trade and financial liberalisation as a rational method of
surmounting stagflation, exchange rate instability and balance of payments
The past 20 years have witnessed that countries that adopted the WC agenda
have performed poorly, and the economic prospect of Latin America in the
context of a world economy engulfed by a severe financial crisis is certainly not
buoyant. Despite the evidence of the last two decades, Williamson (2009:1) went
on to state:

The microeconomic advice that we have given and the report of the Spence
Commission are as valid as ever. It remains sensible to use rather than fight
the market, but to be prepared to alter a laissez-faire outcome when one of
the classic conditions for market failure arises. These are: when monopoly
is unavoidable, when externalities are important, and when the resulting
income distribution offends social norms.

The problem with the above conclusion is that such ‘classic conditions for
market failure’ are ubiquitous in the modern economy; hence the need of an
alternative strategy. The question to be asked is where to go from here. We have
made the case for a developmental industrial policy. Another question to be
raised is how to actually implement such a policy. The main thrust of the WC is
fiscal discipline on the ground that budget deficits produce balance of payments
crises. The orthodox fiscal policy mix (primary fiscal surplus and nominal fiscal
54 I. Perrotini-Hernández et al.
deficit) coupled with high interest rates resulting from targeting low inflation, is
responsible for the stagnant economic path and the growing economic inequality
observed over the last 20 years (Câmara and Vernengo, 2004:340, passim).
A developmental industrial policy calls for a radical change in the practice of
fiscal policy from the WC concept of a primary surplus-dominated macroeco-
nomic discipline to a stabilising-investment-based fiscal policy regime. There is
a fundamental contradiction in mainstream fiscal policy: in a context of financial
liberalisation and inflation targeting, the primary surplus approach benefits bond
holders, reduces public investment, polarises income distribution and hurts eco-
nomic activity. Hence Williamson’s refusal of industrial policy should not come
as a surprise; it is impossible to have both the primary surplus and develop-
mental industrial policy simultaneously.
A fiscal policy regime aimed at stabilising investment, output growth and the
labour market, congenial with developmental industrial policy, puts forth public
investment as the countercyclical tool par excellence. It needs not entail an
explosive accumulation of debt, provided the central bank targets low interest
rates – Keynes’s euthanasia of the rentier (Keynes, 1936:376) – rather than low
inflation, and provided the monetary authority restricts capital movements.11
This alternative policy framework was known to Keynes in the 1930s, and
appears to be consistent with Schumpeterian technological and institutional

1 To quote Adam Smith (1776 [1976, p. 7]: ‘The greatest improvement in the produc-
tive powers of labour, and the greater part of the skill, dexterity, and judgement with
which it is any where directed, or applied, seem to have been the effects of the divi-
sion of labour.’
2 The WC view is that fiscal deficits crowd out private investment, cause recession,
inflation and balance of payments crisis; hence the rejection of fiscal policy as a macr-
oeconomic policy instrument and hence the prescription of perennial primary surplus.
3 Our data in Figure 2.1 are consistent with the growth patterns shown by Solimano and
Soto (2005:9–12, passim). They also assert ‘a substantial slowdown in economic
growth after 1980’ and identify a high number of ‘growth crises’ (defined as negative
growth rates). Solimano and Soto coincide with Titelman et al. (2008), who argue that
while the average frequency of shocks to Latin America’s terms of trade diminished
from six in 1980–1990 to two in 2002–2006, financial shocks became more frequent
between the 1980s and the 1990s. Most importantly, the latter’s magnitude increased
from 0.7 per cent of GDP to 3.5 per cent. The authors conclude that during 1980–2006
adverse financial shocks claimed domestic absorption adjustments equivalent to
almost –7 per cent of GDP, whereas the impact of terms of trade shocks on absorption
declined from 2.25 per cent of GDP in 1980–1990 to 0.40 per cent in 1991–2001 and
0.00 per cent in 2002–2006. Clearly, capital account liberalisation has caused volatil-
ity of real economic activity to increase.
4 [L]ow, stable inflation is important for market-driven growth, and . . . monetary
policy is the most direct determinant of inflation. Further, of all the government’s
tools for influencing the economy, monetary policy has proven to be the most
flexible instrument for achieving medium-term stabilisation objectives.
(Bernanke et al., 1999:3)
Beyond the Washington Consensus 55
5 Pacheco-López and Thirlwall (2006) summarise Rodrik’s concept of growth accelera-
tions as a positive discrepancy of two percentage points or more between eight years
prior to the growth acceleration episode and eight years after the event, with a post-
acceleration growth rate of 3.5 per cent or higher.
6 As Rodrik (2004) put it: ‘The cold fact is that per capita economic growth perform-
ance [in Latin America] has been abysmal during the 1990s by any standards.’
7 Our results are consistent with those in Pacheco-López and Thirlwall (2006).
8 A national innovation system is about government creating institutions to provide
technical education, to promote the diffusion of technological information, to fund
precompetitive research, to provide tax incentives for R&D, to encourage venture
capital, to stimulate the growth of industrial clusters, and so on.
(Williamson, 2004a:11)
9 Industrial policy is defined as a tool that targets specific industrial branches, sectors
and firms with the aim of enhancing productivity, rapidly catching up with technolo-
gical leaders, increasing employment in the formal labour market and accelerating
output growth in line with some clearly defined national economic development goals
(cf. Bresser-Pereira 2007; Cimoli et al., 2009a; Rodrik, 2007).
10 As Khan and Blankenburg (2009:367) put it,
the main effect of liberalisation, across virtually all of Latin America, has been to
reinforce Latin America’s commodity bias in the absence of any attempts at
‘Schumpeterian’ dynamic upgrading into higher-technology, higher-value added
processes and/or products. Put differently, technological improvements have been
limited to certain basic commodities, such as copper concentrates in Chile or iron
in Brazil, but no attempts have been undertaken to upgrade to different processes
(copper smelting) or product (steel).
(emphasis in the original)
11 See also Câmara and Vernengo (2004) for a detailed discussion of the implications of
Keynes’s pragmatic separation of the current budget and the capital budget in the
practice of countercyclical fiscal policy.

Aghion, P. and Howitt, P. (2005) ‘Appropriate growth policy: A unifying framework’,
The 2005 Joseph Schumpeter Lecture, paper presented to the European Economic
Association Congress, Amsterdam, 25 August.
Bernanke, B., Laubach, T., Mishkin, F.S. and Posen, A. (1999) Inflation Targeting:
Lessons from the International Experience, Princeton, NJ: Princeton University Press.
Bresser-Pereira, L.C. (2007) Macroeconomia da Estagnação: Critica da ortodoxia con-
vencional no Brasil pós-1994, São Paulo: Editora 34.
Câmara, N.F.A. and Vernengo, M. (2004) ‘Fiscal Policy and the Washington Consensus:
A Post Keynesian Perspective’, Journal of Post Keynesian Economics, vol. 27, No. 2,
pp. 333–43.
Castaldi, C., Cimoli, M., Correa, N. and Dosi, G. (2009) ‘Technological learning, policy
regimes and growth: The long-term patterns and some specificities of a “globalized”
economy’, in M. Cimoli, G. Dosi and J.E. Stiglitz (eds) Industrial Policy and Develop-
ment: The Political Economy of Capabilities Accumulation, Oxford: Oxford University
Press, pp. 39–77.
Chang, H.J. (2003) Rethinking Development Economics, London: Anthem Press.
Chang, H.J. (2008) Bad Samaritans, New York: Bloomsbury Press.
56 I. Perrotini-Hernández et al.
Cimoli, M., Dosi, G. and Stiglitz, J.E. (2009a) ‘The political economy of capabilities
accumulation: The past and future of policies for industrial development’, in M.
Cimoli, G. Dosi and J.E. Stiglitz (eds) Industrial Policy and Development: The Polit-
ical Economy of Capabilities Accumulation, Oxford: Oxford University Press,
pp. 1–38.
Cimoli, M., Dosi, G. and Stiglitz, J.E. (eds) (2009b) Industrial Policy and Development:
The Political Economy of Capabilities Accumulation, Oxford: Oxford University Press.
Davidson, P. (2003) ‘What is wrong with the Washington Consensus and what should we
do about it?’ paper presented at conference on ‘Reforming the Reforms: What Next for
Latin America’, Rio de Janeiro, 25 July.
Di Maio, M. (2009) ‘Industrial policies in developing countries: History and perspec-
tives’, in M. Cimoli, G. Dosi and J.E. Stiglitz (eds) Industrial Policy and Development:
The Political Economy of Capabilities Accumulation, Oxford: Oxford University Press,
pp. 107–43.
Dosi, G., Pavitt, K. and Soete, L. (1990) The Economics of Technical Change and Inter-
national Trade, London and New York: Harvester Wheatsheaf, New York University
ECLAC (Economic Commission for Latin America and the Caribbean) (2002) Latin
America and the Caribbean in the World Economy, 2000–2001, Santiago: ECLAC.
ECLAC (Economic Commission for Latin America and the Caribbean) (2003) Latin
America and the Caribbean in the World Economy, 2001–2002, Santiago: ECLAC.
Ffrench-Davis, R. (2005) Reformas para América Latina: después del fundamentalismo
neoliberal, Buenos Aires: CEPAL-Siglo XXI.
Galindo, L.M. and Ros, J. (2006) ‘Alternatives to inflation targeting: Central bank policy
for employment creation, poverty reduction and sustainable growth’, Working Paper
no. 7, PERI, University of Amherst, MA, Amherst, September, pp. 1–25.
Imbs, J. and Wacziarg, R. (2003) ‘Stages of diversification’, American Economic Review,
vol. 93, No. 1, pp. 63–86.
Ito, T. and Sato, K. (2007) ‘Exchange rate pass-through and domestic inflation: A com-
parison between East Asia and Latin American Economies’, RIETI Discussion Paper
Series 07-E-040 (May 2007): 1–45.
Keynes, J.M. (1936) The General Theory of Employment, Interest and Money, New York:
Harcourt and Brace.
Khan, M.H. and Blankenburg, S. (2009) ‘The political economy of industrial policy in
Asia and Latin America’, in M. Cimoli, G. Dosi and J.E. Stiglitz (eds) Industrial Policy
and Development: The Political Economy of Capabilities Accumulation, Oxford:
Oxford University Press, pp. 337–77.
Krugman, P. (1995) ‘Dutch Tulips and Emerging Markets’, Foreign Affairs, vol. 74, No.
4, pp. 36–7.
Kuczynski, P.P. and Williamson, J. (eds) (2003) After the Washington Consensus:
Restarting Growth and Reform in Latin America, Washington, DC: Institute for Inter-
national Economics.
Lavoie, M. and Seccareccia, M. (eds) (2005) Central Banking in the Modern World:
Alternative Perspectives, Cheltenham, UK, Northampton, MA: Edward Elgar.
List, F. (1885) The National System of Political Economy, translated from the original
German edition published in 1841 by Sampson Lloyd, London: Longmann, Green
and Co.
McKinnon, R. (1973) Money and Capital in Economic Development, Washington, DC:
Brooking Institution.
Beyond the Washington Consensus 57
McKinnon, R. (1991) The Order of Economic Liberalisation: Financial Control in the
Transition to a Market Economy, Washington, DC: Brooking Institution.
Moreno-Brid, J.C. (1999) ‘Mexico’s Economic Growth and the Balance of Payments
Constraint: A Cointegration Analysis’, International Review of Applied Economics,
vol. 13, No. 2, pp. 149–60.
Ocampo, J.A. (2004) ‘Latin America’s Growth and Equity Frustrations during Structural
Reforms’, Journal of Economic Perspectives, vol. 18, No. 2, pp. 67–88.
Ohlin, B. (1933) Interregional and International Trade, Cambridge, MA: Harvard Uni-
versity Press.
Oreiro, J.L., Coelho Squeff, G. and de Paula, L.F. (2008) ‘A post Keynesian proposal for
a flexible institutional arrangement of inflation targeting regime in emerging econo-
mies’, March, paper prepared for the workshop ‘Inflation Targeting: Is There a Cred-
ible Alternative?’ at Balliol College, Oxford, 4 April 2008, pp. 1–24. The workshop is
organised by the Post Keynesian Study Group, United Kingdom.
Pacheco, P. and Thirlwall, A.P. (2006) ‘Trade Liberalisation, the Income Elasticity of
Demand for Imports and Growth in Latin America’, Journal of Post Keynesian Eco-
nomics, vol. 29, No. 1, Fall, pp. 41–61.
Palley, T.I. (1996) Post Keynesian Economics: Debt, Distribution and the Macroecon-
omy, London: Macmillan.
Prasad, E.S. and Rajan, R. (2008) ‘A pragmatic approach to capital aAccount liberaliza-
tion’, Working Paper 14051, NBER, Cambridge, MA.
Ricardo, D. (1821 [1951]) On the Principles of Political Economy and Taxation, edited
by Piero Sraffa with the collaboration of M.H. Dobb, vol. I of the Works and Corres-
pondence of David Ricardo, Cambridge: Cambridge University Press.
Rochon, L.P and Rossi, S. (2006) ‘Inflation Targeting, Economic Performance, and
Income Distribution: A Monetary Macroeconomic Analysis’, Journal of Post Keyne-
sian Economics, vol. 28, No. 4, pp. 615–38.
Rodrik, D. (2004) ‘Rethinking growth strategies’, WIDER Annual Lecture 8, Helsinki:
United Nations World Institute for Development Economic Research.
Rodrik, D. (2007) One Economics Many Recipes: Globalization, Institutions, and Eco-
nomic Growth, Princeton, NJ: Princeton University Press.
Rojas-Suárez, L. (2003) ‘Monetary policy and exchange rates: Guiding principles for a
sustainable regime’, in P.P. Kuczynski and J. Williamson (eds) After the Washington
Consensus: Restarting Growth and Reform in Latin America, Washington, DC: Insti-
tute for International Economics, pp. 123–55.
Ros, J. (2000) Development Theory and the Economics of Growth, Ann Arbor, MI: Uni-
versity of Michigan Press.
Saavedra, J. (2003) ‘Labor markets during the 1990s’, in P.P. Kuczynski and J. William-
son (eds) After the Washington Consensus: Restarting Growth and Reform in Latin
America, Washington, DC: Institute for International Economics, pp. 213–63.
Schumpeter, J.A. (1934) The Theory of Economic Development, Cambridge, MA:
Harvard University Press.
Schumpeter, J.A. (1943) Capitalism, Socialism, and Democracy, New York: Harper & Row.
Shackle, G. (1983) The Years of High Theory, Cambridge: Cambridge University Press.
Shaffaedin, S.M. (2005) ‘Trade liberalization and economic reform in developing coun-
tries: Structural change or de-industrialisatioon?’ UNCTAD Discussion Papers, No.
179, pp. 1–25.
Smith, A. (1776 [1976]) An Inquiry into the Nature and Causes of The Wealth of Nations,
Chicago, IL: University of Chicago Press.
58 I. Perrotini-Hernández et al.
Solimano, A. and Soto, R. (2005), ‘Economic growth in Latin America in the late 20th
century: evidence and interpretation’, in Serie Macroeconomía del Desarrollo Num. 33,
Santiago, Chile: CEPAL, February, pp. 1–44.
Taylor, L. (1991) Income Distribution, Inflation, and Growth: Lectures on Structuralist
Macroeconomic Theory, Cambridge, MA: MIT Press.
Thirlwall, A.P. (1979) ‘The Balance of Payments Constraint as an Explanation of Inter-
national Growth Rate Differentials’, Banca Nazionale del Lavoro Quarterly Review
(March), pp. 45–55.
Titelman, D., Pérez-Caldentey, E. and Minzer, R. (2008) ‘Comparación de la dinámica de
los efectos de los choques financieros y los choques de términos del intercambio en
América Latina en el período 1980–2006’, Working Paper, ECLAC, Santiago, Chile.
Weisbrot, M. (2006) ‘Latin America: The End of an Era’, Washington, DC: Center for
Economic and Policy Research, pp. 1–23.
Wicksell, K. (1898 [1965]) Interest and Prices: A Study of the Causes Regulating the
Value of Money, New York: Augustus Kelley.
Williamson, J. (1990) ‘What Washington means by policy reform’, in J. Williamson (ed.)
Latin American Adjustment: How Much Has Happened? Washington, DC: Institute of
International Economics, pp. 5–38.
Williamson, J. (2003) ‘Overview: An agenda for restarting growth and reform’, in P.P.
Kuczynski and J. Williamson (eds) After the Washington Consensus: Restarting
Growth and Reform in Latin America, Washington, DC: Institute for International Eco-
nomics, pp. 1–19.
Williamson, J. (2004a) ‘A short history of the Washington Consensus’, paper commis-
sioned by Fundación CIDOB for a conference ‘From the Washington Consensus
towards a New Global Governance’, Barcelona, 24–25 September.
Williamson, J. (2004b) ‘The Strange History of the Washington Consensus’, Journal of
Post Keynesian Economics, vol. 27, No. 2, pp. 195–206.
Williamson, J. (2009) ‘The Washington Consensus and the global crisis’, paper presented
at a conference sponsored by the Johns Hopkins School of Advanced International
Studies and the Center for Global Development, ‘New Ideas in Development Finance
after the Financial Crisis’, Speeches, Testimony and Papers, Peterson Institute for
International Economics, Washington, DC, 22 April.
Wood, A. (1995) North–South Trade, Employment and Inequality: Changing Fortunes in
a Skill-Driven World, Oxford: Clarendon Press.
3 Foreign trade and per capita
New evidence for Latin America and
the Caribbean
Humberto Ríos-Bolívar and Omar Neme-Castillo

3.1 Introduction
Given the growing importance of goods and services inside global exchanges
during the last decades, analysis of relationship between international trade and
economic growth takes a central role on present empirical debate. In theoretical
scope, neoclassic economics shows out that countries’ share of foreign trade
could become a positive strength for stimulating economic dynamism. However,
at empirical level, there is not a consensus about the impact of that share and, in
most cases, it has only found static effects with limited significance on that rela-
tionship. In Mexico and Latin America there are some studies in this sense, nev-
ertheless proxies used for both growth and control variables are limited or
exclude important aspects for growth process, such as human capital or imports
of goods which enlarge production capacities.
The aim of this chapter is twofold. First, it derives a regression equation
directly from a neoclassical growth model incorporating human capital and
international trade elements following Bidlingmaier’s (2007) model. Second, it
applies this equation to the case of Latin America and the Caribbean countries
in the period 1970–2007 using a panel data analysis. For this purpose, in the
next section aspects of endogenous growth theory concerning international
trade are considered with particular attention to underdeveloped countries.
These fundamentals are incorporated into an extended neoclassical growth
model in the third section of the chapter. The model includes measures of trade
openness different from those commonly used, thus establishing whether or
not there is a link between international trade and per capita income level, and
to directly study the channels through which this effect occurs. The fourth
section estimates the mentioned equation in order to test the hypothesis that
establishes a positive relationship between international trade and per capita
income level differentiated among the considered countries in the region.
Finally, conclusions are presented with some implications for international
trade theory and for commercial policy.
60 H. Ríos-Bolívar and O. Neme-Castillo
3.2 Theoretical aspects
Traditional theory of international trade displays possible benefits of goods
exchange among countries, which derives from production specialisation
(Ricardian model). If countries specialise according to their comparative advant-
age, efficient resource allowance would be encouraged; and therefore, welfare of
trading countries improves. However, this is only a level effect on the consump-
tion possibilities, that is, it is a ‘one-time increase’. After achieving complete
specialisation, productivity will not rise again.1 Therefore, this theory is limited
in its explanations of the output growth rate.
More recent theories suggest different effects of international trade. In a
context, where leading companies enjoy scale economies, the classical explana-
tion of trade and its welfare prediction does not hold any more (Gomory and
Baumol, 2001). Consequently, small economies that opened to international
trade in a late or slow way and did not acquire required scale may not be able to
compete successfully against leading economies, as is the case of most of Latin
America and the Caribbean countries.
Another argument that limits the validity of classical international trade
theory is the possibility of disadvantages arising from increasing specialisation,
particularly for developing countries like those of Latin America. If these coun-
tries specialise in sectors with lower productivity growths or with low-income
elasticity of demand (primary and traditional manufacturing), growth rates will
always be below of those of industrialised countries and economic inequality
will be widened, falling into what is known as ‘trap of specialisation’ (Redding,
Besides, Lucas (1988) developed a model of human capital accumulation with
external effects arising from learning processes, noting that human capital is not
only formed by investing in education but also through ‘learning-by-doing’ or
training. The model can achieve different growth rates between countries with no
differences in productivity rates. If the goods are substitutes, the international
exchange of goods increases the comparative advantage through ‘spillover’ effects
on labour, restricting countries’ growth with lower learning rates.
In the model, ‘learning-by-doing’ process shows diminishing returns. This
means that high rates of learning and of human capital formation, and the sub-
sequent economic growth, can only be maintained by reallocating resources to
new activities or through developing new products. With diminishing learning
rates and with no sector showing permanently higher learning rates than any
other sector, a ‘catching up’ of the poor to the rich countries can occur. Never-
theless, if trade liberalisation leads developing countries to import goods of
higher quality without producing locally, then learning and growth rates in those
countries will be lower; but in contrast, if imports are mainly capital goods used
to produce goods of higher quality or goods that, in turn, allow to improve pro-
ductivity level, then the learning and growth rates will be higher.
Likewise, Lucas (1993) pointed out another type of human capital formation
based on the idea of global public good and, particularly, on knowledge. He
Foreign trade and per capita income 61
assumes that knowledge diffuses throughout the world, through scientific publi-
cations, reverse engineering, blueprints and others. Knowledge can be acquired
not only through own efforts of each agent but also through global knowledge
spillover. The nature of knowledge is no rivalry and no partial exclusion, which
means it can be applied by different people at the same time without incurring
additional costs. Economic operators can use new ideas even without contribut-
ing to global research and development (R&D) efforts. These knowledge proper-
ties allow countries with low human capital stocks to accumulate some of that
knowledge quickly through its international adoption and, ultimately, catching
up with richer countries.
Moreover, Romer (1990) proposed a model with differentiated capital goods,
where output level is determined by the number of varieties produced by the
domestic economy. The model assumes that capital goods are not perfect substi-
tutes and are produced in the middle sector which uses patents, obtained from
R&D processes in the technological sector, as input which, in turn, uses human
capital and knowledge for the development of these patents. Because knowledge
can be used without cost by any agent through spillover effects, the R&D sector
does not exhibit diminishing returns. Thus, the higher the stock of human capital
and its use in the R&D sector, the greater the growth rate of an economy. The
model predicts positive outcome of international trade. Trade liberalisation both
generates level effects and growth effects. The growth effects are reached if
knowledge and technology transfer are free of additional costs.
One aspect of relevance to Latin American and the Caribbean countries is that
with a small stock of human capital and limited ability to employ such trained
personnel in the R&D sector restrict its growth, leading to production of only a
small range of capital goods (Rivera-Batiz and Romer, 1991) and, thus, estab-
lishing the basis for their exclusion from the international economic scenario.
This model has two implications for international trade and mainly for the
countries in the Latin American region. First, the number of varieties of differen-
tiated capital goods and total output would increase. In countries with different
factor endowments, trade in goods increases the growth rate of the economy that
grew to a lesser extent and reduces the rate of the most advanced country.
Second, if it is true that ideas flow freely worldwide then stock of knowledge in
each country that could be employed in the research sector would increase, stim-
ulating the growth of technology and the economy as a whole. Moreover,
increasing human capital productivity in the R&D sector would employ more of
this factor, boosting even more the growth rate. Therefore, to the extent Latin
American countries are able to establish centres and programmes of applied sci-
entific research that use those ideas, per capita income of those countries would
be encouraged.
Finally, it must be noted that the set of knowledge, procedures, techniques, etc.,
specific to each firm, generated by persistent R&D activities within the organisa-
tions, find greater barriers to spread both to the rest of the domestic economy of
the firm carrying out R&D activities, as well as to foreign economies. This idea is
supported by the eclectic approach of foreign direct investment (FDI). That is,
62 H. Ríos-Bolívar and O. Neme-Castillo
firms enjoy a property advantage, because they have created that specific know-
ledge. Moreover, it is inferred that profitability of operating in foreign markets is
greater to the sale or licensing of that technology exclusive for the firm, then effi-
ciently internalising that technological advantage.
In other words, part of knowledge, the most advanced or strategic, generated
by firms that invest productively abroad is excludable in higher grade than the
knowledge generated by other organisms (for instance, private research institu-
tions). However, marginal product of innovative activity in the aggregate
domestic economy grows and, that while this new knowledge is not totally
spread to the rest of economy, it must be recognised that it has impacts on

3.3 Model derivation

This section provides a model that incorporates the essence of theoretical elements
noted above. The model begins from the neoclassical basic production function,
which defines output (Y) as a function of capital (K) and labour (L); Y = Ka(A×L)b,
with α + β = 1. Technological progress (dA/dt) is labour augmenting and Y exhibits
constant returns in all production factors. Population growth is given by the exoge-
nous rate n. Capital accumulation is financed by domestic savings (exogenous
saving rate, s = S/Y), that is, dK/dt = I = S = s·Y. Production function is extended by
incorporating human capital (H) and international trade, understood as capital
goods imports (Z) and the participation of multinational companies (MNCs) oper-
ating in countries of the region (W), through their stock of knowledge:


The idea is that capital goods imports imply a direct transfer of technology.
Developing countries with limited capacities for production of capital goods may
import these goods from technologically advanced countries. Imported capital
stock is modelled as a separate production factor because it is financed through
income and net exports and it likely shows higher levels of productivity than
domestic capital (Bidlingmaier, 2007).
In this sense, exports impact on income only through savings that finance
capital goods imports. The change in imports of that kind or goods in time allows
imported capital accumulation (M), which is formulated as: dZ/dt = M = NX = x·Y,
where NX is net exports, x the participation of net exports in total output
(x = NX/Y); XN and x refer only to capital goods. Capital stock directly imported
is increasing in M financed by NX.
Also, the inclusion of a trade integration variable (W) is based on arguments
similar to that of the incorporation of Z. Technological capital stock of develop-
ing countries is limited. The latter are integrated economically to industrialised
countries with higher production and technological capacities. Particularly,
through productive investment flows and a strong commercial relationship, their
domestic production capacities and, ultimately, their output will increase.
Foreign trade and per capita income 63
Thus, integration represents two options. First, national firms within the value
chain of MNCs that invest abroad, update their technological capacities, generat-
ing a ‘catching up’ effect on the domestic economy. Second, stronger trade links
between these economies mean that MNCs from an advanced country that
employ knowledge, processes, organisational methods, distribution networks,
etc., in domestic production, impact to some extent on product growth rate,
without necessarily disseminating that knowledge and, therefore, without con-
tributing to human capital formation of the domestic economy.
W is included in the model because it depends on foreign investment in
capital formation and in systematic R&D activities, which reflects higher pro-
ductivity levels than the domestic ones. W is understood as expenditures made
by other countries in order to build their production capacities, which are never-
theless exported, at least partially, to less developed economies through FDI and
represents the available stock that helps them to produce more efficiently,
without incurring additional costs.
W is interpreted as the international knowledge spillover restricted to related
MNCs. That is, knowledge of these firms, or part of them, is limited; it does not
flow either with speed or with magnitude one could think if it were a pure public
good. Knowledge spillover is partially restricted to links between matrix and
subsidiaries of MNCs, which affect the production capacity of host economies of
FDI and, ultimately, their growth rate.
Thus, trade integration affects income by two different ways. First, to the
extent agents in the foreign economy are capable to accumulate physical and
technological capital financed by savings. Second, by FDI which operates
through MNCs in the domestic economy, namely, to the extent that activities of
such companies grow in relation to the national market, firms will import part of
this capital stock from their country of origin, with a growing effect in host
country output.
Formally, dW/dt = (dK/dt)·Λ, where s and Y are the rates of savings and
income of the foreign economy (i), respectively, which determine technological
capital accumulation in that country; Λ = (FDIij/VDj), is the diffusion rate of this
knowledge abroad, where FDIij is foreign direct investment from country i to j
and VDj is domestic sales in j. Thus, the second term on the right side represents
the share of MNCs that use their ‘know how’ to produce domestically. The mag-
nitude of the ‘spread’ of physical and technological capital by MNCs in foreign
markets depends on accumulated amount available in their country of origin.
The stock of foreign capital ‘imported’ indirectly is increasing in both terms.
On the other hand, human capital can be accumulated by three different ways:
education investment, ‘learning-by-doing’ and international knowledge spillo-
ver. In the first case, human capital formation through investment in education
implies that the share of income a country invests in education infrastructure
creates human capital (H). The same production function of physical capital
applies to human capital (Mankiw et al., 1992). H is accumulated through
investment in education (Ih) which is financed by saving (Sh), that is,
dH/dt = Ih = Sh = sh·Y.
64 H. Ríos-Bolívar and O. Neme-Castillo
In the second case, H is accumulated through learning-by-doing during the
process of production of each good that faces diminishing returns. The only way
to keep learning at high levels, and thus H formation, is through the continued
introduction of new differentiated goods. This idea is incorporated into the
model by multiplying the stock of human capital formed by investment in educa-
tion, sh, by an index of differentiated goods produced domestically (p):
dH/dt = s·Y·(1 + p), with 0 < p < 1. The more differentiated goods an economy pro-
duces, the greater the value of p and, hence greater H.
The third option for H formation is international spillover. With the advance
of information and communication technologies knowledge is disseminated
beyond national borders. Countries that do not make persistent efforts in R&D
can benefit from new ideas, processes or inventions through knowledge dissemi-
nation. Thus, the more technologically old-fashioned a country is, the greater the
possibility of adapting existing ideas through imitation. This offers the possibil-
ity of ‘catching up’ with countries in the technological frontier by a faster human
capital accumulation. This represents a growth strategy for Latin American
countries. This concept is integrated into the model assuming that absorptive
capacity of knowledge of a country allows accelerated human capital accumula-
tion. Technically, this is done similarly to ‘learning-by-doing’ by adding the
knowledge diffusion factor (m): dH/dt = s·Y·(1 + p)·(1 + m), with, 0 < p < 1 and
0 < m < 1.
However, there are barriers to absorption of knowledge between countries. A
weak communications infrastructure can hinder the flow of knowledge. Internet
access is essential for disseminating global knowledge.2 When an economy has
strong barriers to dissemination and application of global knowledge, W takes
greater importance since the hypothesis of restriction on the dissemination of
knowledge and technology within MNCs and related companies seems to be
The inclusion of variables related to product development (p) and knowledge
diffusion (m) allows considering new channels for H accumulation in a neoclas-
sical approach. Multiplying these variables by traditional variables of human
capital investment (Ih) provides a basis for a more rapid accumulation of H and,
consequently, for a higher growth rate. Thus, in this context the steady-state
level of H depends on both the willingness to save for investing in education,
and on the ability to develop new products and the knowledge absorption
This model explains in detail the level of production compared to traditional
neoclassical model; the basic production function is extended in order to include
three types of capital in addition to physical capital (K); human capital (H),
imported capital (Z) and, unlike Bidlingmaier’s (2007) model, technological
capital ‘available’ from abroad (W).3 Thus, from equation (3.1) the income per
unit of efficient labour is obtained:4

Foreign trade and per capita income 65
Assuming the same depreciation rate (δ) for all types of capital, domestic or
foreign, a growth rate of population (n) and constant growth of technology (g),
the evolution of capital stocks is given by:





The additional increase in capital stock is offset by the depreciation, population

growth and technological progress (n + g + δ). The steady-state values are:5





Substituting these equations into (3.2) gives the equation of income per unit
of effective labour of equilibrium and, after applying the logarithmic function
for the econometric estimation, obtains:


where x = 1 – α – β – γ – η. However, since there are no data of income per unit of

effective labour, the previous expression is rewritten as a per capita income
equation. Thus, as y = Y/AL = ŷ/A ( ŷ: per capita income), and considering that
ln( y) = ln( ŷ) – ln(A), it yields:
66 H. Ríos-Bolívar and O. Neme-Castillo


The technology level, A, is expressed as the product of initial stock, A0, and the
exogenous growth rate, g, and a country specific shock, namely, A = A0·egt+ε. A0 is
country-specific factors such as resource endowments, institutions, business
environment, legal framework, climate, etc. With this, the above equation can be
rewritten as:


Equation (3.13) specifies the factors that determine the level of per capita
income of steady-state: initial stock of technology (A0); exogenous technological
growth rate (g); domestic savings for investment in education (sh) and in capital
(sk); product innovation (1 + p); international knowledge spillover (1 + m) that
increase human capital accumulation; net exports (ex) that finance capital accu-
mulation; foreign savings for investment in technological capital (sw) that stimu-
late the restricted transfer of productive capacities through trade integration;
population growth (n) with negative effect on the level of per capita income; and
capital depreciation (δ).
This specification focuses on the contribution of international trade to accu-
mulation. Trade liberalisation, expressed in (1 + m) and (sw), impacts on accumu-
lation of physical and human capital through knowledge that is disseminated on
the whole economy and that is transferred from abroad through multinational
companies towards the domestic economy. Thus, the level of per capita income
depends on the determinants established in equation (3.13) that are econometri-
cally estimated in the next section for the sample of Latin American and the
Caribbean countries considered in this study.

3.4 Empirical application

Equation (3.13) is estimated following the panel data technique. This equation is
a panel regression model with a general structure that has the form:
yit = αit + βkitXkit + uit, where i refers to countries (i = 1, 2, . . ., 21), t represents time
(t = 1, 2, . . ., 31), k is independent regressors and uit is the error term given by
uit = μt + υit. Where uit is the unobserved individual specific effect, and υit is the
rest of the traditional error, identically and independently distributed, that is,
iid ~ (0, σ2υ). The dependent variable, yi, is a vector of per capita income in equa-
tion (3.13).
Foreign trade and per capita income 67
Similarly, the intercept αit, represents the stock of autonomous technology,
ln(A0). Xkit is a matrix containing all independent variables, k, for each Latin
American and the Caribbean country. In this case, these variables are the loga-
rithms of the growth rates of population and technology plus the rate of depreci-
ation (n + g + δ), of capital accumulation (sk), accumulation of ‘freely’ imported
capital (ex), the accumulation of human capital through education (sh), learning-
by-doing (1 + p), knowledge transfer (1 + w) and capital accumulation in the
foreign countries that spreads ‘exclusively’ through related companies in differ-
ent countries (sw).
Several estimation methods can be used, depending on the variation of the
intercept and the coefficients in time and cross-section units. This chapter con-
siders three cases. The first assumes that the intercepts and coefficients are invar-
iant over time and cross-sections (yi = α + βkXkit + uit), which represents the case of
pooled data. The second case allows cross-section units have different intercepts
(yi = αi + βkXkit + uit). The third possibility considers countries have different inter-
cepts and, further, that they respond in different ways to different independent
variables (yi = αi + βkiXkit + uit). Estimating these three options is of interest because
it first considers the restrictive assumption of the same technology and, after
considering the possibility that variables do not affect per capita income in each
country in the same way, it reflects the possibility of heterogeneity in the explan-
atory variables.
In this regard, the assumption of a common intercept for all countries within
the panel data regression is restricted because it prevents an estimation of the
impact of different initial level of technology (A0) for the countries in our
sample. This hypothesis considers that A0 is the same for all countries; which is
amply questionable because factors like resource endowment, climate or institu-
tional features could be included in this term.
The estimates are done in two ways. On the one hand, by the fixed effects
model (FEM) that introduces a dummy variable for each country that incorpor-
ates individual effects. On the other, through the random effects model (REM),
that divides the error term in a single term, which captures the variance of the
dependent variable and a common term, caused by individual random-effect in
each country.6

3.4.1 Data
The study was performed to a set of 21 Latin American and Caribbean countries
(see Table 3.1). The period considered is 1977–2007 for a total of 31 years. The
sample is a balanced panel. The data on GDP per capita (pcgdp) were taken from
the Penn World Table version 6.2. All terms are expressed by purchasing power
parity. Data of population growth were obtained from World Development Indic-
ators (WDI) 2009. Following Mankiw et al. (1992), it is assumed that rates of
depreciation and of technology growth together are approximately 5 per cent
(g + δ = 5 per cent), which are added to n in order to form the variable popdr. The
expected sign of this variable is negative. In addition, the gross capital formation
68 H. Ríos-Bolívar and O. Neme-Castillo
Table 3.1 Latin America and the Caribbean countries considered in the sample

Observation Country Code Observation Country Code

1 Argentina arg 12 Honduras hon
2 Belize bel 13 Jamaica jam
3 Bolivia bol 14 Mexico mex
4 Brazil bra 15 Nicaragua nic
5 Chile chi 16 Panama pan
6 Colombia col 17 Paraguay par
7 Costa Rica cr 18 Peru per
8 Dominican repd 19 Trinidad and tyt
Republic Tobago
9 Ecuador ecu 20 Uruguay uru
10 El Salvador elsal 21 Venezuela ven
11 Guatemala gua

Source: own elaboration.

as a percentage of GDP (gkf_gdp) is used to approximate the investment in phys-

ical capital stock (sk). The data were taken from WDI 2008. It is expected that
the sign of the estimated coefficient should be positive.
The indicator of the rate of accumulation of directly imported physical capital
(ex) is mm_gdp, and was constructed by dividing manufacturing imports by
aggregate GDP. Manufactured imports include sections 5 (chemicals), 6 (basic
manufactures), 7 (machinery and transport equipment) and 8 (miscellaneous
manufactured goods) of the Standard International Trade Classification revision
1 (SITC). The series of imports were obtained from the database BADECEL of
the Economic Commission for Latin American and the Caribbean (ECLAC) and
were combined with those of the UN Comtrade Data Base of the United Nations
(UN) for countries and years that were missing.7
Besides, the literacy rate (litr) is considered as an approximation of H forma-
tion. Clearly, the expected sign should be positive. The data are in the WDI
2008. The second form of human capital accumulation is through ‘learning-by-
doing’, which is approximated by an index of product discovery. The discovery
does not necessarily mean inventing a new product, but when a country begins
to produce a new good inside its territory, regardless if it was already produced
elsewhere, a learning process starts that creates domestic human capital. Given
the limited data for the whole set of countries that allows to calculate a rate of
introduction of new goods or data related to straight evidence of new goods
produced, this study follows the Bidlingmaier (2007) proposal, which employs
exports as proxies of ‘learning-by-doing’. The idea is that a good which is sys-
tematically exported means a ‘permanent’ discovery that allows the accumula-
tion of knowledge. Thus, the higher the export level, the greater the formation
of H.8
Thus, to identify new discoveries we used the series of exports of SITC revi-
sion 1 disaggregated to four digits (about 1,100 groups of goods), obtained from
Foreign trade and per capita income 69
BADECEL and the UN Comtrade Data Base. The methodology to be assumed
as a new discovery with lasting effects is to identify each product that has been
exported for the first time with a value greater than or equal to $100,000 and that
they have been exported at least during two consecutive years. The variable is
labelled dis and is normalised to obtain values between zero and one. The closer
to the unit dis is, the greater the expected impact on pcgdp.
The third concept of H is knowledge transfer, associated with the notion of
absorptive capacity of a country. It uses two types of proxies of absorptive
capacity: (1) the number of telephone lines and cellular, Internet users and
number of personal computers (indicator of infrastructure) and (2) coefficient of
openness (total trade share in GDP). Data were collected from Social Statistics
and Indicators (BADEINSO) of ECLAC. The variables were normalised between
zero and one, and combined into an index of knowledge absorption capacity
Finally, to partially represent excludable knowledge that is appropriated by
the companies with R&D activities and that only spread to related companies
abroad, a technological capital stock is built for a set of economies with which
the listed countries maintain relevant economic relations. The argument is that
part of technological capital accumulation in advanced economies spreads out to
economies in the region through MNCs.9 That indicator is constructed by
weighting the foreign technological capital stock by the share of each country’s
trade in total trade of a foreign country in order to obtain ‘available’ foreign
technological capital.10 The latter is multiplied by the share of FDI, which flows
from these economies, in the value of domestic sales. Thus the stock of ‘availa-
ble’ foreign technological capital (saftk) entering the domestic economy is
obtained through FDI. Formally:


where stkj is the stock of technological capital in foreign country j, Cij is the total
trade between the i home countries and the j foreign country, Cj is the total inter-
national trade of j, FDIji is foreign direct investment flowing from j to i and DSi
is the domestic sales of i. The data on R&D investment of foreign economies are
derived from the OECD Stan Data Base and are complemented by data from
WID 2008 and national offices. The series of FDI statistics are taken from
UNCTAD and domestic sales of BADEINSO of ECLAC and national statistical

3.5 Results
Three different options were estimated using the pooled, fixed and random
effects models. The Hausman test indicates that the best model is the fixed
effects (eliminating the random effects model). We report the results for the fol-
lowing estimates: (1) the pooled model (Model I), (2) different constants but
with common coefficients (Model II) and (3) different constant and intercepts
70 H. Ríos-Bolívar and O. Neme-Castillo
(Model III). Robust results were obtained for the latter two cases, after correc-
tions to address autocorrelation and heteroskedasticity problems.
Although panel data models tend to show lesser problems of multicollinearity
among the explanatory variables, the possible presence of multicollinearity was
assessed using the correlation coefficient matrix. In 40 per cent of relationships,
the highest correlation coefficient is less than 0.50. Further, only in 45 per cent
of relationships, half of the coefficients is above 0.50, and less than 10 per cent
is higher than 0.85. Then, it is accepted that there are no serious problems of
multicollinearity: the correlation coefficient is less than 50 per cent for 40 per
cent of the relationships, above 50 per cent in less than 25 per cent of the rela-
tionships and higher than 85 per cent for less than 10 per cent of them. There-
fore, we conclude that there are no serious problems of multicollinearity.
Given that panel data estimations often violate the condition of independent
errors and identically distributed with constant variance, an autocorrelation and
heteroskedasticity test was conducted. We applied the Wooldridge test for diag-
nosis of autocorrelation, whose null hypothesis is the non-existence of autocor-
relation. In fact, this test rejects the null, indicating that there is a problem of
autocorrelation.12 Also, to determine the presence of heteroskedasticity two dif-
ferent tests were considered. First, the Modified Wald test for heteroskedasticity
test indicates the rejection of the null hypothesis of constant variance and, there-
fore, the existence of a heteroskedasticity problem.13 This result was confirmed
by the LM test suggested by Greene (2000), which rejected the null of no
In order to solve both problems encountered, transformation of Prais–Winston
by country was used following the Panel Corrected Standard Errors Method
(PCSE). In this respect, Beck and Katz (1995) have demonstrated that standard
errors of PCSE are more accurate than errors estimated with Feasible General-
ised Least Squares (FGLS). After this correction the problems were eliminated.
In general, the explanatory power of the three models is high, implying that other
factors not included in the model have limited effects on the income levels of
Latin American and the Caribbean countries. On average, the regressors con-
sidered explain more than 91 per cent of income changes.
Results of the estimations (1) and (2) are shown in Table 3.2. All variables in
both models have statistical significance but not all of them show the expected
sign. In Model I, fixed effects with common constants and slopes across coun-
tries, all variables are significant at 99 per cent. In Model II, fixed effects with
equal slopes but differentiated constants across economies, all variables are
significant at 99 per cent except the literacy rate which is significant at 95 per
cent and the rate of population growth, which is not significant.
In Model I, for all countries, population growth, technological progress and
depreciation jointly affect negatively the per capita income, as expected. By con-
trast, the sign found in Model II is positive, though with not statistically signific-
ant contribution. The accumulation of physical capital is an element that
positively determines this variable in both models. In the first model, it is by far
the variable that contributes the most to income; in the second specification it
Table 3.2 Panel data regressions of per capita income

Variable Models Country Differentiated SE Country Differentiated SE

(code) constants1 (code) constants1

C 8.0474* 8.4405* arg 9.6643 0.0585 hon 7.4955 0.0404

(0.0523) (0.0176) bel 8.2154 0.0791 jam 8.1606 0.0331
popdr –0.3701* 0.0190** bol 7.4571 0.0360 mex 9.1042 0.0403
(0.0188) (0.0090) bra 9.3644 0.0468 nic 7.2945 0.0395
gkf_gdp 0.3060* 0.0467* chi 8.9427 0.0510 pan 8.7266 0.0506
(0.0242) (0.0110) col 8.3309 0.0268 par 7.8788 0.0646
mm_gdp –0.1032* 0.4057* cr 8.5613 0.0325 per 8.4592 0.0433
(0.0115) (0.0099) rdom 8.4048 0.0475 tyt 9.3529 0.0566
litr –0.0711* –0.0019*** ecu 8.0375 0.0358 uru 9.1477 0.0440
(0.0037) (0.0013) elsa 7.9374 0.0375 ven 8.9786 0.0398
dis –0.0675* –0.0080* gua 7.7354 0.0346
(0.0049) (0.0014)
ikac 0.2349* 0.1629*
(0.0128) (0.0053)
saftk –0.0179* 0.0084*
(0.0021) (0.0012)
R2 adjusted 0.844 0.993
F 503.27 3264.49
Observations 651

Source: own elaboration.

1 Differentiated coefficients by countries in accordance with Model II and all constants are significant at 1%.
Model I: same constants and coefficients for all countries, and Model II: differentiated constants.
Model I: pooled data model, and Model II: fixed effects model with different intercepts.
*, ** and *** are the levels of significance at 1%, 5% and 10%, respectively. SE: standard errors that appear between brackets and below coefficients.
72 H. Ríos-Bolívar and O. Neme-Castillo
appears as the third variable with the largest contribution. In general, this reflects
the central role of capital accumulation in the level of per capita income.
The coefficient of foreign physical capital, employed in the domestic
economy and purchased directly, has an ambiguous effect on income. In Model
I, the sign is contrary to the expected one, which means import of capital goods
has a negative effect on income levels but relatively low level. This may be due
to two reasons: (1) countries with low levels of development (low per capita
income) have limited capacity to absorb the new technology (low human capital)
and this situation is greater than the positive effect on the more developed coun-
tries and (2) the variable is not a good proxy for physical capital, so it can be
replaced by more specific variables such as more disaggregated sections of SITC
included in the manufacturing imports. However, the results using OLS esti-
mates do not produce the best linear unbiased estimator, due to the specific indi-
vidual effects, which unequally affect individual countries and are invariant over
time. For this reason the model is estimated by fixed effects.14
In contrast, for Model II, capital goods imports are, as expected, significantly
positive for income. Slightly less than half of the efforts of technology purchas-
ing from abroad are reflected in the income of each country. Therefore, no matter
the size of each economy or of individual capacities of countries, the use of more
advanced technology seems to bring positive effects for the economic develop-
ment of countries. In any case, the relationship between these two variables
should be examined more extensively to eliminate the degree of ambiguity found
in these two models.
Furthermore, increase of literacy rates, contrary to expectations, negatively
affects the income level of Latin American and the Caribbean countries, suggest-
ing that favourable impact of human capital and in particular the level of educa-
tion, is contained in other factors related to H as ‘learning-by-doing’ (dis) or
knowledge transfer (ikac).
The negative sign means that increasing education of general population does
not recover the investment made therein. More people who can read and write
means, on average, lower income levels, suggesting the existence of a ‘literacy
trap’. Improvements in the way reading and writing is learned, on one side, and
acquisition of other skills related to education and work, on the other, could
eliminate this perverse circle. The above argument is subject to further explora-
tion using other proxies, and carrying out other studies at sector and individual
country levels, which could reduce the level of that trap.
Moreover, the variable dis was introduced to measure the impact of ‘learning-
by-doing’ in per capita income. The sign found in both models is negative, indi-
cating that for all countries of the region, the discovery of new activities
adversely affects these economies: the greater the number of products exported,
the lower the per capita income. One possible reason for this result lies in the
way this indicator was built. It is difficult to determine exactly when there is a
new discovery. The new exported goods could be produced domestically for
some time before they were exported for the first time, implying that learning
rates were reached, laying the bases and skills to start the export process, so that
Foreign trade and per capita income 73
the effect of discovery could be dissipated over time. Moreover, the range of
$100,000 and two years was set arbitrarily, and does not include adjustments for
inflation, which could consider a larger number of discoveries when in fact they
are not reflected in the sign found. This result is subject to further study consid-
ering, for example, individual differences in dis for each country.
In contrast to the two previous proxies, the composite index of technology
transfer, ikac, formed by elements of technology infrastructure and international
trade, has the expected positive sign and is statistically significant in both
models. The variables within ikac are crucial for the dissemination, acquisition
and international knowledge absorption and therefore increase the stock of
human capital and income of countries. Certainly, not all aspects of technology
transfer are in ikac, such as institutional factors, however a significant portion is
captured in such a way that is meaningful and is, in fact, the second largest con-
tribution to income of Latin American and the Caribbean countries in either
Finally, contrary to theoretical approach, the coefficient of the stock of ‘avail-
able’ foreign technological capital (saftk) appears in the first specification with a
negative sign, which is due to a number of reasons. First, the knowledge gener-
ated abroad gathered in this variable measures the part which is used in domestic
economies just for foreign companies and that, at least for a while, kept without
diffusion to other firms, creating a temporary monopoly or some distortion in the
market structure which impacts negatively on aggregated production. Second, in
the case of Latin America, FDI may not be a disseminator of knowledge and, on
the contrary, widens the technological gap vis-à-vis advanced countries, because
FDI is oriented to production stages which are not capital or knowledge inten-
sive, generating, ultimately, lower levels of income.
However, in the model with differentiated intercepts, the coefficient saftk is
significant and positive, albeit low. Anyway, in general terms, the exclusive
advantage MNCs own in their home market (skills, knowledge, resources, spe-
cific technology, etc.) and that used in some degree in their branches abroad,
tend to improve income in the Latin American and the Caribbean countries.
While the estimated coefficient is low (0.0084), it should be noted that this
technological stock is not built with a view to benefit foreign economies and,
nevertheless, it ultimately favours income. Finally, Model II shows that the
higher constants are linked to the major economies, while lower intercepts are
associated with smaller economies.15 It should be recognised that one limitation
of the current model is that all constants are significant and of similar
Since the panel regression results are not totally satisfactory we tried to miti-
gate the problem of the assumption of a common intercept among countries by
introducing macroeconomic dummies. The idea is that initial technology, repre-
sented by term A0, could be similar within the region but different across coun-
tries, considering the market size and the necessities derived from this.
The countries sample was divided into three groups for each year, depending
on the level of GDP (in current dollars) and the average GDP of each group was
74 H. Ríos-Bolívar and O. Neme-Castillo
obtained. Those countries above the average of the first group were assigned a
value of 3 (D4); those above average of the second group but below the average
of the first group were assigned a value of 2 (D3); countries below the second
average but above the third average were given a value of 1 (D2); and countries
below the average of group three were given a value of 0 (D1). Only the dummy
for D1 is statistically significant at 90 per cent of confidence with positive sign
(results not reported). It could be that the variance of initial technology, related
to market size, in the other groups was so big that these variables did not produce
significant results. Nevertheless, with the additional macroeconomic dummy the
fit of the model did not improve at all.
Consequently, a third model was considered: a fixed effects model with dif-
ferentiated constant and intercepts. In general, results remain roughly equal to
that of Model II (see Table 3.3). The explanatory power is high (R2 is 0.992) and
a high number of coefficients are statistically significant (less than 20 per cent of
the regressors are not significant). In this sense, the overall significance is good.
For any variable, at least 67 per cent of the estimated coefficients are significant.
The coefficient of the population growth rate (popdr) has the expected sign
only in four countries (Bolivia, Costa Rica, Panama and Peru). In contrast, the
positive sign of this coefficient for most countries in the region is based on the
fact that incorporation of labour to productive activity generates a more than
proportional increase in the product. Thus, accumulation of capabilities by
workers (L) favours income. Note that popdr has no significant effect in Brazil
and Chile, two of the largest economies in the area.
Also, for most countries the capital stock has negative repercussions on the
pcgdp. Argentina is a case in point, with an estimated ratio of –0.828, probably
influenced by the strong imbalances experienced at the beginning of this decade.
In addition, for Brazil and Trinidad and Tobago, the economies with higher
levels of GDP and GDP per capita, respectively, this variable is not significant,
likely as a consequence of the relatively inadequate levels of capital goods
In contrast, trade liberalisation, represented by imports of capital goods, is
remarkably significant for the economies of the region. With the exception of
Bolivia, Nicaragua and Paraguay, the directed imported capital stock contributes
positively to per capita income. In this respect Chile and Panama excel, which
means that their economies have assimilated to a greater extent the impact of this
foreign technology through ‘spillover’ effects.
The result regarding the literacy rate by this specification is similar to that
found in Model II. Most of the coefficients show negative signs, although of low
level. In this sense, the existence of a ‘trap of literacy’ is confirmed. Countries
that have managed to escape from this are Chile, Ecuador, Mexico, Nicaragua
and Paraguay.16
Another element that contributes to formation of H is ‘learning-by-doing’
(dis), this is statistically significant with a negative sign for 10 out of 21 coun-
tries, among which are Chile, Mexico and Venezuela. It can be argued, however,
when the relatively large participation of foreign capital in these countries is
Foreign trade and per capita income 75
considered over the period, that foreign companies do not generate positive
externalities for domestic productivity.
However, this indicator does not appear to be a good approximation of learn-
ing effects, since the results are significant with positive sign, as indicated by the
theory, just for four economies (Belize, Costa Rica, Panama and Paraguay).
Anyway, it seems that there is not a strong relationship (positive) between the
discovery of new activities and per capita GDP in the countries of the region.
The third element inside the stock of human capital is technological transfer
measured by ikac; it includes technological infrastructure and commercial
exchange with other countries. Results are varied: six coefficients are not signi-
ficant (Colombia, Ecuador, Guatemala, Honduras, Jamaica and Venezuela) while
five are negative (Argentina, Brazil, Mexico, Nicaragua and Paraguay). This
result is the effect of two factors: first, inadequate capacities to meet the techno-
logical needs of the productive and social sectors in these countries (compara-
tively, the first three countries maintain the lowest levels of penetration in mobile
phones, number of computers and of Internet users); and, second, as they keep
trade openness rates relatively high it means that these economies have not
absorbed purchases of foreign capital goods into local production processes or,
on the other hand, sales to abroad are of low technological content or are focused
on markets of low dynamism that, ultimately, means few processes of technolo-
gical transfer and of human capital formation. In contrast, for ten countries this
index is significant and positive, indicating the favourable effect on pcgdp.
Finally, the effect of the stock of foreign technological capital ‘available’ for
the Latin American and Caribbean countries through the MNCs and that of the
stock the latter keep on their property and exploit directly, seem to be ambiguous
but significant. Only one coefficient has no statistical significance (Chile). Half
of coefficients have a low positive effect on income (the highest coefficient is
0.20). Among the beneficiaries are Brazil, Colombia and Mexico. This verifies
the idea that the more closely linked domestic economies are to foreign econo-
mies the greater the profits in the domestic economy. That is, higher inflows of
FDI, international trade and presumably intra-industry trade in medium and
high-tech sectors (as in the specialisation of Brazil and Mexico) have real effects
on income.
In contrast, this technological stock affects negatively the income of Argen-
tina and Costa Rica, suggesting that foreign firms in these markets internalise the
advantage of technological knowledge they have in their country of origin. In
other words, the technological stock of MNCs enables them to achieve extra-
ordinary benefits in these countries, possibly through mergers and acquisitions,
affecting the market structure of these countries, by crowding out productive
investments of domestic firms and, consequently, the level of per capita income.

3.6 Conclusions
This chapter derived an extended neoclassical growth model that incorporates
factors such as population, physical capital, domestic and imported, human
Table 3.3 Panel data regressions of per capita income

Differentiated constants and coefficients: Model III

Country c popdr gkf_gdp mm_gdp litr dis ikac saftk

arg 9.7078* 0.2433* –0.8277** 0.4090* 0.0256 –0.0113 –0.1220* –0.0050*
(0.2929) (0.0834) 0.1423 (0.03666) (0.0212) (0.0119) (0.0449) (0.0140)
bel 9.1324* –0.0393 –0.0923* 0.3136* –0.0238* 0.1206* 0.5117* –0.2038*
(0.2141) (0.0344) (0.1085) (0.1087) (0.0092) (0.0283) (0.0403) (0.0106)
bol 7.2814* –0.2144* –0.3725* 0.4959P –0.0088* –0.0030 0.1066* –0.0124*
(0.1351) (0.0147) (0.0632) (0.0291) (0.0026) (0.0055) (0.0214) (0.0044)
bra 8.8924* 0.0313 –0.9477 0.4249* –0.0113* –0.0071 –0.0418*** 0.0926*
(0.2976) (0.0340) (0.1835) (0.0307) (0.0042) (0.0100) (0.0309) (0.0133)
chi 10.0206* –0.0298 –0.0730* 0.9837* 0.0253* –0.0554* 0.3247* –0.3146
(0.1803) (0.0250) (0.1034) (0.0256) (0.0054) (0.0066) (0.0289) (0.0089)
col 9.3485* 0.0399** 0.0437* 0.4583* –0.0060*** 0.0004 0.0090 0.0018***
(0.1159) (0.0201) (0.0555) (0.0234) (0.0038) (0.0031) (0.0248) (0.0047)
cr 9.6667* –0.0906* 0.3300* 0.5304* –0.0156*** 0.0411* 0.0705* –0.0244***
(0.1951) (0.0275) (0.0884) (0.0300) (0.0104) (0.0080) (0.02143) (0.0161)
repd 9.0590* 0.1045* 0.3147 0.5996* 0.0079 –0.0156 0.0857** –0.0150*
(0.3106) (0.0356) (0.1319) (0.0570) (0.0103) (0.0205) (0.0460) (0.01005)
ecu 7.1244* 0.1205* –0.3452* 0.3794* 0.0217* –0.0114** –0.0146 0.0814*
(0.1680) (0.0226) (0.0577) (0.0446) (0.0051) (0.0051) (0.0229) (0.0050)
elsal 7.5904* 0.2263* –0.5013*** 0.6179* –0.0231* –0.01230** 0.1628* –0.0548*
(0.1417) (0.0225) (0.0655) (0.0268) (0.0042) (0.0064) (0.0190) (0.0057)
gua 6.5755* –0.0743 –0.3831* 0.3574* 0.0041 –0.02556* 0.0012 0.0569*
(0.1530) (0.0180) (0.0721) (0.0327) (0.0053) (0.0060) (0.0212) (0.0186)
hon 7.5288* –0.0195 –0.3700* 0.2315* –0.0022 –0.0047 0.0349 –0.1313*
(0.2145) (0.0166) (0.0633) (0.0448) (0.0095) (0.0076) (0.0327) (0.0179)
jam 8.3161* 0.0113 –0.6976 0.7180* –0.0150* –0.0058 0.1568 –0.0841*
(0.1754) (0.0187) (0.0785) (0.0339) (0.0051) (0.0105) (0.0177) (0.0140)
mex 9.2575* 0.2576* 0.7463** 0.3189* 0.0184** –0.0186*** –0.1305* 0.1541*
(0.3605) (0.0342) (0.1457) (0.0199) (0.0109) (0.0122) (0.0188) (0.0232)
nic 6.8427* 0.1779* 0.1879* –0.0552 0.0358* –0.0262* –0.0864* 0.0325*
(0.1032) (0.0264) (0.0680) (0.0803) (0.0076) (0.0079) (0.0200) (0.0023)
pan 9.0973* –0.0734* –0.4125* 0.8416* –0.0919* 0.0245** 0.4103* –0.0745*
(0.1741) (0.0277) (0.1318) (0.0711) (0.0123) (0.0117) (0.0460) (0.0107)
par 7.9701* 0.5099* 0.1991* –0.006 0.0739* 0.0227*** –0.3591* 0.1246*
(0.3238) (0.1502) (0.2714) (0.0322) (0.0141) (0.0176) (0.0306) (0.0100)
per 8.8726* –0.1100* –0.2470* 0.6510* –0.0399* –0.0853* 0.1736* –0.0373*
(0.1598) (0.0271) (0.1224) (0.0376) (0.0149) (0.0114) (0.03035) (0.0084)
tyt 9.2489* 0.0452* –0.1570 0.7440* –0.0545* –0.0122*** 0.2666* 0.0461*
(0.1300) (0.0178) (0.0838) (0.0495) (0.0138) (0.0087) (0.0399) (0.0095)
uru 9.9290* –0.0091 0.2891* 0.7674* –0.0840* –0.0961* 0.0996* 0.0262*
(0.0877) (0.0369) (0.0429) (0.0315) (0.0120) (0.0133) (0.0325) (0.0092)
ven 8.6731* 0.1304* –0.2806** 0.5766* 0.0069 –0.0408* 0.0452 0.0400*
(0.2284) (0.0389) (0.1261) (0.0558) (0.0147) (0.0111) (0.0498) (0.0075)
R2 adjusted 0.992
F 497.56
Observations 651

Source: own elaboration.

*, ** and *** are the levels of significance at 1%, 5% and 10%, respectively. Standard errors appear between brackets and below coefficients.
78 H. Ríos-Bolívar and O. Neme-Castillo
capital (direct learning, ‘learning-by-doing’ and product discoveries) and techno-
logical capital that foreign firms employ in the Latin American and the Carib-
bean markets as a consequence of international trade. This is an alternative way
to determine the importance of international trade based on neoclassical func-
tions of income. This model was applied to the study of 21 countries in Latin
America and the Caribbean during 1977–2007 following a panel data
Regarding human capital, a strong link with income, when structural differ-
ences across countries are considered, is observed. With regard to foreign
technological stock, it appears that, on average, exclusive technology of foreign
firms located in the Latin American and Caribbean countries boosts per capita
GDP. However, differentiated estimations are ambiguous, although it seems that
the positive effect tends to be present in the larger economies or with greater
levels of openness.
About the hypothesis that international trade impacts on income levels, it is
fulfilled in general terms in the countries of the region, although this assertion
must be weighted by country and variable. A notable result is that for most coun-
tries the stock of physical capital negatively affects pcdgp, mainly in Argentina.
By contrast, imports of capital goods, positively impact on the economies of the
region (except for Bolivia, Nicaragua and Paraguay), particularly in Chile and
Panama, which have assimilated foreign technology to a greater extent. Thus,
the argument that countries can import capital goods they are unable to produce
domestically and then benefit in terms of productivity is confirmed.
In addition, a negative sign in the relationship between the literacy rate and
income was found, so there is a ‘trap of literacy’, except for Chile and Mexico.
In contrast, the variable ‘learning-by-doing’, approximated by the new products
exported, is statistically significant with a negative sign for half of the countries,
including Chile, Mexico and Venezuela. Then it can be argued that discovery of
new products or processes (by mere imitation), reduces per capita income in
these countries. Regarding intangible transfer of knowledge, which affects pro-
ductive and organisational processes, it plays an important role in the accumula-
tion of knowledge, human capital and the subsequent income level. It should be
remarked that capabilities such as infrastructure and international trade that
allow absorption of intangible transfer of knowledge also play an important role.
In conclusion, international trade has effects on income, although not all the
effects identified by theory.
Finally, the inclusion of a variable derived from the region’s opening to
foreign productive capital allowed expanding the model to measure the effect of
foreign technological capital stock ‘available’ to the region. The results indicate
a strong relationship with income but the sign of that relationship is not clear.
They prove that the greater the openness to foreign economies, the greater the
effect on domestic income.
Foreign trade and per capita income 79
1 Growth literature calls this situation level-effect, in contrast with growth-effect that is
achieved when A, on production function, is growing steadily.
2 Other ways of knowledge dissemination are FDI and imports of advanced technology
goods. Thus, the openness of a country, in terms of removing barriers to FDI or
tariffs, is essential for global knowledge dissemination within an economy. However,
the perspective in this document implies that some knowledge of MNCs is not dis-
seminated to domestic economies, so the impact they may have on human capital for-
mation is lower.
3 The function exhibits neoclassical conventional characteristics like diminishing
returns in all production factors and constant returns to scale.
4 kt = K/AL is the stock of physical capital per effective unit of labour, ht = H/AL denotes
the stock of human capital per unit of efficient labour, zt = Z/AL is the stock of directly
imported capital per unit of efficient labour, wt = W/AL is the foreign technological
capital ‘available’ to produce from the domestic economy for each unit of efficient
labour and, yt = Y/AL denotes the income per unit of efficient labour, all valid at time t.
5 The steady-state values of k, h, z and w are derived from equations (3.3), (3.4), (3.5)
and (3.6) setting them to zero and solving simultaneously. Thus, a system of four
equations with four unknown variables is formed. Solving this system of equations
yields the steady-state values (k*, h*, z*, w*) expressed as a function of the independ-
ent variables sk, sh, sw, and of parameters (1 + p), (1 + m), n, g and δ.
6 The choice between FEM and REM is made using the Hausman test. There is a third
specification that can be estimated by restricting pooled panel data, where errors, uit,
are independent between time, and individual units have zero mean and constant vari-
ance, which would represent the traditional case of regression model that can be esti-
mated by ordinary least squares. The test for determining whether the best
specification is a pooled model or FEM is the F test; while the LM test is used for
determining the best model between the pooled and REM ones.
7 For Venezuela and Paraguay there is no available data on BADECEL for the years
2006, 2007 and 2008, while for Trinidad and Tobago the series was discontinued
from 2004.
8 It should be noted that there is a time lag between discovery and the start of export
activity, but it can be argued that when a good is exported for the first time, the country
went through a learning process to produce more efficiently, with lower costs, best qual-
ities or different varieties that ultimately, enables the country to produce domestically.
9 The natural way to represent this knowledge partially excludable by multinational
companies are data related to the presence of MNCs in Latin American economies.
However, these series are neither available for all countries nor for the entire period
of interest.
10 The stock of technological capital is calculated through the perpetual inventory
method using the expenditures on R&D, that is: stkt = (1 – δ)stkt–1 + It–1, δ is the depre-
ciation rate, It–1, the investment in R&D in the previous period and skt–1 the stock of
technological capital in t – 1 that is obtained as skt–1 = [(1 – δ)sk0 + i]/[(1 + υ)], where i is
the ratio of investment in R&D to product (GDP), υ the growth rate of investment and
sk0 the initial stock of technological capital.
11 Countries included to create the foreign technological stock are: USA, Germany,
France, UK, Spain, Italy, Netherlands, Belgium, Portugal, Greece, Switzerland,
Sweden, Norway, Finland, Denmark, Australia, New Zealand, Korea and Japan; Latin
American countries register a higher degree of trade integration with these
12 The value of the test was 117.64 and the p-value was 0.0038.
13 For the first test, the estimated value of the test was 1.6E12 and the linked p-value was
nearly 0.00006 and for the latter, the estimated value was 2062.5.
80 H. Ríos-Bolívar and O. Neme-Castillo
14 In fact, a per capita income equation was estimated, in addition to the pooled model,
following the fixed and random effects models. The Breusch-Pagan test was per-
formed to discriminate between REM and OLS, rejecting the null hypothesis which
points out there are not significant differences between these models. Thus, the relev-
ance of the random effects is accepted. Also the F test was applied in order to deter-
mine the significance of fixed effects; the results rejected the null of absence of fixed
effects. Both tests indicate that the model is candidate to be estimated taking into
account fixed or random effects. Accordingly, the Hausman test was performed,
finding that the best estimate is the fixed effects one. The test results are not reported
but are available upon request via email.
15 As the constants reflect the effect of variables not included that precisely make the
difference across countries, it is argued that probably variables such as domestic
market size, scale economies, poverty and inequality levels, and internalisation,
among others, play a leading role in the level of per capita income; thus, an extension
of this document should go in that direction.
16 The estimated coefficient is not statistically significant for Argentina, Dominican
Republic, Guatemala, Honduras and Venezuela, so that the ‘trap of literacy’ does not
apply in these cases either.

Beck, N. and Katz, J.N. (1995) ‘What To Do (and Not To Do) with Times-Series Cross-
Section Data’, American Political Science Review, vol. 89, no. 3, pp. 634–647.
Bidlingmaier, T. (2007) ‘International Trade and Economic Growth in Developing Coun-
tries’, Dynamics, Economic Growth, and International Trade, DEGIT – XII. Mel-
bourne, Australia, June 2007.
Gomory, R. and Baumol, W. (2001) Global trade and conflicting national interests. Cam-
bridge, MA: MIT Press.
Greene, W. (2000) Econometric analysis (5th edn). New York: Macmillan Publishing.
Lucas, R. (1988) ‘On the mechanics of economic development’, Journal of Monetary
Economics, vol. 22, no. 7, pp. 3–42.
Lucas, R. (1993) ‘Making a miracle’, Econometrica, vol. 61, no. 2, pp. 251–272.
Mankiw, N., Romer, D. and Weil, D. (1992) ‘A contribution to the empirics of economic
growth’, Quarterly Journal of Economics, pp. 407–437.
Redding, S. (1999) ‘Dynamic comparative advantage and the welfare effects of trade’,
Oxford Economic Papers, vol. 51, no. 1, pp. 15–39.
Rivera-Batiz, L. and Romer, P. (1991) ‘International trade with endogenous technological
change’, European Economic Review, vol. 35, no. 4, pp. 971–1,004.
Romer, P. (1990) ‘Endogenous technological change’, Journal of Political Economy, vol.
98, no. 2, pp. 71–102.
4 Regional integration and its
effects on inward FDI in
developing countries
A comparison between North–South
(Mexico) and South–South (Brazil)
Thomas Goda

4.1 Introduction
During the last 15 years, there has been a proliferation of regional integration
agreements (RIAs),1 mostly because of massive increases in North–South2 agree-
ments but also because of more South–South3 agreements (Fiorentino et al.,
2007). There are many different motives why more and more RIAs have been
signed. Probably the most important motivation for developing countries is the
hope to develop economically by achieving higher growth rates, i.e. to catch up
with already developed countries. According to the new growth theory, higher
growth rates might result from openness to trade and foreign direct investment
(FDI) inflows, among other factors. Thus, for developing countries one import-
ant reason to engage in a RIA, besides market access, is the expectation that
foreign direct investment inflows will increase. In contrast, important motives
for developed countries to promote RIAs with developing countries seem to be
the aim to lock-in liberalisation reforms in these countries (i.e. to get market
access and resource access) as well as to circumvent multilateral negotiations
which show slow progress at the moment (Schiff and Winters, 2003).
These new motives have led to a new type of RIA, which involves not only
deep trade but also deep investment provisions – to attract FDI inflows – and is
part of a process called ‘new regionalism’. Another characteristic of ‘new
regionalism’ is that, although the provisions favour member countries, external
barriers to trade, services, and capital often are also lowered, with the aim of
integrating the region in international production networks. Therefore, the focus
is not any longer exclusively regional (as it has been in the ‘first regionalism’
agreements) but also global (Schlageter, 2005).
Although regulations in North–South and South–South integration agree-
ments are similar, there is a dispute in the literature about whether North–South
integration is more beneficial than South–South integration. According to
UNCTAD (2007a) there are good reasons to disfavour the engagement of devel-
oping countries in a North–South integration agreement. Perhaps the most
82 T. Goda
important problem is the negative ‘impact of these agreements on . . . [the ability
of developing countries] to use alternative policy options and instruments in the
pursuit of a longer term development strategy’ (ibid., p. 65). Thus, developing
countries should cautiously calculate the gains and risks from North–South inte-
gration to avoid rushing into integration agreements with possibly overrated
One argument why North–South integration might be preferable is that FDI
inflows in a North–South agreement can be expected to be higher than in a
South–South agreement (Schiff and Winters, 2003). However, even though com-
prehensive literature exists on (1) FDI and its possible impacts on developing
countries and (2) the effects of RIAs in general, unfortunately, not much research
has questioned whether the signing of a North–South agreement has different
consequences on inward FDI for developing countries than the signing of a
South–South integration agreement. Most of the existing literature about the
effects of RIAs is devoted to the impacts on trade, especially trade diversion and
The objective of this chapter is to analyse and discuss which impact North–
South and South–South RIAs probably have on FDI inflows in developing coun-
tries, i.e. to find out if North–South agreements really attract more FDI as
conventional wisdom declares. As an example of a developing country in a
North–South agreement Mexico – member of the North American Free Trade
Agreement (NAFTA) – is taken, and as an example for South–South integration
Brazil – member of the Mercado Común del Sur (Mercosur). Both countries are
suited very well for a comparison because they have a huge growing internal
market, cultural similarities, similar endowments, and they experienced rela-
tively high FDI inflows before they signed and implemented an integration
agreement in the 1990s.4 In addition, Mexico and Brazil witnessed structural
adjustment programmes in the 1980s and 1990s, respectively.
This chapter is structured as follows: in the first section likely impacts of
North–South and South–South RIAs on inward FDI in developing countries and
more specifically on Brazil and Mexico are discussed. In the second section, an
empirical evaluation on the basis of recent FDI inflow and FDI stock data for
Brazil and Mexico is made, before the chapter ends with a conclusion of the

4.2 Theoretical considerations: regional integration

agreements and possible effects on inward FDI

4.2.1 The possible static effects

As already discussed in the introduction, ‘new regionalism’ is expected to lock-
in reforms and to push further liberalisation. On the one hand, RIAs contain pro-
visions which increase the security for investors who are located inside the
region and as a result promote FDI flows within the region. These investment-
friendly regulations normally consist of most favoured nation rules, national
North–South and South–South comparison 83
treatment for regional investors, protection of intellectual property rights, unre-
stricted capital flows, rules regarding dispute settlement, and the non-existence
of performance requirements (Medvedev, 2006). On the other hand, ‘new
regionalism’ often also lowers discriminatory barriers against the rest of the
world (ROW), i.e. external barriers to trade and capital. As a result, MNCs are
more likely to integrate the region in their global production chain and build new
affiliates to exploit the locational advantage of the region (te Velde and Bezemer,
Accordingly, Medvedev (2006, p. 2) states: ‘[T]he most immediate impact of . . .
[RIAs] on foreign investment is the direct response of FDI to the implementation
of various investment provisions in an agreement’. While Dunning (1997) argues
that the integration of regions, i.e. the harmonisation of standards and increased
investor security, is necessary for countries to attract FDI in the future. This is due
to the fact that companies – in order to become more efficient and competitive –
increasingly undertake joint ventures. Thus, countries/regions that do not support
this new alliance character of capitalism run the risk of being left out.
However, there is an emotional discussion if investor protection and the lock-
in of liberalisation reforms in the course of RIAs are really that important to
attract FDI because a more predictable investment climate seems to be only deci-
sive for new investment inflows if ex ante the conditions for investors have been
unpredictable (te Velde and Bezemer, 2004) and in most cases restrictions for
investors have been lowered considerably and reforms have been undertaken
before RIAs have been signed. Accordingly, liberalisation policies and provi-
sions regarding investor protection in the course of RIAs perhaps are the least
important conditions for increasing inward FDI flows (UNCTAD, 1998). Never-
theless, after the implementation of RIAs the liberalisation of some sectors might
accelerate because countries that have deregulated less than other member coun-
tries might be pressured to open up some sectors more radically Therefore, RIAs
might be important for FDI inflows because they can open up the service sector
– which in most cases consist of non-tradables and accordingly only can be
served by foreigners through FDI (Medvedev, 2006).
Another reason why inward FDI might increase is that production processes
for goods that are aimed to serve the domestic market are different from produc-
tion processes for goods which serve a regional market.5 Therefore, due to
regionalism, such FDI inflows are encouraged which are aimed at the restructur-
ing, i.e. modernisation and enlargement, of existing plants (IADB, 1998). In
addition, because of harmonisation and lower trade barriers regionalism contains
better possibilities for investors to build clusters for the production of similar
goods or goods that consist of similar inputs, respectively. In other words,
agglomeration economies can take place regionally (i.e. between bordering
countries) and are not limited to one country anymore. Hence, regions become a
more attractive location for FDI (UNCTAD, 1998).
However, the static effects of RIAs on FDI depend on the pre-existing situ-
ation (like pre-existing barriers) and factor endowments. Therefore, the effect
might be negative for single countries within the region, while for the region as a
84 T. Goda
whole it is likely to be positive (UNCTAD, 1998). For example, if a country has
no locational advantage in comparison to the other member countries and ex ante
was a host to a lot of ‘tariff-jumping’ FDI – which tried to circumvent tariff bar-
riers – the removal of such barriers would most probably lead to a decreasing
stock of FDI in that country. In contrast, if a country has attractive factor endow-
ments and the economy was not closely linked with the new partners before, the
positive effect of a RIA on new FDI inflows presumably is significant (Blom-
ström and Kokko, 1997). Furthermore, some member countries might witness
the relocation of existing FDI stocks to other members, if ex ante MNCs had
market-seeking FDI in more than one member country and ex post relocate their
investment to one location which they use as platform to serve the regional
market to exploit economies of scale (Levy Yeyati et al., 2003).
Next, with respect to factor endowments and the pre-existing situation, it
should be noted that investment provisions, the opening-up of sectors, non-tariff
barriers within the region, and external barriers can be seen as important deter-
minants for FDI inflows and changes in FDI stocks. In addition, the regional
transport and communication infrastructure is a crucial factor because without
sufficient infrastructure transaction costs are so high that platform investment
will not take place and (regional) production networks will not emerge
(UNCTAD, 1998). In addition, the size of the individual market and plant-level
economics of scale still play an important role in the decision making process if,
and to what extent, this relocation will take place (Levy Yeyati et al., 2003).
However, if one wants to examine the effects of RIA on inward FDI more
closely, it makes sense to divide the effects into intra-regional effects6 and inter-
regional effects7 because ‘insiders’ and ‘outsiders’ will be treated differently
after the RIA comes into existence.

4.2.2 The intra-regional effects

There are two views about the relationship of trade and intra-regional invest-
ment. One view postulates that FDI and trade are substitutes. MNCs might serve
the regional market through FDI because of barriers to exports, therefore, the
abolition of tariffs is supposed to lead to disinvestment of ‘tariff-jumping’
market-seeking FDI from members of the RIAs because they might prefer to
serve the whole regional market through exports from their home location to
reap economies of scale, i.e. it becomes more efficient to export (Blomström and
Kokko, 1997, Ledermann et al., 2003, and Cuevas et al., 2005).
However, the other view postulates that FDI and trade can be seen rather as
complements than as substitutes because without regional tariff barriers com-
panies are likely to undertake investment in locations with a locational advant-
age (e.g. best access to all markets of the member countries) and use this location
as a platform to serve the whole regional market through exports (Balasubra-
manyam et al., 2002). Thus, in some countries disinvestment might take place
while in countries which have the potential to serve as a platform for the regional
market strong intra-regional horizontal FDI inflows can be expected.
North–South and South–South comparison 85
Another argument in regard to the view that trade and FDI are rather comple-
ments than substitutes is that RIAs can lead to the establishment of regional pro-
duction networks. Due to lower transaction costs it is possible to produce only a
part of the product in one location and afterwards send it to another location in
the region for further processing or assembling. Therefore, regional production
networks emerge if factor endowments between member countries and factor
requirements in certain stages of production differ (Levy Yeyati et al., 2003).
The exploitation of these comparative advantages might require the internalisa-
tion of production, i.e. to engage in vertical FDI (te Velde and Bezemer, 2004).
In addition, rules of origin (RoO)8 might lead to increased vertical FDI within
the region. These rules – which state that a certain part of the production (inputs)
needs to take place within (come from) the integration area to have preferential
access to the whole regional market – prevent that inputs are sourced from affili-
ates outside the region. Thus, ‘insider’ MNCs might reallocate production pro-
cesses that have been undertaken ex ante outside the region ex post to member
countries, ensuring preferential access to the markets of the partner countries

4.2.3 The inter-regional effects

In contrast to the intra-regional effect, market-seeking ‘tariff-jumping’ FDI that
flows into the region might increase if the level of outside protection goes up or,
alternatively, foreign investors fear that the protection will increase in the future.
In addition, an increase of market-seeking FDI can be expected because, in con-
trast to the local markets, the regional market might be big enough to make an
investment profitable, i.e. it is possible to achieve economies of scale and bear
the fixed costs (Blomström and Kokko, 1997). Countries in a RIA tend to share
common values and beliefs as well as regional infrastructure and transportation
networks. Hence, investment within a RIA has the advantage that it is possible
to serve a huge regional market from within the region with low transaction
costs. As a result, the region is becoming more attractive as a location for FDI
(Sethi et al., 2003).
Supplementary to market-seeking FDI, vertical FDI might increasingly flow
into the region. As a result of lower discrimination due to ‘new regionalism’,
MNCs possibly integrate the region in their global production chain and build
new affiliates or extend and modernise existing affiliations to exploit the loca-
tional advantage of the region (te Velde and Bezemer, 2004). Rising discrimina-
tion against the ROW ex post through the emergence of non-tariff barriers like
RoO, quantitative restrictions on imports, and custom procedures, can lead to
less vertical FDI or more vertical FDI inflows. The latter two make exporting
from outside impossible or more costly. The former means that it is not suffi-
cient to engage only in ‘tariff-jumping’ FDI but also a certain amount of inputs
need to come from inside the region (Tuman and Emmert, 2004).
If adequate inputs cannot be purchased from local producers RoO might lead
to investment with deep production processes (only assembling is not possible)
86 T. Goda
or, alternatively, MNCs might ‘force’ their suppliers to produce as well in the
region. Both arguments imply that the amount of FDI inflows will be higher than
without RoO (Cuevas et al., 2005). However, RoO could also lead to less FDI
inflows into the region. MNCs might not be able or willing to source their inputs
from within the region and therefore serve the regional market through exports
from outside the region or not at all (Sanguinetti and Bianchi, 2006).

4.2.4 Possible dynamic effects

Besides the static effects RIA might also lead to dynamic effects which could
‘prevent’ that FDI inflows decrease after a while (e.g. after the FDI stock has
adjusted),9 although these effects are less clear-cut. As discussed above, one
motivation to engage in regionalism is the belief that RIAs will raise the eco-
nomic growth rate of member countries in the medium or long term. If this is the
case, one can argue that in the medium and long term also more FDI will flow
into the member countries. On the one hand firms within the region will grow
stronger and will increase business and on the other hand new companies will
emerge (Medvedev, 2006).
In addition, the removal of trade barriers supposedly leads to more competi-
tion, and as a consequence companies most probably will engage on a larger
scale in R&D by which they become more efficient. Along with the new com-
petition situation this might motivate companies to engage in M&As and stra-
tegic alliances, i.e. more FDI will occur in the region. The growth of the
economies will also lead to an increased market size which makes the region
more attractive for FDI. Hence, spillover effects might occur, and specialisation,
economics of scale, and agglomeration effects are likely to take place which
results in further improvements of growth rates (Blomström and Kokko, 1997).
These dynamic effects could lead to a virtuous cycle: the RIA leads to growth,
which brings forth new (foreign) investment, which leads to more growth, which
attracts additional (foreign) investment . . .
However, empirical evidence is not very robust that RIA necessarily increases
the growth rate of the member countries (Berthelon, 2004). Moreover, FDI
might not lead to positive spillover effects and may even dampen economic
growth (e.g. through crowding-out of local investment). Furthermore, there is
the danger that some countries be left out and even receive less FDI in the
medium and long term due to agglomeration effects within the region
(UNCTAD, 1998). Profit repatriation, i.e. a lower net contribution of FDI on
gross capital formation, may disturb the dynamic effect as well (Ramirez, 2003).
Thus, countries should not count on dynamic effects as a result of the establish-
ment of a RIA.

4.2.5 Possible effects for Mexico

The common opinion in the literature is that North–South integration will lead to
larger FDI inflows in the developing partner countries than would be the case
North–South and South–South comparison 87
within a South–South agreement. It is argued that if the members of a RIA differ
in their factor endowments, the agreement most probably stimulates vertical
intra-regional FDI. More specifically in the case of Mexico – which is a member
of NAFTA, i.e. a North–South RIA between the US, Canada, and Mexico – the
two developed partner countries are likely to shift labour-intensive production to
Mexico because preferential access to the markets of the US and Canada is
ensured, capital flows are liberalised, and investor protection regulations are part
of the agreement (Cuevas et al., 2005). Hence, NAFTA is expected to bring
forth more specialisation due to different competitive advantages between the
member countries. As a result, a clustering of activities can be expected, i.e. the
forming of regional production networks (IADB, 1998). Another scenario could
be that the North acts like a kind of flying goose and shifts older technologies to
the South (UNCTAD, 2007a). In both cases, theory expects higher flows of ver-
tical FDI from the US and Canada to Mexico, which has much lower labour
Furthermore, NAFTA is seen to have positive impacts on Mexico in terms of
increased stability (i.e. the lock-in of economic reforms) and accelerated liberali-
sation (Blomström and Kokko, 1997). Accordingly, higher market-seeking FDI
flows to Mexico are expected to take place due to an opening up of the service
sector which was embedded in the NAFTA negotiations (UNCTAD, 1998).
Moreover, the location advantages of Mexico make it likely that extra-bloc
investors will invest in Mexico to produce cheaply and gain access to the
markets of the US and Canada, i.e. ‘tariff-jumping’ FDI (Schiff and Winters,
2003). Intra-regional investors as well are likely to use Mexico as a platform for
exports in the region and hence to restructure the production processes of former
market-seeking affiliates (IADB, 1998), rather than to close these affiliates as
expected by Cuevas et al. (2005).
However, Kim (2007) developed a model which expects less beneficial con-
sequences. While he agrees that the less developed country can be expected to
receive larger efficiency and resource-seeking intra-regional flows, he believes
that – if the technology gap is large – inter-regional market-seeking FDI tends to
flow into the countries with more sophisticated technology.10 In addition, he
argues that technologically intensive production will relocate from the South to
the North. This argument is in line with the finding of UNCTAD (1998) that
agglomeration will take place.

4.2.6 Possible effects for Brazil

In contrast to Mexico, Brazil is engaged in a RIA with developing countries, i.e.
in the Mercosur together with Argentina, Uruguay, and Paraguay. However, this
South–South agreement is also seen to attract higher inward FDI flows to Brazil.
Schiff and Winters (2003) argue that the impact of South–South RIAs on FDI
flows to middle-income countries seems to be positive.11 This might be because
agglomeration effects (centripetal forces) tend to shift industrial investment to
the most developed country within the South–South agreement. However,
88 T. Goda
according to them, Brazil is only likely to attract market-seeking ‘tariff-jumping’
FDI that serves the regional market, while vertical FDI most probably will not
increase as a response to the Mercosur agreement.
Levy Yeyati et al. (2003) confirm this proposition. They state that due to the
fact that Mercosur has increased the size of the regional market, more ‘tariff-
jumping’ FDI is supposed to take place because now it is more attractive to cir-
cumvent the common external tariff than to export from outside the bloc.
Furthermore, as for Mexico, the opening up of formerly restricted sectors
(mainly the service sector) and the further enhancement of structural reforms are
also likely to have led to increased market-seeking FDI inflows to Brazil (Cas-
tilho and Zignago, 2005).
However, while the IADB (1998) confirms that the privatisation of the service
sector was a very important source for FDI inflows to Brazil, it disagrees with
the opinion that Brazil will only attract market-seeking FDI as a consequence of
the forming of Mercosur. As discussed earlier, ‘new regionalism’ is not only
inwardly but also outwardly oriented and lowers the external barriers, which in
combination with better investor protection and the liberalisation of capital flows
can lead to the growth of the manufacturing industry and the integration of
foreign affiliates in global production networks. Nevertheless, intra-regional ver-
tical FDI is seen to develop only slowly as a result of Mercosur.
Accordingly, common wisdom is that Mexico should have received consider-
ably more vertical inward FDI as a consequence of the forming of NAFTA than
Brazil due to the forming of Mercosur and hence Mexico should have attracted
more inward FDI in total as well.

4.2.7 Other possible explanations for increased inward FDI

So far the argument was that developing countries should engage in regionalism to
attract higher amounts of inward FDI. RIAs might lead to increasing liberalisation
(including the opening up of protected sectors), investor protection, access to an
increased regional market, the building of (at least) regional production networks,
and dynamic effects which attract further FDI. However, it is questionable if RIAs
are necessary to attract FDI. East Asian countries are an example in which huge
amounts of inward FDI have been attracted due to competitive strength and the
lowering of trade barriers within the region, despite the missing of extensive multi-
lateral agreements (UNCTAD, 1993). Therefore, it can be expected that regional
dynamics can lead to production networks and higher inward FDI even when the
countries have not signed RIAs (UNCTAD, 2007a).
Accordingly, other factors might be more important than the signing of RIAs.
These other determinants could be resource endowments, infrastructure, a rela-
tively huge domestic market, and strong growth of domestic industries (ibid.).
Tuman and Emmert (2004) add to this list per capita growth, the real exchange
rate, education, inflation, political stability, and openness to trade. According to
Ramirez (2003) unit labour costs are also an important factor to explain higher
FDI inflows.
North–South and South–South comparison 89
Another argument why FDI flows might have increased regardless of RIAs is
that as a result of structural adjustment programmes, privatisation, the opening
up of sectors, and debt conversion programmes would have happened anyway –
all can be seen to have contributed significantly to FDI inflows in Mexico and
Brazil (Nunnenkamp, 1996). Most of these changes concern the services sector
for which a huge regional market does not matter so much because ‘[g]iven the
non-tradable nature of most services, FDI in this sector is . . . almost entirely
directed towards local markets’ (IADB, 1998, p. 22). Considering the huge
market share of Spanish and Portuguese FDI in the Latin American service
sector, FDI therefore might be influenced rather by cultural affinity than by RIAs
(Guedes and Gómez Olivarez, 2005).
Another argument why inward FDI perhaps would have been higher in
Mexico and Brazil even without the signing of NAFTA and Mercosur, respec-
tively, could be that since the beginning of the 1990s the structure of worldwide
capital flows to developing countries has changed in favour of FDI (UNCTAD,
2005). The reason for this change could be that due to competition-linked con-
siderations, companies more often engage in cooperative agreements (joint ven-
tures) to become more efficient (Dunning, 1997). In addition, companies are
seeking to organise their business in a more effective way and accordingly sepa-
rate their functions like R&D, purchasing, assembling, book keeping, etc., to
carry them out at the most efficient location (Chudnovsky and López, 2004).
As UNCTAD (2002, p. 121) has phrased it:

[Due to better telecommunication and transportation possibilities and] . . .

falling barriers to international transactions . . . international production
systems have emerged within which TNCs locate different parts of the pro-
duction processes, including various services functions, across the globe, to
take advantage of fine differences in costs, resources, logistics and markets.

In addition, low interest rates, increasing share prices on the stock markets, and
relatively high growth rates in the USA – which constitutes the lion’s share of
Latin American FDI flows – might have played an important role in the emer-
gence of increased FDI flows to Latin America, i.e. more funds have been avail-
able to be invested in profitable locations (Ramirez, 2003).
To sum up, it can be said that RIAs might influence some variables positively
like per capita growth, the regional infrastructure, openness to trade, stability,
liberalisation policies, and the emergence of (regional) production networks.
However, it is not clear to which extent these factors would have developed
without the creation of a RIA because ‘[i]n a liberalizing world of falling bar-
riers, the location advantage of a large regional market may not be what it used
to be’ (UNCTAD, 1998, p. 122). Accordingly, RIAs might be less influential in
regard to inward FDI than one thinks and they are neither a necessary nor a suf-
ficient condition to attract increasing inward FDI to developing countries (Leder-
man et al., 2003).
90 T. Goda
4.3 Empirical evaluation of inward FDI in Mexico and

4.3.1 Changes in inward FDI in Mexico and Brazil

Changes in FDI inflows and the FDI stock

First of all, it is interesting to compare the amounts of FDI that have flown12 into
Mexico and Brazil. In 1985 Brazil ($1.4 billion) starts at a lower level than
Mexico ($2 billion). This level stays lower until 1995 (Brazil $4.9 billion and
Mexico $9.5 billion) – with the exception of 1988. Then from 1996 to 2000 FDI
flows to Brazil skyrocket to about $33 billion (Mexico about $18 billion). From
2001 to 2005 Mexico again has higher inflows than Brazil (around $5 billion p.a.
higher). In 2006 both countries end up with FDI inflows of around $19 billion
(Figure 4.1).
Besides the amount it is interesting to take a look at trends of FDI inflows. In
Brazil a massive increase in the mid to late 1990s followed a slump in the
2001–2003 period with a subsequent recovery. However, in 1995 and 2001
changes in the collection method of FDI stock and flow data took place in
Brazil13 which had at least some significant influence on the measurement of
inward FDI stocks. Thus, it is not clear how far the increase in 1995 can be
attributed to the change in the methodology. In addition, it can be expected that
the slump after 2000 would have been even higher without the second methodol-
ogy change.
Also notable is that in 2000 an exceptionally large M&A between ‘Tele-
fonica’ and ‘Telecomunicaçoes de Sao Paulo’ took place (around $10 billion).
Without this transaction the trend of FDI inflows would have been negative
already after 1999. However, FDI flows to Brazil are consistently higher after
1995 than before. Therefore, it seems to be clear that FDI flows to Brazil in

Figure 4.1 Foreign direct investment, net inflows (in billion US$), comparison with
ROW (sources: WDI (2007); own calculation).
North–South and South–South comparison 91
general have been higher after Mercosur’s (imperfect) common market came
into existence, which is in line with the theoretical expectation. That being said,
the trend has been similar in the ROW and therefore the increased FDI inflows
might rather reflect a global pattern, not so much the consequence of Mercosur.
Mexico clearly has a less volatile upward trend in FDI inflows than Brazil.
Between 1987 and 1999 this trend is very similar to other middle-income coun-
tries14 (Figure 4.2). The most obvious difference in this period is a peak in 1994,
when NAFTA came into existence. However, this peak perhaps can be explained
partly by the fact that: ‘The methodology for compiling FDI statistics changed
significantly in 1994, so figures for the years before are not directly comparable’
(UNCTAD, 2004, p. 377). The massive increase in 2001 can be explained by
one big acquisition of $12.5 billion (in this year the City Group acquired
Banacci). Without this M&A Mexico would have had lower FDI inflows in 2001
than in 2000 and accordingly the trend after 1999 would be more similar to other
middle-income countries which had a slump in FDI inflows at the beginning of
the twenty-first century – possibly due to the stock market crisis at that time.
In any case, the overall trend of FDI inflows after the establishment of
NAFTA in 1994 clearly is positive until 2004. As a result, FDI inflows have
been much higher than before (although the difference perhaps would have been
lower without the methodology change in 1994). This is congruent with the
expectations of Section 4.2. However, other middle-income countries also had
an upward trend during this period and, therefore, as for Brazil the increased FDI
inflows to Mexico might rather reflect a global pattern and not the effect of
Another way to compare FDI inflows, which is probably more useful, is the
FDI inflows to GDP ratio (UNCTAD, 2002). The advantage of this measurement
is that the size of the host economy is taken into account and hence countries
with a different size can be better compared. Furthermore, this ratio is a good

Figure 4.2 Foreign direct investment, net inflows (in billion US$), comparison with other
middle-income countries (sources: WDI (2007); own calculation).
92 T. Goda
instrument to see if inflows are over proportionally higher compared to the
increase in GDP, or if the increase in inflows is owed to inflation.15 In any event,
the trends of the data in Figure 4.3 are similar to the trends from above. Brazil’s
FDI inflows as percentage of GDP have increased tremendously in the period
1995 to 2000 (in 2000 they have reached 5 per cent of GDP). Afterwards there is
a sharp downtrend which stops 2003. The movement of Mexico’s FDI/GDP ratio
is also relatively similar to the trend of the FDI net inflows in US$ – beside the
increase in 1995, which can be explained by a slump in GDP due to the ‘tequila
crisis’ (the GDP decreases by over 6 per cent in that year (WDI, 2007)).16
If one makes a cross-comparison of the FDI/GDP ratio it can be stated that
the trend of the ROW is similar to Brazil’s from 1992 to 2003, although the
sharp increase starts one year later and the decline in 2001 is even sharper –
from 1986 to 1995 the ROW has a higher ratio than Brazil and then, from 1996
to 2004 (with one exception in 2000) the ratio is higher in Brazil. These data
give some support to the view that Mercosur has led to higher inflows in Brazil.
But, when one compares Brazil with other middle-income countries, which is
the more interesting reference group because of similar characteristics, Brazil’s
FDI/GDP ratio is much below this group’s from 1989 to 1997. Then from 1998
onwards (i.e. three years subsequent the full implementation of Mercosur) the
ratio is significantly higher until 2003 when the ratio again is lower. Hence, in
the 16 years after the implementation of Mercosur the FDI/GDP ratio in compar-
ison to other middle-income countries only is higher in five years, i.e. less than
one-third of the time.
In contrast, Mexico outperforms other middle-income countries in most of the
post-NAFTA period. Only in the periods 1998–1999, 2005–2006, and in the year
2003 have other middle-income countries had a higher FDI/GDP ratio than
Mexico. This means that after 1994 Mexico’s FDI/GDP ratio was higher
than that of other middle-income countries 60 per cent of the time. When one

Figure 4.3 Foreign direct investment as percentage of GDP (sources: WDI (2007); own
North–South and South–South comparison 93
compares Mexico with the ROW or Brazil its superior performance becomes
even more obvious (in 1999–2000 and 2006 it has a lower FDI/GDP ratio than
the ROW, and in 1998–2000 lower than Brazil). However, when one leaves
aside the peaks in the 1994–199517 period and in 2001, Mexico’s FDI/GDP ratio
is very similar to other middle-income countries.
To sum up, it can be said that after the signing of NAFTA and Mercosur the
amount of FDI inflows and the level of the FDI/GDP ratio in Mexico and Brazil
have been higher than before (with the exceptions of 1992 for Brazil, and 2003
and 2006 in Mexico) – interestingly an anticipation effect is not obvious.18 This
gives support to the theoretical expectation that FDI inflows in middle-income
countries increase after RIAs have been signed. On average the FDI/GDP ratio
has been higher in Mexico (average 1994–2006: 3 per cent) than in Brazil
(average 1991–2006: 2.2 per cent) after the corresponding RIAs came into
effect.19 These data can be seen as a confirmation of the expectation that North–
South RIAs lead to higher FDI inflows in middle-income countries than South–
South RIAs. But, Brazil started from a much lower level (in 1990: 0.3 per cent)
than Mexico (in 1993: 1.1 per cent). Thus, the increase of the average FDI/GDP
ratio has been similar after Mercosur and NAFTA came into existence.
Beside FDI flows, it is useful to consider the changes in the inward FDI stock.
Changes in inflow figures ‘do not necessarily reflect changes in the attractiveness
of a location . . . [but instead might show only] exceptional one-off investments’
(IADB, 1998, p. 230). Furthermore, inward stocks can be used to see if higher
inflows to Mexico and Brazil are ‘just’ signs of stock adjustment. However, as
for the flow figures, a comparison with other countries is difficult because differ-
ent methodologies are used. Furthermore, the changes in methodology most
likely have changed the FDI stock to GDP ratio significantly in Mexico and
Brazil. Nevertheless, the data from Table 4.1 do not confirm that stock adjust-
ment in Mexico has taken place; rather the changes seem to reflect a global
pattern. The same is true for Brazil; however, the case is less clear.
After the signing of Mercosur in 1991 the FDI stock in Brazil increases to the
same level as developing economies as a group. But, in 1995 the FDI stock as a
percentage of GDP falls drastically and accordingly is then below the ratio of
developing economies – this decrease most likely is explained by the change in
statistics. Afterwards, the FDI stock/GDP ratio steadily increases until 2004
(with the exception 2002). This trend is very similar to the group of developing
economies and the South and Central American region. But, the ratio is much
lower until 2003. Hence, one could argue that stock adjustment was responsible
for the high inflow figures of FDI to Brazil after 1995. However, the data from
2002, 2005, and 2006 do not fit in this picture. Furthermore, in the literature FDI
stock adjustment refers to higher FDI inflows in a rather short period.
Mexico’s FDI stock in relation to GDP decreases in 1994, the first year of
NAFTA – most probably due to the methodology change in this year, but in the
subsequent year the FDI stock/GDP ratio increases dramatically. However, this
increase is likely explained by the fall of GDP after the ‘tequila crises’. If one
leaves aside the exceptional years 1994 and 1995 and the period 1999–2001 the
Market Liberalism, Growth, and
Economic Development in Latin

The principal themes pursued in this book emerge from the great transformation
that the Latin American and the Caribbean economies experienced in the after-
math of both the foreign debt crisis of 1982 and the macroeconomic stabilization
policies that vividly and painfully produced the so-called ‘lost decade’ of the
Latin America implemented an economic liberalization process during the
late 1980s and the 1990s. The main policy reforms involved in that course can
be summarized as privatization of state owned firms, trade openness, deregula-
tion of the foreign direct investment regime, and fiscal discipline. Latin Amer­
ican countries have also embarked in regional trade agreements, the most
important ones being Mercosur and the North American Free Trade Agreement.
This book compares results from the experience of North–South and South–
South moulds of integration. Thus, the impacts of these policies on growth,
development, technological progress, poverty, and inequality are analysed.
Orthodox and heterodox economic policies and theories are discussed along with
relevant empirical evidence with a view to assess, on the one hand, the relative
merits of the various policy reforms applied by different countries in the region,
and, on the other, the experience of integration into the global economy.
There are 13 chapters in this collection linked in varying ways to the series of
economic reforms introduced in the region in the last decades. The book will be
of interest to academics, researchers, students, and policymakers interested in the
study of economic development in emerging economies and in particular in
Latin America.

Gerardo Angeles-Castro is Head of Research and Postgraduate Studies at the

School of Economics at Instituto Politécnico Nacional, México.

Ignacio Perrotini-Hernández is Professor and Chair of the Graduate Faculty of

Economics at Universidad Nacional Autónoma de México.

Humberto Ríos-Bolívar is Research Economist at the School of Economics at

Instituto Politécnico Nacional, México.
Table 4.1 Inward FDI stock as a percentage of GDP, by host region and economy, 1990–2006

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

World 8 8 8 9 9 9 10 12 14 16 18 20 21 22 23 23 25
Developing 10 10 10 11 12 12 13 17 20 25 26 27 26 27 27 26 27
South and Central 9 10 10 10 10 10 11 14 16 21 21 26 28 31 30 29 27
Brazil 8 9 10 11 10 6 6 8 11 16 17 24 22 26 27 24 21
Argentina 6 6 7 8 9 11 12 14 16 22 24 30 42 37 33 30 27
Paraguay 9 9 10 11 12 8 9 11 15 17 19 18 18 20 17 17 19
Uruguay 8 7 6 5 6 6 6 6 7 9 10 13 11 16 16 18 23
Mexico 9 10 10 10 8 14 14 14 15 16 17 23 24 27 28 27 27
Canada 20 20 19 19 20 21 22 22 24 27 30 30 31 34 33 31 30
United States 7 7 7 7 7 7 8 8 9 10 13 13 13 13 13 13 14

Source: WIR Annex Tables (2007).

North–South and South–South comparison 95
FDI stock/GDP ratio is similar to developing countries and/or South and Central
America, before and as well as after NAFTA came into existence.

Changes in inward FDI by sector

For Mexico, in addition to the total FDI inflows and the FDI stock to GDP ratio
it is interesting to see if the sectoral distribution of FDI has changed. In the case
of Mexico the expectation is that after the implementation of NAFTA both FDI
flows into the secondary and tertiary sector should increase. In the secondary
sector, this is mainly because of higher vertical FDI inflows from Canada and
the United States – which are likely to shift labour intensive production to
Mexico as discussed in Section 4.2. In addition, extra-bloc countries might
invest in the secondary sector with the aim to produce cheaply in Mexico and
use the country as an export platform for the regional market. In the tertiary
sector the expectation as well is that investment should increase because
Mexico was forced to open up its service sector after the accession to NAFTA
(UNCTAD, 1998).
The best ‘indication of the geographic and sectoral distribution of FDI’ in
general is provided by data about FDI stocks (IADB, 1998, p. 230). Unfortu-
nately, the OECD, the UNCTAD or the Economic Secretariat of Mexico are not
publishing stock figures after 2000. Therefore, for the period 2000–2006 inflow
figures are taken. These flow figures do not display changes in the stock one to
one – because of depreciation and revaluation of capital (IADB, 1998) – but at
least they give some hindsight as to how the FDI stocks might have changed
after 2000. Another problem with the data is that, as aforementioned, a method-
ology change took place in 1994. However, it looks as if this change had no
great impact on the sectoral distribution of the inward FDI stock because the
figures of 1994 are similar to 1993 (Figure 4.4).
Anyhow, the available data show that the share of the secondary sector as
a percentage of the total FDI stock has increased after NAFTA came into

Figure 4.4 Mexico inward FDI stock, 1990–2000 (% of total stock) (sources: OECD
(2004, 2005); own calculation).
96 T. Goda

Figure 4.5 Mexico inward FDI flows, 2001–2006 (% of total inflows) (sources: Secre-
taría de Economía Mexico (2008); own calculation).

existence. From 1994 to 1998 the share increases, after a decrease between 1990
and 1994. Consequently, despite the decrease in 1999 and 2000, the share of the
secondary sector is much higher in 2000 (57 per cent) than in 1993 (44 per cent).
The inflows from 2001 to 2006 suggest that, despite higher investments in the
tertiary sector at the beginning of the twenty-first century (especially because of
the Biancci acquisition in 2001), the level of the secondary sector in 2006 is
likely to be still higher than in 1993 (Figure 4.5). Thus, the effect of NAFTA on
vertical FDI seems to be most important. These data give support to the view
that the lock-in of liberalisation reforms as a result of North–South integration is
not that important because many restrictions have been lowered already before
the RIAs have been signed.
In the case of Brazil the expectation is that, as a response to Mercosur,
market-seeking tariff-jumping FDI in the secondary sector should increase. In a
later phase and to a much lower extent the same probably is true for vertical
FDI. Moreover, the opening up of some sectors as a result of Mercosur should
lead to higher inflows, mainly in the tertiary sector. However, it is not obvious in
which sector the effect should be stronger according to theory.
Unfortunately, like for the data of Mexico, inward FDI stock data is only
available until 2000. Therefore, flow figures are demonstrated for the time after
2000. The existing data show that the share of the secondary sector in the FDI
stock in Brazil has halved between 1991 (68 per cent), when Mercosur was
signed, and 2000 (34 per cent). The only exception in this downward trend is the
year 1995 (Figure 4.6), but it is very likely that the methodological change in
that year has led to this deviation. Although the flow figures between 2001 and
2006 indicate that the share of the secondary sector as a percentage of the total
stock has gone up again, it has most probably not reached near the high level of
1990 (Figure 4.7). Thus, it can be said that in general the tertiary sector has
attracted more FDI than the primary and secondary sector in Brazil since Merco-
sur was signed.
North–South and South–South comparison 97

Figure 4.6 Brazil inward FDI stock, 1990–2000 (% of total stock) (sources: UNCTAD
(2008a); own calculation).

Figure 4.7 Brazil inward FDI flows, 2001–2006 (% of total) (sources: UNCTAD
(2008a); own calculation).

Changes in inward FDI by home country/region

According to theory and the findings from above Mexico should have received
more (vertical) FDI from North America since NAFTA was signed and there-
fore, the share of North America’s inward FDI stock should have increased. This
is true for 1994 when the share increased by seven percentage points. However,
because of the statistic change in this year and because the sum of regional FDI
stocks is bigger than 100 per cent, this increase does not seem to be very mean-
ingful (Figure 4.8). After 1994 the share of North America decreases every year
except 2000 and hence the share of 2000 (59 per cent) is much lower than that of
1993 (66 per cent). As for the industry data, there are no stock figures available
after 2000. Hence, flow figures are used to consider changes in recent years. If
we leave aside the year 2001 – when the Banacci acquisition by the US based
City Group in a way distorts the picture – the unweighted average of the inflow
figures suggest that the figures have not been changed much since then (Figure
98 T. Goda

Figure 4.8 Mexico inward FDI stock by regions, 1990–2000 (% of total stock) (sources:
OECD (2004, 2005); own calculation).

Figure 4.9 Mexico inward FDI flows by region, 2001–2006 (% of total flows) (sources:
Secretaría de Economía Mexico (2008); own calculation).

4.9). Thus, surprisingly Europe instead of North America has increased its share
significantly after NAFTA came into existence.
If one takes a look at country data (OECD, 2004 and 2005; Secretaría de
Economía Mexico, 2008; own calculations) it becomes apparent that Canada
increased its share between 1994 (1.4 per cent) and 2000 (2.5 per cent), while
the United States – even though it remains the country which has by far the
highest FDI stock in Mexico – lost shares after 1993 (minus eight percentage
points by 2000). This is highly surprising as theory expects that investment from
the United States should be over-proportional in comparison to extra-bloc coun-
tries.20 The flow figures after 2000 do not change this picture because even with
an exceptional year in 2001 the US stock as a percentage of the total stock most
likely has not increased back to old levels before NAFTA came into existence.
In contrast, Canada probably could hold its percentage of the total stock at the
North–South and South–South comparison 99
level of around 2.5 per cent and accordingly its share has likely increased signifi-
cantly since NAFTA.
Regarding extra-bloc countries it is interesting to notice that in 2006 the share
of Japan most probably is much below its pre-NAFTA amount. This is surpris-
ing as nearly one-third of Japan’s exports have flown into the NAFTA area since
it came into existence (UNCTAD, 2007b; own calculation) and accordingly
Japan should be expected to invest in Mexico to produce cheaply and then use
the country as an export platform to circumvent tariff barriers. Perhaps, this is
not the case because Japan prefers to have regional production networks in Asia,
or because of RoO (e.g. for automobiles), or because the external barriers are not
high ‘enough’ (e.g. for automobiles). Within Europe most notably Spain21 and
the Netherlands improved their shares. Both countries can be expected to have
shares around or above 10 per cent. In contrast, Germany’s FDI stock shares in
Mexico are most probably lower in 2006 than they were in 1993 (4 per cent).
This is just as surprising as the decrease in Japan’s share, even though the share
of Germany’s exports to NAFTA has been much lower than that of Japan with
an average of 10.5 per cent after 1994 (ibid., own calculation).
In contrast to Mexico, Europe has the biggest FDI stock in Brazil. The Euro-
pean stock has decreased by eleven percentage points until 1997, but then recov-
ers to reach 50 per cent in 2000, which is as high as it was in 1990 (Figure 4.10).
The flow figures in the subsequent years suggest that the share should be close to
this level in 2006 (Figure 4.11). The inward stock of North America develops
similarly to Mexico. In the first two years after 1990 the share of the North
American stock increases, but then it shrinks constantly and as a result is much
lower in 2000 (26 per cent) than in 1990 (34 per cent). The flow figures indicate
that the share will have even decreased slightly more by 2006. Hence, because
the market shares of other developing countries have shrunk too, the only region
able to gain significant shares after Mercosur came into existence is Latin
America and the Caribbean. However, the increase of the share from this region

Figure 4.10 Brazil inward FDI stock by regions, 1990–2000 (% of total stock) (sources:
UNCTAD (2008a); own calculation).
100 T. Goda

Figure 4.11 Brazil inward FDI inflows by regions, 2001–2006 (% of total inflows)
(sources: UNCTAD (2008a); own calculation).

mainly comes from higher investment from the Caribbean22 which will be
excluded from the analysis.
If one takes a look at the countries with the most important shares of the FDI
stock in Brazil it becomes visible that from 1990 to 2000 Spain (from nearly 0 to
12 per cent), the Netherlands (from 3 to 11 per cent), and Portugal (from nearly
0 to 4 per cent) increased their shares significantly, while Germany’s share
shrunk dramatically from 15 per cent in 1990 to 5 per cent in 2000. Spain, the
Netherlands, and Portugal have all invested mostly in the tertiary sector at the
end of the 1990s. However, according to FDI inflow data from 2001 to 2006,
Spain can be expected to have lost some of its stock share it gained during the
1990s. In contrast, the Netherlands seems to have increased its share, while
Germany and Portugal should have kept their shares more or less stable.
In 2000, the Canadian, US, and Japanese FDI stock shares are well below
their 1990 shares. But, after 2001 Japan’s and Canada’s shares seem to have
recovered a little bit, while the share of the United States is likely shrinking
further. In contrast, the shares of Argentina and Uruguay increase until 2000
(from nearly 0 to 1 and 2 per cent, respectively). However, the average of the
inflows after 2000 suggests that Argentina and Uruguay have not maintained
these shares (data from UNCTAD, 2008a; own calculations). Thus, the data con-
firms the expectation that FDI within the region has hardly increased as a result
of Mercosur.

4.3.2 Possible other explanations for increased inward FDI

As aforementioned the increase in global FDI flows and stocks could be the main
driver behind the increase of inward FDI in Mexico and Brazil. In addition, as
discussed in the theoretical section, changes in inward FDI might have many
other different influences besides the signing of a RIA. Therefore, in this section
North–South and South–South comparison 101
of the chapter, some other factors that might have led to higher FDI inflows will
be discussed very briefly. They are as follows: (1) a change in the potential to
attract FDI inflows, (2) privatisation and the opening up of the service sector, (3)
changes in western stock markets, and (4) changes in the exchange rate.
This chapter refers to the UNCTAD inward potential index to see how far the
potential of Mexico and Brazil to attract FDI has changed. The potential index
accounts for some (quantifiable) factors23 that are supposed to influence FDI
inflows – the signing of RIAs is not included as a factor. According to this index,
the potential of Brazil to attract FDI increases significantly after 1991 and
reaches its highest point in 1997–1999 (when FDI inflows also increased dra-
matically). In Mexico, after a dip in 1994–1996, the potential index is also higher
than before 1994 (data from UNCTAD, 2008b). Therefore, one could argue that
FDI flows to Brazil after 1991 and to Mexico after 1996 have increased because
the overall endowments of both countries have been more attractive for investors
than before. However, the potential index is just a benchmark because many
factors are not quantifiable (e.g. specific skills of the workforce, quality of local
suppliers, etc.) and in the last 15 years the potential index has increased for many
other countries as well.
Another explanation why inflows might have increased regardless of the RIA
status relates to privatisation and the opening up of sectors in which foreigners
could not invest before. UNCTAD (2004) states that privatisation in the service
sector has led to temporary higher inflows of FDI in Latin America: the telecom-
munication sector is given as example. It is estimated that in Brazil $64 billion
of private capital has been invested in the telecommunications sector. In Mexico
the amount has been considerably lower ($38 billion) and most of this invest-
ment has come from national sources. Hence, privatisation seems to be more
important for Brazil than for Mexico in explaining higher FDI inflows. The data
from Table 4.2 confirms these findings. While Brazil saw approximately $31.6
billion of FDI inflows in the late 1990s as a result of the privatisation, Mexico
only received around $2.5 billion because of privatisation efforts.
Unfortunately, there are no data available as to how much the opening up of
formerly closed sectors in general has contributed to FDI inflows in Mexico and
Brazil. That Brazil has received much more FDI in the tertiary sector than
Mexico and that the share of M&As in total FDI inflows in Brazil between 1990
and 2006 have been much higher than in Mexico (Table 4.3) are hints, however,
that the opening up of service sectors has likely been a driving force for FDI

Table 4.2 Transaction values of cross-border M&As of privatised firms (in US$ billion)

1995 1996 1997 1998 1999 Total

Brazil – 2.9 6.0 19.9 2.8 31.6

Mexico – 0.1 2.1 – 0.3 2.5

Source: authors’ elaboration with UNCTAD information.

Table 4.3 Share of cross-border M&A sales in total FDI inflows in Mexico and Brazil (%)

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

Brazil 22 14 8 48 17 40 61 64 102 33 70 31 36 52 37 38 53
Mexico 88 0 22 42 17 8 16 62 24 6 22 62 37 8 29 21 11

Source: authors’ elaboration with UNCTAD information.

North–South and South–South comparison 103
inflows to Brazil as compared to Mexico. This might explain as well why after
2000 – by which time the major privatisation programmes were completed – the
flows of FDI to Brazil have decreased more drastically than in Mexico
(UNCTAD, 2004).
Another reason for a decrease of FDI flows to Latin America might be the
devaluation of Brazil in that time and the stock market crises after 2000, which
led to liquidity constraints (UNCTAD, 2004). As already discussed in the theo-
retical section, the exchange rate and the stock market might be important vari-
ables in explaining FDI flows. If we compare the average of the composite Dow
Jones Index24 with FDI inflow indices from Brazil and Mexico it becomes visible
that FDI flows to Brazil have reacted similarly to the stock market in the US,
while a correlation is much less visible for Mexico (Figure 4.12).
Regarding the exchange rate a clear correlation between the FDI inflows to
GDP ratio and changes in the exchange rate is visible neither in Brazil nor in
Mexico (Figure 4.13 and 4.14). Therefore, the major reason for decreasing flows
to Mexico after 2000 – if one excludes the Banacci M&A – and the stagnation
after 2005 might be that countries with lower labour costs than Mexico have
become favourable for investors to serve as an export platform to the US (espe-
cially China).25 However, an analysis as to how far and for what reasons home
countries have shifted their FDI to other locations is beyond the scope of this

4.4 Conclusions
The relationship between regional integration and FDI is complex and the influ-
ence of RIAs on FDI inflows depends on many determinants like factor endow-
ments, the pre-existing situation, investment provisions, the opening up of
sectors, non-tariff barriers within the region (e.g. RoO), external barriers, and the

Figure 4.12 Dow Jones Index composite average (US$) and FDI inflows in Brazil and
Mexico (indices, 1990 = 100), 1990–2006 (sources: Dow Jones Indexes
(2008); WDI (2007); own calculation).
104 T. Goda

Figure 4.13 Brazil FDI inflows (as % of GDP) and exchange rate (LCU per US$)
(source: WDI (2007)).

Figure 4.14 Mexico FDI inflows (as % of GDP) and exchange rate (LCU per US$)
(source: WDI (2007)).

regional transport and communication infrastructure. Moreover, it is not clear

how far other determinants besides the signing of a RIA might influence FDI
inflows positively. Nevertheless, theory expects both Mexico and Brazil to
receive higher FDI inflows as a result of their accession to NAFTA and Merco-
sur, respectively. While Brazil should receive mainly more market-seeking FDI
in theory, Mexico is predicted to receive significantly more market-seeking and
vertical FDI (especially from the US and Canada). Thus, in theory NAFTA
should have been more beneficial for Mexico in attracting FDI than Mercosur
for Brazil.
North–South and South–South comparison 105
The data confirm that in Brazil both FDI inflows and the FDI/GDP ratio are
much higher than they were before Mercosur came into existence. The biggest
share of this investment has come from European countries and has taken place
in the tertiary sector. Therefore, privatisation and the opening up of the service
sector seem to have played a decisive role in higher FDI flows to Brazil.
However, it is unfortunately not clear how far Brazil would have privatised and
opened up sectors without the existence of Mercosur. But, it is very likely that
the degree of liberalisation would have been similar without Mercosur because
during the 1990s the government of Brazil was in favour of liberalisation pol-
icies (Gardini, 2008).
In addition to liberalisation, the increase in FDI inflows might have taken
place because of changing global patterns, i.e. a worldwide increase in FDI
flows. Furthermore, some factors that are believed to attract FDI have changed
for the better in Brazil which might explain as well why flows to Brazil have
been higher. Above all, the stock market bubble and bust between 1996 and
2002 seems to have been an important determinant for the amount of FDI
inflows that Brazil received during this time.
In Mexico, FDI inflows also increased significantly after the signing of
NAFTA, mainly due to investment activities in the secondary sector – for
Mexico privatisation seems to be much less important in attracting FDI than it
was for Brazil. Therefore, one can argue that NAFTA has had a bigger impact
on Mexico than Mercosur on Brazil. However, the changes in the FDI/GDP ratio
do not confirm this view. Interestingly, the share of the inward FDI stock (as a
percentage of the total stock) from the United States has decreased, while the
European share of the total stock has increased. This finding is not expected by
theory and is contrary to the results of previous studies.
As for Brazil, there are other factors besides the RIA that might explain the
increasing flows of FDI to Mexico, e.g. the FDI inward potential index of
Mexico is higher after 1997. The most important factor next to the RIA,
however, seems to be the increase of FDI flows to other middle-income coun-
tries during the NAFTA period. Mexico’s FDI inflows/GDP ratio is very much
in line with these countries if one leaves out years with exceptionally high
inflows (which can be explained by other determinants better than by NAFTA).
Therefore, one can argue that the increase in FDI flows to Mexico might have
taken place due to a global trend of higher investment in middle-income coun-
tries and not so much as a result of NAFTA.
Ultimately, it should be kept in mind that for developing countries RIA and
attracting FDI are only instruments to achieve economic development. If we
have a look at some factors in this regard, the performance of Mexico in the last
14 years is rather poor. As UNCTAD (2007a, p. 66) puts it: ‘[NAFTA] does not
appear to have helped accelerate output growth, nor does it seem to have con-
tributed significantly to employment growth or to much higher standards of
living of the Mexican people’. To achieve economic development, it is import-
ant to attract the ‘right’ FDI that brings along beneficial effects for the economy.
This may require some regulations. Therefore developing countries should
106 T. Goda
cautiously calculate the gains and risks (i.e. the loss of alternative policy options)
that are involved in North–South integration and keep in mind that regional inte-
gration is neither a necessary nor a sufficient condition to attract FDI.

1 RIA is a synonym for: preferential trade agreements, free trade agreements, customs
unions, common markets, and economic unions (for a distinction between these agree-
ments please have a look at Bergstrand et al., 2008).
2 North–South integration refers to an integration scheme where at least one developed
and one developing country is involved.
3 South–South integration is a synonym for a RIA between developing countries
(including countries in transition).
4 NAFTA is a free trade agreement and came into effect on 1 January 1994. Mercosur
was founded on 26 March 1991 as a free trade agreement but has been extended to a
common market until the 31 December 1994 (although an imperfect one).
5 The label of the product might need to contain different languages, customers in dif-
ferent markets may have different preferences and for this reason the product needs to
have different appearances, or rules and regulation within the region are different and
hence the production process needs to deviate slightly.
6 The reaction of the investment behaviour from MNCs that are located within the
7 The reaction of the investment behaviour from MNCs that are located outside the region.
8 Rules of origin are implemented to prevent trade deflection and will only exist if the
RIA is not as deep as a customs union (for more information please consult Sangui-
netti and Bianchi, 2006).
9 Stock adjustment means that the FDI stock of countries is below those of countries
with comparable attributes and hence FDI inflows will be above the average level for
some years to reach equilibrium (Buch et al., 2001). As example for FDI stock adjust-
ment often Portugal and Spain are mentioned: ‘[S]tock adjustment took place after the
countries joined the EC, with world investors rebalancing their portfolios in favour of
Spain and Portugal in a process that lasted several years’ (Lederman et al. 2003).
10 In the case of NAFTA into the US or Canada.
11 While South–South regionalism is unlikely to have positive effects on FDI inflows
into the poorest countries of the agreement.
12 All inflow data are given on a net basis.
13 Before 1995 data were compiled from the Brazilian Central Bank according to regis-
try. In 1995 the system changed and data were compiled from foreign exchange trans-
actions. In the first years after the change, only exchange transactions with the
minimum level of $10 million were included in the statistics but this policy was abol-
ished after 2000 (UNCTAD, 2004, p. 161).
14 Other middle-income countries are all countries that are classified from WDI (2007)
as middle-income countries minus Brazil and Mexico.
15 In all sources FDI flows are presented only in current US$.
16 For detailed information about the ‘tequila crisis’ please have a look at Taylor (1998).
17 As said before, the increase in 1994 can possibly be explained by the change in meth-
odology and the increase in 1995 by a slump in GDP due to the ‘tequila crisis’. The
peak in 2001 can be explained by the acquisition of Banacci.
18 Quite the contrary is true. Both countries have had a lower FDI/GDP ratio in the year
before the RIA came into existence.
19 However, unfortunately a comparison is difficult as the changes in methodology in
both countries might have had a significant influence on the data. Furthermore, both
countries have different methodologies to measure the inflow of FDI.
North–South and South–South comparison 107
20 Some empirical studies found that Mexico has received higher FDI inflows as a
response to NAFTA solely because of higher flows from the United States and
Canada (e.g. Tekin-Koru and Waldkirch, 2008).
21 Spain has increased its share mainly because of huge investments in the banking and
telecommunications sector (Guedes and Gómez Olivarez, 2005).
22 The FDI that comes from the Caribbean offshore centres (i.e. British Virgin Islands
and Cayman Islands), can be expected to be investment from the United States or
Europe that is only channelled through the Caribbean with the aim to lower tax pay-
ments (IADB, 1998).
23 The variables are the rate of growth of GDP; per capita GDP; share of exports in
GDP; telephone lines per 1,000 inhabitants; commercial energy use per capita;
share of R&D expenditures in gross national income; share of tertiary students in
the population; and country risk.
(UNCTAD, 2002, p. 24)
The better the unweighted average of these factors in a three year period (to counter-
act annual fluctuations) the higher is the FDI potential index.
24 The Dow Jones Index is taken, because it is the world’s leading stock market.
25 For example, in the ‘maquiladoras’ factories in Mexico over 200,000 people have
been dismissed in the period 2000–2001 (UNCTAD, 2003).

Balasubramanyam, V.N., Sapsford, D., and Griffiths, D. (2002): ‘Regional Integration
Agreements and Foreign Direct Investment: Theory and Preliminary Evidence’. Man-
chester School, Vol. 70(3), pp. 460–482.
Bergstrand, J.H., Estevadeordal, A., and Evenett, S.J. (2008): ‘Introduction: The Sequenc-
ing of Regional Economic Integration’. World Economy, Vol. 31(1), pp. 1–4.
Berthelon, M. (2004): ‘Growth Effects of Regional Integration Agreements’. Central
Bank of Chile Working Paper No. 278, Santiago de Chile.
Blomström, M. and Kokko, A. (1997): ‘Regional Integration and Foreign Direct Invest-
ment’. NBER Working Paper No. 6019, Cambridge.
Buch, C.M., Kokta, R.M., and Piazolo, D. (2001): ‘Does the East Get What Would Other-
wise Flow to the South? FDI Diversion in Europe’. Kiel Working Paper No. 1061, Kiel.
Castilho, M. and Zignago, S. (2005): ‘Foreign Direct Investment, Trade and Regional
Integration in Mercosur’. In: Graham, E.M. (ed.): Multinationals and Foreign Invest-
ment in Economic Development. Basingstoke: Palgrave Macmillan, pp. 145–162.
Chudnovsky, D. and López, A. (2004): ‘Transnational Corporations’ Strategies and
Foreign Trade Patterns in MERCOSUR Countries in the 1990s’. Cambridge Journal of
Economics, Vol. 28(5), pp. 635–652.
Cueveas, A., Messmacher, M., and Werner, A. (2005): ‘Foreign Direct Investment in
Mexico since the Approval of NAFTA’. World Bank Economic Review, Vol. 19(3),
pp. 473–488.
Dow Jones Indexes (2008): ‘Dow Jones Averages: Composite’. Retrieved 26 June 2008
Dunning, J.H. (1997): ‘Reconfiguring the Boundaries of International Business Activity’.
In Boyd, G. and Rugman, A.M. (eds): Euro-Pacific Investment and Trade. Aldershot:
Edward Elgar, pp. 1–18.
Fiorentino, R.V., Verdeja, L., and Toqueboeuf, C. (2007): ‘The Changing Landscape of
Regional Trade Agreements: 2006 Update’. WTO Discussion Paper No. 12, Geneva.
108 T. Goda
Gardini, G.L. (2008): ‘Who Invented Mercosur?’ Diplomacy and Statecraft, Vol. 18(4),
pp. 805–830.
Guedes, C.M. and Gómez Olivares, M. (2005): ‘Portuguese Investment in Brazil: The
Challenges of an Iberian Logic’. In: Graham, E.M. (ed.): Multinationals and Foreign
Investment. Basingstoke: Palgrave Macmillan, pp. 212–233.
IADB (1998): Foreign Direct Investment in Latin America: Perspectives of the Major
Investors. Madrid: IADB.
Kim, Y.-H. (2007): ‘Impacts of Regional Economic Integration on Industrial Relocation
through FDI in East Asia’. Journal of Policy Modeling, Vol. 29(1), pp. 165–180.
Lederman, D., Maloney, W.F., and Servén, L. (2003): Lessons from NAFTA for Latin
American and Carribean (LAC) Countries: A Summary of Research Findings. Wash-
ington, DC: The World Bank.
Levy Yeyati, E., Stein, E., and Daude, C. (2003): ‘Regional Integration and the Location
of FDI’. Inter American Development Bank Research Department Working Paper No.
492, Washington, DC.
Medvedev, D. (2006): ‘Beyond Trade: The Impact of Preferential Trade Agreements on
Foreign Direct Investment Inflows’. World Bank Policy Research Paper No. 4065.
Nunnenkamp, P. (1996): ‘The Changing Pattern of Foreign Direct Investment in Latin
America’. Kiel Institute for the World Economy Working Paper No. 736, Kiel.
OECD (2004): ‘International Direct Investment Statistics Yearbook: 1990–2001’.
Retrieved 3 June 2008 from:
OECD (2005): ‘International Direct Investment Statistics Yearbook: 1992–2003’.
Retrieved 3 June 2008 from:
Ramirez, M.D. (2003): ‘Mexico under NAFTA: A Critical Assessment’. Quarterly
Review of Economics and Finance, Vol. 43(5), pp. 863–892.
Sanguinetti, P. and Bianchi, E. (2006): ‘Implementing PTAs in the Southern Cone Region
of Latin America: Rules of Origin’. In: Cadot, O., Estevadeordal, A., Suwa-Eisenmann,
A., and Verdier, Thierry (eds): The Origin of Goods: Rules of Origin in Regional Trade
Agreements. Oxford: Oxford University Press, pp. 213–237.
Schiff, M. and Winters, L.A. (2003): Regional Integration and Development. Washing-
ton, DC: The World Bank.
Schlageter, K.M. (2005): Strategien der Automobilindustrie in Südamerika: Eine kritische
Diskussion vor dem Hintergrund des neuen Regionalismus. Lohmar: Eul.
Secretaría de Economía Mexico (2008): ‘Dirección General de Inversión Extranjera’.
Retrieved 8 June 2008 from:
Sethi, D., Guisinger, S.E., Phelan, E., and Berg, D.M. (2003): ‘Trends in Foreign Direct
Investment Flows: A Theoretical Empirical Analysis’. Journal of International Busi-
ness Studies, Vol. 34(4), pp. 315–326.
Taylor, L. (1998): ‘Lax Public Sector, Destabilizing Private Sector: Origins of Capital
Market Crises’. CEPA Working Paper Series III, Working Paper No. 6, New York.
te Velde, D.W. and Bezemer, D. (2004): ‘Regional Integration and Foreign Direct Invest-
ment in Developing Countries’. Retrieved 30 April 2008 from:
Tekin-Koru, A. and Waldkirch, A. (2008): ‘North–South Integration and the Location of
Foreign Direct Investment’. MPRA Paper No. 6912, Munich.
Tuman, J.P. and Emmert, C.F. (2004): ‘The Political Economy of U.S. Foreign Direct
Investment in Latin America: A Reappraisal’. Latin American Research Review, Vol.
39(3), pp. 9–28.
North–South and South–South comparison 109
UNCTAD (1993): World Investment Report 1993: Transnational Corporations and
Integrated International Production. New York: United Nations.
UNCTAD (1998): World Investment Report 1998: Trends and Determinants. New York
and Geneva: United Nations.
UNCTAD (2002): World Investment Report 2002: Transnational Corporations and
Export Competitiveness. New York and Geneva: United Nations.
UNCTAD (2003): World Investment Report 2003: FDI Policies for Development:
National and International Perspectives. New York and Geneva: United Nations.
UNCTAD (2004): World Investment Directory: Volume IX Latin America and the Carib-
bean 2004 Parts 1 and 2. New York and Geneva: United Nations.
UNCTAD (2005): World Investment Report 2005: Transnational Corporations and the
Internationalisation of R&D. New York and Geneva: United Nations.
UNCTAD (2007a): Trade and Development Report, 2007: Regional Cooperation for
Development. New York and Geneva: United Nations.
UNCTAD (2007b): UNCTAD Handbook of Statistics 2007 On-line. Retrieved 20 June
2008 from:
UNCTAD (2008a): ‘FDI/TNC Database’. Data received 12 February 2008 via email from
Kotte, D. (Head, Macroeconomic and Development Policies Branch, UNCTAD).
UNCTAD (2008b): ‘Inward FDI Potential Index’. Retrieved 7 June 2008 from: www.
WDI (2007): ‘The World Bank Group: World Development Indicators Online’. Retrieved
May and June 2008 from:
WIR Annex Tables (2007): ‘Key Data from WIR Annex Tables’. Retrieved 3 June 2008
5 Trade blocs as determinants of
trade flows in South American
An augmented gravity approach
Clemente Hernández-Rodríguez

5.1 Introduction
In the last years trade flows have been evaluated in the literature, analysing the
effects of such flows on the trade agreement itself and the influence on the multi-
lateral trade system.1 Those works focus on the evaluation of the good and bad
results of the trade agreements and identify the determinants of trade flows. As a
facet of this, is the need to ascertain just how effective trading blocs are in pro-
moting trade. To answer such questions, gravity models2 have been extensively
used in the past (see, for example, Tinbergen (1962), Linnemann (1966), Aitken
(1973), Frankel et al. (1995), Martinez-Zarzoso and Nowak-Lehmann (2003)).
Gravity equation is the most frequent utilised approach in order to evaluate
the determinants of the trade flows. Gravity equation assumes that the volume
traded between two trade partners is an increasing function of their income and a
decreasing function of the transportation cost. Distance between the economic
centres of the countries involved is considered to approximate the transportation
McCallum (1995), Frankel (1997), Breuss and Egger (1999), Egger (2000),
Soloaga and Winters (2001), Feenstra et al., (2001), Carrillo and Li (2002), and
Martinez-Zarzoso and Nowak-Lehmann (2003), figure among the many empiri-
cal works that have used the gravity equation to evaluate the international trade
Latin American countries have embarked in regional trade agreements: the
two most important being the Common Market of the South (Mercosur) and the
Andean Community (CAN) in South America. In this work the impact of
regional integration in South American countries is studied. The Mercosur and
the CAN countries are chosen to verify if the pattern of the trade flows is similar
or different in both regions.
On 26 March 1991 the Heads of State of Argentina, Brazil, Paraguay, and
Uruguay met in the Paraguayan capital to sign the Treaty of Asunción, with a
view to forming the Mercosur. The Treaty of Asunción (1991) established Mer-
cosur’s objectives: the liberalisation of intraregional trade; a common external
tariff; harmonisation of laws and regulations concerning rules of origin; and the
mutual consultation on macroeconomic policies. The rationale for Mercosur was
Trade blocs as determinants of trade flows 111
the notion of ‘Open Regionalism’. Unlike the protectionist, inward-oriented
model of regional integration of the 1960s and 1970s, Open Regionalism is often
presented as part of a broader agenda of the neoliberal economic reforms.
On the other hand, the Andean Community, which began with the Cartagena
Agreement in 1969, as the Andean Pact, is one of the oldest free trade move-
ments still existing today. The original membership – Bolivia, Chile, Colombia,
Ecuador, and Peru – has changed slightly with the inclusion of Venezuela (1973)
and the break of Chile (1976), but the Pact remains generally intact. For nearly
20 years, its goals of economic growth and the creation of a regional common
market floundered. Over the past several years, however, changes to democratic
governance and developments in trade and investment liberalisation have
prompted renewed interest and actions among Pact members. These actions have
brought the Andean Pact to the forefront of the integration movement in Latin
America. The Andean Pact has progressed beyond a free trade agreement, as it
begins to implement a customs union similar to the European Union.
This chapter seeks to evaluate the extent to which the establishment of the
Mercosur and the CAN has led to an improvement of intra-regional trade, as
promised by Open Regionalism, avoiding at the same time the reduction of the
inter-regional trade. This analysis is interested only in the static effects in the
trade flows. It does not contemplate other aspects as the negotiation process,
policy coordination, credibility of the members, and the dynamic gains obtained
from the trade flows that may be of interest.
A gravity approach is utilised because this approach: (a) can explain the vari-
ations in the trade flows among a variety of countries and in different periods of
time, (b) includes the geographic dimension, (c) improves the theoretical founda-
tions using the imperfect substitutes theory, and (d) helps to examine the con-
venience of trade agreements, and its effect on the trade pattern, level, and
direction of trade flows. The sample of data, 1987–2000, includes some years
before the beginning of Mercosur (1991), and then allows the chance to see if
the level and direction of the trade flows changed. Regarding CAN countries, it
gives us the opportunity to see if the pattern of trade flows was altered with the
renovation of the Andean Pact in 1988.
One limitation of this analysis is that it is constrained to the South American
region, excluding other regions such as the European Union. However, at the
same time, that is an advantage because both the Mercosur and the CAN region
share a similar political, social, economic, and even cultural centennial tradition.
One of the contributions of this chapter is that it shows that the establishment
of Mercosur and the CAN led to an increase of intra-regional trade, as the Open
Regionalism view proposes, but did not lead to a contraction of extra-regional
This chapter consists of four additional sections. The next section shows the
impact of the regionalisation in the Mercosur and CAN countries, observing
the direction of the trade flows in the past decade. The third section presents the
theoretical setting of the gravity model and provides a literature review about the
gravity approach. In the fourth section, using an augmented gravity model and
112 C. Hernández-Rodríguez
Panel Analysis, we identify the determinants of the trade flows. The last section

5.2 Impact of the regionalisation in the level and direction of

the trade flows in the Mercosur and CAN countries
This work also compares results from the experience in the Mercosur and the
CAN. So the extent to which the establishment of the Mercosur and the CAN
has led to an improvement of intra-regional trade, as promised by Open Region-
alism, avoiding at the same time the reduction of the inter-regional trade, is
evaluated. Therefore, I present an analysis of the static effects in the trade flows.
However, I first introduce some historical facts about the trading blocs under

5.2.1 The Mercosur countries

Argentina, Brazil, Paraguay, and Uruguay signed the Mercosur agreement in
1991 (see Table 5.1). Mercosur went into effect in 1995, becoming a customs
union. Following the entry into force of the common external tariff (CET) on 1
January 1995, the Mercosur countries must maintain a common commercial
policy. Bolivia and Chile are associated countries of Mercosur without full mem-
bership status. Bolivia and Chile signed the association agreements with Merco-
sur in 1995 and 1996, respectively. Mercosur has also been trying to promote
Chile’s full membership and inclusion into the Mercosur customs union in 2000.
Hugo Chavez’s Venezuela has also been negotiating an inclusion to an expanded
Mercosur is considered an emerging market offering good investment oppor-
tunities, with a population over 200 million people (it represents half of the
population, 58 per cent of the GDP, and 40 per cent of the total foreign trade of

Table 5.1 Mercosur and the CAN

Agreement Membership evolution Type

Common Market of the 1991: Argentina Customs union
South (Mercosur)
1991: Brazil
1991: Paraguay
1991: Uruguay
Andean Community (CAN) 1988: Bolivia Preferential trading
1988: Colombia
1988: Ecuador
1988: Peru
1988: Venezuela

Source: the list of countries, and regional groupings, is given in an Appendix in Frankel et al. (1995).
Trade blocs as determinants of trade flows 113
Latin America and Caribbean). Mercosur is the third largest trading bloc in the
world (after the European Union (EU) and the North American Free Trade
Agreement (NAFTA)). The four Mercosur countries embrace an area larger than
the continental United States. In 1998 the EU accounted for some 33 per cent of
Mercosur’s imports and 39 per cent of its exports. In 2002, the EU imported five
times more from Mercosur than the US, making it the group’s main trading
partner. Trade in goods between EU and Mercosur has risen considerably in the
last years, with the total value of trade flows between the two blocs rising from
€18.8 billion in 1990 to €42.5 billion in 1998, an increase of almost 125 per
When compared with NAFTA, Mercosur is a distinct model of regional inte-
gration. Mercosur seeks to establish a European-style common market with a
CET and coordinated macroeconomic policies. NAFTA, by contrast, is essen-
tially a free trade area, and does not contemplate adopting a CET.
By 31 December 1994, Mercosur was a free trade area covering 95 per cent
of intra-regional trade. The increase from US$5.2 billion in 1991 to US$20.3
billion in 1997 in intra-Mercosur trade (fourfold increase) showed that in spite of
divergent exchange rate policies it can be an established vibrant and successful
free trade agreement.
Due to the crisis in Brazil which resulted in a devaluation of the real in 1999,
Argentina, Paraguay, and Uruguay sought exceptions from the CET. This devel-
opment not only weakened Mercosur as a customs union but also had a negative
effect on future negotiations. Especially the Argentinian crisis, which led to even
more exceptions from the CET, has left doubts regarding the stability of Merco-
sur as a customs union and the solvency of Argentinian importers.
Figure 5.1 consists of a set of different graphs. In those graphs we can
observe that in the period 1987–2000 Mercosur increased the flows at the intra-
region level. Argentina changed its exports from the EU countries to the Merco-
sur countries. Even Brazil increased its exports to the Mercosur countries,
though its exports to other regions (NAFTA and the EU) are more important in
the exports share. Paraguay increased its exports share to Mercosur countries,
substituting export to the EU. Uruguay exhibits a similar evolution.

5.2.2 The CAN countries

An economic agreement was forged in 1969 between several South American
countries (Bolivia, Chile, Columbia, Ecuador, and Venezuela) to assist in redu-
cing trade barriers and fostering the economic development of the members. The
document that formalised the establishment of the Andean Pact is the Cartagena
Agreement. Venezuela became part of the Andean Pact in 1973.3 In the mean-
time, Chile withdrew from the Andean Pact in 1976 to pursue more liberal trade
The initial stage of the Andean Pact was characterised by the Closed Region-
alism Model. Intra-regional trade only increased from 1.7 per cent of total
exports in 1970 to 4.5 per cent in 1979. This early stage of the Andean Pact
Figure 5.1 (a) Graphic evolution of exports by country member of Mercosur (source:
Excel output using data from the IMF (International Monetary Fund), CEPAL
(Comisión Económica para América Latina y el Caribe), and national
CAN is Andean Community; MS is Mercosur; NAFTA is the North America Free Trade Agreement;
EU is the European Union.

Figure 5.1 (b)

Trade blocs as determinants of trade flows 115

Figure 5.1 (c)

Figure 5.1 (d)

became very inefficient and failed for several reasons: many products were
exempted from the tariff liberalisation process; a clear consensus about the
common external tariff was lacking due to significant differences in the level of
protection of each Andean country; the production requirements established by
the Andean Pact did not match the trade needs of each country, especially after
the foreign debt crisis; the market was too small; and trade activity was directed
mainly to the members of the Andean Pact. Therefore, the Andean countries
were limited in their capacity to generate new foreign exchange, which became
very important for paying the increasing foreign debt. In 1985, the Andean Pact
was practically moribund. Intra-regional trade did not follow the initial industrial
116 C. Hernández-Rodríguez
planning band, only about one-third of the investment programmes (metal-
mechanic, petrochemical, and automobile sectors) were approved.
The external debt crisis that Latin America experienced during the 1980s led
Andean countries to apply adjustment policies. These new policies reduced the
trade preferences that had been established among the Andean countries and,
thus, reduced trade during the mid-1980s. However, the Andean Pact reactivated
itself with the Quito Protocol, which was signed in 1987 and later modified over
the course of several Presidential Meetings. The members eventually established
the CAN in which trade restrictions between members have been reduced. The
CAN is a preferential agreement signed in 1988 by Bolivia, Colombia, Ecuador,
Peru, and Venezuela, only Bolivia being landlocked (see Table 5.1).
The most important modification, the Trujillo Protocol of 1996, resulted in
the name being changed from the ‘Andean Pact’ to the ‘Andean Community of
Nations’ (CAN), a new structural organisation, and a shift in emphasis from
Closed Regionalism (inward integration) to Open Regionalism (outward integra-
tion) with the rest of the world.5 The Declaration of Santa Cruz of the CAN,
signed in January 2002, introduces a new structure of the CET. The CAN
expected to complete the definition of the CET in 2003 and of the Andean
Common Market in 2005.
The CAN is a sub-regional organisation of Latin America with a total population
120 million, an area of 4.7 million square kilometres, and a GDP of US$260 billion
(in 2002). The region is rich in petroleum, minerals, agricultural and forestry
resources. The CAN has a CET (average 11.7 per cent, in 2001) for imports from
third countries. The total imports of the Community in 2000 were US$40 billion
and exports US$57.5 billion. The Community coordinates the position of member
countries and speaks with one voice in the World Trade Organization and other
international and regional forum. In keeping with the Open Regionalism approach,
the CAN has supported the agenda of becoming part of wider economic agreements
such as Mercosur and the attempt of the Free Trade Area of the Americas (FTAA).
Intra-regional trade has already increased tremendously with the existing free
trade zone agreement. Trade between Colombia and Ecuador reportedly
increased by 76 per cent from 1992 to 1993; meanwhile, Venezuela–Colombia
trade grew from $1 billion in 1992 to nearly $2.5 billion in 1993, an astounding
increase of 250 per cent. Similar gains are observed for Peru, with full reinser-
tion into the Community. From 1992 to 1993, Peru experienced an increase of
29 per cent in intra-regional CAN trade.
Figure 5.2 is composed of a set of different graphs corresponding to members
of the CAN. In those graphs we can observe that from 1987 to 2000 the CAN
steadily increased the flows at the intra-region level. However, the most import-
ant trade flows still are placed in regions outside the CAN, at the inter-regional
level. For the entire sample, NAFTA is the first target of their exports, followed
by the EU. The CAN occupies the third place in the exports share, and Mercosur
occupies the last place.
Figure 5.1 and Figure 5.2 show that the establishment of the CAN led, on the
one hand, to an increase of intra-regional trade, as the Open Regionalism view
Figure 5.2 (a) Graphic evolution of exports by country member of the CAN (source:
Excel output using data from the IMF (International Monetary Fund), CEPAL
(Comisión Económica para América Latina y el Caribe), and national
CAN is Andean Community; MS is Mercosur; NAFTA is the North America Free Trade Agreement;
EU is the European Union.

Figure 5.2 (b)

118 C. Hernández-Rodríguez

Figure 5.2 (c)

Figure 5.2 (d)

proposes, and, on the other hand, that there is no contraction of extra-regional

trade in both cases.

5.3 Gravity model

In the gravity model, the antecedents of which are found in Tinbergen (1962),
Pöyhönen (1963), and Linnemann (1966), trade between two countries is ana-
logous to the gravitational force between two objects: directly related to the
countries’ size (or income), and inversely related to the distance between them.
The general hypothesis of the gravity model is that the trade flows between two
countries is an increasing function of their income (GDP) and population, and it
is a decreasing function of the distance between them.
Trade blocs as determinants of trade flows 119

Figure 5.2 (e)

The hypotheses of the augmented gravity model are:

a more income and the population in a pair of countries, increase trade flows;
b more distance between two countries diminish trade flows;
c higher indexes of price in a pair of countries, increase trade flows;
d a common border encourage trade flows in a pair of countries;
e a trade agreement common to a pair of countries encourage trade flows;
f the most important determinants of trade flows in a pair of countries are
income and distance between the two of them.

Since Tinbergen (1962) and Pöyhönen (1963) applied the gravity equation to
analyse international trade flows, the gravity model has become a popular instru-
ment in empirical foreign trade analysis. The model has been successfully
applied to flows of varying types such as migration, foreign direct investment,
and more specifically to international trade flows. According to this model,
exports from country i to country j are explained by their economic sizes (GDP
or GNP), their populations, direct geographical distances, and a set of dummies
incorporating some kind of institutional characteristics common to specific
Theoretical support of the research in this field was originally very poor, but
since the second half of the 1970s several theoretical developments have
appeared in support of the gravity model. Anderson (1979) made the first formal
attempt to derive the gravity equation from a model that assumed product differ-
entiation. Bergstrand (1985, 1989) also explored the theoretical determination of
bilateral trade in a series of papers in which gravity equations were associated
with simple monopolistic competition models. Helpman and Krugman (1985)
used a differentiated product framework with increasing returns to scale to
120 C. Hernández-Rodríguez
justify the gravity model. More recently Deardorff (1998) has proven that the
gravity equation characterises many models and can be justified from standard
trade theories. Finally, Anderson and Wincoop (2003) derived an operational
gravity model based on the manipulation of the CES expenditure system that can
be easily estimated and helps to solve the so-called border puzzle. The differ-
ences in these theories help to explain the various specifications and some diver-
sity in the results of the empirical applications.
There are a huge number of empirical applications in the literature of interna-
tional trade, which have contributed to the improvement of performance of the
gravity equation. Some of them are closer regarding this work. First, in recent
papers, Mátyás (1997, 1998), Breuss and Egger (1999), Egger (2000), and Cheng
and Wall (2002), improved the econometric specification of the gravity equation.
Second, Bergstrand (1985), Helpman (1987), Wei, (1996), Limao and Venables
(1999), Bougheas et al. (1999), and Soloaga and Winters (2001), among others,
contributed to the refinement of the explanatory variables considered in the anal-
ysis and to the addition of new variables.
According to the generalised gravity model of trade, the volume of exports
between pairs of countries, Xij, is a function of their incomes (GDPs), their popu-
lations, their geographical distance, and a set of dummies. For estimation pur-
poses, the gravity model, in log-linear form for a single year, is expressed as:


where, Xij are the exports from country i to country j, i is the exporter, j is the
importer; Yi is the GDP in country i; Yj is the GDP in country j; Li is the popula-
tion of country i; Lj is the population of country j; Dij measures the distance
between the two countries’ (country i and country j) capitals (or economic
centres); dij is a vector of dummy variables, they take the value one when a
certain condition is satisfied (e.g. belonging to a trade bloc), zero otherwise; eij is
the error term of the bilateral trade flow; aij is the constant in the bilateral trade
flow, representing any other factors aiding or preventing trade between pairs of
countries; and bk, for k = 1, . . ., 6, is the corresponding coefficient of the explana-
tory variable or vector of variables.6
The gravity specification, like many models of trade, contains a role for
income, asserting that countries with higher income will trade more. A high level
of income in the exporting country indicates a high level of production, which
increases the availability of goods for exports. Therefore we expect b1 to be
positive. The coefficient of Yj, b2, is also expected to be positive since a high
level of income in the importing country suggests higher imports. The coeffi-
cient estimate for population of the exporters, b3, may be negative or positive
signed, depending on whether the country exports less when it is big (absorption
effect) or whether a big country exports more than a small country (economies
of scale). The coefficient of the importer population, b4, has also an ambiguous
Trade blocs as determinants of trade flows 121
sign, for similar reasons. The distance coefficient, b5, is expected to be negative
since it is a proxy of all possible trade cost sources.7 While the basic gravity
specification relates trade to income and distance, the ‘full’ gravity model (using
Frankel’s (1997) terminology) also allows a role for income per capita, sharing a
common border, and membership in a trading bloc. I also incorporated dummy
variables for trading partners sharing a common border as well as dummy varia-
bles for trading blocs evaluating the effects of preferential trading agreements.
The coefficients of all these trade variables are expected to be positive.
In this research an extended gravity specification, for a single period t (as in
Frankel et al. 1995), is specified as follows:


where, (Xit + Xjt) is the trade volume measured in 1995 US dollars from country i
to country j. This is the sum of Xit, the exports from country i, to country j at
time t, plus Xjt, the exports from country j, to country i at time t; Yit/Yjt is the rela-
tive economic size, in terms of GDP, of country i and country j in the year t,
measured in 1995 US dollars; Lit/Ljt is the relative size, in terms of population, of
country i and country j in the year t; Dij is the distance between the capital (eco-
nomic centres) of country i to the capital (economic centre) of country j; Pit/Pjt
are the relative prices, in terms of consumer price indexes, of country i and
country j in the year t; Aij is a dummy variable for adjacency, it is equal to one if
country i and country j share a common border, zero otherwise; Tijt is a dummy
variable equal to one if country i and country j are members of the same trade
bloc in the year t, zero otherwise; eijt is the error term of the bilateral trade flow
in time t; aijt is the specific effect associated to the trade flow. These effects
allow us to control for omitted variables in the bilateral trade flow; bk, for
k = 1, . . ., 6, is the corresponding coefficient of the explanatory variable.
Income enters positively in the equation, implying that trade between two
medium-sized countries should exceed trade between a small and a large country
(Cyrus 2002). Such an outcome would result from a Helpman and Krugman-
type (1985) model of monopolistic competition, but Deardorff (1998) has shown
that a standard Heckscher–Ohlin framework can produce the same outcome, so
the empirical success of the gravity specification cannot be used as evidence to
support a particular theory of trade.8 Distance is a proxy for transportation costs
and should have a negative coefficient. The dummy variables for the common
border and for trading bloc membership help to determine to what extent trade is
due to geographic or political, as opposed to economic, factors.

5.4 Empirical research

Gravity models are almost always estimated using ordinary least squares (OLS).
This specification contains a potential problem, however: the causality between
122 C. Hernández-Rodríguez
income and trade is not clear-cut. The gravity equation suggests that high income
causes high trade, but perhaps it is trade that instead causes income to be high.
Alternatively, it is also possible that another factor, such as free-market govern-
ment policies, pushes up both income and trade. In that case, the gravity equa-
tion is mis-specified, for income will be correlated with the error term in the
regression; thus, OLS will not provide consistent estimates. Instead, the OLS
estimates will overstate the importance of income. In addition, the bias to the
income coefficient may also bias the other coefficients in the equation; it cannot
be assumed that any of the OLS coefficients can be reliably estimated.
There are two important factors that we want to capture. First, heterogeneity
across countries in trade flows. Second, how the business cycle, or time, will affect
bilateral trade flows. To identify these effects, and hence correctly specify the
econometric model, one requires a pooled time-series of cross-sections (panel
data) of the countries of interest, in this case countries of the Mercosur and CAN.
In the context of the gravity model, we can identify the business cycle or local
or exporting country effects, by treating and estimating them as constants in a
fixed effects model. Although gravity models have been criticised for their lack
of theoretical underpinning, empirically (especially in forecasting) they seem to
perform particularly well, and are therefore well suited for policy analysis.
In constructing the empirical model I consider a sample of nine countries, four
countries from Mercosur (Argentina, Brazil, Paraguay, and Uruguay) and five from
the CAN (Bolivia, Colombia, Ecuador, Peru, and Venezuela). The time period under
study goes from 1987 to 2000. Countries are ordered by pairs in order to capture the
bilateral flows. The data set includes the period 1987–2000. I use the database built
by Frankel et al. (1995), to extract the sample by restricting on observations with no
missing values over the period 1987 to 2000. I obtain a database covering nine coun-
tries, and 36 country-pairs. We thus have a balanced panel structure.
I estimated the gravity model of trade described in equation (5.2), in a panel
data framework. The use of panel data methodology has several advantages over
cross-section analysis. First, panels make it possible to capture the relevant rela-
tionships among variables over time. Second, a major advantage of using panel
data is the ability to monitor the possible unobservable trading-partner pair indi-
vidual effects. When individual effects are omitted, OLS estimates will be biased
if individual effects are correlated with the regressors.
I constructed three different balanced panels. The first panel includes the 36
combined pairs, which includes both the Mercosur and CAN countries. The
second panel includes only pairs of Mercosur countries. The third panel includes
only pairs of the CAN countries. The last two panels are constructed to test the
gravity equation in the intra-regional level.
In these panels, I include six variables belonging to the gravity equation.
Those variables are: (1) Exports, (2) Population, (3) Distance, (4) Price Indexes,
(5) Adjacency, and (6) Trade Agreement. On the one hand, there is Exports
(endogenous variable), and, on the other hand, there are GDP, Population, Dis-
tance, and Price Indexes (exogenous variables). Additionally, there are two other
exogenous variables: those are the dummy variables Adjacency and Trade
Trade blocs as determinants of trade flows 123
Agreement. The dummy variable Adjacency is not included in the second panel
given that adjacency is only present in a pair of countries (Paraguay–Uruguay).
Trade Agreement is included only as an explanatory variable in the first panel. It
does not make sense to include it in the other two panels given the definition of
those panels.

5.4.1 Results
Equation (5.2) is estimated taking into account both fixed and random effects.9
The results are presented in Table 5.2. Fixed effects turn to be the best approach
given that the variance for the estimated coefficients within groups is less than
the one for the estimated coefficients between groups. In the fixed effects
approach we can see the increase in the commercial flows over time.

Table 5.2 Gravity equation for the panels of Mercosur and CAN: random effects

Variables Coefficient (fixed effects) Coefficient (random effects)

GDP (Y) 0.37** (0.07) 0.43** (0.08)

Population (L) 0.12 (0.08) 0.06 (0.09)
Distance (D) 0.97** (0.14) 1.00** (0.16)
Price indexes (P) 0.27* (0.15) 0.15 (0.17)
Adjacencies (A) 1.74** (0.16) 1.78** (0.19)
Trading bloc (T) 2.06** (0.19) 2.09** (0.22)
Constant-α – 2.20 (1.29)
87-α 1.61 –
88-α 1.46 –
89-α 1.56 –
90-α 1.83 –
91-α 2.13 –
92-α 2.33 –
93-α 2.45 –
94-α 2.56 –
95-α 2.98 –
96-α 2.99 –
97-α 3.09 –
98-α 3.26 –
99-α 3.01 –
00-α 3.10 –
R2 0.76 0.53
Adjusted-R2 0.75 0.52
Durbin Watson Statistic 2.00 1.53
Observations 504 504

Sources: Eviews output using data from the IMF (International Monetary Fund), CEPAL (Comisión
Económica para América Latina y el Caribe), and national sources.
Standard error is in parentheses; ** significance of 5%, * significance of 10%.
124 C. Hernández-Rodríguez
Table 5.2 shows that GDP, distance, CPI, and the two dummy variables are signi-
ficant. The signs of the coefficients are the expected signs, with the clear exception
of distance (distance was expected to be negative in the gravity equation). The posit-
ive sign in distance means that distance does not affect inversely the increase in the
trade flows. Then the sign of the variable distance does not undermine the theories
of international trade, but challenges the traditional gravity equation.
However, population is not significant. In order to clarify the non-significance
of population, we look at the correlation matrix and see that population is corre-
lated to GDP ( ρ = 0.84). Collinearity is another problem in the gravity equation,
and population seems to be the cause.
Table 5.2 in the fixed effects column shows that the establishment of the
South American trade blocs (the Mercosur and the CAN) led to an increase of
total regional trade (from the positive sign in the Trading Bloc dummy variable),
as the Open Regionalism view proposes.
The second and third panels are estimated to test the gravity equation for each
Trade Agreement. Table 5.3 shows the results of the fixed effects approach.

Table 5.3 Gravity equation for the panels of Mercosur and CAN: fixed effects

Variables Mercosur CAN

GDP (Y) 0.69** (0.39) –0.07 (0.10)

Population (L) –0.79* (0.38) 0.46** (0.07)
Distance (D) 0.54* (0.19) 0.11 (0.49)
Price indexes (P) 0.61 (0.75) 0.14 (0.48)
Adjacencies (A) – 2.00** (0.41)
87-α 8.48 10.26
88-α 8.47 9.29
89-α 8.87 9.21
90-α 8.79 9.56
91-α 8.92 10.11
92-α 9.11 10.42
93-α 9.49 10.46
94-α 9.74 10.82
95-α 9.99 11.16
96-α 10.11 11.06
97-α 10.21 11.33
98-α 10.25 11.62
99-α 10.01 11.49
00-α 10.17 11.67
R2 0.46 0.50
Adjusted-R2 0.33 0.49
Durbin Watson Statistic 3.26 0.43
Observations 84 140

Sources: Eviews output using data from the IMF (International Monetary Fund), CEPAL (Comisión
Económica para América Latina y el Caribe), and national sources.
Standard error is in parentheses; ** significance of 5%, * significance of 10%.
Trade blocs as determinants of trade flows 125
Table 5.3 shows that, for Mercosur, GDP, population (L), and distance (D)
were significant. But price index (P) was not significant due to collinearity with
the GDP. The only expected sign belongs to the GDP. The coefficient of popula-
tion (L) was expected to be positive, and is negative. One possible explanation is
the high correlation with the GDP ( ρ = 0.92). Distance was expected to have a
negative coefficient and had a positive sign.
Table 5.3 shows that, for the CAN countries only population (L) and adjacen-
cies (A) were significant. The coefficient of population has the expected positive
sign. The coefficient of GDP is negative but not significant. The coefficient of the
dummy adjacency is highly correlated with distance ( ρ = 0.81). But price index (P)
was not significant due to collinearity with the GDP. The only expected sign
belongs to the GDP. The coefficient of population was expected to be positive, and
is negative. One possible explanation is the high correlation with the GDP (0.92).
Distance was expected to have a negative coefficient and also had a positive sign.
Given the results showed in Table 5.3, the gravity equation fails to support
that the main determinants of the trade flows are income and distance. In both
panels the sign of the coefficient of distance is positive and in the CAN panel
income is negative and not even significant.
Figure 5.3 shows a graphic evolution of the fixed effects for both South
American trading blocs. This figure confirms the Open Regionalist view that the
establishment of a trading bloc (in this case the Mercosur and the CAN) leads to
an increase of intra-regional trade flows.

Figure 5.3 Graphic evolution of gravity equation fixed effects for Mercosur and CAN
(source: Fixed effects coefficients (Eviews output) using data from the IMF
(International Monetary Fund), CEPAL (Comisión Económica para América
Latina y el Caribe), and national sources).
126 C. Hernández-Rodríguez
5.4.2  Why is the coefficient of distance positive? Robustness and 
specification tests
One hypothesis for the positive coefficient on distance is that distance is working
as an instrumental variable. To test if distance is such an instrumental variable
we use the Hausman Exogeneity Test. Thus, the next step is to determine
whether distance is an instrumental variable and if it is valid. The first require-
ment of good instruments is that they be highly correlated with the variable for
which they are instrumenting.
The Hausman Exogeneity Test consists in the contrast of a regression in
which distance is the only regressor yt = α0 + α01Dt + εt, and yt = α + βDt + δut + εt,
where u are the residuals of the regression of the variable D on the instrumental
variable DR, ut = D – (θ0 + θ1DRt). This contrast will be done assuming that the
regressor is non-stochastic (null hypothesis). The alternative hypothesis is that
the regressor is stochastic (only the instrumental variables have this property).
The first-stage regression in Table 5.4 shows that distance is indeed highly
significant in explaining trade flows. Hausman Tests were performed for the
coefficients and regressions. Examining Hausman Tests for particular coeffi-
cients allows us to see whether we can trust the OLS results in all cases. The
rejection of the null hypothesis implies that we use distance as an instrument.
The purpose of the Hausman Test is to determine whether there is indeed corre-
lation between GDP (or, more generally, the regressors) and the error term. The
null hypothesis is that there is no correlation, so OLS provides consistent and
efficient estimates; if this is true, then the Instrumental Variables (IV) estimates
should be similar to the OLS estimates (the IV estimates should equal the OLS
estimates plus noise). So, the Hausman Test is a test of equality between the
OLS estimates and the IV estimates. In this case, the null hypothesis is rejected
at a 90 per cent level; in other words, we do not fail to reject the OLS specifica-
tion. It implies that distance must be an endogenous variable.
The second requirement of good instruments is that they be uncorrelated with
the error term. To determine this, tests of over-identifying restrictions were run.
This test involves regressing the residual of the OLS regression on the instru-

Table 5.4 Exogeneity Hausman Test for distance in the gravity equation

Variable Coefficient Std. Error t-Statistic Prob.

DR 1.632954 0.306643 5.32526 0.0000

RESID01 –1.624571 0.370665 –4.38285 0.0001
C –3.697832 2.351061 –1.572836 0.1253
R-squared 0.462191 Akaike AIC 2.923188
Adj. R-squared 0.429596 Schwarz SC 3.055148
F-statistic 14.18001
Prob. (F-statistic) 0.000036 Durbin–Watson stat 2.505928

Sources: Eviews output using data from the IMF (International Monetary Fund), CEPAL (Comisión
Económica para América Latina y el Caribe), and national sources.
Trade blocs as determinants of trade flows 127
ments. I analysed the individual explanatory capacity of the variable distance. It
is very difficult for the goodness of fit of this regression to be highly
Table 5.5 shows that distance is a relevant variable to explain the trade flows.
Interestingly in this case the regression shows the expected negative sign. It may
perhaps be better to examine the adjusted-R2 of this regression, which is never
above 0.100; this indicates that, while the individual instrument is significant, it
has virtually no power to explain the error term.
Recently Bougheas et al. (1999) showed that transport costs are a function
not only of distance but also of public infrastructure. They augmented the gravity
model by introducing additional infrastructure variables (stock of public capital
and length of motorway network). Their model predicts a positive relationship
between the level of infrastructure and the volume of trade, which is supported
using data from European countries. Martinez-Zarzoso and Nowak-Lehmann
(2003) took a further step in this direction by introducing a new infrastructure
index (taking information on roads, paved roads, railroads, and telephones) and
differentiating between exporter and importer infrastructure as explanatory vari-
ables of bilateral trade flows. Due to the unavailability of data in this case I was
unable to reproduce a similar index.

5.5 Conclusions
In order to study the impact of regional integration in South America – which
includes the Mercosur and the CAN – an augmented gravity model approach
was utilised. Data consist of a balanced panel data of 36 trading pairs during 14
years, in the period 1987 to 2000. A gravity equation for trade flows was esti-
mated. Trade flows from country i to country j was used as the dependent vari-
able. Explanatory variables are GDP for both source and destination country
(market size), population of both countries, distance between the two countries,
and dummy variables for adjacency and trade agreement.
The specification is log-linear and the procedure estimation identified colline-
arity, caused by price index (P) and GDP. This could cause some coefficients

Table 5.5 Distance as the only regressor in the gravity equation

Variable Coefficient Std. Error t-Statistic Prob.

C 19.2276495 1.05685158 18.1933299 0.0000

D –1.10523307 0.14774535 –7.48066234 0.0000
R-squared 0.10029443
Adj. R-squared 0.09850218
F-statistic 55.960309
Prob. (F-statistic) 0 Durbin–Watson stat 0.894435

Sources: Eviews output using data from the IMF (International Monetary Fund), CEPAL (Comisión
Económica para América Latina y el Caribe), and national sources.
128 C. Hernández-Rodríguez
like population to have a sign not expected. One of such counterintuitive coeffi-
cients is a negative sign for distance. It seems that in this aspect – a positive
coefficient in the distance variable – the gravity model does not fit the empirical
evidence, at least for this panel. The econometric results, nevertheless, suggest
that the factors that influence the most the flows of trade are geographic factors
(adjacency), followed by size (income).
This work is aimed at exploring how these attempts of regional integration
have increased trade volume and trade flows. The following results are found: the
establishment of Mercosur and the CAN led to an increase of intra-regional trade,
and it has not resulted in a contraction of extra-regional trade. In terms of volume,
with the exception of Brazil, for countries that are members of Mercosur, their
trading bloc is the most important partner. NAFTA is the most important trading
bloc partner for countries of the Andean Pact, and then the EU and Mercosur.
The estimations reveal that both Mercosur and the CAN have increased the
total regional trade as the Open Regionalism view proposes. For the Mercosur
case, there is an increase in the trade flows and trade volume both at the intra-
regional level and the extra-regional level. These findings are in line with the
importance of Mercosur, the customs union, at the intra-regional level.
However, for the CAN extra-regional trade flows are more important, pointing
to the fact that there is little evidence of a significant effect of this Preferential
Agreement in the intra-regional trade as may be expected. In this context, it
seems that the South American trade blocs’ characteristics must be studied to
understand the reasons of the success of Mercosur when compared to the

1 Trade flows are the trade transactions among countries, and are represented by
imports and exports.
2 The reason for the name of ‘gravity model’ is the analogy to Newton’s law of gravity:
Just as the gravitational attraction between any two objects is proportional to the
product of their masses and diminishes with distance, the trade between any two
countries is, other things equal, proportional to the product of their GDPs and
diminishes with distance.
(Krugman and Obstfeld, 2009: 14)
3 The Andean Pact was formed in 1969 to reverse the stagnation of the Latin American
Association of Free Trade and to address the development needs of the Andean coun-
tries (Venezuela, Colombia, Chile, Ecuador, Peru, and Bolivia).
4 The Pact sought to harmonise policies, define a common external tariff, liberalise
intra-regional trade, regulate foreign direct investment in the region, and to organise
production across member Andean countries by encouraging the development of
promising industries. This strategy was consistent with the Import Substitution, or
Closed Regionalism, Model that predominated in Latin America during the 1970s.
According to this model, the government must coordinate economic policies and
regional development plans in order to direct the market towards proposed goals. The
consequence of this model is that protected rent activities develop, mainly in the
industrial sector, which are financed in part by the resources generated by primary-
resource-intensive exports.
Trade blocs as determinants of trade flows 129
5 The establishment of the Andean Free Trade Zone (AFTZ) in 1993 and the Andean
tariff union or the Andean common external tariff in 1995 gave rise to private initia-
tives and innovative rent-seeking activities – instead of protected rent activities – that
aimed at achieving an efficient allocation of resources and exploiting the competitive
advantages of the region. This increasing efficiency and innovation is the main reason
behind the shift towards the Open Regionalism model. The next step for the CAN is
the establishment of the Andean Common Market in the year 2005, as ratified in the
presidential meeting of the Andean countries in June of 2000. This market would
enable the free movement of goods, services, capital, and people.
6 An alternative formulation of equation (5.1) uses per capita income instead of popula-
tion, where YHi (YHj) are the exporter (importer) GDP per capita. This alternative
formulation is equivalent to the formulation in (5.1). The specification given by equa-
tion (5.1) is often used to estimate aggregated exports (Endoh, 2000).
7 Traditionally, the gravity model uses distance to model transport costs.
8 Income per capita is included in a gravity equation in order to gauge the importance
of development rather than mere size. Even if bigger countries trade more, it also
seems that richer countries, whatever their size, engage in more trade, so we expect
the coefficient on GDP per capita also to be positive.
9 Under fixed effects we control years. That is, we try to see the individual effect for
each pair of countries in each year.
10 Since the high number of observations, even a tiny amount of correlation between the
residual and the instruments will cause the instruments to appear invalid.

Aitken, N.D. 1973. ‘The Effect of the EEC and EFTA on European Trade: An Empirical
Cross Section Analysis’, American Economic Review 63 (5): 881–892.
Anderson, J.E. 1979. ‘A Theoretical Foundation for the Gravity Equation’, American
Economic Review 69 (1): 106–116.
Anderson, J.E. and Wincoop, E.V. 2003. ‘Gravity with Gravitas: A Solution to the Border
Puzzle’, American Economic Review 93 (1): 170–192.
Bergstrand, Jeffrey H. 1985. ‘The Gravity Equation in International Trade: Some Micro-
economic Foundations and Empirical Evidence’, Review of Economics and Statistics
67 (3): 471–481.
Bergstrand, Jeffrey H. 1989. ‘The Generalized Gravity Equation, Monopolistic Competi-
tion, and the Factor-Proportions Theory in International Trade’, Review of Economics
and Statistics 71 (1): 143–153.
Bougheas, S., Demetriades, P.O., and Morgenroth, E.L.W. 1999. ‘Infrastructure, Trans-
port Costs and Trade’, Journal of International Economics 47 (1): 169–189.
Breuss, F. and Egger, P. 1999. ‘How Reliable are Estimations of East–West Trade Poten-
tials based on Cross-Section Gravity Analyses?’ Empirica 26 (2): 81–95.
Carrillo, C. and Li, C.A. 2002. ‘Trade Blocks and the Gravity Model: Evidence from
Latin American Countries’, Working Paper (downloaded), Department of Economics,
University of Essex, UK.
Cheng, I-Hui and Wall, Howard J. 2002. ‘Controlling for Heterogeneity in Gravity
Models of Trade and Integration’, revision of Federal Reserve Bank of St Louis
Working Paper 1999-OlOC, p. 33.
Cyrus, Teresa L. 2002. ‘Income in the Gravity Model of Bilateral: Does Endogeneity
Matter? International Trade Journal 16 (2), 161–180.
Deadorff, Alan V. 1998. ‘Determinants of Bilateral Trade: Does Gravity Work in a
130 C. Hernández-Rodríguez
Neoclassical World?’ In: Jeffrey A. Frankel (ed.), The Regionalization of the World
Economy, University of Chicago Press: Chicago, IL, pp. 7–31.
Egger, P. 2000. ‘A Note on the Proper Econometric Specification of the Gravity Equa-
tion’, Economics Letters 66 (1): 25–31.
Endoh, M. 2000. ‘The Transition of Post-War Asia-Pacific Trade Relations’, Journal of
Asian Economics 10 (4): 571–589.
Feenstra, R. 2003. Advanced International Trade: Theory and Evidence, Princeton Uni-
versity Press: Princeton, NJ, Chapter 5.
Feenstra, R., Markusen, James A., and Rose, Andrew K. 2001. ‘Using the Gravity Equa-
tion to Differentiate among Alternative Theories of Trade’, Canadian Journal of Eco-
nomics 34 (2): 430–447.
Frankel, J. 1997. Regional Trading Blocs in the World Economic System, Institute for
International Economics: Washington, DC.
Frankel, J., Stein, Ernesto, and Wei, Shang-Jin. 1995. ‘Trading Blocs and the Americas:
The Natural, the Unnatural, and the Super-Natural’, Journal of Development Eco-
nomics 47 (1): 61–95.
Hamilton, C. and Winters, L.A. 1992. ‘Opening up International Trade with Eastern
Europe’, Economic Policy 14 (April): 77–116.
Helpman, E. 1987. ‘Imperfect Competition and International Trade: Evidence from Four-
teen Industrial Countries’, Journal of Japanese and International Economics 1 (1):
Helpman, E. and Krugman, P. 1985. Market Structure and Foreign Trade, MIT Press:
Cambridge, MA.
Krugman, P. and Obstfeld, M. 2009. International Economics: Theory and Policy,
Pearson-Addison Wesley: Boston, MA.
Limao, N. and Venables, A.J. 1999. ‘Infrastructure, Geographical Disadvantage and
Transport Costs’, Policy Research Working Paper 2257, World Bank.
Linnemann, H. 1966. An Econometric Study of International Trade Flows, North-
Holland: Amsterdam.
Martinez-Zarzoso, Inmaculada and Nowak-Lehmann, Felicitas. 2003. ‘Augmented
Gravity Model: An Empirical Application to Mercosur-European Union Trade Flows’,
Journal of Applied Economics 6 (2): 291–316.
McCallum, John. 1995. ‘National Borders Matter: Canada–US Regional Trade Patterns’,
American Economic Review 85 (3): 615–623.
Mátyás, L. 1997. ‘Proper Econometric Specification of the Gravity Model’, World
Economy 20 (3): 363–368.
Mátyás, L. 1998. ‘The Gravity Model: Some Econometric Considerations’, World
Economy 21 (3): 397–401.
Pöyhönen, P. 1963. ‘A Tentative Model for the Volume of Trade Between Countries’,
Weltwirtschaftliches Archiv 90 (1): 93–99.
Soloaga, I. and Winters, L.A. 2001. ‘Regionalism in the Nineties: What Effect on Trade?’
North American Journal of Economics and Finance 12 (1): 1–29.
Tinbergen, J. 1962. Shaping the World Economy: Suggestions for an International Eco-
nomic Policy, Twentieth Century Fund: New York.
Wei, Shang-Jin. 1996. ‘Intra-National versus International Trade: How Stubborn Are
Nations in Global Integration?’ NBER Working Paper 5531
Part II

Trade reforms and

development experience
Case studies in Latin America
6 Downhill or the long agony of
Argentinian development
Alcino Ferreira Câmara-Neto and Matías Vernengo

6.1 Introduction
The performance of the Argentine economy during the twentieth century is
usually seen as an atypical case of persistent economic decline.1 Income per
capita fell from levels equivalent to Western Europe during the Belle Époque, or
from 80 per cent of the income of the United States to around one-third in more
recent times (see Figure 6.1). Yet, it is important to note, from a long period his-
torical perspective, that decline was not uniform and was basically associated to
two specific events with different causes.
The extraordinary performance of the economy in the last decades of the
nineteenth and the beginning of the twentieth century was the result of the
increased integration with the centre based on the exports of primary commodi-
ties (beef and grains) (Cortés, 1998), and in a very simplified way the decline

Figure 6.1 Income per capita (% of US GDP per capita) (source: Maddison (2001), IFS/
IMF and authors’ calculations).
134 A. Ferreira Câmara-Neto and M. Vernengo
that ensued can be seen as the result of the deconstruction of the process of glo-
balisation that tied the Argentine economy to the world through trade and capital
flows. However, it is important to note that the collapse of the primary-export
model of development in Argentina, as much as in other parts of Latin America
and the peripheral world, were the result of the collapse of the system in its
centre. It was the collapse of the hegemonic position of the United Kingdom that
disorganised the international division of labour, and the Gold Standard system
that underpinned and gave sustainability to British domination. In other words,
the internal conditions in Argentina, in contrast to other countries in the region
(e.g. Mexico, that suffered a revolution), were not central for the change in the
development strategy.
In contrast to conventional wisdom, the second relative decline of income per
capita in Argentina cannot be attributed to the failure of the import substitution
industrialisation (ISI) model, in the economic plane, or to the failures of the pop-
ulism of Peronism or the developmentalist programme of Frondizi, in the polit-
ical plane, simply because the economic performance in the immediate post-war
period up to the 1970s is rather satisfactory.2 Only in the 1980s, the second
important decline in the relative income of Argentina actually took place, that is,
well after the import substitution model was abandoned by the liberalising pol-
icies of the last military government (1976–1983). The policies initiated in 1976
were designed to restore the primary-export model of the Belle Époque, promote
growth and reduce what was perceived as the excessive power of the trade
unions. In contrast to the first change in development strategy, the move away
from ISI was a policy decision of the local elites.
In other words, the relative decline in income per capita took place in two
distinctive periods, associated to the crisis of the export commodity and the ISI
model. For that reason it seems more promising to treat the Argentine story not
as one of continual decline, but as one in which for external and internal circum-
stances the model of development is overturned. The results of the new develop-
ment strategy might be reduced levels of growth and divergence of income per
capita with respect to the centre, but that might not be the overarching reason for
the change in the first place. In the second case, reduced power for trade unions
and lower levels of social conflict might have been seen as more relevant than
high rates of growth, for example.
In this chapter we seek to analyse the relative performance of the Argentine
economy during the ISI period and the subsequent neoliberal model imple-
mented after 1976 and, with the truncation of the Alfonsín government
(1983–1989), maintained during the 1990s up to the collapse of Convertibility in
2001–2002. Finally, the period of recovery between 2003 and 2008 is shown to
fall short of a full break with the neoliberal model implemented after 1976.
The rest of this chapter is subdivided in four sections. The next section dis-
cusses the general tendencies of output, investment and productivity for the
whole period starting in 1950. The following one deals with the macroeconomic
policies. The third section analyses the coordination of the process of investment
and the international insertion of the Argentine economy, while the fourth deals
Argentinian development 135
with income distribution. In the conclusion we discuss the current Argentine
development strategy, and the effects of the international economic crisis that
started in 2008.

6.2 From the hegemonic tie to the commodity boom

In the period that goes from the first government of Juan Domingo Perón
(1946–1955) up to the last military government, which is approximately the
same as the so-called Golden Age of capitalism, the economic performance of
the Argentine economy is not stellar, if compared to some other peripheral coun-
tries, but not dismal either. The rate of growth of income per capita was around
2 per cent (see Table 6.1).
During the whole period, as we have seen, the income per capita remains con-
stant at around 50 per cent of the American one. There is no convergence with
the centre, but the disparities that had increased during the Great Depression at
least levelled off. Given the difficulties imposed by the commodity-export model
and the inevitability of industrialisation after the Great Depression, which had
limited the ability to import manufactured goods, the economy performance was
quite reasonable.
The economic policies associated to the developmental programme of expan-
sion of the domestic market allowed for the acceleration of labour productivity.
The explanation of that is associated to the so-called Kaldor-Verdoorn’s Law
that implies that demand expansion drives labour productivity. Also, the acceler-
ation principle indicates that capacity adjusts to the level of activity. Both rela-
tions seem to be supported by Table 6.1 in which there is a clear correlation
between output growth and productivity and investment behaviour.
It must be noted that the post-war period was an extremely volatile period in
political terms, in many respects worse than in other countries in the region, that
were also quite convoluted. In the Argentine case, the reasons for political insta-
bility are often associated to what Juan Carlos Portantiero (1973) referred to as
the hegemonic tie, that produced a sort of stop-and-go cycle, as suggested by
Kalecki (1971 [1943]) for developed economies, but also more profound ques-
tioning about the pertinence of the strategy of development. The political insta-
bility is reflected in the economic fluctuations as shown in Figure 6.2.

Table 6.1 Growth and productivity

Growth and productivity

1950–1975 1976–2002 2003–2008

Rate of investment (%) 3.1 1.3 9.0

GDP growth (%) 3.4 1.7 8.5
GDP per capita growth (%) 2.1 –0.1 7.5
Labour productivity growth (%) 2.1 0.6 1.6

Source: World Bank and Maddison (2001).

136 A. Ferreira Câmara-Neto and M. Vernengo

Figure 6.2 Business cycle (source: Maddison (2001) and World Bank).

The consolidation of Peronism in the second half of the 1940s led to increas-
ing political force for the process of industrialisation, which according to Portan-
tiero lacked economic force. The incredible economic power of the sectors
associated with primary exports, given their economic success in the past, gave
them sufficient veto power, so to speak, on any development strategy. The
tension between the trade unions and the emerging industrial bourgeoisie, on the
one hand, and the old agrarian elite, on the other, were reflected in the cycles of
political rupture and democratic restoration. The political instability might be
one of the factors that explain the relatively less dynamic performance of the
economy in this period when compared with Brazil and Mexico.3
The change in economic strategy in 1976 despite its relative success, in par-
ticular during the decade that preceded the military coup, indicates that the
decision to arrest the process of industrialisation and weaken the social groups
connected to that process should be seen as a political and not an economic
It must be noted that Argentina had already done the transition associated
with the easy stage of ISI, and the industrial sector occupied a preponderant
place in the economic system in the 1930s. In other words, it could be argued
that faced with the difficult stage of ISI, associated with the installation of a
capital goods sector, the increasing balance of payments problems5 during the
1960s, and the exacerbated class conflict, the captains of industry, some with ties
with the old agrarian elite, decided that the industrial project in Argentina was
not viable.
Argentinian development 137
The economic performance in the subsequent period is usually seen as infe-
rior to the ISI period, but the reasons are not uniquely associated to the change in
the development strategy and the turn towards neoliberalism in 1976. The 1970s
were associated to the disorganisation of the international monetary system,
competitive devaluations, the oil shocks and the acceleration of inflation. In
Argentina, after the so-called ‘Rodrigazo’ – named after Celestino Rodrigo the
Finance Minister at the time – in June 1975, with a maxi-devaluation of 160 per
cent of the peso in nominal terms, and an increase in utilities and energy prices
of around 180 per cent, in addition to nominal wage adjustments, inflation accel-
erated out of control.6
After the mid-1970s the problem of stabilisation dominated all macroeco-
nomic debates and long run questions were relegated to a secondary plane until
the Convertibility Plan in 1991. In that sense, the liberalising policies of Mar-
tínez de Hoz starting in 1976 were defended fundamentally as a part of a stabili-
sation plan, and the change in the development strategy were seen as
instrumental in promoting an environment less prone to inflationary pressures.
In addition the debt crisis, which originated in the Mexican default of August
1982, and exacerbated in the Argentine case by the appreciation of the currency
that was part of the stabilisation plan, led to a collapse of economic growth. In
terms of income per capita the economy actually contracted in this period.
During the 1980s, after redemocratisation, the inability to renegotiate the debt on
a sustainable level imposed a severe external constraint on the economy, which
led, as in the rest of Latin America, to the so-called ‘lost decade’. The external
restriction and the need for permanent devaluations in a context of wage indexa-
tion explain a great deal of the inflationary acceleration during the period.
In the early 1990s after the renegotiation of the external debt, the re-entry in
international financial markets and the stabilisation of prices, economic growth
picked up, but it was a short lived expansion. The frequent crises in Mexico,
Asia, Russia and Brazil starting in 1995 and the straitjacket of Convertibility
implied moderate rates of growth and a profound recession starting in 1998. For
the period as a whole, which goes from 1976 to the collapse of Convertibility in
the 2001–2002 biennium, the rate of growth of income per capita was nil (see
Table 6.1). If we break the period in two, in the first part from 1976 to 1989
income per capita was negative 1.4 per cent, while the second sub-period from
1990 to 2002 it was of only 0.9 per cent per year on average.
The crisis of the Convertibility Plan opened up a new chapter in Argentine
economic policy, and the performance of the economy, up to the 2008 crisis,
was exceptional by historical standards. The rate of growth of income per capita
of more than 7 per cent was higher than the rates associated with the commodity-
export model. Ironically, even though growth was associated to the increased
utilisation of excess capacity, the expansion was ultimately correlated to the
commodity boom, which represents a partial return to the old commodity-export
138 A. Ferreira Câmara-Neto and M. Vernengo
6.3  External constraint, inflation and financialisation
Macroeconomic policy in peripheral countries is more often than not determined
by structural factors associated to the balance of payments, than to short run
issues related to the smoothing of the business cycles, and that is also the case in
Argentina. The overriding short run preoccupation is often related to price
stability. Demand management policies are, then, often limited by the current
account performance, since the recurrent external deficits would lead to external
debt accumulation and foreign exchange crises.
The decomposition of the elements of demand permits to understand whether
internal or external components of demand are central for the expansion of
output.8 The decomposition used here follows the methodology presented in
Lance Taylor (2006). Aggregate supply (X) is defined as the sum of consump-
tion (C), investment (I ) and exports (Ex). The national rate of savings (s) is
defined as income minus consumption over aggregate supply and the propensity
to import (m). The internal and external stances are, then, given by:


The decomposition is shown in Figure 6.3, and it is shown that Argentina

almost always grew below the balance of payments constraint, with external
demand being central for expansion. With the exception of two periods in the
late 1950s/early 1960s and the 1990s internal demand was of secondary impor-
tance, or at least it never expanded significantly to the point that it threatened
the balance of payments position. This suggests that the balance of payments
imposed a severe restriction on the possibility of expanding domestic demand,
and, as a result, of the domestic markets itself. In other words, growth was
always limited by the external constraint and demand management policies
were fundamentally used to stay within the limits imposed by the balance of
It should be noted that the decomposition does not explain the rhythm of
growth, but only its composition between internal and external forces. The evo-
lution of output (X ) does, however, show the rate of growth of the economy. As
it can be seen in Figure 6.3, output expands relatively fast up to the 1980s, stag-
nates for about a decade after that, to briefly recover and completely collapse at
the end of the last century, and shows a brisk upturn in the current century. In
that sense, the 1980s, when growth was pushed by external forces, can be seen
as a decade of export-led stagnation. In other words, the fact that the external
market was the dynamic component of demand does not imply that the economy
was growing fast.
In contrast with East Asia, in Argentina, as in the rest of Latin America,
exports were necessary to service foreign debt obligations, while imports col-
lapsed during the 1980s, and the external market was incapable of promoting
accelerated expansion of the economy. Only after the Brady Plan, with the
Argentinian development 139

Figure 6.3 Output decomposition (source: authors’ calculations).

re-entry into external financial markets, did the economy start to recover.
However, the recovery was short lived with the economy plunging into its worst
crisis in history, even counting the Great Depression, the debt crisis of 1982 and
the 1989 hyperinflation.
To a great extent, from a macroeconomic point of view, the inability to break
with the external restriction was associated to exchange rate policy. During the
ISI period there were extensive foreign exchange controls, and multiple
exchange rates, with pressures from groups associated with the process of indus-
trialisation for appreciated exchange rate in order to facilitate the importation of
capital and intermediary goods. Devaluations tended to be contractionary, and to
favour exporters, often linked to traditional commodity producing sectors (Díaz-
Alejandro, 1965). The conflict between those favouring an appreciated versus a
depreciated exchange rate produced a great political instability, creating
exchange rate volatility with bouts of appreciation followed by maxi-
devaluations (see Figure 6.4).
In addition, the foreign exchange rate was often used as an anchor in stabili-
sation programmes. Figure 6.4 shows two such processes of exchange rate
appreciation linked to stabilisation plans. In the mid-1970s the military regime
introduced a system of pre-announced devaluations (the infamous tablita), and
after April 1991, the Convertibility Plan, which collapsed in 2002, tied the peso
to the dollar.
It should be noted that after the collapse of Bretton Woods it became consid-
erably more difficult to manage the exchange rate, and the pressures from the
140 A. Ferreira Câmara-Neto and M. Vernengo

Figure 6.4 Real exchange rate (source: authors’ calculations).

International Monetary Fund (IMF ) and the United States for a more open finan-
cial account became stronger. It should not be a surprise that the use of an
exchange rate anchor as an instrument of stabilisation, in the 1970s and 1990s,
was accompanied by financial liberalisation and speculative capital inflows that
proved to be excessively volatile to maintain the exchange rate arrangement in
the long run.
In the same vein, the option for a more open financial account also reflects the
interests of the economic groups that prefer an economy more integrated with
international markets, and with a smaller role for domestic industries in the com-
position of the economy, with a reduced role for trade unions. It should be noted
that the process of liberalisation and deindustrialisation did not imply, at least
not initially, a reduced role for the state in fiscal terms.
Table 6.2 shows the primary and global fiscal balances and the debt servicing
spending throughout the decades, beginning with the 1960s. It is quite clear that
in the transition from a developmentalist to the neoliberal model (1970s and
1980s) the primary and global deficits increase, and that a significant adjustment

Table 6.2 Fiscal policy (% GDP)

Primary balance Global balance Interest

1961–1970 –3.4 –4.0 0.6

1971–1980 –6.0 –7.0 1.0
1981–1990 –5.1 –7.0 1.9
1991–2000 0.1 –2.1 2.2
2001–2008 2.4 0.1 2.3

Source: Damill et al. (2003) and ECLAC.

Argentinian development 141
only takes place in the 1990s. Further, the fiscal adjustment is tightened in the
current century, after the collapse of Convertibility.
Table 6.2 also shows the increasing financialisation of public spending with
more than 6.2 per cent of GDP being transferred as interest payments to owners
of public bonds, mostly banks and corporations, and their owners.9 This shows
that the nature of the state’s intervention changed after the debt crisis and the
opening up of the financial account of the balance of payments, and that the
transfer of resources to rentier groups became a central part of public policy.
Finally, it should be noted that monetary policy was passive for most of the
time that exchange rate policy was a central element of stabilisation policy that
was at last successful in the 1990s, in the context of global price stability.10 In
that sense, exchange rate policy had always a short run bias related to stabilisa-
tion instead of a long run preoccupation with external competitiveness.

6.4 The state, foreign capital and spurious competitiveness

The characteristics of the model of development in the post-war period are rela-
tively well known. Fundamentally, the economy was less open, characterised by
higher tariffs, quantitative controls and bureaucratic restrictions to trade, favour-
ing capital imports and discouraging imports of superfluous consumer goods.
Also, a larger participation of the state in the economy, through direct produc-
tion of goods and services, more active purchasing policies and through the
public financial sector, was the norm. For example the Banco de Crédito Indus-
trial Argentino (BCIA), created in 1944, provided at its peak 80 per cent of all
the credit to the manufacturing sector.11
In several areas, the entry of foreign direct investment was stimulated as a
way to promote transfer of technology, in particular, in the petrochemical and
metal-mechanic complex, like the automobile industry (Sourrouille et al., 1985,
p. 39). It must be noted that the relative openness with respect to foreign capital
began in 1953, still during Perón’s government, allowing foreign firms greater
freedom to repatriate profits to their country of origin. It is in this period of the
1950s that several foreign multinational groups like Fiat, Mercedes-Benz,
Siemens, Bayer, established their local affiliates.
Further, the national oil company, Yacimientos Petroliferos Fiscales (YPF ),
signed, in this period, contracts for exploitation of oilfields with foreign com-
panies, in particular Standard Oil, which reveals that, in contrast to Brazil during
Vargas or Mexico during Cardenas, the participation of foreign capital in the
energy sector during the government of Perón was significant. In this sense, it is
important to avoid the typical simplification that equates the Peronist period with
the dominance of national developmentalist groups with populist tendencies.
Even though the influence of foreign capital was smaller than in the so-called
developmentalist government of Arturo Frondizi (1958–1962), it is still not true
that foreign capital was excluded in the Peronist period.
The industrialisation process in Argentina, as much as in the rest of Latin
America, was more dependent on foreign capital and imported technology than
142 A. Ferreira Câmara-Neto and M. Vernengo
in other peripheral regions. The participation of foreign capital was not central in
terms of its volume, as noted by Altamir et al. (1967), but for its strategic char-
acter connected to sectors with high value added and complex technologies.12
The boom in foreign direct investment (FDI) in the 1990s is related to the
process of privatisation rather than capital formation per se.13
Table 6.3 shows the participation of public capital in the process of invest-
ment during the ISI period, and its subsequent decline after liberalisation. It is
important to note that the participation of the public sector in gross capital for-
mation in the last period, associated to the commodity boom, did not lead to a
recovery of the levels of the ISI period. Another important element of the tempo-
ral trajectory of investment is that the relative fall after 1976 can be entirely
attributed to the decline in public investment. In fact, in the last period private
investment increased, but the change in private investment was insufficient to
compensate the fall in public capital formation.
Finally, another characteristic of investment in Argentina during the ISI
process is related to the dimension of the firms that, when compared with enter-
prises of other countries, were relatively small if measured by number of
employees or by energetic capacity per establishment (Vitelli, 1999). Part of the
problem of smaller firms is that for scale and scope reasons they tend to be less
technologically dynamic. The larger dependence on foreign capital and the larger
number of small domestic firms might be part of the explanation for the reduced
innovativeness of Argentine firms.
The process of industrialisation until the 1970s did not lead to a heavy burden
of foreign debt accumulation. The disproportional growth of foreign debt, and
the use of public firms as a vehicle for external borrowing and of the central
bank for the nationalisation of private debt were a phenomenon of the 1970s and
of the international context associated with the recycling of the petrodollars. The
liberalisation policies of this period exacerbated the problems by facilitating
imports and the movements of funds, which eventually materialised in capital
In Table 6.4 we can see that the external debt as a proportion of gross national
income was approximately 19.1 per cent in 1970 and grew to 85 per cent during
the 1970s, more than double its growth during the following decade. Addition-
ally, we can see that debt not only grew in the 1970s, but continued to grow in
the 1980s, after the debt crisis, and 1990s after further liberalisation policies had

Table 6.3 Investment composition

1960–1975 1976–2002 2003–2006

Public investment 7.2 4.3 2.3

Private investment 15.3 15.5 17.1
Total investment 22.5 19.8 19.4

Source: World Bank and Indec.

Argentinian development 143
Table 6.4 Debt sustainability indicators

1970 1980 1990 2000 2007

Debt/GNP 19.1 35.6 46.0 53.3 50.0

Debt/exports ND 242.4 373.7 380.4 174.0
Interest/exports ND 37.3 37.0 70.5 13.0

Source: World Bank.

been implemented.14 It is only after the 2002 default that the burden of debt eases
up, and debt servicing falls from 70 per cent to 13 per cent of exports.
In the last period, the fall in the interest to export ratio, not only results from
the default and renegotiation of the debt by the Kirchner government, but also
from the export performance. Table 6.5 shows the performance of manufactur-
ing exports and of the exports of high technological content within total manu-
facturing exports.
An increase in manufacturing exports took place in the 1980s and it was
maintained in the following decade, and an increase in the export of manufac-
tures intensive in the use of technology. It must be noted that the majority of
Argentinian exports are still of traditional products, and that the exports of goods
intensive in the use of more complex technology are small when compared to
other peripheral countries.
To some extent, the lack of export dynamism reflects the difficulties con-
nected to what Fajnzylber (1989) called spurious competitiveness. In his view,
competitiveness resulted less from technological capabilities than with the
advantages associated with the specialisation in natural resources and primary
commodities, low wages, devalued exchange rates and the combination of
domestic protection with excessive export subsidies. However, it is important to
qualify Fajnzylber’s argument for the Argentine case. While it is true that the
Argentine economy is still very much specialised in commodity exports, and that
income distribution had an impact in the changes in the productive structure, it is
far from clear that the exchange rate or commercial policies were active in the
expansion of exports.
Arguably, Argentinian governments made an explicit effort to return to the
old commodity export model, in which the comparative advantages would be
explicitly associated to the Ricardian competitiveness of the agro-pastoral

Table 6.5 Manufacturing exports (%)

1980 1990 2000

Manufactures 35.0 52.2 52.1

High technology 26.8 23.6 38.5

Source: authors’ calculations.

144 A. Ferreira Câmara-Neto and M. Vernengo
sector.15 In that sense, the question of distribution, and lower wages, should not
be separated from the strategy of reprimarisation of the Argentine economy.

6.5 Inequality and reprimarisation

If there were a question that shows the negative consequences of the productive
transformation of the last 30 years that would be the evolution of income distri-
bution. Income per capita was at its highest at the beginning of the twentieth
century, but only at the end of the ISI period that income distribution reached
levels that were comparable with those of more equalitarian developed
Figure 6.5 shows the evolution for Argentina and the average of selected
Latin American countries (Brazil, Chile, Colombia, Mexico, Peru and Vene-
zuela). In the 1950s the Gini coefficient was around 40, but by the 1970s it had
fallen to around 35, not very different from Southern European countries.
However, after the coup d’état of 1976, there is a significant increase in the Gini,
converging to the Latin American average over time. The Gini coefficient
usually shows the difference between wages, rather than the evolution of wages
in total income.
Figure 6.6 shows the evolution of wages with respect to income per capita. It
is in the 1950s that we find the highest remuneration for workers. Also, the graph
suggests that the inflationary acceleration of the 1970s was instrumental for
wage contraction. In that sense, the second Peronist government in the 1970s
was incapable to promote wage redistribution.
It is important to note that the recovery of wages since 2003 was insufficient
to bring them back to the levels of the 1960s. The recovery in wages should be
seen as a modest concession, by the Kirchner administration, on a model of
development that still favours an open economy that is integrated mainly through

Figure 6.5 Gini coefficients (source: authors’ calculations).

Argentinian development 145

Figure 6.6 Wage share (source: Llach and Gerchunoff (2004) and Indec, and authors’

the exports of commodities and maintains a relatively devalued currency to facil-

itate those exports.
Finally, it should be noted that distribution of land in Argentina remains rela-
tively unequal. The Gini coefficient for land distribution is of approximately 80,
a level significantly higher than that of more equalitarian countries like Cuba
after the Revolution, with a Gini of slightly less than 60, or Mexico after Cárde-
nas’ agrarian reform, also below 60 (IFAD, 2001). In that sense, the pro-agrarian
bias of the development strategy tends to maintain and exacerbate the deeply
entrenched inequalities of the Argentine society.

6.6 Concluding remarks

The recent economic history of Argentina can be seen as composed of three dis-
tinctive periods. The first period can be described as a successful, in terms of
growth but not necessarily in terms of income distribution, integration to the
world economy on the basis of commodity exports, which was not sustainable
after the disintegration of the international division of labour during the inter-
war period. In particular, after the Great Depression it was clear that the export
commodity model was incapable of incorporating surplus labour, in part because
it was highly productive.
The second period was also very successful in terms of growth, albeit less than
the previous, but considerably more equalitarian, and more capable of incorporat-
ing surplus labour. The limits to the import substitution development strategy that
characterised the second period were fundamentally political in nature. The inher-
ent difficulties of the process of industrialisation were exacerbated by the strength
of labour, the increasing social conflicts of the 1960s, and the pressures of the Cold
War and the fears of the Cuban Revolution, leading in the 1970s to a change in the
model of development and the implementation of the liberalisation strategy.
146 A. Ferreira Câmara-Neto and M. Vernengo
The new development strategy presupposed that a return to the old export
commodity model was feasible and desirable, in spite of the significant changes
in the global economy. For example, while for the first globalisation the integra-
tion was with a hegemon that imported commodities, the United Kingdom, in
this renewed phase the hegemonic power is a competitor that produces commod-
ities, the United States. In addition, it seems that the countries that succeeded, at
least to some extent, to close the gap with the developed world have followed
the path of manufacturing exports and the development of the domestic market
(Amsden, 2001).
From that point of view, even if the change in development strategy is
comprehensible, the possibilities for success are limited at best. The good per-
formance in the 2003 to 2008 period should not be seen as a new phase, but
simply as the result of the functioning of a development strategy that was imple-
mented in the 1970s, and that worked fundamentally as the result of the positive
terms of trade shock. In this case one must hope for prices of commodities to
continue to go up forever. Hope springs eternal!

1 According to Della Paolera and Gallo (2003, p. 373) the Argentine decline remains a
puzzle, ultimately explained by micro and macro institutional failures.
2 Diaz-Alejandro’s (1970, p. 129) classic book is paradigmatic in its enthronisation of
Peronism as the villain of the Argentine process of development. According to him
the Peronist government was keener on redistribution and higher rates of consumption
for the masses that were pursued at the cost of lower rates of capital accumulation.
These views are much favoured by current Argentine historiography (e.g. Llach and
Gerchunoff, 2004).
3 According to Basualdo (2006), even though there might have been a hegemonic tie
until the early 1960s, by the mid to late part of that decade it had dissipated. That
would, in part, explain the accelerated industrial expansion between 1966 and 1974.
4 Jorge Schvarzer (1983, p. 15) argues that not only in Argentina, but also in the whole
Southern Cone, the political logic overpowered economic reasoning in the determina-
tion of the development strategy.
5 Jorge Katz and Bernardo Kosacoff (1989, p. 16) noted the relevance of the inelasticity
of primary exports as a permanent restriction to economic growth.
6 See Rapoport (2005, pp. 571–573) for a discussion of the economic policies in this
7 See Pérez Caldentey and Vernengo (2008a) for a discussion of the current model of
development in Latin America.
8 The assumption is that the level of output is determined by the autonomous com-
ponents of demand along Keynesian lines. The idea of the external constraint to accu-
mulation is based on Prebisch (1949), and has been more recently formalised by
Thirlwall (1979).
9 It should be clear that the simple analogy between developmentalist and Keynesian
policies, the latter in the conventional sense of favouring deficits, lacks support. For a
more accurate description of the role of fiscal deficits in the process of economic
development, and Keynes’ views on the issue see Câmara-Neto and Vernengo
10 For different interpretations of the Convertibility Plan see Della Paolera and Taylor
(2001) and Pérez Caldentey and Vernengo (2008b).
Argentinian development 147
11 For example, the imports of non-durable consumption goods fell from 23.3 per cent to
less than 6 per cent in the first government of Perón. On the other hand, the imports of
capital goods increased from 3 per cent to 17 per cent, at the peak of the investment
process in the post-war period (Rapoport, 2005, p. 358).
12 During the period of the commodity export model, foreign capital was certainly
higher. The participation of foreign capital in investment fell from 38 per cent
between 1900 and 1909 to around 3 per cent in 1953. After the fall of the first Per-
onist government in 1955, there was a significant increase in foreign participation in
investment reaching 13 per cent in 1959 (Altamir et al., 1967).
13 Chudnovsky and López (2002) provide a lengthy discussion of the role of multina-
tional firms in the economy during the 1990s.
14 It must be noted that the debt crisis in Argentina, as in the rest of Latin America, was
the result of the international situation characterised by large liquidity stocks and
higher rates of interest in the United States. Internal conditions were considerably less
relevant (Calcagno, 1988, p. 45).
15 It must be noted that since the 1970s there has been an agriculturisation of the Argen-
tine primary sector, with an increasing role for grain production, particularly soy-
beans, and a reduced role for cattle. The good performance of the agribusiness in the
1990s took place despite the overvaluation of the currency, and of the increase in
wages and retentions to exports in the current decade (Barsky and Gelman, 2001).

Altamir, O., Santamaría, H. and Sourrouille, J. (1967) ‘Los instrumentos de promoción
industrial en la posguerra’, Desarrollo Económico, 7(27), pp. 361–376.
Amsden, A. (2001) The rise of the rest, New York: Oxford University Press.
Barsky, O. and Gelman, J. (2001) Historia del agro argentino, Buenos Aires: Sudameri-
Basualdo, E. (2006) Estudios de Historia Económica Argentina, Buenos Aires: Siglo
Calcagno, A.E. (1988) La perversa deuda, Buenos Aires: Legasa.
Câmara-Neto, A.F. and Vernengo, M. (2004–5) ‘Fiscal policy and the Washington Con-
sensus’, Journal of Post Keynesian Economics, 27(2), pp. 333–343.
Chudnovsky, D. and López, A. (2002) ‘The strategies of the multinationals in the 1990s
Argentina’, Revista de la CEPAL, No. 76, April, pp. 151–166.
Cortés Conde, R. (1998) Progreso y declinación de la economía argentina, Buenos Aires:
Fondo de Cultura Económica.
Damill, M., Frenkel, R. and Juvenal, L. (2003) ‘Las cuentas públicas y la crisis de la con-
vertibilidad en la Argentina’, Desarrollo Económico, 43(170), pp. 203–229.
Della Paolera, G. and Gallo, E. (2003) ‘Epilogue: the Argentine puzzle’, in G. Della
Paolera and A. Taylor (eds) The new economic history of Argentina, Cambridge: Cam-
bridge University Press.
Della Paolera, G. and Taylor, A. (2001) Straining at the anchor, Chicago, IL: University
of Chicago Press.
Díaz-Alejandro, C.F. (1965) Exchange-rate devaluation in a semi-industrialized country:
the experience of Argentina, Cambridge: MIT Press.
Díaz-Alejandro, C.F. (1970) Ensayos sobre la historia económica argentina, Buenos
Aires: Amorrortu.
Fajnzylber, F. (1989) ‘Industrialización en América Latina: de la “caja negra” al “casill-
ero vacío” ’, Cuadernos de la CEPAL, No. 60, Santiago de Chile.
148 A. Ferreira Câmara-Neto and M. Vernengo
IFAD (International Fund for Agricultural Development) (2001) Rural Poverty Report,
Rome: IFAD.
Kalecki, M. (1971 [1943]) ‘Political aspects of full employment’, in Selected essays on
the dynamics of capitalist economies, Cambridge: Cambridge University Press, pp.
Katz, J. and Kosacoff, B. (1989) El proceso de industrialización en la Argentina: evolu-
ción, retroceso y prospectiva, Buenos Aires: CEAL/CEPAL.
Llach, L. and Gerchunoff, P. (2004) Entre la equidad y el crecimiento: Ascenso y caída
de la economía argentina, 1880–2002, Buenos Aires: Siglo XXI.
Maddison, A. (2001) The world economy: a millennial perspective, Paris: OECD.
Pérez Caldentey, E. and Vernengo, M. (2008a) ‘A tale of two monetary reforms: Argen-
tinean convertibility in historical perspective’, Studi e Note di Economia, 12(2),
pp. 139–170.
Pérez Caldentey, E. and Vernengo, M. (2008b) ‘Back to the future: Latin America’s
current development strategy’, Ideas Working Paper Series, No. 07/2008.
Portantiero, J.C. (1973) ‘Clases dominantes y crisis politica en la Argentina actual’, in O.
Braun (ed.) El capitalismo argentino en crisis, Buenos Aires: Siglo XXI, pp. 73–118.
Prebisch, Raúl (1949) ‘O desenvolvimento da América Latina e seus principais proble-
mas’, Revista Brasileira de Economia, 3(3), pp. 47–111.
Rapoport, M. (2005) Historia económica, política y social de la Argentina, Buenos Aires:
Schvarzer, J. (1983) Martínez de Hoz: La lógica política de la política económica,
Buenos Aires: CISEA.
Sourrouille, J. (1985) Trasnacionalisación y Política Económica en Argentina, Buenos
Aires: Centro Editor de América Latina.
Taylor, L. (2006) ‘External liberalization, in Asia, socialist Europe and Brazil’, in L.
Taylor (ed.) External liberalisation, in Asia, socialist Europe and Brazil, New York:
Oxford University Press.
Thirlwall, A. (1979) ‘The Balance of Payments Constraint as an Explanation of Interna-
tional Growth Rate Differences’, Banca Nazionale del Lavoro Quarterly Review,
32(128), pp. 45–53.
Vitelli, G. (1999) Los dos siglos de la Argentina, Buenos Aires: Prendergast.
7 The determinants of FDI in Chile
A gravity model approach
Matteo Grazzi1

7.1 Introduction
Since 1980, foreign direct investment (FDI) flows have grown at remarkable
rates. From 1991 to 2000 the outflows averaged a yearly rise of over 28 per cent.
Even if the FDI trends slowed considerably in the early 2000s, in the last years
FDI has started to increase at substantial rates again. In 2005 the growth was of
29 per cent and in 2006 FDI was 38 per cent, reaching a considerable amount of
US$1.3 trillion (UNCTAD, 2007). According to preliminary estimates, global
FDI flows in 2007 reached the record level of US$1.8 trillion (ECLAC, 2007),
which would correspond to an increase of 29 per cent with respect to the previ-
ous year.
These figures reflect not only a greater level of cross-border mergers and acqui-
sitions (M&As) among developed countries, but also an impressive surge in flows
towards developing countries. With a net flow of over $300 billion in 2006, FDI
has become the most important source of foreign financing for emerging econo-
mies. Furthermore, the overall benefit of FDI is well documented in literature.
Given a correct host-country policy and a minimum level of development, most
scientific work shows that FDI generally triggers technology spillovers, helps to
create human capital, contributes to international trade integration, leads to a more
competitive business environment and enhances enterprise development. All these
factors stimulate economic growth, which is a prerequisite for alleviating poverty
in developing countries. Furthermore, beyond the strictly economic benefits, if
addressed correctly, FDIs may help to improve environmental and social con-
ditions in the host country, for example by transferring less polluting technologies
and leading to more socially responsible corporate policies (OECD, 2002).
In the global situation described above, it is of primary importance for devel-
oping countries to find ways to look more and more attractive in the eyes of
investors, in order to receive higher FDI flow amounts, especially of the kind
that will bring the greatest benefit to a country in terms of investment levels, job
creation, higher-value-added activities and innovation. Consequently, an urgent
research topic for the development economist is to analyse the foreign invest-
ment flows into developing countries and clearly identify their main determi-
nants, in order to offer useful indications for policy-makers.
150 M. Grazzi
The aim of this chapter is to analyse the determinants of FDI inflows into
Chile from 1985 to 2005 through the estimation of a gravity equation. Chile is a
relatively small and resource-rich country, and has been highly successful in
attracting FDI. In fact, FDI has represented an important contribution to the sus-
tained economic development of Chile in the last decades. Since 1990, when the
country returned to democracy, Chile has undertaken an active foreign policy
targeted to the full integration in the international arena. A main pillar of this
strategy has been constituted by the signature of a high number of bilateral eco-
nomic agreements, such as Bilateral Investment Treaties (BITs), Double Taxa-
tion Treaties (DTT) and Free Trade Agreements (FTAs). The validity of such
treaties as instruments to raise the level of FDI into a country has been ques-
tioned by several scholars and, to date, there is still controversial empirical evid-
ence. This work contributes to this debate.
The chapter is structured as follows. The second section analyses trends and
characteristics of FDI inflows into Chile, with a particular attention on the geo-
graphical origin and the sector distribution. In the third section the empirical
methodology is presented. The fourth section contains data and variables
description. Then, the results are presented and commented in the fifth section.
Finally, some conclusions and policy implications are provided in the last

7.2  FDI to Chile: facts and figures
With a population of 16.3 million people as of 2005, Chile is often referred to as
one of South America’s most stable and prosperous nations. However, despite
being Latin America’s fastest-growing economy during the 1990s, the country
experienced moderate economic downturns at the end of the past decade, as a
result of the unfavourable economic conditions that the Asian financial crisis
brought with it.
Mining is the dominant sector of the national economy, accounting for the
largest share of GDP: in 2006, copper mining alone2 contributed for 21.6 per
cent of GDP while manufacturing industry for 12.8 per cent. The agricultural
sector accounts for a mere 5.5 per cent of GDP, while services and the industrial
sectors account respectively for 47.7 per cent and 46.8 per cent of GDP (World
Bank, 2005). Mine products are the main export product as they account for
about half of the total export value. Apart from minerals, Chile also exports
wood products, fish and fishmeal, fruits and wine. Main imports are petroleum,
wheat, capital goods, spare parts and raw materials. Its chief trading partners are
European Union nations, the United States, Japan and Brazil. Specifically
between the years of 1998 and 2006, the share of export of goods and services of
Chilean GDP has impressively increased (from 26.3 per cent to 45.4 per cent),
growing not only in quantity but also in variety (Economist Intelligence Unit,
2007). Although the mining sector remains prevalent, with a dominating copper
industry (43 per cent of merchandise export earnings in 2002–2006), the diversi-
fication of trade has stimulated the developing of new export-oriented industries,
FDI in Chile 151
such as cellulose, salmon, fruit, meat, wine, ethanol. In fact, the dependence of
the economy on mineral prices, together with the production of adequate food
for its population, is the major economic problem of the country.
Since the 1970s, Chile has based its national development strategy on open-
ness to foreign investment. The Chilean Constitution grants no discrimination
against foreigners, and national treatment is generally guaranteed in the legisla-
tion. Nevertheless, it is only since the return of full democracy in the country, in
1990, that investment flows have markedly increased. The country’s business-
friendly environment, based on certainty of law and transparency, coupled with
political stability and the signing of numerous investment treaties effectively
attracted a large amount of foreign capital during the 1990s. The FDI inflow was
$287 million in 1980, 661 in 1990, 4,860 in 2000 and 6,667 in 2005. Today, at
$475.76 FDI per capita, Chile is the third largest recipient of foreign investment
in Latin America (after Brazil and Mexico) in absolute terms and the second in
per capita figures after Trinidad and Tobago.
According to the 2006 World Investment Report, published by the United
Nations Conference on Trade and Development (UNCTAD), the FDI stock in
Chile passed from 30.0 per cent of GDP in 1990 to 64.6 per cent in 2005. These
numbers are impressive, especially if compared with the average world percent-
age of 22.7 per cent in 2005 and that for developing countries, 27.0 per cent.
Between 1974 and 2005, the total gross materialized FDI in Chile was US$78.1
billion, 89 per cent of which entered the country after 1990. If we consider net
inflows, US$52.1 billion entered the country during this period. In the 1990s,

Figure 7.1 FDI flows into Chile 1985–2006 (in nominal US$ million) (source: author’s
elaboration on data provided by the Chilean Foreign Investment Committee).
152 M. Grazzi
FDI gross inflows represented an annual average of 6.4 per cent of Chile’s GDP,
rising to an annual average of 8 per cent between 1995 and 2000. After this unin-
terrupted surge in FDI, the amount of foreign investment entering Chile
decreased significantly in 2000 as a consequence of difficult international eco-
nomic conditions, which affected global FDI flows in almost all countries. The
amount of FDI inflows dramatically fell from the record high of US$9.9 billion
in 1999 to a negative peak of US$5 billion. In the early 2000s, the negative trend
continued as a consequence of the global collapse of the M&As market, the main
driving force of FDI flows around the world and in Chile in the previous decade.
Many reasons explain such a phenomenon: not only the global economic uncer-
tainty, but also a drop in share prices, lower corporate earnings and a cut in the
budget of multinational companies for expansion. Furthermore, various deep
financial crises provoked in these years heavy losses to a broad range of inves-
tors across several Latin American countries. As a result the region was per-
ceived as a high risk investment area exactly in the moment when risk-adverse
shareholders were pushing multinational firms to perform safer investments.
The Chilean Foreign Investment Committee (2006) indicated that, in the case
of Chile, a greater use of the local capital market by foreign investors could have
distorted recent FDI figures. The high liquidity and the dynamism of the Chilean
financial sector, combined with historically low interest rates, encouraged a
growing number of foreign companies to raise finance locally, through either
borrowing from local banks, issuing bonds on the domestic market or reinvest-
ing locally their profits. Such a trend could be reflected negatively in the FDI
inflows into the country. However, starting from 2004, FDI inflows into Chile
showed an increase that reflected a renewed interest in mergers and acquisitions
and the development of new projects in mining, telecommunications and infra-
structure. In 2006, the provisional figures for gross FDI inflows had reached
US$5.9 billion.
Table 7.1 presents Chilean FDI gross inflows by sector of destination, from
1974 to 2006.3 Considering the entire period, mining accounted for 33.1 per
cent; services for 19.6 per cent; electricity, gas and water industries for 19.2
per cent; manufacturing for 12.7 per cent; transport and communication for 11.6
per cent; construction for 2.3 per cent and agriculture, forestry and fishing for
0.5 per cent each. However, on evaluating the variation of percentages over time,
it is interesting to note that the mining sector, traditionally the most important
recipient sector by far, represented almost 47 per cent of total inflows until 1990,
but progressively lost importance thereafter. In the 2001–2005 period its share
fell to 22.04 per cent, and in 2006 FDI gross inflows directed to the mining
sector amounted to 23.1 per cent of the total. This decrease has been offset by
higher inflows into the industries involved in the privatization process: the trans-
port and communication sector (28 per cent in the 2001–2005 period, even if just
16 per cent in 2006) and the electricity, gas and water sector (28.9 per cent in the
2001–2005 period and 28 per cent in 2006). From 1997 to 2001, because of the
privatization process, Chile saw a dramatic surge in M&As activity, mainly in
the services, electricity and telecommunications sectors.
Table 7.1 FDI inflows into Chile by sector, 1974–2006 (in nominal US$ million and percentages)

Period 1974–1989 1996–2000 2001–2005 2006

Sector $ % $ % $ % $ %

Agriculture and livestock 79,626 1.56 88,004 0.31 13,164 0.06 101,168 0.21
Forestry 10,695 0.21 105,930 0.37 4,590 0.02 110,520 0.22
Fishing and aquaculture 17,853 0.35 152,803 0.54 25,239 0.12 178,042 0.36
Mining and quarrying 2,398,769 46.92 6,766,117 23.89 4,622,950 22.04 11,389,067 23.1
Food, beverages and tobacco 215,471 4.22 1,115,662 3.94 520,144 2.48 1,635,806 3.32
Wood and paper products, printing 100,152 1.96 243,005 0.86 613,909 2.93 856,914 1.74
and publishing
Chemical, rubber and plastics 524,511 10.26 1,313,636 4.64 918,210 4.38 2,231,846 4.53
Other manufacturing industries 315,902 6.18 470,693 1.66 289,105 1.38 759,798 1.54
Electricity, gas and water supply 0 0 7,675,990 27.1 6,067,620 28.92 13,743,610 27.87
Construction 122,953 2.41 657,763 2.32 598,763 2.85 1,256,526 2.55
Wholesale and retail trade 153,659 3.01 750,290 2.65 313,173 1.49 1,063,463 2.16
Transport and storage 21,061 0.41 276,832 0.98 198,329 0.95 475,161 0.96
Communications 283,110 5.54 1,826,667 6.45 5,688,993 27.12 7,515,660 15.24
Financial services 779,061 15.24 4,069,726 14.37 292,413 1.39 4,362,139 8.85
Insurance 37,015 0.72 1,410,331 4.98 391,499 1.87 1,801,830 3.65
Engineering and business services 22,963 0.45 270,938 0.96 169,052 0.81 439,990 0.89
Sewage, sanitation and similar services 299 0.01 505,946 1.79 5,845 0.03 511,791 1.04
Other services 28,887 0.57 626,736 2.21 245,864 1.17 872,600 1.77
Total 5,111,987 100 28,327,069 100 20,978,862 100 49,305,931 100

Source: author’s elaboration on data provided by the Chilean Foreign Investment Committee.
154 M. Grazzi
After 2001, the trend in FDI changed again, shifting towards projects that
require smaller amounts of capital but have a high impact in terms of job crea-
tion and technology transfer. These smaller, high-impact projects range from
software development initiatives, call centres and shared services centres, to new
investment in the manufacturing and agribusiness sectors. Unfortunately, a large
part of these investments do not pass through the mechanism of the DL600,
therefore are not considered in the Chilean Foreign Investment Committee data
(see Note 3). From 2004, in line with the regained dynamism of the M&As
market on the global stage, the Chilean market experienced a new wave of
acquisitions of local companies by foreign firms.
As for geographical origin of the investments, in the period between 1974 and
2006, materialized FDI through D.L.600 mechanism has been mainly originated
in the European Union (41 per cent), in the United States (25 per cent), in
Canada (16 per cent), Australia (5 per cent) and Japan (3 per cent).
However, considering data on disaggregated time periods, the share of invest-
ment originated in the EU is growing, while that from the United States is
decreasing. In 1995–1998 the percentages of the European Union and of the US
were similar: 35.33 per cent and 35.58 per cent (see Table 7.2). Yet in
1999–2002 the gap became consistent, with 51.99 per cent for the EU against
21.94 per cent for the US. In 2003–2006, while the EU maintained its invest-
ment level (51.79 per cent), the United States accounted only for 7.78 per cent of
the total investment that arrived in the country. Within the European Union, the
main investor is Spain, followed by the United Kingdom. Finally, it is worth

Figure 7.2 FDI inflows into Chile by geographical origin 1974–2006 (percentages)
(source: author’s elaboration on data provided by the Chilean Foreign
Investment Committee).
FDI in Chile 155
Table 7.2 FDI inflows into Chile by geographical origin, 1974–2006 (in nominal US$
million and percentages)

Period 1995–1998 1999–2002 2003–2006

Origin $ % $ % $ %

European Union 4,631,610 35.33 10,747,520 51.99 5,620,957 51.79

United States 4,665,013 35.58 4,536,305 21.94 844,658 7.78
Canada 2,279,577 17.39 2,316,572 11.21 2,476,485 22.82
Australia 330,942 2.52 847,965 4.10 677,905 6.25
Japan 339,270 2.59 509,425 2.46 148,117 1.36
South America 319,219 2.43 247,652 1.20 0 0.00
Others 544,834 4.16 1,467,834 7.10 1,085,943 10.00
Total 13,110,465 100.00 20,673,273 100.00 10,854,065 100.00

Source: author’s elaboration on data provided by the Chilean Foreign Investment Committee.

noting that the investment originated from other South American countries is
very low, and it has decreased over time.

7.3  Empirical methodology
The gravity equation is a common well-established formulation for statistical
analyses of bilateral flows between different geographical entities (Head, 2003).
It is based on the relationship described in the ‘Law of Universal Gravitation’
postulated by Isaac Newton in 1687, which states that the attractive force
between two point masses is proportional to the product of the two masses and
inversely proportional to the square of the distance between them. Tinbergen
(1962) was the first to propose that the same functional form could be utilized to
describe international trade flows and, since, the gravity equation has gained
increasing popularity thanks to its remarkable explanatory capacity.
The most commonly used version of the gravity equation (Bergstrand, 1985)
is presented in equation (7.1):

All displayed equations in this chapter should have thin spaces around 
mathematical symbols.


where Xij,t is the amount of exports from country i to country j, at time t. The
variable Yi,t is the GDP of country i at the time t, while Yj,t is the GDP of country
j at the time t. Dij is the distance between the two countries i and j. The variable
Aij represents various factors that may either stimulate or reduce trade between
country i and country j. Finally, ζij,t is a log-normally distributed error term, with
E (ln (ζij,t)) = 0.
156 M. Grazzi
In this elementary version, the equation indicates that the volume of export
between two countries depends positively on their economic size and negatively
by the transport costs, captured by the absolute distance between their economic
centres. This specification is usually presented in a logarithm format, where log-
arithms are all natural logarithms. Therefore, the coefficients of the independent
variables represent the elasticities of the export flows to host and source coun-
try’s GDPs and distance between the countries. Even this basic version of the
gravity equation is a quite simple but powerful instrument to explain bilateral
flows between countries. However, a large part of variation in these flows
remains unexplained. In order to refine the estimation of the gravity equation,
most scholars ‘augment’ it, adding other variables, with various theoretical justi-
fications. The most commonly used include: income per capita, adjacency,
common language, colonial links and border effects.
In the last 40 years the gravity equation has been one of the most popular
techniques to analyse bilateral trade flows, but only recently it has been applied
to the analysis of cross-border capital movements or cross-border multinational
activities. A plausible reason appears to be the fact that, while in the trade liter-
ature the gravity model has a robust conceptual basis, the use of this model for
the case of FDI is still somewhat ad hoc (Stein and Daude, 2007), although there
have been recent developments in laying the theoretical underpinning of the
gravity equation applied to cross-border investment (e.g. Kleinert and Toubal,
2005). However, given the empirical similarity of FDI trends with those of trade
flows, the gravity instrument has often been employed in estimating bilateral
FDI flows, usually with good results (e.g. Brainard, 1997; Brenton et al., 1998;
Eaton and Tamura, 1994).
In this chapter, the gravity model is used to predict the volume of FDI flows
into Chile, and it implies a particular version of the basic gravity equation, where
the dependent variable is now specified as inward FDI flows into Chile, while
the variables regarding the host country on the right hand side do not vary by
country, but just over time. In this first basic model, we add just the per capita
GDP of the investor country as an additional variable, to proxy for the develop-
ment level of a country. So the gravity equation becomes:


where the dependent variable is the flows of FDI into Chile from country j at
time t and the independent variables are: Chilean GDP at time t, GDP of country
j at time t, GDP per capita of country j at time t and geographical distance
between Chile and country j.
A main problem of this specification is that FDI flows between two countries
are often equal to zero and they may also be negative (e.g. due to repatriation of
profits). Yet the gravity equation predicts that flows between countries are
always positive, even if they may be small. Furthermore, the usual natural loga-
rithm transformation cannot operate on zero and negative values. Several strat-
egies have been suggested in the literature to handle the presence of zero flows.4
FDI in Chile 157
A common solution is to simply reduce the considered sample to the positive
observations, in order to avoid the estimation problem related to zero and negat-
ive values (e.g. Rose, 2000). But this approach does not consider the fact that
also zero flows may convey important information for the analysis.
A second possible approach suggests substituting the non-positive values with
a small constant, in order to be able to utilize these observations in the log-linear
model. Wang and Winters (1991) and Raballand (2003), among others, followed
this strategy. However, this approach seems generally unsatisfactory, because
the inserted value is arbitrary and does not necessarily reflect the underlying
expected value (Linders and de Groot, 2006). Moreover, in the case of bilateral
trade, zero flows mostly occur between very small or very distant countries, spe-
cifically for FDI, zero or negative values are much more frequent. For example,
in our sample, 281 observations are non-positive and omitting them from the
analysis could seriously underestimate the effects of variables with a negative
impact on the flows.
A third solution implies the application of the so-called ‘Inverse Hyperbolic
Sine Function’ to the dependent variable, instead of the natural logarithm func-
tion. Such a transformation, first proposed by Johnston (1949), does not truncate
or eliminate values of the dependent variable and it allows performing the
regression on the entire available sample. This way of imposing the inverse
hyperbolic sine (IHS) function to the dependent variable while imposing natural
logarithm on the dependent variables has been used in studies on household
wealth (Burbidge et al., 1988 and Carroll et al., 1999) and it has been recently
proposed by Kristjánsdóttir (2005) as applicable in gravity modelling.
In this investigation, the estimation strategy will consist of two different
stages. In the first one, the basic equation (7.2) is estimated using the three pro-
cedures5 described above in a panel-estimation framework. Even if the large
majority of studies estimate the gravity equation using ordinary least squares
(OLS) either on cross-sectional or pooled cross-sectional data, two distinct
advantages in using a panel data model rather than the traditional linear regres-
sion model can be identified. First, a panel data model captures both cross-
section and time-series variation of the dependent variable. Second, it allows the
measurement not only of the effects that observable variables have on the
dependent variable, but also of the effects of relevant unobservable or non-
measurable variables.
While observable variables are normally considered into the model, the
unobservable variables are incorporated into the model depending upon
whether a fixed-effect (FE) or random-effects (RE) model is used in estima-
tion. In the RE model, the unobservable factors that differentiate cross-section
units are assumed to be characterized as randomly distributed variables. Even
if Matyas (1997) and Egger (2000) claim that the correct econometric specifi-
cation should be estimated as a fixed effect model rather than as a random
model one, given that the cross-sectional units of this analysis are the countries
which are the source of FDI flows into Chile and that these partners vary con-
siderably by culture, religion, political system and many other factors, it seems
158 M. Grazzi
quite reasonable to assume here that the differences between them are ran-
domly distributed. Among others, Bevan and Estrin (2004) used random
effects to estimate a gravity equation for FDI flows. Moreover, in all our speci-
fications we perform the Breusch-Pagan test and in all cases the result is a
rejection of the null hypothesis of no-random effect. We have performed also
the Hausman test to assess the independence of the random effects from the
explanatory variable and the result is the acceptance of the null hypothesis.
Considering that, the adoption of the random-effects estimation model method
is justified.
An additional econometric problem arises from the fact that we are estimating
a single-country panel model. In fact, the variable GDPC,t is an exclusively time-
variant regressor, and then the direct estimation of the gravity equation would
prevent us to identify the real impact of this variable on FDI inflows. A possible
solution, offered by Földvári (2006) is to include in the regression a polynomial
time trend, in order to capture the time-variant effects without losing the possi-
bility to identify the coefficients of the only time-variant regressor. Con-
sequently, a polynomial time trend of fourth order is included in the regression
Then, applying the natural logarithm transformation6 and adding the time
trend, the basic equations to be estimated becomes for each of the three speci-
fications, respectively:




where zj,t is replaced by ej,t, so that E(ln zj,t) = E(ej,t) = 0.

The estimation of equations (7.3), (7.4) and (7.5) constitutes the first step of
our empirical analysis, while the second step is the estimation of an ‘augmented’
version of the basic equation, designed to take into consideration the impact of
the entry into force of different models of international economic agreement,
such as the FTAs, DTTs and BITs, and the membership of APEC (Asian-Pacific
Economic Cooperation). Moreover, we control for trade flows, classic ‘gravity
dummies’ (common language and sharing a common border)7 and a set of ‘trade
blocs’ dummy, which indicate the membership of the partner country in a
regional economic agreement (EU, NAFTA, CAN and Mercosur). Thus, the
augmented equations for the three specifications are:
FDI in Chile 159




7.4  Data and variables description
Data on FDI used in this chapter have been provided by the Chilean Foreign
Investment Committee. These data cover FDI flows into Chile from 47 different
countries over a 16 years period, from 1990 to 2005, covering almost 98 per cent
of the total inflows through the Decree 600 Mechanism. The data cover the
annual FDI flows into Chile from: Argentina, Australia, Austria, Belgium,
Bolivia, Brazil, Canada, China, Colombia, Costa Rica, Denmark, Dominican
Republic, Ecuador, Egypt, Finland, France, Germany, Greece, Honduras,
Iceland, India, Ireland, Israel, Italy, Japan, Republic of Korea, Liberia, Luxem-
bourg, Malaysia, Mexico, Netherlands, New Zealand, Norway, Panama, Para-
guay, Peru, Portugal, Romania, Singapore, South Africa, Spain, Sweden,
Switzerland, United Kingdom, United States, Uruguay and Venezuela. Inflows
from Bahamas, Barbados, Bermuda, Cayman Islands and Netherlands Antilles
have not been considered, because they are presumably not the real source of the
investment but just a channel used for fiscal reasons. The number of observa-
tions is therefore 47 times 16, equal to 752.
As for trade data, the chapter makes use of the IMF data set ‘Direction of
Trade’, which provides bilateral merchandise trade between over 180 countries
from 1948 to 2004. The FDI and trade values are reported in current US dollars,
which should approximate a correction for the different exchange rates across
countries. In order to obtain real figures, these values have been deflated by
using the US GDP deflator furnished by the Global Development Finance data-
base of the World Bank, whose base year (i.e. where GDP deflator value is equal
Table 7.3 Variable description

Variable Description Unit of measurement

FDIj,t FDI flows into Chile from country j in time t through DL600 Mechanism Billions of US dollars (2000 base)
ln (FDIj,t) Natural log transformation of FDIj,t Natural logarithm
Sinh–1 (FDIj,t) Inverse hyperbolic sine transformation of FDIj,t
GDPj,t GDP of country j in time t Trillions of US dollars (2000 base)
ln(GDPj,t) Natural log transformation of GDPj,t Natural logarithm
GDPPCj,t GDP per capita of country j in time t US dollars (2000 base)
ln (GDPPC) Natural log transformation of GDPPCj,t Natural logarithm
GDPC,t GDP of Chile in time t Trillions of US dollars (2000 base)
ln (GDPC,t) Natural log transformation of GDPC,t Natural logarithm
Dj Distance between Santiago de Chile and the most important city in country j Kilometres
ln (Dj) Natural log transformation of Dj Natural logarithm
Adjj Dummy variable = 1 if country j shares a border with Chile Dummy variable
Lanj Dummy variable = 1 if the language of country j is Spanish Dummy variable
FTAj,t Dummy variable = 1 if an FTA is in force between Chile and country j in time t Dummy variable
DTTj,t Dummy variable = 1 if DTT is in force between Chile and country j in time t Dummy Variable
BITj,t Dummy variable = 1 if a BIT is in force between Chile and country j in time t Dummy variable
EUj,t Dummy variable = 1 if country j was member of the EU in time t Dummy variable
CANj,t Dummy variable = 1 if country j was member of the CAN in time t Dummy variable
Mercosurj,t Dummy variable = 1 if country j was member of the Mercosur in time t Dummy variable
NAFTAj,t Dummy variable = 1 if country j was member of the NAFTA in time t Dummy variable
APECj,t Dummy variable = 1 if country j and Chile were both member of APEC in time t Dummy variable
IMPj,t Trade flows from country j into Chile in time t Thousands of US dollars (2000 base)
ln (IMPj,t) Natural log transformation of IMPj,t Natural logarithm
EXPj,t Trade flows from Chile into country j in time t Thousands of US dollars (2000 base)
ln (EXPj,t) Natural log transformation of EXPj,t Natural ;ogarithm
t, t2, t3, t4 Polynomial time trend

Source: author’s elaboration.

FDI in Chile 161
Table 7.4 Summary statistics

Variable Obs. Mean Std. Dev. Min. Max.

FDIj,t 752 73.49 301.15 –390.36 4,720
Ln (FDIj,t) 471 9.03 2.72 0.69 15.37
Ln (FDIj,t +1) 752 5.63 4.85 0 15.35
Sinh–1 (FDIj,t) 752 5.83 5.68 –13.57 16.06
GDPj,t 752 563 1,440 121 11,100
ln(GDPj,t) 752 25.47 1.93 18.61 30.04
GDPPCj,t 752 13,512.36 11,625.46 56.52 49,979.78
ln (GDPPC) 752 8.87 1.38 4.03 10.82
GDPC,t 752 67.9 15.3 40.8 93.9
ln (GDPC,t) 752 24.91 0.24 24.43 25.27
Dj 752 9,899.73 4,761.72 1,128.32 19,079.88
ln (Dj) 752 9.01 0.71 7.03 9.86
IMPj,t 752 328 626 0 5,540
ln (IMPj,t) 721 18.21 2.03 8.64 22.44
EXPj,t 752 305 629 0 4,540
ln (EXPj,t) 716 18.1 2.14 9.06 22.24

Source: author’s elaboration on own calculations.

to 100) is 2000. Population and real GDP data have been obtained from the
World Bank’s World Development Indicators. Distance refers to the geographi-
cal absolute distance between the most important cities/agglomerations (in terms
of population) of two countries. It has been obtained from the CEPII (Centre
d’Etudes Prospectives et d’Informations Internationales) database. Information
on common borders and common language has been taken from the same data-
base. As for the list of BITs, DTTs and FTA signed by Chile and currently in
force, information has been obtained from the Chilean Foreign Investment Com-
mittee. Variables description is summarized in Table 7.3, while descriptive stat-
istics are provided in Table 7.4.

7.5 Results
In Table 7.5 the main empirical results obtained from the estimation of the basic
gravity model (equations 7.3, 7.4 and 7.5) are presented. The results obtained are
generally consistent with the theoretical expectations, and they are generally not
dependent on the adopted estimation methodology. Coefficients of source coun-
tries’ GDP and GDP per capita and distance are positive and highly significant
in all the three columns, while the coefficient of Chilean GDP is not significant,
irrespective of methodology.
The variable distance is statistically significant at the conventional level and
negative in all the estimations. The estimates suggest that a 1 per cent increase in
distance leads, ceteris paribus and on average, to a decline in FDI that varies
between 0.79 per cent and 1.66 per cent, depending on the estimation methodol-
ogy. The fact that transport costs impact negatively on the level of investment
162 M. Grazzi
Table 7.5 Random effects estimation of the baseline equation

Dep. variable Ln (FDIj,t) Ln (FDIj,t +1) Sinh–1 (FDIj,t)

Variables (1) (2) (3)

ln(GDj,t) 0.827*** 1.256*** 1.313***

(0.12) (0.23) (0.26)
ln (GDPPC) 0.596*** 0.909*** 0.929***
(0.18) (0.32) (0.36)
ln (GDPC,t) –5.109 –5.521 –2.612
(5.66) (7.83) (10.7)
ln (Dj) –0.795** –1.664*** –1.453**
(0.31) (0.57) (0.63)
tt 0.321 1.316* 1.228
(0.58) (0.79) (1.08)
tt2 0.112 –0.0629 –0.11
(0.12) (0.16) (0.21)
tt3 –0.0153 –0.00383 0.000815
(0.013) (0.016) (0.022)
tt4 0.000521 0.000225 0.0000895
(0.00041) (0.00053) (0.00072)
Observations 471 752 752
Number of partner 47 47 47
Adjusted R-squared 0.4197 0.4288 0.3411
Breush-Pagan 318.12 923.9 504.61
Prob> chi2 (1) 0 0 0
Hausman Test 7.4 7.29 4.27
Prob> chi2 (6) 0.1925 0.2946 0.6401

Source: author’s elaboration on own calculations.

Robust standard errors in parentheses.
* Significant at 10%; ** significant at 5%; *** significant at 1%.

flows into Chile confirms the theoretical expectations. In fact, Chile is a natural-
resource economy and the main recipient of foreign investment is the mining
sector. Therefore, one could expect that the major share of the FDI is ‘vertical’
in nature and that transport costs should affect FDI negatively.
The coefficients for the market size of the source economies are highly statisti-
cally significant and also in line with the theoretical predictions. An increase in
the source country market size of 1 per cent would increase FDI inflows by the
range 0.82 per cent and 1.31 per cent, depending on the estimation methodology.
A positive coefficient that is statistically significant is also obtained for the esti-
mates regarding the GDP per capita of the source countries in all the three
columns. The theoretical assumption that more developed economies engage in
more FDI is therefore confirmed from the data. Naturally, richer countries gener-
ally source international investments. As for the coefficients corresponding to the
variable related to the size of the host country market (Chilean GDP), they are not
statistically significant. Therefore the hypothesis that foreign investors are
FDI in Chile 163
attracted to a larger domestic market is rejected. It strengthens the argument that
FDI in Chile has been basically resource-seeking rather than market-seeking.
Further confirmation of these findings is offered by the results of the estima-
tion of the augmented model, which are presented in Table 7.6. In fact, the coef-
ficients for GDP and GDP per capita of the source country maintain a positive
sign and high significance independently of the estimation methodology adopted.
Moreover, the dimension of their impact on FDI flows is very similar to that
estimated for the basic model: the coefficient for source country GDP now varies
from 0.92 to 1.3, while that regarding source country GDP per capita varies from
0.82 to 1.07.
The negative sign of the coefficient on distance (not significant only in the
case of the estimation performed in the restricted sample of positive observation)
and the non-significance of Chilean GDP also in the augmented model estima-
tion confirm the hypothesis that FDI into Chile is basically a vertical, resource-
seeking investment. The coefficients for the trade blocs dummies are not
significantly different from zero in all the three regressions. It implies that it is
not possible to identify an impact on the FDI flow into Chile deriving from
belonging to a particular regional organization. Not statistically significant are
also the coefficients regarding the border dummy and the language dummy. It
seems reasonable to assume that in the case of capital flows, adjacency is not as
important as in trade flows. As for the coefficients of trade flows, neither the
coefficients on Chilean imports nor those on Chilean exports are statistically
significant at conventional levels. The lack of significance of the import’s coeffi-
cient is consistent with the hypothesis of prevalent ‘vertical’ FDI,8 while one
would have expected a significant positive relation between FDI inflows and
Chilean exports. In fact, if the firms choose to invest abroad to seek resources or
lower factor prices and the investment is designed to supply the source country’s
market, the theoretical prediction is a positive and significant coefficient.
Finally, we analyse the coefficients related to the ‘Chilean Economic Foreign
Policy’. The entry into force of a BIT between Chile and a FDI source country is
found to have a considerable significant and positive impact in both the estima-
tion methodologies which use the entire sample. The predicted positive variation
is given, on average and ceteris paribus, by [exp(coefficient) – 1]*100. The size
of this impact is then computed to be either an increase of 144.73 per cent or
216.76 per cent.
A small set of papers has empirically assessed the impact of BITs on FDI.
UNCTAD (1998) has not found any statistical evidence that they could attract
FDI in addition to traditional determinants, using a cross-section analysis based
on about 100 countries. Hallward-Driemeier (2003), using a 20 years panel
dataset, confirmed the lack of an independent effect of BITs of FDIs, after
having controlled for other determinants of country attractiveness (especially
institutional quality). However, Egger and Pfaffermayer (2004) in a recent study
provided a more optimistic vision, finding a positive impact.
In this chapter, given the single-country approach adopted, the institutional
quality is not as important as in a multi-country analysis, and our results are in
Table 7.6 Random effects estimation of the augmented equation

Dep. variable Ln (FDIj,t) Ln (FDIj,t +1) Sinh–1 (FDIj,t)

Variables (4) (5) (6)

ln(GDP j,t) 0.916*** 1.280*** 1.309***

(0.26) (0.31) (0.38)
ln (GDPPC) 0.825*** 0.969*** 1.073***
(0.25) (0.35) (0.4)
ln (GDPC,t) –5.309 –3.666 –0.218
(5.75) (8.13) (11.2)
ln (D j) –0.77 –2.384** –2.858**
(0.689) (1.09) (1.32)
ln (IMP j,t) –0.114 –0.0155 –0.0525
(0.18) (0.18) (0.23)
ln (EXP j,t) 0.208 0.087 0.193
(0.14) (0.17) (0.22)
Adj j –0.289 –2.505 –3.571
(1.06) (1.76) (1.98)
Lan j 0.744 0.43 0.14
(0.8) (1.25) (1.43)
EU j,t –0.284 –0.394 –0.783
(0.58) (0.93) (1.07)
NAFTA j,t 0.473 1.504 1.351
(0.66) (1.07) (1.41)
Mercosur j,t –0.268 –1.91 –3.290*
(0.93) (1.38) (1.83)
CAN j,t –0.77 –1.529 –1.682
(0.92) (1.44) (1.65)
APEC j,t –0.436 –0.586 –0.464
(0.42) (0.59) (0.8)
BIT j,t 0.0937 0.895** 1.153**
(0.28) (0.38) (0.51)
DTT j,t 0.408 1.194* 0.973
(0.46) (0.61) (0.84)
FTA j,t –0.0372 0.0354 0.315
(0.4) (0.5) (0.69)
tt 0.284 1.021 0.94
(0.61) (0.87) (1.2)
tt2 0.14 –0.0454 –0.112
(0.13) (0.17) (0.23)
tt3 –0.0189 –0.00449 0.00212
(0.013) (0.017) (0.024)
tt4 0.000642 0.000228 0.00006
(0.00042) (0.00055) (0.00076)
Observations 450 701 701
Number of partner 45 47 47
Adjusted R-squared 0.45 0.51. 0.42
Breush-Pagan 215.13 327.53 151.15
Prob> chi2 (1) 0.00 0.00 0.00
Hausman Test 18.2 32.73 21.28
Prob> chi2 (15) 0.1502 0.112 0.1281

Source: author’s elaboration on own calculations.

Robust standard errors in parenthesis
* Significant at 10%; ** significant at 5%; *** significant at 1%.
FDI in Chile 165
line with the findings of Egger and Pfaffermayer. A possible area for further
research in the field is the study of the impact of signing a BIT on the probability
of establishing an FDI flow, passing from zero to a positive investment. With
regards to other economic treaties, no statistical significance for the entry into
force of a FTA has been found, while there is a small evidence of a positive
impact of DTT, with only one of the adopted estimation methodologies. Finally,
the simultaneous membership of Chile and the source country in the APEC
Forum does not seem to have a positive impact on the amount of FDI flows.

7.6 Conclusions
We have examined the determinants of FDI inflows into Chile from 1990 to
2005 through the estimation of a gravity model. First, the obtained results
confirm the validity of the gravity equation as a valuable instrument to analyse
not only bilateral trade flows but also bilateral capital flows. FDI is found to be
negatively affected by distance and positively affected by the source country
GDP and GDP per capita. Chilean market size, proxied by GDP, does not influ-
ence FDI inflows. These findings confirm the hypothesis that investments in
Chile are mostly of the vertical type, i.e. resource-seeking, rather than market-
Second, I have used the gravity instrument to evaluate the impact of the
Chilean Economic Foreign Policy, that is, the signature of bilateral economic
agreements, such as BITs, DTTs and FTAs, on the amount of FDI flows. The
entry into force of a BIT is found to have a positive and significant impact on the
FDI inflows, while there is little or no evidence of a significant effect of DTTs
and FTAs. The finding that the signature of a BIT with a country increases the
investment flows originated in that country is particularly interesting for the con-
nected strong policy implication, not only for developing countries, but also for
industrialized countries devoted to global development. In fact, BITs represent
one possible way for a developed country to boost private investment flows
towards poor countries. But ‘rich countries do not have many direct policy
instruments to improve the amount of FDI received by poor’ (Mayer, 2006),
because it usually implies policy measures that need to be implemented in the
host country rather than in the source country. Consequently, BITs are a valuable
instrument to drive foreign investment flows.

1 The contents are the sole responsibility of the author, and do not necessarily reflect the
position of the Inter-American Development Bank.
2 Chile is the world’s largest producer of copper with a mine output in 2009 of nearly 5.4
million tonnes, corresponding to over one-third of world copper mine production
(ICSG, 2010).
3 FDI can enter into Chile through different legal mechanisms. The detailed sectoral data
used in this section covers only the investment made through the Decree Law 600
(DL600). Under this regime, whose use is optional, foreign investors bringing capital,
166 M. Grazzi
physical goods or other forms of investment into Chile may ask to sign a foreign
investment contract with the state of Chile. Although it is the channel through which
most of the investment enters the country, this fact could potentially bias the analysis:
investment in sectors where using this legal mechanism is particularly favourable may
be overestimated, while other sectors where a large part of the inflows arrives through
other legal channels may be underestimated.
4 Besides the solutions described in this section, other methodologies used in the literat-
ure to solve the problem are: the estimation of a TOBIT model (Eaton and Tamura,
1994; Wei, 1998); a Heckman selection model (Razin et al., 2003; Razin et al., 2005)
and of a Poisson model (Silva and Tenreyro, 2006).
5 Following Razin et al. (2003), a one-dollar value (with the log equalling zero) as a
common low value for the value of the FDI flows in the zero or negative FDI observations.
6 The natural logarithm transformation is applied both on the sample constituted just by the
positive observation and on the entire sample, having substituted the zero values with
one. In this case, the 14 negative observations have been considered as zero values.
7 The common colonization dummy has not been considered, because, in the case of
Chile, it is equal to the common language (Spanish) dummy.
9 In case of ‘horizontal’ FDI, countries would substitute exports to the partner market by
localizing the production in that country.

Bergstrand, J. (1985), ‘The gravity equation in international trade: some microeconomic
foundations and empirical evidence’, Review of Economics and Statistics, 67 (3):
Bevan, A. and Estrin, S. (2004), ‘The determinants of foreign direct investment into Euro-
pean transition economies’, Journal of Comparative Economics, 32 (4): 775–787.
Brainard, S.L. (1997), ‘An empirical assessment of the Proximity-Concentration Trade-
off between Multinational Sales and Trade’, American Economic Review 87 (4):
Brenton, P., Di Mauro, F. and Lucke, M. (1998), ‘Economic integration and FDI: an
empirical analysis of foreign investment in the EU and in Central and Eastern Europe’,
Empirica, 26 (2): 95–121.
Burbidge, J.B., Magee, L. and Robb, A.L. (1988), ‘Alternative transformations to handle
extreme values of the dependent variable’, Journal of the American Statistical Associ-
ation, 83 (March): 123–127.
Carroll, C., Dynan, K. and Krane, S. (1999), ‘Unemployment Risk and Precautionary
Wealth: Evidence from Households’, Balance Sheets. Federal Reserve Board, Finance
and Economics Discussion Paper No. 1999–15.
Chilean Foreign Investment Committee (2002), online, available at: http://ww.foreignin- (accessed October 2009).
Eaton, J. and Tamura, A. (1994), ‘Bilateralism and regionalism in Japanese and U.S. trade
and direct foreign investment patterns’, Journal of the Japanese and International
Economies, 8 (4): 478–510.
Egger, P. (2000), ‘A note on the proper econometric specification of the gravity equa-
tion’, Economics Letters, 66 (1): 25–31.
Egger, P. and Pfaffermayer, M. (2004), ‘The effect of bilateral investment treaties on
FDI’, Journal of Comparative Economics, 32 (4): 788–804.
ECLAC (2007), ‘Foreign Investment in Latin America and the Caribbean’, Santiago de
FDI in Chile 167
Economist Intelligence Unit (EIU) (2007), ‘Country Report Chile’, London.
Földvári, P. (2006) ‘The economic impact of the European integration on the Nether-
lands: A quantitative analysis of foreign trade and foreign direct investments’, Tekst,
Proefschrift Universiteit Utrecht.
Hallward-Driemeier, M. (2003), ‘Do bilateral investment treaties attract foreign direct
investment? Only a bit . . . and they could bite’, World Bank Policy Research Working
Paper No. 3121.
Head, K. (2003), ‘Gravity for beginners’, online, available at:
(accessed October 2009).
International Copper Study Group (ICSG) (2010), ‘The World Copper Factbook, 2010’.
Kleinert, J. and Toubal, F. (2005), ‘Gravity for FDI’, CEGE Discussion Paper no. 46,
Göttingen University.
Kristjánsdóttir, Helga (2005), ‘Determinants of foreign direct investment in Iceland’, CAM
Centre for Applied Microeconometrics Working Paper no. 15, University of Copenhagen.
Johnston, N.L. (1949), ‘Systems of frequency curves generated by methods of transla-
tion’, Biometrika, 36 (1/2): 149–176.
Linders, G.M. and de Groot, H.L.F. (2006), ‘Estimation of the gravity equation in the
presence of zero flow’, Tinbergen Institute Discussion Paper 2006-072/3.
Mayer, T. (2006), ‘Policy coherence for development: A background paper on foreign
direct investment’, OECD Centre for Development Working Paper No. 253.
Matyas, L. (1997), ‘Proper econometric specification of the gravity model’, World
Economy, 20 (3): 363–368.
OECD (2002), ‘Foreign direct investment for development: Maximising benefits, mini-
mising costs’, CIME Report.
Raballand, G. (2003), ‘Determinants of the negative impact of being landlocked on trade:
An empirical investigation through the Central Asian case’, Comparative Economic
Studies, 45 (4): 520–536.
Razin, A., Rubinstein, Y. and Sadka, E. (2003), ‘Fixed costs and FDI: The conflicting
effects of productivity shocks’, NBER Working Paper 10864, National Bureau of Eco-
nomic Research, Inc.
Razin, A., Sadka, E. and Tong, H. (2005), ‘Bilateral FDI flows: Threshold barriers and
productivity shocks’, NBER Working Papers 11639, National Bureau of Economic
Research, Inc.
Rose, A. (2000), ‘One money, one market: The effect of common currency on trade’,
Economic Policy, 15 (30): 7–46.
Silva, J.M.C. and Tenreyro, S. (2006), ‘The log of gravity’, Review of Economics and
Statistics, 88 (4): 641–658.
Stein, E. and Daude, C. (2007), ‘Longitude matters: Time zones and the location of
foreign direct investment’, Journal of International Economics, 71 (1): 96–112.
Tinbergen, J. (1962), Shaping the World Economy, New York: Twentieth Century Fund.
UNCTAD (various issues), ‘World Investment Report’.
Wang, Z.K. and Winters, L.A. (1991), ‘The trading potential of Eastern Europe’, CEPR
Discussion Paper 610.
Wei, S. (1998), ‘How taxing is corruption on international investors?’ NBER Working
Paper 6030.
World Bank (2005), ‘Country Profile for Chile’, online, available at:,,pagePK:180619~theSiteP
K:136917,00.html (accessed October 2009).
8 Assessment of the distributive
impact of trade reforms in
Fernando Borraz, Daniel Ferrés and Máximo Rossi

8.1 Introduction
Marked trade liberalisation and deeper regional integration have characterised
Uruguay’s trade policy since the early 1970s. As Uruguay entered Mercosur in
the early 1990s, the country embarked in a plan for tariff reduction at the
regional and extra-zone levels. The coming years may imply further trade
opening, particularly as the WTO Doha Development Agenda, the EU–Mercosur
Association Agreement and other initiatives under negotiation will enter the
implementation phase.
Even in the textbook case, traditional trade theory acknowledges that although
the gains from trade might be positive for a country as a whole, they might not
be distributed evenly across all the groups. There is nowadays an increasing
concern throughout the region over the asymmetric distribution of costs and
benefits of trade integration. One of the initial objectives of Doha was to amelio-
rate inequalities between rich and poor countries. In this context, it is fundamen-
tal to determine whether trade integration can be regarded as poverty reduction
policy or, on the contrary, if it may be associated with intensified poverty
Regressive outcomes are more likely in the absence of complementary
domestic reforms and policies that would help maximise gains from trade,
protect the most vulnerable from transitional costs and ensure an equitable distri-
bution of net gains. In order to design a domestic complementary agenda, it is
therefore of the utmost importance to generate empirical evidence to determine
the distributional impacts of trade liberalisation.
Trade reforms cause direct changes in local relative prices, which indirectly
affect household’s income, expenditure and welfare. On the expenditure side, net
effects depend on product structure of the consumption basket and on whether
individuals are net producers or net consumers. Changes in household’s income
are explained by the fact that the trade reforms imply a reallocation of resources
between sectors, resulting in changes in factor prices, particularly wages. As we
analyse both changes in prices and variations in income, we are able to deter-
mine the overall change in household welfare. Recently, promising trade eco-
nomics literature is attempting to precisely measure the net effect of trade
Trade reforms in Uruguay 169
integration on income distribution and poverty, taking into consideration both
income and expenditure effects (Giordano and Florez, 2007).
By trade reforms we mean both national and foreign trade reforms. We con-
sider that national trade reforms imply the removal of tariff protection on Uru-
guayan imports. Foreign trade reforms refer to the possibility of local exports to
access those markets in the developed countries (or elsewhere). For small open
economies, like Uruguay, theory indicates that changes in world prices translate
immediately to local price levels. Therefore, when tariff reductions and import-
quotas removals take place in third countries, the price of Uruguayan exports to
developed countries is positively affected. But trade liberalisation plus enhanced
market access does not necessarily equal poverty reduction. As a means to
measure the effect of trade liberalisation on poverty, we plan to evaluate the
impact of both national and foreign trade reforms on the head count ratio. The
objective of this technical research is to assess the linkages between trade,
poverty and inequality by analysing the impact of trade liberalisation through
two main transmission channels: prices and income. Following the methodology
developed by Porto (2006), the study first assesses the implications of a given
trade shock, that is, a national or a foreign trade reform, in relative domestic
prices of traded goods (imports and exports). Second, the study will analyse the
response of labour income and consumption channels at the household level.
This leads to the third step, which is the induced change in the head count
poverty ratio. This methodology will allow us to identify the new income that
individuals would earn as a result of a policy change, in order to determine to
which extent trade liberalisation contributes to poverty reduction. Detailed data
at the household level will be used to assess how inequality and poverty have
evolved over time, across regions (e.g. urban areas compared to the rest of the
country) and across different household types (e.g. ranked according to the edu-
cation level). Obtained results evidence that (1) the decrease of tradable goods’
prices largely benefited the lower-income segment of the Uruguayan population,
(2) the dynamics of the non-tradable goods’ prices had a clear pro-rich impact
and (3) trade liberalisation had a clear positive impact for both the highly paid
and for those with the lower positions in the salary distribution.

8.2 Trade reform in Uruguay

The trade liberalisation process that took place in Uruguay since 1980 can be
classified in four periods. Average formal tariff decreased from a 30 per cent
level in 1980 to around 16 per cent in 1983. From 1983 to 1990 it remained quite
constant (it actually grew on a year-to-year basis in 1985 and 1990). The 1990s
marked an intense period of tariff reduction: average tariff went from 14 per cent
in 1990 to 8 per cent in 1994. Since then, the average tariff level has been stable.
We can conclude that in the last 25 years Uruguay has experienced two periods
of marked tariff reduction (1980–1983 and 1991–1994) and two periods of tariff
stability (1985–1990 and 1995–2005). When we observe the trade statistics, it is
direct to conclude that Uruguayan trade flows show a sharp increase in the last
170 F. Borraz et al.
Table 8.1 Trade openness coefficient

In constant terms (%)

1970–1979 39.6
1980–1989 47.4
1990–1999 76.8
1995–2004 80.5

Source: Central Bank of Uruguay.

35 years. As shown in Table 8.1, trade has particularly intensified in the last
10–15 years. It is interesting to note that while imports were 12 per cent higher
in 2004 than in 1995, exports grew by 28 per cent in the same period.
In 1991, Uruguay entered Mercosur, a Trade Agreement signed between
Argentina, Brazil, Paraguay and Uruguay (Treaty of Asuncion). The creation of
Mercosur marked the acceleration in the fall of import tariffs and the long-term
commitment that Uruguay would continue the liberalisation process. At the early
years of Mercosur, the administration in place made strong emphasis on finally
concluding the liberalisation process (started in the mid-1970s), removing the
remnants of the protectionism apparatus. For example, since 1991 the number of
tariff categories was reduced from five to three. In this scenario, the Uruguayan
trade policy imaged those requirements of the country’s regional partners.
From January 1995, Mercosur began to operate like an imperfect customs
union. Table 8.2 shows the intra and extra Mercosur trade flows. Ideally, Merco-
sur would enable Uruguay to obtain preferential access to a large and close
market. But at present there is a level of disenchantment with the integration
process at Mercosur. Many Uruguayans feel that the integration process has been
slow-paced, responding to specific interests from industrial lobbying groups
from Brazil and Argentina. As an example, the proliferation of non-tariff barriers
shows the low level of commitment to trade disciplines. This phenomenon has
operated in a way that production specialisation has not really occurred in
Uruguay. Additionally, another sign of Mercosur modest results at the extra-
zone level is related to the low number of trade agreements signed with third
parties (countries or regions). In fact, it was only in 2007 that Mercosur (as a
group) signed its first trade agreement with an extra-zone party (Israel).

Table 8.2 Intra and extra Mercosur trade flows (US$): simple average

1995–2000 2001–2006

Intra-Mercosur trade 35,464,482 34,620,294

Extra-Mercosur trade 148,903,829 202,954,670
Total Mercosur trade 184,368,311 237,574,964
Intra-Mercosur trade (%) 19% 15%

Source: ALADI.
Trade reforms in Uruguay 171
8.3 Inequality and poverty in Uruguay: the stylised facts
It is important to make clear that income inequality and poverty are different
concepts. While income inequality refers to income distribution (a relative term),
poverty refers to the relationship between (absolute) individual income and the
poverty line. Poverty reduction may be associated to either higher income ine-
quality or a more equal income distribution. It is broadly accepted that economic
researchers and policy-makers should be concerned about both indicators of
social welfare, when evaluating alternative policies.
Various studies have described the stylised facts of income distribution and
poverty for Uruguay in the time horizon that we are considering. Bucheli and
Rossi (1994) concluded that inequality was quite constant during the period
1984–1992. Moreover, Rossi (2001) examined the evolution of inequality and
poverty in Uruguay between 1989 and 1997.2 His results show that wage ine-
quality increased since 1991 and poverty levels increased between 1993 and
1997. This result is confirmed both for Montevideo and the rest of the urban
country, especially since 1991. Also, Miles and Rossi (1999) and Gradin and
Rossi (2000) obtained results indicating higher levels of inequality in Uruguay
during the 1990s.
With respect to poverty, the evolution for the 1989–1997 period, based on the
headcount ratio, shows that poverty decreased before 1993 and then (slightly)
increased (Rossi, 2001). Authors attributed that increase in poverty levels during
the 1990s could be related to growth problems, increased openness of the Uru-
guayan economy or to the process of decentralisation in wage negotiation. Men
and women show similar evolutions but women have an increase in their poverty
relative to men. Finally, poverty is more intensive in cities outside Montevideo
than in the capital city. The poverty (headcount ratio) shows stability between
1997 and 2001, an increasing trend until 2004 and then a decreasing trend up to
now (Vigorito 2007).
The figures show that the period of relative stability in tariffs (the 1980s)
coincides with a period of relatively unchanged income distribution. As trade
liberalisation occurred (the 1990s), wage and income inequality grew. Addition-
ally, trade liberalisation coincided with a period of increase in poverty in
Uruguay. It is fundamental that we may investigate whether there is a statistical
relation between changes in import tariffs and changes in income distribution
and poverty. In particular, proper econometric analysis will allow us to identify
the direction in which causality runs.

8.4 Methodology

8.4.1 Effects of national trade reform

From a theoretical perspective, the impact of trade on wage inequality could go
in either direction. In a Heckscher-Ohlin model, workers should see wages
increase relative to capital owners’ rents (alternatively, unskilled wages should
172 F. Borraz et al.
go up relative to skilled wages) in a developing country relatively well endowed
with labour (or unskilled labour). In that case, workers would benefit relative to
capital owners (or more skilled workers) and income distribution would improve.
Under a specific factors model, however, workers that are unable to relocate to
labour-intensive industries would lose, and the distributional impact of trade lib-
eralisation is ambiguous. Moreover, empirical studies show that the wage gap
between skilled and unskilled workers may increase after trade and investment
reform. This could occur, for example, if foreign-owned firms that begin operat-
ing in a developing country bring with them technology that increases the
demand for skilled workers. In that case, the distributional impact is adverse.
The project will study the link between trade, poverty and inequality by ana-
lysing the impact of trade liberalisation through two main transmission channels:
prices and income. The first possibility is that the new tariff levels that result
from trade reforms explain price changes. Price changes may affect individuals
in different ways, for example, depending on the share of each good in their con-
sumption basket, as suggested earlier, or if individuals are net producers (as in
the case of farmers) or net consumers. A second possibility is changes in house-
hold income. This effect is explained by the fact that trade liberalisation implies
a reallocation of resources between sectors, resulting in changes in factor prices
in the process.
In this study we restrict the analysis to four trade goods: food and beverages
(FB), clothing and footwear (CF ), house equipment and electronics (HQ), other
traded goods (OT); and four non-traded goods: health and education (HE), trans-
port and communications (TC), housing (HO) and other non-traded goods (ON).
To analyse the distributional impact of Mercosur on Uruguayan households
we use a model based on Dixit and Norman (1980). The variation in exogenous
income (Y0) needs to compensate household i to keep the same utility after a
change in the price of trade good k (k = 1, . . ., 4) because of the trade reform can
be approximated by the following equation:


where Yi0 is the exogenous income of households i, tk is the tariff for traded good
k, Sik is the budget share spent on the good k by household i, Pk is the price of
trade good k, Pn is the price of non-traded good n, Sin is the budget share spent
by household i, ewiPk is the wage price elasticity with respect to traded good k and
qwi is the share of labour income in total household income.
The first term in equation (8.1) shows that for a given increase in the price of
the trade good k, the higher the share the higher will be the income necessary to
compensate the consumer. The budget share approximates the consumption
effect. The second term of (8.1) shows the compensation generated by the
change in the price of non-traded good that is explained by the trade reform.
Their importance is related also to the share spent on non-traded goods. The first
Trade reforms in Uruguay 173
and second term in (8.1) approximate the consumption effect of the Mercosur.
Finally, the last term is the labour effect. The trade reform, change the price of
trade goods that change household wages. In order to assess the distributional
effect to Mercosur we have to estimate the three terms of the previous equation.

Impact of tariffs on prices of traded goods

Initially, the project will estimate the impact of tariffs on prices. Following
Deaton (1997) it is possible to approximate the change in consumption explained
by the changes in prices using the expenditures shares of each of the goods.
Therefore, only the direct impact will be considered and not other indirect
effects. In order to quantify the distributional effects of these price changes there
are two possibilities. The first one consists in the estimation of price indices for
each individual in the survey, based on pre-trade reform expenditures shares
with both prices. In a second step, the effects on individuals of the price change
that is explained by the reforms will be quantified. The second approach follow-
ing Deaton (1997) consists in a non-parametric estimation of expenditure shares
across the entire distribution of consumption, and computing average market
shares for different incomes. When using the second approach, results are highly
dependent on a proper choice of the bandwidth.
In particular, the induced change in the price of trade good k after the trade
reform is:


where slk is the expenditure share of the sub-category l in traded good k, δlm is the
fraction of imports of good l coming from Mercosur and δkrw is the fraction
coming from the rest of the world. Equation (8.2) estimates the price change of
traded goods from Mercosur.

Impact of prices of traded goods on the price of non-traded goods

In order to estimate the impact of the prices of traded goods on the prices of non-
traded goods we will estimate the following translog equation:


We regress the prices on traded goods on monthly prices of the traded goods and
their interactions. In order to avoid a spurious regression we check for cointegra-
tion between the variables included in equation (8.3).
174 F. Borraz et al.
Impact of prices on income
Some of the papers in this literature focus only on distribution effects of price
changes after the reforms, without considering some import effects on the factor
markets. This proposal seeks to quantify the impact of openness on total income.
In addition, the wage–price elasticity will be estimated. In particular we will
regress the log of the real wage earned by person i against completed years of
schooling (s), variables (z) such as age, marital status, children at home, region,
etc., and the log prices of traded goods interacted with schooling and region.


8.4.2 Effects of external trade reform

In order to analyse the impact of external trade reforms over the Uruguayan
economy, we focus on a major exported good: beef. Although we only analyse
how changes in the global market for beef affect specific variables of the local
economy, we believe that these results could be generalised to other exportable
goods items. Specifically, we will quantify the impact of trade liberalisation in
the global beef markets over labour income, employment and poverty levels in
Uruguay. First, we estimate how the change in global price impacts the price
level in the local market. We follow a methodology similar to that developed by
Porto (2006) in order to identify different effects across education levels and
industries. Second, using results obtained in the first stage, we estimate the
impact of alternative scenarios of trade liberalisation over labour income,
employment and poverty.
We study the link between trade, poverty and inequality by analysing the
impact of trade liberalisation through two main transmission channels: prices
and income. The first possibility is that the new tariff levels that result from trade
reforms explain price changes. Price changes may affect individuals in different
ways, for example, depending on the share of each good in their consumption
basket, as suggested earlier, or if individuals are net producers (as in the case of
farmers) or net consumers. A second possibility is changes in household income.
This effect is explained by the fact that trade liberalisation implies a reallocation
of resources between sectors, resulting in changes in factor prices in the process.

The impact of changes in international prices on domestic prices

In this section, we aim to estimate the impact of variations in international prices
on local price levels: what fraction of the change in global prices is transmitted
to the local price levels? And, how long does the transmission process take? In
this respect, we will test the long-term cointegration between international and
domestic prices.
Trade reforms in Uruguay 175
Given the fact that Uruguay is a geographically small and homogeneous
country, we do not consider that it is necessary to work with prices per region.
So, we work with an average national price. We estimate the following


Equation (8.5) allows us to identify the long-term relationship between local and
international prices. β1 allows us to determine the referred relationship. In order
to estimate cointegration, we conduct the Augmented Dickey–Fuller (ADF ) test
over equation (8.5) residuals. Also, we are interested in testing the short-term
price dynamics so that we can identify the duration of the transition process. We
do this by estimating the following error correction model:


where local prices vary between t–1 and t due to changes in international prices
for that period (response is indicated by δ) and due to the adjustment to the ‘long
term equilibrium’ level with a velocity of γ. In case a cointegration relationship
exists, equation (8.6) is valid since it deals only with stationary variables.
Based on equations (8.5) and (8.6) we obtain the local prices adjustment after
a change in global prices (in an n-months time horizon). The interpretation is as
follows: as world prices increase by 1 per cent, local prices vary by δ per cent. In
the second period, a term for error correction (γ) is considered. The time horizon
for the adjustment of local prices after a shock in the world prices can be esti-
mated as follows:


The impact of changes in domestic prices on labour income

Some of the papers in this literature focus only on distribution effects of price
changes after the reforms, without considering some import effects on the factor
markets. In our work, we seek to quantify the impact of openness on total
income. In addition the wage–price elasticity will be estimated. In particular we
will regress the log of the real wage earned by person i against completed years
of schooling, individual variables, specific variables indicating geographic loca-
tion of the household (per Department), and the log prices of traded goods inter-
acted with a sub-group of independent variables.
We estimate the following model at the individual level:

176 F. Borraz et al.
where wi is the logarithm of real wage per hour, p indicate domestic beef prices,
D indicates geographic variables (per Department) and X is idiosyncratic indi-
vidual variables. We indicate whether the individual is the household head, edu-
cation level, employment status (and industry), marital status, number of children
in the household with age six or below, number of people in the household with
age between six and 14.
Since the dependent variable, wi, is a zero-censored variable the estimation of
(8.4) should not be conducted using OLS. In that case, we would have obtained
biased and inconsistent estimators of the impact of beef prices and of individual
and geographic variables over labour income. Instead, we estimate the bias
selection correction factor based on a Probit model in order to estimate labour
market participation. Then incorporate the referred term into equation (8.8) but
only for those wage levels that are strictly greater than zero.

8.5 Estimation

8.5.1 Estimation of the effects of national trade reform

Impact of tariffs on traded goods

Table 8.3 shows the evolution of tariff in Uruguay since 1985. By the mid-1980s,
the tariff levels in Uruguay ranged between 43 per cent and 55 per cent. The early
years of the 1990s implied a drastic reduction in tariffs: rates went down to an
average of 22 per cent in 1992.3 In 1996, Mercosur imposed a sharp reduction in
the intra-zone tariff and a slight decrease in the non-Mercosur tariff. The most
significant decrease in the intra-zone tariff rate was in FB category (from 21 per
cent in 1992 to 5 per cent in 1996). The reduction in the ‘other traded goods’ cat-
egory was the smallest, from 23 per cent in 1992 to 11 per cent in 1996. In 1999,
only four years after the initial Mercosur intra-zone tariff reduction, for all the
categories of goods the intra-zone tariff was approximately zero. There were only
a few exemptions like the sugar sector. Mercosur was an effective regional trade
agreement to rapidly eliminate almost all intra-zone tariffs.
The situation is different with respect to the common external tariff (extra-
zone tariff ), where the reduction was minor. Only for FB we observed an import-
ant reduction in the extra-zone tariff following Mercosur, from 21 per cent in
1992 to 15 per cent in 1996. We expect this decrease to cause an important
improvement in income distribution and poverty alleviation because poor house-
holds have a higher consumption share for FB than the rich households. For the
other three categories of goods we observe a minor reduction (around three
points) in the extra-zone tariff. However in 1999, we observe a reversal in the
trend of reduction of the extra-zone tariff and the tariff for CF, HE and OG
return to their pre-Mercosur levels. There is a reversal in the trend toward inte-
gration to the world economy. In the FB sector the extra-zone tariff increased
only two points from 15 per cent in 1996 to 17 per cent in 1999. In this case we
expect an effect of the extra-zone tariff reduction on income inequality and
Trade reforms in Uruguay 177
poverty because of the small tariff reduction. In 2006 Uruguayan extra-zone
tariff was 14 per cent for FB and approximately 18 per cent for the other cat-
egories of goods.
In Table 8.4 we estimate the induced change in tradable prices after Mercosur
for the four categories of traded goods considered. We estimate the price change
for the 1992–1996 period. Mercosur causes a decrease in the price of the four
traded goods considered. It is remarkable that the price reduction was very
similar across goods. The highest decrease was for the other traded goods (6.1)
and the lowest was for house equipment (4.7 per cent).
Figure 8.1 shows the consumption effect for each of the traded good categor-
ies. Estimations are made as a Kernel regression. The effect is positive for all of
the individuals. However, for FB, HE and OG the consumption effect is pro-
poor. For the poor individuals the consumption gain is higher than for richer
individuals. Figure 8.2 shows the pro-poor consumption effect of traded goods.

Impact of tariffs on non-traded goods

To avoid the spurious regression problem we apply the Engle–Granger cointegra-
tion test (based on residuals) to determine the long-term equilibrium cointegrating

Figure 8.1 Consumption effect: compensating variation as per cent of income by income
distribution ($U) (source: authors’ estimations).
Table 8.3 Tariff structure: Uruguay

Simpled average

Food and beverages Clothing and footwear House equipment and electronics Other traded goods
1985 43 55 53 49
1992 21 23 21 22
1996 4 7 5 11
1999 0 0 0 0
2006 0 0 0 0
1985 43 55 53 49
1992 21 23 21 22
1996 14 21 19 19
1999 15 22 21 22
2006 12 19 18 17
Weighted average by expenditure shares

Food and beverages Clothing and footwear House equipment and electronics Other traded goods
1985 44 53 48 50
1992 21 24 21 23
1996 5 9 6 11
1999 0 0 0 0
2006 0 0 0 0
1985 44 53 48 50
1992 21 24 21 23
1996 15 21 18 18
1999 17 23 21 21
2006 14 20 18 17

Source: ALADI and Secretaría del Mercosur.

Table 8.4 Price changes from Mercosur

Category Tariff Consumption share Intrazone tariff Extrazone tariff Price changes from Mercosur

1992 1994–95 1996 1996

Food and beverages 21 62 5 15 –5.1

Clothing and footwear 24 15 9 21 –4.8
House equipment 21 13 6 18 –4.7
Other traded goods 23 10 11 18 –6.1

Source: authors’ estimations.

The price change in the last column is computed using equation (8.2).
180 F. Borraz et al.



Figure 8.2 Compensating variation as per cent of income by income distribution ($U):
traded good (source: authors’ estimations).

relationship between each of the prices of non-tradable goods and the prices of
the traded goods.
In the first step, we use the ADF unit root test to analyse the stationarity of
the prices. Table 8.5 indicates that all the price variables are non-stationary with
a unit root. Next, we proceed to estimate the equation (8.3) by OLS and check
for stationarity of the residuals. The result of the Engle-Granger based on resid-
ual cointegration tests is shown in Table 8.6: prices of non-traded and prices of
traded goods are cointegrated. In other words, there is a stable long-run relation-
ship between both prices.
Figure 8.3 shows that the consumption effect of non-traded goods is pro-rich.
This fact can be explained by the effect of the change of the price of traded
goods in the housing price.

Wage–price elasticities
Since it is likely that there is a large number of individuals who do not work
(especially women) and therefore report zero wages, it would not be appropriate
to estimate equation (8.4), the wage equation, using OLS. Since the dependent
variable is censored at zero, we only observe the wages of the employed indi-
viduals and estimation of the wage equation by OLS will simply yield inconsist-
ent estimates. We allow the impact of the price of traded goods on wages to vary
according to individual characteristics including schooling, age and geographical
location of the household. This implies that the elasticities of wage and labour
market participation with respect to prices vary from one individual to another,
according to her age, schooling and geographic location. This is mandatory to
Table 8.5 Unit-root test: tradable and non-tradable prices (ADF performed with 12 lags)

Level Tradable goods Non-tradable goods

Constant and trend –1.73 –2.11 –1.50 –1.67 –2.43 –1.42 –1.40 –1.66
Constant –2.30 –2.10 –1.74 –1.69 –2.77* –3.29** –1.08 –1.80
None 0.03 –11.00 0.41 0.38 0.91 1.52 –0.58 0.19
Log difference
Constant and trend –3.05 –2.57 –3.90** –2.08 –3.65** –3.21* –1.75 –3.23*
Constant –1.86 –3.43*** –4.13*** –4.39*** –3.74*** –3.13** –2.76* –4.99***
None –2.81*** –4.33*** –4.59*** –5.38*** –4.70*** –3.89*** –4.48*** –6.43***

Source: authors’ estimation.

* Statistically different from 0 at the 10% level or better; ** statistically different from 0 at the 5% level or better; *** statistically different from 0 at the 1% level or
182 F. Borraz et al.
Table 8.6 Prices co-integration

Engle–Granger Cointegration Test

ADF performed with 12 lags

Constant and trend

Health and education –6.07***

Transport and communications –4.25***
Housing –4.16**
Other non-tradable –4.85***

Source: authors’ estimation.

*** Statistically different from 0 at the 1% level.



Figure 8.3 Compensating variation as per cent of income by income distribution ($U):
non-tradable goods effect (source: authors’ estimations).

estimate the impact of changes in prices on household wages at different points

of the whole income distribution.
We estimate a Heckman selection model using maximum likelihood. All
regressions include year and geographic location dummies. Estimates from this
model allow us to calculate the impact of the price of traded goods on labour
income and the impact of changes in prices of traded goods on the labour market
participation of each individual in the sample. We also take into consideration
the fact that men and women’s labour market rewards may differ and we there-
fore separately estimate wage equations by gender. Our wage equations are
limited to individuals aged 18 through 55.
Figure 8.4 shows that the labour effect is pro-poor. This fact can be explained
by the effect that the change of the price of traded goods has the highest impact
on the wage of the low-income individuals.
Trade reforms in Uruguay 183



Figure 8.4 Compensating variation as per cent of income by income distribution ($U):
labour income effect (source: authors’ estimations).

Estimation of total effect

Figure 8.5 presents the estimation of the consumption and labour income effects.
Trade liberalisation had a clear positive impact for both the highly paid and those
with the lower positions in the salary distribution.



Figure 8.5 Compensating variation as per cent of income by income distribution ($U):
total effect (source: authors’ estimations).
184 F. Borraz et al.
Poverty and inequality effects
We use the wage–price elasticities estimated above to quantify the change in the
head count ratio and income inequality indicators after Mercosur. In Tables 8.7a
and 8.7b we observe a reduction in poverty for low educated persons located in
the border and central regions of Uruguay. We do not observe differences by
gender. There are no significant changes in income inequality after reform. It is
interesting to note that we observe a decrease in poverty but income inequality
remains constant.

8.5.2 Estimation of the effects of external trade reform

First, we present results related to the price transmission. Second, we show
results related to the labour market participation and labour income.

Price transmission
We aim to determine whether there is a permanent and long-term relationship
between domestic prices (paid to producers) and global prices in the beef sector.
We conducted a unit-root analysis, using the ADF test.
Table 8.9 presents the ADF results for variables expressed in levels and in
differences. We analysed both a model incorporating constant and trend and an
alternative model without constant.
Results indicate that we cannot reject the null hypothesis of existence of unit-
root for the following series: the log of the price paid to the beef producer; the
log of the export price in Brazil; the log of the export price in New Zealand. So,
we conducted ADF test for the growth rates of the prices levels. At this time, we

Table 8.7a Poverty: before and after trade reform

Headcount ratio (P0), poverty gap index (P1) and squared poverty gap index (P2)

Total (men + women) Change P0 Change P1 Change P2

Total –0.018 (**) –0.004 (**) -0.002 (**)

Education ≤6 years –0.028 (***) –0.008 (*) –0.002 (**)
Education 7–12 years –0.017 (**) –0.003 (**) –0.000
Education >12 years –0.002 –0.000 –0.000
Montevideo –0.006 (**) –0.001 (**) –0.000
Border –0.041 (**) –0.001 (**) –0.003 (**)
South –0.017 (**) –0.003 (**) –0.001 (**)
Central –0.036 (**) –0.007 (**) –0.002 (**)

Source: authors’ estimations.

* Statistically different from 0 at the 10% level or better; ** statistically different from 0 at the 5%
level or better; *** statistically different from 0 at the 1% level or better. Poverty line = half of mean
laboural income
Trade reforms in Uruguay 185
Table 8.7b Poverty: before and after trade reform

Headcount ratio (P0), poverty gap index (P1) and squared poverty gap index (P2)

Change P0 Change P1 Change P2

1 Men
Total –0.020 (**) –0.000 (**) –0.002 (**)
Education ≤6 years –0.036 (**) –0.009 (**) –0.004 (**)
Education 7–12 years –0.018 (**) –0.004 (**) –0.001 (**)
Education >12 years –0.004 (**) –0.000 –0.001
Montevideo –0.008 (**) –0.002 (**) –0.001 (**)
Border –0.049 (**) –0.011 (**) –0.004 (**)
South –0.018 (**) –0.005 (**) –0.002 (**)
Central –0.043 (**) –0.010 (**) –0.004 (**)
2 Women
Total –0.015 (**) –0.103 (**) –0.001 (**)
Education ≤6 years –0.027 (**) –0.006 (**) –0.001 (**)
Education 7–12 years –0.013 (**) –0.002 (**) –0.001 (**)
Education >12 years –0.001 (**) –0.000 –0.000
Montevideo –0.005 (**) –0.001 (**) –0.000 (**)
Border –0.039 (**) –0.009 (**) –0.002 (**)
South –0.014 (**) –0.003 (**) –0.001 (**)
Central –0.029 (**) –0.005 (**) –0.001 (**)

Source: authors’ estimations.

** Statistically different from 0 at the 5% level or better. Poverty line = half of mean laboral income.

Table 8.8 Change in income inequality

Before and after trade reform Gini index

Total No effect Total No effect

1 Men 2 Women
Total No effect Total No effect
Education ≤6 No effect Education ≤6 No effect
Education 7–12 No effect Education 7–12 No effect
Education >12 No effect Education >12 No effect
Montevideo No effect Montevideo No effect
Border No effect Border No effect
South No effect South No effect
Central No effect Central No effect

Source: authors’ estimation.

186 F. Borraz et al.
Table 8.9 Unit-root test: ADF

Levels Differences

Domestic price paid to producer –3.01 –5.47***

Export prices –1.63 –3.86***
Export prices in Brazil –1.99 –7.43***
Export prices from New Zealand to USA –1.6 –10.48***

Source: authors’ estimation.

*** Null hypothesis is rejected at 1% significance level. Critical values are –3.14 at 10%, –3.43 at
5% and 4.00 at 1%. The number of lags was estimated using Akaike

were able to reject the null hypothesis at the 1 per cent significance level. We
conclude that series (in level) are integrated of order 1. This is to say that we are
dealing with no stationary time series. So we proceeded to analyse the cointegra-
tion hypothesis between domestic and international prices.
We estimated cointegration for three relationships: domestic prices and export
prices in Uruguay; domestic prices and export prices in Brazil; domestic prices
and export prices in New Zealand; in Table 8.10 we present results for the
cointegration test of (real) domestic prices and the (real) export prices in
Uruguay. We find that both prices are cointegrated at 1 per cent significance
level. We conclude that both international prices and domestic prices move
together. Although the transmission is not perfect, β1 from equation (8.1) is 0.76,
the relationship is statistically significant at 1 per cent level.
We also analysed the short-term price dynamics. We find that adjustment to
the long-term equilibrium price level takes four years. We note that after three
months that the external shock has appeared, only one-third of the total impact
has occurred; only two-thirds of the total impact takes place after one year from
the shock. We conclude that price adjustment occurs, but definitely at a pretty
low pace.

Selection models estimation

We used Heckman models for estimating wages for both men and women
(and for the entire sample). Obtained results have the expected signs and are
statistically significant. Interestingly, results suggest that impact of global
beef prices over wage levels have opposite signed for the cases of men and
women. Export prices negatively affect female average wage levels while they
positively impact wage levels among men (the impact over male wage is mar-
ginally more important). In both cases results are statistically significant. As
we pay attention to the selection variable, we find that the impact is positive
both for men and women. Again, the marginal effect is stronger among men.
Total effect is also affected by interactions of different variables and beef
Trade reforms in Uruguay 187
Table 8.10 Engle-Granger: cointegration test

Dependent variable: average domestic price paid to producer

Variable Coefficient Std. Error Ratio-t Prob.

Constant 0.35 0.10 3.45 0.0007

Export prices in Uruguay 0.76 0.04 21.35 0
ADF test on residuals –6.10 0.0000
Critical value at 1% level is 4
Error Correction Model
Constant 0.00 0.00 0.89 0.38
Short-term adjustment 0.22 0.06 3.91 0.00
Long-term adjustment –0.06 0.04 –1.69 0.09
Speed of adjustment
3 months 0.32
6 months 0.40
1 year 0.51
1 and a half years 0.59
2 years 0.65
3 years 0.71
4 years 0.75

Source: authors’ estimation.

Global price variations: simulations

In this chapter, we consider three scenarios suggested by Olarreaga (2006).

• Scenario 1: Doha Agreement (favourable), price increase of 3.9 per cent.

• Scenario 2: Free Trade Agreement with Europe, price increase of 5.6 per
• Scenario 3: Free Trade Agreement with USA, price increase of 7.6 per cent.

We analyse the impact of an increase in the price of beef under the above men-
tioned scenarios. In general terms, we find that only in Scenarios 2 and 3 there
are statistically significant effects. For example, under Scenario 3 (Free Trade
Agreement with USA) the impact of beef price changes over men and women
are qualitatively different. In the case of men, there is a positive impact in labour
income across education levels and industry. In particular, those who are highly
educated and work in the agro-industrial sector are highly benefited in this sce-
nario. In the case of women, there is a positive effect for labour income for those
working in the agricultural sector and related industries. Still, that effect is
clearly lower than in the case of men. For women working in other industries,
Scenario 3 implies a marked negative impact on wage levels.
We conducted simulations in order to evaluate the impact of variations in
prices over poverty and inequality levels. We found that none of the considered
scenarios imply a clear impact on poverty levels (Table 8.11).4 Instead, we
observe changes on income inequality indicated by changes in the Gini and Theil
188 F. Borraz et al.
Table 8.11 Poverty and inequality effects of liberalisation in the beef international trade

Alternative scenarios

Base scenario Scenario 1 Scenario 2 Scenario 3

3.9% 5.6% 7.6%
Gini men 0.4052 0.402 0.401* 0.399*
Theil men 0.2662 0.262 0.260* 0.258*
Gini women 0.3513 0.355 0.3565* 0.359*
Theil women 0.2004 0.204 0.206* 0.209*
Poverty ratio (P0) 0.296 0.296 0.296 0.297
Poverty gap (P1) 0.116 0.117 0.117 0.117
Poverty gap – squared (P2) 0.060 0.061 0.061 0.061

Source: authors’ estimation.

* Indicates that result is statistically different from the base scenario at 5% significance level.

coefficients under Scenarios 2 and 3. In those cases, income becomes more

evenly distributed among men and more concentrated in the case of women.
Table 8.12 presents the change in the probability of being employed under
Scenario 3 (Free Trade Agreement with USA). In general, we observe that
variations are minor (less than 1 per cent). For the case of men employed in the

Table 8.12 Change in the probability of being employed after a Free Trade Agreement
with USA

By activity sector and educational level


Low Medium High

Agroindustrial –0.28 –0.24 –0.13
Beef industry 0.15 0.05 0.00
Other industry 0.30 0.20 0.13
Other sector 0.22 0.15 0.14
Agroindustrial 0.29 0.15 0.03
Beef industry 0.47 0.34 0.09
Other industry 0.61 0.50 0.28
Other sector –0.04 –0.13 –0.24

Source: authors’ estimation.

Trade reforms in Uruguay 189
agricultural sector we see a negative change in the probability of being
employed. This result is also observed for the case of women across a broad
range of economic activities.

8.6 Conclusions and policy implications

Although it is commonly believed that trade liberalisation results in higher GDP,
little is known about its effect on poverty and inequality. As many developing
countries embrace trade integration as the remedy for all diseases, it is funda-
mental that liberalisation could be analysed from a broad range of perspectives
(GDP growth, employment, poverty, inequality, etc.).
In our work we focused on the poverty and inequality effects of tariff reduc-
tion in Uruguay for the 1986–2006 period. We measured the variation in income
needed to compensate each household to keep the same utility after a change in
the price of tradable goods. A positive change in the referred variable means that
the household has improved when compared to the pre-liberalisation scenario.
We analysed the impact of trade integration on households’ welfare through
various transmission channels: (1) reduced tariffs affect the price of tradable
goods, (2) reduced tariffs impact the prices of non-tradable goods and (3)
reduced tariff cause a reallocation of productive resources and changes on labour
income. As said, when interpreting results, it is important to bear in mind that
while intra-zone tariffs were slashed after Mercosur was in place, extra-zone
tariffs slightly decreased in the 1992–2006 period. Also, note that while tariffs
for the ‘food and beverage’ category were drastically reduced in the initial Mer-
cosur years, tariffs affecting other industrial sectors experienced a more ‘gradual’
Obtained results evidence that: (1) the decrease of tradable goods’ prices
largely benefited the lower-income segment of the Uruguayan population, (2) the
dynamics of the non-tradable goods’ prices had a clear pro-rich impact and (3)
trade liberalisation had a clear positive impact for both the highly paid and for
those with the lower positions in the salary distribution.5 Going further, one
could say that the evolution of the prices of housing, health and education nega-
tively affected the lower income population, while the decrease of the ‘food and
beverages’ prices positively affected them. We think that these findings could
have clear policy implications: as tariffs are reduced, the price of non-tradable
goods became burdensome for the poor; if public authorities aim to develop pro-
poor policies, then efforts should target the housing, health and education
We also analyse results at the aggregate level (when changes of the prices of
tradable and non-tradable goods and labour income are considered together).
Results show that average income (actually, compensating income – as defined
in equation (8.1)) increased along the liberalisation process across the entire
income distribution. We think that this result is important, indeed. For the case
of Uruguay, talking about the income effect of trade liberalisation should not be
associated with the typical ‘winners and losers’ scheme. Evidently, specific
190 F. Borraz et al.
groups obtained higher benefits than others, but we could not find any evidence
about absolute losers resulting from Mercosur. In sum, the question about the
impact of trade liberalisation over poverty and income can be answered with a
common place: (mild) gains from trade. While not evenly distributed among the
income distribution, benefits from trade spread in every Uruguayan household.
With respect to external trade reform we focused on the most important Uru-
guayan export: the beef sector. The adjustment of local beef prices after an exter-
nal shock to the worldwide price levels is imperfect. Estimations indicate that 76
per cent of a certain shock to the export prices is transmitted to the price paid to
the local producers. Short-term local price dynamics show that the transmission is
pretty low paced. One year after the shock, only two-thirds of the long-term effect
has occurred. Price changes after trade liberalisation (under alternative scenarios)
imply that men become better off (in terms of earned real wages), in particular
those who are highly educated and work in the agricultural sector. For the case of
women, increases in labour income after trade liberalisation are mild (or negative,
in specific cases). We do not observe poverty impacts after trade liberalisation (in
the three proposed scenarios). Additionally, changes in employment levels are
almost immaterial. Yet, we find that there are specific inequality effects under
some scenarios. In particular, for Scenarios 2 and 3, we conclude that income con-
centration is lower in the case of men and higher for the case of women.

Appendix: data
To undertake this study we used the annual Uruguayan national household
survey, Encuesta Continua de Hogares (ECH), conducted by the Instituto
Nacional de Estadística (INE). We used ECH data for estimating the price–wage
elasticity for the 1990–2001 period.
We also used data from Encuesta Nacional de Ingresos y Gastos de los
Hogares (ENIGH), the national household expenditure and income survey (we
used the 1996 wave). This survey identifies the consumption structure of an
average family in Uruguay. ENIGH also contains socio-economic information
about Uruguayan households. This fact is crucial for us, because it allows us to
identify the consumption structure of households of the same socio-economic
group. We used this information in order to assess the impact of change in prices
on changes in the value of the consumed basket of each household.
Asociación Latinoamericana de Integración (ALADI) and Uruguay’s Minis-
try of Finance (MF ) provided historical information about the Mercosur common
external tariffs for the period between 1986 and 2006. Secretaría del Mercosur
(SM) provided data about intra-zone tariff levels (for the same time horizon).
Both ALADI and SM provided raw data at a per-item desegregation level. Addi-
tionally, ALADI and the Central Bank of Uruguay (BCU) sourced our informa-
tion about trade flows for the four-product categories with Mercosur and the rest
of the world. We used this information in order to determine the impact of
change in tariffs on prices of tradable and non-tradable goods. Information about
price levels comes from the Consumer Price Index, constructed by INE.
Trade reforms in Uruguay 191
1 We thank Marcela Arnaiz for excellent research assistance. We also thank the Trade
and Poverty Trust Fund of the Inter-American Development Bank for financial support.
2 Rossi used the Gini coefficient, the Theil index and the coefficient of variation to
measure inequality.
3 In those years, tariff rates differentials across imported goods became significantly
more uniform.
4 We calculated the poverty line by dividing the average income of the referred year by
two (for each scenario).
5 This explains the U-shaped curve in Figure 8.5.
6 The negative impact for the poor through the non-tradable goods’ prices is explained
by the evolution of the housing prices.

Bucheli, M. and Rossi, M. (1994) ‘Distribución del ingreso en Uruguay: 1984–1992’,
Facultad de Ciencias Sociales, Universidad de la República.
Deaton, A. (1997) The analysis of household surveys: a micro econometric approach to
development policy, Baltimore, MD: John Hopkins University Press.
Dixit, A. and Norman, V. (1980) Theory of international trade: a dual general equilib-
rium approach, Cambridge: Cambridge Economic Handbooks.
Giordano, P. and Florez, V. (2007) Assessing the trade and poverty nexus in Latin
America, Washington, DC: Inter-American Development Bank.
Gradín, C. and Rossi, M. (2000) ‘Income distribution in Uruguay: the effects of economic
and institutional mimeo’, Facultad de Ciencias Sociales, Universidad de la República.
Miles, D. and Rossi, M. (1999) ‘Geographic concentration and structure of wages in
developing countries: the case of Uruguay’, Working Paper No. 13/99, Economics
Department, Social Sciences Faculty.
Olarreaga, M. (2006) ‘Estimating changes in the export price of meat received by Uru-
guayan exporters under different trade shocks’, The World Bank Group.
Porto, G. (2006) ‘Using survey data to assess the distributional effects of trade policy’,
Journal of International Economics, 70:140–160.
Rossi, M. (2001) ‘Poverty in Uruguay: 1989–1997’, Facultad de Ciencias Sociales, Uni-
versidad de la República.
Vigorito, A. (2007) ‘Estadísticas de Distribución del Ingreso y Pobreza’, Mimeo Instituto
de Economía.
Part III

Economic liberalization,
development and growth in
9 Economic liberalisation and
income distribution
Theory and evidence in Mexico
Gerardo Angeles-Castro

9.1 Introduction
In Mexico, the debt crisis of 1982 signalled the end of the import-substitution
industrialisation model (ISI) and the predominance of protectionist policies.
Over the subsequent years a number of structural reforms and market-oriented
policies were undertaken. In 1985 the government eliminated some import
licences and reduced the number of tariff categories. In 1987 the elimination of
import licences was extended, the degree of tariff dispersion was reduced, and a
stabilisation programme was put in place. Between 1988 and 1990 the govern-
ment liberalised the financial system, reformed the FDI regime, eliminated some
restrictions to portfolio investment, and opened the stock market and the money
market to foreign investors; in addition, the external debt was renegotiated. The
privatisation process was initiated in 1982 and was intensified during the late
1980s and early 1990s. Negotiations on the North America Free Trade Agree-
ment (NAFTA) commenced in 1990 and it became effective in 1994.
On the basis of the Stolper-Samuelson theorem (SST) we can expect that
trade liberalisation in Mexico can increase demand for unskilled labour, as this
is considered an abundant factor in this country. The introduction of trade
reforms therefore should lead to a rise in the relative return to unskilled labour
and to a narrowing of inequality. However, the empirical evidence shows that
income distribution worsened in Mexico following economic liberalisation (Feli-
ciano, 2001; Cortez, 2001; Tanski and French, 2001; Ros and Bouillon, 2002).
Globalisation is sometimes presented in the relevant literature as a cause for
the deterioration of income distribution in recent decades across developed coun-
tries (Smeeding, 2002). Furthermore, some empirical studies show a positive
relationship between the increase in trade and income dispersion (Baldwin and
Cain, 2000; Haskel and Slaughter, 2001). This trend in many developed coun-
tries is in keeping with the SST. On the other hand, several studies attribute the
rise in inequality to the skill-biased technological change (SBTC) (Berman et
al., 1998; Acemoglu, 2002). According to this argument, countries tend to
experience a fall in relative demand for unskilled labour and an increase in that
for skilled labour, due to an acceleration of technical change over the past few
decades, this process is expected to exacerbate inequality. Both explanations
196 G. Angeles-Castro
(trade and technical change) dominate the literature dealing with the study of
inequality in industrialised countries and have been dubbed the ‘transatlantic
Some evidence from the developing world is also consistent with the idea that
trade openness can lead to more income inequality, despite the opposite SST
prediction (Litwin, 1998; Flemming and Micklewright, 2000; Ros and Bouillon,
2002; Mah, 2002). An approach to explain this relationship is the idea that
greater competition leads to a reduction of producer rents in the traded sector; to
the extent that these rents are shared with workers, wages will decline post-
liberalisation. The skill enhancing trade hypothesis (SETH), based on empirical
evidence, is another explanation about the expansion of the wage gap in devel-
oping countries (Robbins, 1996, O’Connor and Lunati, 1999). It takes arguments
that, to some extent, can be similar to those used in the ‘transatlantic consensus’.
This hypothesis claims that economic liberalisation and the intrinsic adoption of
new technologies are accompanied by a relative increase in demand for skilled
labour, which can worsen inequality. The relevant literature offers substantial
support to both approaches (Feenstra and Hanson, 1997; Arbache et al., 2004).
Thus, trade and technological change are also relevant causes of inequality in
developing countries.
The rise of service argument has also been advanced to explain the rise in
income dispersion. It holds that globalisation fosters demand for specialised
services; this process can increase income dispersion, as the service sector can
be considered skill-biased in developing countries (Gordon and Gupta, 2003;
Sinha, 2005).
In this chapter we explore whether trade, technological change, and the rise
of services can be a cause of increasing inequality in post-reform Mexico. On
the other hand, there is some evidence that since the late 1990s inequality has
levelled and even decreased slightly. In this respect, we also explore whether
these three arguments can remain in force during the period of income distribu-
tion improvement in Mexico.
The data source, the Encuesta Nacional de Ingresos y Gastos de los Hogares
(ENIGH), is a household income and expenditure survey produced by the
Mexican government’s statistical office, Instituto Nacional de Estadística,
Geografía e Informática (INEGI). It has been carried out in 1984, 1989, and sub-
sequently every two years since 1992 in randomly selected households. The
period of analysis is from 1984 to 2002; this time frame allows us to examine
the sub-period of rising inequality and the later sub-period in which inequality
We find that income inequality worsens after liberalisation, mainly because
of an increase in skill premium, an expansion of the income gap between the
service and the agricultural sectors, and a negative relationship between market
openness and wages in the traded sector. On the other hand, there is evidence
that inequality decreases after 1998 and, potentially, the factors driving this trend
are the decrease in returns to skill and a weaker effect of trade on wages in the
traded sector. In this study we also identify two main factors that help to mitigate
Theory and evidence in Mexico 197
inequality along the period, transfer income and the re-composition of house-
holds, whereas the deterioration of the agricultural sector is a persistent source
of inequality.
The Mexican case is particularly interesting for the following reasons. First,
this country has long been known for its unequal distribution of income.1
Second, in a few years Mexico moved from protectionism to market liberalism;
moreover, it has signed a number of free trade agreements and was the first
developing country to implement one (NAFTA) with two developed countries.
Finally, the era of market openness in Mexico has now lasted for more than two
decades. Therefore the Mexican case offers a good time span for assessing
whether market-oriented policies can reduce high levels of inequality or produce
a different effect.
The chapter is organised as follows. Section 9.2 gives a discussion of theoret-
ical issues supporting distributional effects under conditions of market openness
and also discusses contesting arguments. Section 9.3 explores individual income
distribution and wage inequality through a descriptive approach. The analysis is
extended in Section 9.4 by using parametric methods. Section 9.5 explores addi-
tional forms of income distribution, at the level of households and at the level of
income source. Finally concluding remarks are provided in Section 9.6.

9.2 Theoretical debate and complementary arguments

9.2.1 Standard theory

Since the 1980s a number of developing countries, especially in the Meso-south
American subcontinent, have adopted an economic model that places special
emphasis on market forces. The set of policies involved in this development
paradigm can be summarised as deregulation, privatisation, liberalisation of
markets, and macroeconomic discipline. This prescription is intended to create
preconditions for the expansion of trade and flow of investment across countries
and finds theoretical support in familiar neoclassical theory (Jones and Barry,
1988: 30–33; Corden, 1993), which claims that trade, investment, and the market
mechanism in general boost growth and facilitate development.
Proponents of the model maintain that improvements in income distribution
can be achieved for two main reasons. First, emphasis on outward-looking
growth fosters exports, employment, and output, and thus provides additional
resources for redistribution. Second, economic liberalisation facilitates the oper-
ation of market forces and the price mechanism, which allows resources to be
allocated more efficiently.
The basis for tracing distributional effects of market liberalism in developing
countries is the SST (FitzGerald, 1996: 32; Litwin, 1998: 3). Within this two-
factor (capital and labour) neoclassical model, liberalisation of foreign trade
increases demand for the abundant factor, as exports and imports adjust accord-
ing to the orthodox principle of comparative advantages, and redirects demand
away from the scarce factor. This mechanism increases the return of the factor
198 G. Angeles-Castro
which is relatively most used in the export sector and which is also more abun-
dant – this factor is conventionally assumed to be low-wage, unskilled labour in
developing countries – and leads towards factor price equalisation; by the same
token income distribution improves.

9.2.2 Contesting arguments

The skill-enhancing trade hypothesis

According to this proposition, increasing openness in developing countries can
accelerate inflow of foreign technology due to a rise in imports and FDI. Robbins
(1996) finds that the skill gap tends to widen in a sample of developing countries
and shows that there is a high correlation between increasing demand for skill
and imports of technology. He calls this effect ‘skill-enhancing trade hypothesis
(SETH)’ and argues that trade liberalisation may sometimes widen wage disper-
sion instead of compressing relative wages, as more openness permits or encour-
ages the acceleration of imported physical capital stock.
In this sense, Arbache et al. (2004: 76–77) argue that the new inflowing tech-
nology can be skill biased because it is designed through relatively skill-
intensive methods in more industrialised countries and because its
implementation and operation involves new procedures and techniques. As a
result, technological change can increase demand for skilled workers. Moreover,
they point out that the reduction in demand for skilled labour predicted by ortho-
dox theory can be surpassed by this process depending on the magnitude of the
shift. New technology is not only considered skill biased in developing countries
but also in developed economies (Berman et al., 1998).

The rise of services

An alternative argument undermining basic predictions of neoclassical theory is
the idea that the service sector, which has traditionally been considered a sector
with higher wages than some of the conventional economic activities in develop-
ing countries, is likely to expand faster than other sectors, under conditions of
economic liberalisation, and can also be considered skill-biased. This is because
the globalisation and internationalisation of the economy brings with it increas-
ing demand for financial, communication, IT, transport, and business services
among others. These activities clearly require workers relatively more qualified,
on average, than workers in some of traditional economic activities in develop-
ing countries, such as primary production and labour-intensive manufacturing.
Gordon and Gupta (2003) show that factors playing an important role in acceler-
ating services growth are high-income elasticity of demand, increased input
usage of services by other sectors, and deregulation and economic reforms.
Sinha (2005) shows that although employment in the service sector in India
during the 1990s remained steady, its share of GDP rose substantially. She also
stresses that the pattern and composition of growth acceleration of services
Theory and evidence in Mexico 199
creates further inequality between rural and urban areas, and between the skilled
and the unskilled.

Reduction of rents in the traded sector

Arbache et al. (2004) hold that the reduction or elimination of trade barriers and
tariffs turns protected markets into more contestable ones, which induces lower
prices and therefore a reduction of producer rents; if rents are shared with
employees it is expected that wages fall after liberalisation. They show that con-
trary to the predictions of the SST, wages fell substantially in the traded sector
after trade liberalisation in Brazil, consistent with the reduced rents argument, as
industries faced greater competition.

Temporary adverse effects

Some authors have analysed the view that when a country begins to adjust to a
more competitive environment serious dislocations are encountered as the
economy adapts to the shifting patterns of employment and resources. As a con-
sequence, income dispersion may widen and absolute poverty increases in the
short run. However, this effect is considered to be temporary because as the
period of adjustment continues, individuals adapt and markets react, boosting
investment in the traded sector and generating greater employment. In addition,
growth provides additional resources for poverty alleviation and redistribution.
Eventually, there may be a decrease in unemployment and income gap, and
inequality begins to decrease in the long run (Jacobsen and Giles, 1998:
419–420; FitzGerald, 1996: 32).
Even leading globalisers have stressed that in the short run, structural reforms
may increase unemployment and worsen inequality; however, as long as these
policies are consistent with sustainable economic growth they can reduce
poverty and improve equity over the longer term (Camdessus, 1998: xiv–xv).
From a more theoretical approach Pissarides (1997) illustrates that the rising
gap between skilled and unskilled wages observed in developing countries, that
have adopted market-oriented policies, can be explained by increasing trade that
acts as a means for the transfer of technology across countries. The difference to
the skill-enhancing hypothesis is that either the new technology or the importa-
tion and assimilation process can be skill biased and give a temporary and rela-
tive advantage to skilled labour that leads to higher relative wages only during
the period of transition towards a higher level of technology. He also argues that
the response of relative supply of skilled and unskilled labour to trade openness
can also explain a temporary increase of wage differentials. In addition, Goldin
and Katz (1998) hold that within firms, demand for skill rises when new technol-
ogies are introduced, but it declines once the other workers learn to use the new
equipment; hence, this process can follow a technological cycle.
200 G. Angeles-Castro
9.3 Individual income distribution
The data source is ENIGH by INEGI, as outlined earlier. We use data from six
periods 1984, 1989, 1994, 1998, 2002, and 2006. Initially, the selected sample
comprises individuals reporting monetary income, aged between 16 and 65
inclusive, and there are no restrictions for the number of hours employed in the
corresponding economic activity. In a first stage the analysis involves the main
source of income only and is not restricted to labour earnings. In other words,
individuals whose main source of income comes from entrepreneurial and finan-
cial activities, property rents, and transfers are also included in the sample. The
hourly income is computed as monthly income in the respondent’s main eco-
nomic activity, divided by weekly hours employed in the corresponding eco-
nomic activity multiplied by 4.33. The hourly income is deflated by the
consumer price index based in 2002 pesos in order to obtain real hourly income.
Between 1984 and 1998 the Gini coefficient for individuals increased from
0.512 to 0.632 and then dropped to 0.548 between 1998 and 2006. To under-
stand the forces driving this pattern we present a decomposition of income by
educational levels, economic sectors, and deciles, and conduct a comparative
analysis between the period of increasing inequality and the latter one. In addi-
tion, a parametric analysis is conducted using labour income data.

9.3.1 Returns to education

This section analyses whether skill premium is likely to increase after economic
liberalisation and also explores whether returns to skill can decrease over a
longer period. We present the average real hourly income for three different
levels of education; primary, secondary, and tertiary.2 From panel 1 in Table 9.1
we observe that the average hourly income for both primary and secondary edu-
cation tends to decline along the whole period. As for tertiary education, this
indicator has a substantial increase between 1984 and 1994 but then shows a
sharp fall over the last two periods. The average hourly income for the three edu-
cational levels has a slight recovery in 2006. It is worth noting that the percent-
age change between 1984 and 2006 is negative for the three educational levels
but it decreases less in the tertiary level, as illustrated in panel 2.
In order to explore how these changes on average income of educational levels
have affected income dispersion between skilled and unskilled individuals, Table
9.1 presents the ratio of average hourly income in panel 3. We observe that marginal
returns to tertiary education in relation to primary and secondary levels increased
between 1984 and 1998 and declined between 1998 and 2006, but remained above
their original levels. On the other hand, although returns to secondary education
fluctuated, they actually decreased slightly in relation to the first period. Therefore,
only income premium paid to high-skilled individuals has expanded, although there
is evidence that this trend has reversed over the last few years.
Table 9.1 also reports share of the three education categories under two dif-
ferent considerations. Panel 4 presents labour share weighted by hours3 and
Theory and evidence in Mexico 201
Table 9.1 Average real hourly income (2002 pesos) per level of education

1984 1989 1994 1998 2002 2006

(1) Income
Primary 23.12 22.91 20.77 16.82 16.76 17.03
Secondary 39.16 32.65 34.97 29.57 27.08 27.43
Tertiary 63.04 61.92 73.01 66.92 59.09 59.68
(2) Income variation 2006 vs 1984
Primary –26.35%
Secondary –29.95%
Tertiary –5.33%
(3) Ratio of income
Primary * 1.69 1.43 1.68 1.76 1.62 1.61
Secondary ** 1.61 1.90 2.09 2.26 2.18 2.18
Tertiary *** 2.73 2.70 3.52 3.98 3.53 3.50
(4) Share of hours
Primary 82.38 73.07 75.49 72.99 69.04 67.31
Secondary 10.23 15.01 14.31 16.42 17.43 18.44
Tertiary 7.39 11.92 10.19 10.6 13.53 14.25
(5) Share of income bill
Primary 68.21 55.29 51.31 49.75 47.19 45.87
Secondary 14.71 17.38 17.7 18.54 19.54 20.11
Tertiary 17.08 27.33 30.99 31.71 33.27 34.02

Source: author’s computation with information from INEGI (various years).

* Comparison between secondary and primary education; ** comparison between tertiary and sec-
ondary education; *** comparison between tertiary and primary education.

panel 5 displays income bill share of individuals. In both panels we observe that
the share of tertiary education increased along the whole period, the share of sec-
ondary education also increased, but the variation is more moderate; in contrast,
the share of primary education fell gradually.
Table 9.2 makes comparisons between the rising inequality period and the
later period, by exploring annualised changes in the labour share weighted by
hours, income bill share of individuals, and in the average hourly income for
both primary and tertiary educational categories. Adopting Autor et al. (1998)
assumptions,4 and following Airola and Juhn (2005), we interpret changes in
income bill share as relative demand shifts.
As for the first period, the simultaneous increase in relative income and rela-
tive supply of individuals with tertiary education suggests that demand for highly
educated individuals also increased. In fact, labour share increased at an annual
pace of 2.57 per cent, whereas income bill share expanded at an annual rate of
4.38 per cent. Although highly skilled labour supply increased over the first
period, its expansion was not enough to meet the larger increase of demand;5 this
fact explains the rise in income of this educational category.6 On the other hand,
202 G. Angeles-Castro
Table 9.2 Changes in average hourly income and share of educational levels (annualised
log change × 100)

1984–1998 1998–2006

Average income –2.26 –0.08
Hours –0.86 –1.43
Income bill –2.25 –1.34
Average income 0.43 –3.06
Hours 2.57 6.02
Income bill 4.38 1.24

Source: author’s computation with information from INEGI (various years).

changes in labour share and income bill share of the least educated group show
that relative unskilled labour demand decreased faster than relative supply. Not
surprisingly, average income of this category fell at an annual pace of 2.26 per
During the second period, income bill share of the most educated individuals
continued to increase but slowed to the rate of 1.24 per cent per year, and labour
share continued its expansion and even accelerated to the pace of 6.02 per cent.
Consequently, income of the tertiary education category decreased at an annual
pace of approximately 3.06 per cent. Finally, changes in income bill share and
labour share of the primary education category continued to decrease; however,
the former slowed to the rate of 1.34 per cent and the latter accelerated to the
pace of 1.43 per cent; as a result, average income of the least educated group
continued to fall but at a negligible annualised rate of 0.08 per cent.

9.3.2 Decomposition of income by economic sectors

Here we evaluate whether the service sector is likely to raise post-liberalisation
in relation to other income sources and its evolution over further periods. Income
is decomposed into three main economic sectors – agriculture, manufacturing,
and services. Initially, panel 4 in Table 9.3 illustrates changes in the sectoral
composition of employment by reporting labour shares weighted by hours. It can
be observed that the share of agriculture declined throughout the period and the
decrease is more severe between 1998 and 2006. The manufacturing sector
remained more or less steady. In contrast, the share of the service sector
increased and the largest variation is registered during the latest period too.
Average real hourly income per sector is presented in panel 1 and the percent-
age change throughout the period is displayed in panel 2. We observe that indi-
vidual income fell 7.95 per cent in the service sector whereas it fell 24.37 and
31.63 per cent in the manufacturing and agricultural sectors respectively.
Panel 3 presents ratios of average hourly income. This indicator shows that
income dispersion between the service and the other two sectors expanded
Theory and evidence in Mexico 203
Table 9.3 Average real hourly income (2002 pesos) and educational attainment per

1984 1989 1994 1998 2002 2006

(1) Income
Agriculture 21.78 19.59 19.46 16.75 14.61 14.89
Manufacturing 27.98 25.91 24.69 19.46 20.94 21.16
Services 30.57 33.57 34.05 28.62 27.9 28.14
(2) Income variation 2006 vs 1984
Agriculture –31.63%
Manufacturing –24.37%
Services –7.95%
(3) Ratio of income
Agriculture* 1.40 1.71 1.75 1.71 1.91 1.89
Manufacturing** 1.09 1.30 1.38 1.47 1.33 1.33
(4) Share of hours
Agriculture 22.73 20.49 22.11 19.77 13.82 13.17
Manufacturing 17.99 18.29 17.18 17.68 18.01 18.26
Services 59.28 61.21 60.71 62.55 68.17 68.57
(5) Share of income bill
Agriculture 16.76 13.13 13.57 12.05 7.56 7.03
Manufacturing 18.9 16.86 16.14 16.76 16.29 16.56
Services 64.34 70.01 70.28 71.19 76.15 76.43
(6) Years of education
Agriculture 2.94 3.54 3.52 4.41 3.85 3.93
Manufacturing 6.69 7.61 7.2 7.58 7.83 8.14
Services 7.13 8.30 8.00 8.36 8.94 9.30

Source: author’s computation with information from INEGI (various years).

* Comparison between the service sector and the agricultural sector; ** comparison between the
service sector and the manufacturing sector.

between 1984 and 1998; however, during the last period the income gap with
respect to the manufacturing sector fell, whereas it continued to increase in rela-
tion to the agricultural sector. Panel 5 illustrates that income bill share of the
service sector rose along the whole period. It remained more or less steady in the
manufacturing sector after a decrease between 1984 and 1989; in contrast,
income bill share of the agricultural sectors dropped permanently. Finally, panel
6 shows that individuals in the service sector have higher educational attainment
and their skill upgrading is faster than in the other two sectors. In contrast, indi-
viduals in the agricultural sector have the lowest educational attainment and their
skill upgrading is the slowest.
Table 9.4 separates income bill shares for individuals by sector and education
category and reveals that in the service sector skill demand increased perman-
ently, whereas demand for unskilled individuals fell in relative terms. Relative
204 G. Angeles-Castro
Table 9.4 Income bill share per sector and level of education

1984 1989 1994 1998 2002 2006

Primary 15.63 11.98 12.30 10.66 6.52 6.02
Secondary 0.64 0.42 0.65 0.77 0.43 0.57
Tertiary 0.49 0.73 0.62 0.62 0.61 0.71
Primary 12.82 10.00 9.47 9.20 9.16 8.63
Secondary 2.28 3.07 2.90 3.00 3.53 3.37
Tertiary 3.79 3.79 3.78 4.57 3.61 4.17
Primary 39.76 33.31 29.97 29.88 31.51 29.90
Secondary 11.78 13.89 14.22 14.77 15.58 16.06
Tertiary 12.80 22.81 26.10 26.53 29.06 30.57

Source: author’s computation with information from INEGI (various years).

demand for unskilled labour was expected to increase in the manufacturing and
agricultural sectors, but it fell gradually in the former and substantially in the
latter. From these two sectors, relative skill demand seems to remain steady
throughout the period.
From the descriptive analysis provided above, changes in inequality can be
explained as follows. Between 1984 and 1998 both employment and skill
demand increased in relative terms in the service sector. In contrast, relative
employment and demand for unskilled individuals did not increase in the manu-
facturing and agricultural sectors, as predicted by standard theory; this fact can
explain the increasing income gap between the service sector and the other two
sectors, and can be consistent with the rise of services argument. Following the
Arbache et al. (2004) industry classification, we can consider the agricultural
and the manufacturing sectors as the traded industry and the service sector as the
non-traded industry; in this sense, we observe a sharp fall of income in the traded
industry relative to the non-traded industry, and this pattern is in keeping with
the argument supporting the reduction of rents in the traded sector. Furthermore,
in this period demand for skill increased faster than supply, whereas relative
demand for unskilled individuals dropped; this fact can explain the upturn in
skill premium and is consistent with the SETH. Hence, the increase of relative
income in the service sector, the sharp fall of income in the traded sector, espe-
cially in the agricultural sector, and the expansion of the skill premium can con-
tribute to explain the growth of overall inequality between 1984 and 1998.
However, it is important to explore whether the non-traded sector keeps a rela-
tive increase in average income, once education and other variables are control-
led for.
Between 1998 and 2006 the increase in overall skill demand slowed down
and educational attainment, on average, increased faster, although the increase
Theory and evidence in Mexico 205
does not necessarily occur among individuals with relative low income and low
educational achievement.7 As a result, skill premium declined and this fact seems
to be an important reason explaining the decrease in inequality in this period.
Both employment and skill demand continued to increase in relative terms in the
service sector, whereas in the manufacturing sector relative employment had
negligible improvements and relative demand for unskilled individuals fell
slightly. Bearing this in mind, we should expect further income dispersion
between these two sectors; nevertheless the income gap dropped, the most plau-
sible reason for this drop is thus a reduction of skill premium. However, as noted
before, it is important to explore whether, allowing for education and other vari-
ables, the changes of relative income between sectors persist.

9.4 Econometric analysis with disaggregate data (labour

This section extends the preliminary analysis by conducting parametric methods.
It uses labour income data and applies standard Mincerian earning functions, in
which the log of real monthly wages are regressed on personal characteristics
and different variables in order to analyse the effect of the skill premium, returns
to labour by sector, and openness on wage dispersion. So as to explore the
effects of economic liberalisation over different stages in time, the analysis
follows a before–after (liberalisation) approach as in Arbache et al. (2004), and
also splits the sample in different periods. A (0,1) dummy variable is created; it
takes a value of one for the post-liberalisation period, which is defined as after
1984. In addition, the impact of liberalisation is explored separately for the dif-
ferent sectors; we also focus on returns to education and the effect of market
openness pre- and post-liberalisation by applying the corresponding interactions.

9.4.1 Returns to labour by sectors

Column 1 of Table 9.5a shows an OLS regression which decomposes the log
wage between sectors (agriculture and manufacture vs services) and distin-
guishes trade regime. On average, workers in the agricultural and manufacturing
sector were paid 50.80 per cent less and 6.37 per cent more respectively than
those in the service sector before liberalisation.8 However, as anticipated in
panels 1 and 2 of Table 9.3, in the post-liberalisation period average wages in
agriculture and manufacturing fell more than those in services.9 In agriculture
average wages dropped 20.18 per cent and in manufacturing they dropped 26.85,
whereas they fell 13.80 per cent in services.10
Once age, gender, education attainment, and unionisation are controlled for in
column 2, we observe that higher human capital and higher unionisation rates in
the service sector contribute to increase average wages compared to those in the
other sectors. By comparing columns 1 and 2 we notice that before liberalisation
the wage gap between the agricultural and service sectors changes from –50.80
per cent to –42.61 per cent and between the manufacturing and service sectors it
206 G. Angeles-Castro
changes from 6.37 per cent to 11.02 per cent. Moreover, after liberalisation the
drop of average wages in the service sector is larger and the fall of average
wages in the manufacturing sector is more moderate. Nevertheless, column 2
illustrates that average wages post-liberalisation increase in relative terms in the
service sector, as they fall 19.62 per cent, whereas wages in agriculture and man-
ufacture drop 21.11 and 24.13 per cent respectively. In addition, the wage gap of
agriculture widens slightly from –42.61 per cent to –43.68 per cent and the wage
gap of manufacturing decreases from 11.02 per cent to 4.79 per cent, between
the pre- and post-liberalisation periods.11 This result is consistent with the rise of
services argument; it is also consistent with the reduction of rents in the traded
sector argument, if we consider both the agricultural and the manufacturing
sectors as the traded sector.
We also observe that there is an inverted U-shaped age-earning profile with a
peak at around 45 years, women earn 28.40 per cent less than men with similar
age and education, union workers earn 31.30 per cent more than equivalent non-
union workers, and returns to education increase with higher education levels.
This wage equation explains 38 per cent of total variation in earnings between
Using the before and after methodology the post-liberalisation period is dis-
aggregated in five sub-periods in order to examine any differential effect of lib-
eralisation over time; results are presented from column 3 to column 7 in Table
9.5b. The sharpest fall in wages in the immediate post-liberalisation period
occurs in the manufacturing sector and the most moderate occurs in the agricul-
tural sector. Over the subsequent periods wages continue to fall, there is some
recovery in the manufacturing and service sectors between 1998 and 2002, but

Table 9.5a Performance of sectors (labour income)

(1) (2)

Pre-lib. Post-lib. Pre-lib. Post-lib.

Services –0.148 –0.218

Agriculture –0.709 –0.935 –0.555 –0.793
Manufacture 0.062 –0.251 0.105 –0.172
Age 0.080
Age2 –0.001
Female –0.334
Union 0.272
Secondary 0.424
Tertiary 0.913
Constant 7.968 6.637
Observations 78,815 78,815
R2 0.11 0.38

Source: author’s computation with information from INEGI (various years).

Table 9.5b Performance of sectors (labour income)

(3) 1984, 1989 (4) 1984, 1994 (5) 1984, 1998 (6) 1984, 2002 (7) 1984, 2006

Pre-lib. Post-lib. Pre-lib. Post-lib. Pre-lib. Post-lib. Pre-lib. Post-lib. Pre-lib. Post-lib.

Services –0.113 –0.091 –0.375 –0.275 –0.269

Agriculture –0.593 –0.683 –0.536 –0.694 –0.523 –0.877 –0.552 –0.962 –0.547 –0.963
Manufacture 0.103 –0.059 0.117 –0.085 0.109 –0.308 0.099 –0.222 0.106 –0.209
Age 0.083 0.083 0.084 0.080 0.080
Age2 –0.001 –0.001 –0.001 –0.001 –0.001
Female –0.324 –0.345 –0.340 –0.328 –0.308
Union 0.185 0.263 0.350 0.326 0.310
Secondary 0.420 0.503 0.473 0.394 0.382
Tertiary 0.770 1.045 0.988 0.868 0.842
Constant 6.538 6.469 6.461 6.567 6.501
Observations 16,137 17,108 15,311 22,281 27,535
R2 0.33 0.41 0.40 0.38 0.40

Results corrected for heteroskedasticity, all coefficients are significant at the 1% level.
208 G. Angeles-Castro
wages do not return to their pre-liberalisation levels in any of these two sectors.
In the agricultural sector the fall is permanent.
When education and other variables are controlled for, we observe that a
source of inequality between the pre- and post-liberalisation periods is the
change in the wage gap of agriculture relative to services, as it widens from
–42.61 per cent to –43.68 per cent, although the variation is slight. On the other
hand, the relative increase in wages in the service sector tends to equalise wages
in relation to the manufacture sector, since manufacturing wages are originally
higher but then the wage gap between these two sectors falls from 11.02 per cent
to 4.79 per cent. Thus, the rise of services and the relative reduction of rents in
the traded sector have an unequalising effect in relation to the agricultural sector
The evolution of returns to labour by sectors does not contribute to explain
the reversal of inequality between 1998 and 2000, because the wage gap relative
to services continues to widen in agriculture and remains relatively stable in
Table 9.6 focuses on the returns to education pre- and post-liberalisation. The
first column illustrates the results obtained from the whole sample and the last
five columns show the results obtained once the post-liberalisation period is dis-
aggregated over time. Three main findings emerge from this analysis. First,
average income tends to be lower in every level in the post-liberalisation periods
and this is consistent with decreasing real wages as noted previously. Further-
more, as anticipated in panels 1 and 2 of Table 9.1, average wages for the
primary and secondary levels fall relative to the tertiary level. In the post-
liberalisation period workers with primary and secondary education are paid
21.20 per cent and 25.82 per cent less respectively, whereas workers at the
highest educational level are paid 9.22 per cent less.12
Second, the marginal returns to education – comparing each education level
with those below – tend to be greater along the post-liberalisation periods only
for high-skill workers or those with tertiary education, but not for those with sec-
ondary education. The point estimate of the marginal return to tertiary level rises
from 119.10 per cent to 152.42 per cent and from 34.74 per cent to 64.89 per
cent in relation to the primary and secondary levels respectively, between the
pre- and post-liberalisation periods.13 This finding confirms the trend observed in
panel 3 of Table 9.114 and is in keeping with the skill-enhancing trade hypothe-
sis.15 Finally, the marginal returns to tertiary education peak by 1994 and then
decline but remain higher than in the pre-liberalisation period.
When controlling for sectors, unionisation, and personal characteristics, we
observe that the evolution of skill premium post-liberalisation is a factor that has
a clear effect on changes in inequality because returns to tertiary education
increase after 1984 and this fact widens the income gap between skilled and
unskilled labour. This trend is due to a faster increase in skill demand than
supply, whereas relative unskilled demand decreased faster than supply between
1984 and 1998, as shown in Table 9.2. Hence, the evidence illustrates that during
periods of economic liberalisation and its intrinsic technological change, relative
Table 9.6 Returns to education (labour income)

(1) (2) 1984, 1989 (3) 1984, 1994 (4) 1984, 1998 (5) 1984, 2002 (6) 1984, 2006

Pre-lib. Post-lib. Pre-lib. Post-lib. Pre-lib. Post-lib. Pre-lib. Post-lib. Pre-lib. Post-lib. Pre-lib. Post-lib.

Age 0.080 0.083 0.083 0.084 0.080 0.080

Age2 –0.001 –0.001 –0.001 –0.001 –0.001 –0.001
Female –0.335 –0.325 –0.348 –0.344 –0.330 –0.310
Union 0.271 0.184 0.261 0.348 0.321 0.305
Primary –0.238 –0.098 –0.160 –0.407 –0.308 –0.298
Secondary 0.486 0.188 0.508 0.295 0.494 0.347 0.483 0.066 0.455 0.077 0.475 0.081
Tertiary 0.784 0.688 0.800 0.663 0.786 0.928 0.782 0.643 0.760 0.580 0.781 0.574
Agriculture –0.571 –0.576 –0.585 –0.509 –0.655 –0.659
Manufacture 0.052 0.068 0.038 0.081 0.064 0.060
Constant 6.394 6.526 6.523 6.486 6.590 6.524
Observations 78,815 16,137 17,108 15,311 22,281 27,535
R2 0.37 0.33 0.41 0.40 0.38 0.40

Source: author’s computation with information from INEGI (various years).

Results corrected for heterosedasticity, all coefficients are significant at the 1% level.
210 G. Angeles-Castro
demand for skill tends to increase and this pattern supports the SETH. Moreover,
variations in skill premium also contribute to explain the fall in overall inequal-
ity after 1998 because skill premium tends to fall, especially after this year. The
cause of this trend is also shown in Table 9.2, where we observe that the increase
in skill demand slows down whereas the increase in supply accelerates, and
unskilled supply falls faster than demand. Hence the rise in skill premium is tem-
porary and cyclical.
We also use the data-set comprising all income sources in the analysis of edu-
cational levels and sectors, and find that the general conclusions are similar to
those using labour income only.

9.5 Additional forms of income distribution (all income


9.5.1 Household inequality vs individual inequality

Initially, a simple comparison between households and individuals in terms of
income and Gini coefficients is presented in Table 9.7. From the first panel we
observe that household Gini is lower than individual Gini and the last column
reveals that the former grew slower than the latter throughout the period. More-
over, the rise of household Gini started to reverse slightly after 1994, whereas
individual Gini started to drop after 1998. The second panel shows that real
hourly individual income declined 12.08 per cent, whereas real monthly house-
hold income fell 0.45 per cent over the whole period. It is worth noting that
household income increased 18.57 per cent when it is expressed in per capita
An important reason for mitigation of inequality and income fall among
households is presented in the bottom panel. We observe that the average
number of members per household dropped 17.41 per cent, whereas the number
of income receivers increased 33.87 per cent. As a result, the proportion of
income receivers per household increased, from 31.66 per cent to 51.31 per cent
between 1984 and 2002. Although the upper quintiles have kept a higher propor-
tion of income receivers over time, the lower quintiles have increased the pro-
portion faster and therefore the percentage of income receivers tends to converge
across income levels. This families’ reaction counteracts the increase in inequal-
ity and the general trend of declining real income; this in fact raises per capita
household income.

9.5.2 Gini decomposition by income source

Table 9.8 presents the decomposition of the household Gini coefficient by three
main income sources – labour, transfers, and business and finance (B&F ) –
applying the Yao (1999) method. The first panel reveals that the transfer income
is the most equally distributed and its Gini has fallen markedly, as recorded in
column 7. In contrast, the Gini of B&F income is the largest and has expanded
Table 9.7 Average real monthly income (2002 pesos), Gini, and composition of income receivers per household and individuals

1984 1989 1994 1998 2002 2006 Change %

2006 vs 1984

(1) Gini
Household Gini 0.485 0.530 0.553 0.549 0.515 0.504 3.90
Individual Gini 0.512 0.551 0.587 0.632 0.560 0.548 7.02
(2) Income
Monthly income per household 6,441 7,146 6,928 5,859 6,348 6,412 –0.45
Monthly income per member 1,270 1,437 1,469 1,331 1,499 1,505 18.57
Hourly individual income 28.03 29.30 29.02 24.40 24.55 24.65 –12.08
(3) Household composition
Household members 5.07 4.97 4.72 4.40 4.23 4.19 –17.41
Receivers per household 1.61 1.68 1.86 2.00 2.10 2.15 33.87
Receivers per household % 31.66 33.86 39.52 45.51 49.63 51.31 62.08
Receivers per quintile %
1st 23.03 22.71 29.82 37.90 46.50 47.75 107.37
2nd 27.21 29.30 33.93 41.59 44.50 46.17 69.69
3rd 31.41 35.12 40.13 45.53 47.81 49.14 56.45
4th 36.95 42.27 46.78 52.11 54.31 55.11 49.13
5th 47.84 47.59 54.98 56.42 59.26 59.93 25.28

Source: author’s computation with information from INEGI (various years).

212 G. Angeles-Castro
sharply over time. As a result, the second and third panel illustrate that the con-
tribution of transfer income to total income is larger and has increased faster
over time (column 8) than its contribution to the overall Gini, whereas the con-
tribution of B&F income to total income is lower and has decreased more than
its contribution to the overall Gini. As for labour income, its income share has
increased slightly more than its Gini share.
Consequently, transfer income, which is mainly composed of remittances
from emigrant workers and social government expenditure, helps to reduce
household inequality for the following reasons. First, its Gini is reasonably
smaller than the overall Gini and the gap has expanded over time (panel 4).
Second, although the smallest income source out of the three categories, it has
increased gradually. In this context and to a lesser extent, labour income helps to
reduce household Gini too. On the other hand, B&F income drives inequality up
as its Gini is higher than the overall household Gini and the gap has tended to
increase over time (panel 4).
The last panel summarises the impact of every income category on the overall
household income inequality by displaying the ratio of Gini share to income
share. If the ratio is greater than one, it means that the corresponding income
source can increase inequality, otherwise it helps to decrease the household Gini
coefficient. We observe that the contribution ratio of the B&F income is greater
than one and has increased over time (column 7), which suggests that this
income source is a driving force of household inequality. Labour income is rela-
tively neutral. Finally, the contribution ratio of transfer income is the lowest and
has decreased sharply along the period, which indicates that transfer income is
an important factor to reduce the household Gini coefficient.

9.6 The effects of market openness on income dispersion

The before–after methodology, conducted so far, compares variations in the
income structure before and after trade liberalisation, but it does not disentangle
the effect of market openness. In order to address this problem, we apply a
second strategy in this section. The strategy includes an industry-level index of
openness in the analysis; the index allows for intensity and timing in the process
of liberalisation across the traded sector. However, due to the nature of the anal-
ysis we restrict the sample to the traded sector only.
The industry-level index of openness is expressed as follows:

where ejt denotes the effective rate of protection in industry j at time t.

There is an inverse relationship between the measure of openness and e
because higher tariffs reflect more protectionism, while lower tariffs reflect a
more competitive environment.
Table 9.9 presents the impact of trade reforms on log real wages using the
measure of openness in 17 traded industries across time. In all columns we
Table 9.8 Decomposition of household Gini by income source

(1) 1984 (2) 1989 (3) 1994 (4) 1998 (5) 2002 (6) 2006 Change

(7) % (8) Diff.

2006 vs 1984 2006–1984

(1) Gini
Business and finance 0.515 0.624 0.612 0.616 0.601 0.587 14.08
Labour 0.471 0.476 0.547 0.526 0.496 0.472 0.30
Transfer 0.448 0.481 0.384 0.452 0.378 0.369 –17.72
Total 0.485 0.530 0.553 0.549 0.515 0.503 3.79
(2) Contribution to income
Business and finance 35.73 35.98 32.20 33.89 29.40 30.07 –5.66
Labour 56.63 56.70 59.53 56.14 60.00 58.76 2.13
Transfer 7.64 7.32 8.28 9.97 10.60 11.17 3.53
(3) Contribution to Gini
Business and finance 37.94 42.38 35.37 38.03 34.34 35.26 –2.68
Labour 55.00 50.97 58.52 53.77 57.87 57.12 2.12
Transfer 7.07 6.64 6.10 8.20 7.79 7.62 0.55
(4) Gini variation % (income source vs overall)
Business and finance 6.18 17.79 10.64 12.21 16.78 16.70
Labour –2.89 –10.10 –1.00 –4.22 –3.55 –6.16
Transfer –7.46 –9.21 –30.52 –17.76 –26.46 –26.64
(5) Contribution ratio (gini/income)
Business and finance 1.06 1.18 1.10 1.12 1.17 1.17 10.44
Labour 0.97 0.90 0.98 0.96 0.96 0.97 0.10
Transfer 0.93 0.91 0.74 0.82 0.74 0.68 –26.28

Source: author’s computation with information from INEGI (various years).

214 G. Angeles-Castro
Table 9.9 The effect of market openness on income (traded sector)

(1) (2) 1984–1998 (3) 1998–2006

Open –0.227 –0.275 –0.164

Age 0.071 0.078 0.07
Age2 –0.001 –0.001 –0.001
Female –0.337 –0.339 –0.335
Union 0.269 0.254 0.278
Agriculture –0.387 –0.367 –0.394
Secondary education 0.422 0.419 0.426
Tertiary education 0.919 0.927 0.886
Constant 6,478 6,523 6,469
Observations 17,928 9,143 8,785
R2 0.39 0.41 0.38

Source: author’s computation with information from INEGI (various years).

control for educational levels, age, gender, and unionisation. The first column
reveals that greater openness is associated with lower income. This result repre-
sents support for the reduction of rents in the traded sector argument, which
stresses that more competition can reduce rents and therefore can reduce wages
too. A 10 per cent increase in the index of openness, ceteris paribus, reduces
income by 2.03 per cent.
Columns 2 and 3 split the whole time period in two sub-periods, 1984–1998
and 1998–2006, which are the periods of rising inequality and decreasing
inequality, respectively, as commented before. In the first period we observe that
the effect of trade liberalisation on income is more adverse; an upturn of 10 per
cent in the openness index, ceteris paribus, reduces income by 2.40 per cent;
whereas in the period of decreasing inequality the reduction is smaller, 1.51 per
Two main conclusions emerge from these results. First, market openness
reduces income in the traded sector and this pattern helps to explain the income
polarisation between this sector and the service sector. Second, over the longer
run the adverse effect of market openness on income tends to decline and this
pattern helps to explain the reduction of income inequality after 1998. To some
extent this pattern gives support to the temporary adverse effects argument,
which claims that in the longer run markets react and individuals adapt to a more
competitive environment, and therefore the dislocations that occur in a country
during its early economic liberalisation period tend to banish in subsequent

9.7 Concluding remarks

Due to market-oriented reforms in Mexico since the mid-1980s, and on the basis
of the SST we might expect a rise in the relative return to low-income, unskilled
labour, or an increase in individual income in activities such as agriculture and
Theory and evidence in Mexico 215
labour-intensive manufacturing, and therefore a reduction of income inequality.
However, in the post-liberalisation period skill premium and income differential
between low- and high-income individuals expanded,16 and relative income in agri-
culture and manufacturing dropped. Furthermore, overall individual inequality
increased, although there is some evidence that has tended to decline after 1998.
These trends undermine orthodox theory and provide room for contesting
The analysis, finds various factors driving inequality between 1984 and 1998.
An important reason for income dispersion is the fact that marginal returns to
education increased, which is consistent with the SETH. Note, however, that the
hypothesis applies to tertiary education in particular.
In the service sector relative income, employment, and demand for skill
increased; consequently, the evidence corresponds with the rise of service argu-
ment. This pattern contributes to explain income dispersion, in the sense that the
wage gap between the service and the agricultural sectors expanded.
Relative income in the traded sector fell following liberalisation, and this is in
keeping with the view that market-oriented reforms increased the degree of com-
petition and therefore reduced rents. Income also dropped in the non-traded indus-
try, indicating either a degree of spillover, or the effect of other reforms such as
privatisation or deregulation.17 However, the relevant finding is that income in the
traded sector fell in relative terms, which is another reason of income dispersion.
The evidence also corresponds with the decline of labour market institutions
argument to the extent that average wages, union density, and union premium
fell. However, changes in the wage gap between union and non-union workers
cannot contribute to explain an increase in income dispersion, as the gap
decreased in average in the post-liberalisation period. Only around 1998 union
premium was higher than its position pre-liberalisation. Nevertheless, the fact
that a large number of workers moved away from unions and entered a non-
union sector, characterised by diverse and flexible wages and higher Gini coeffi-
cient, represents a source of inequality.
The rise in income Gini coefficient reversed between 1998 and 2002 and so did
the income gap between upper and lower deciles. The factors that can explain this
variation are summarised as follows. The upturn in skill premium started to reverse
around 1994 and the downturn was faster around 1998. In addition, by 1998 the
wage gap between union and non-union workers had peaked and fell afterwards
and the fall in the unionisation rate stopped and reversed slightly.
In this respect, some authors have stressed the possibility that income distri-
bution can follow cycles under conditions of market openness and technological
change. One of these approaches explains that when a country begins to adjust to
a more competitive environment serious dislocations are encountered as the
economy adapts to the shifting patterns of employment and resources. As a con-
sequence, income dispersion may widen and absolute poverty increases in
the short run. However, this effect is considered to be temporary because as the
period of adjustment continues markets stabilise and individuals adapt to the
prevailing conditions. Eventually, there may be a decrease in unemployment and
216 G. Angeles-Castro
income gap, and inequality may begin to decrease in the longer run (Jacobsen
and Giles, 1998: 419–420, FitzGerald, 1996: 32).
In keeping with this approach, evidence in the Mexican case shows that over
the longer term, individuals react by achieving higher educational attainment or
increasing movements towards higher income activities.18 In addition, transition
and adjustment in labour unions seem to come to an end, or at least changes are
less marked. Finally we observe that individuals tend to increase the number of
income receivers and to reduce the number of members in their households,
which leads to higher per capita income, especially in low-income sectors.
In terms of technological change Pissarides (1997) shows that in developing
countries, that have adopted market-oriented policies, the importation and assim-
ilation process of new technology can be skill biased and give a temporary and
relative advantage to skilled labour only during the period of transition towards
a higher level of technology. He also argues that the response of relative supply
of skilled and unskilled labour to trade openness can also explain a temporary
increase of wage differentials. In addition, Goldin and Katz (1998) hold that
within firms, demand for skill rises when new technologies are introduced, but it
declines once the other workers have learned to use the new equipment. Around
1998 the evidence starts to correspond with these ideas since we observe higher
levels of educational achievement and an acceleration of skill supply in relation
to previous periods, whereas skill demand falls substantially.
Although we have found factors that can contribute to lessen inequality in the
longer term, there are adverse effects lasting the whole period of study, for
instance the deterioration of the agricultural sector. We also found two main
factors that can contribute to mitigate adverse effects; they are the recomposition
of households and transfer income.
The study identifies reactions of individuals that can help to reduce inequal-
ity. However, the results suggest that solutions for income inequality can also
rely on government action. Some of the main policies implied are to increase
expenditure in the form of transfers, to take strategic action to develop the agri-
cultural sector, and to facilitate access to education, especially to the vulnerable
and those at low-income levels. Furthermore, the boost of employment in
unskilled, labour-intensive activities, combined with the reduction of supply of
unskilled individuals by increasing educational levels can encourage factor price
equalisation. However, heavy reliance on low-wage employment is not a desir-
able long-term solution as it does not encourage domestic markets and sustained
growth; in this context, gradual and strategic industrialisation can be a com-
plementary strategy. Finally, income redistribution can be encouraged by intro-
ducing a progressive taxation policy at the highest income levels.

1 From a sample of 49 countries including different definitions of Gini coefficients over
time, Li et al. (1998) show that the Mexican average Gini, 54.59, is the second highest
of the sample.
2 The classification is conducted under the following criteria. Primary level comprises
Theory and evidence in Mexico 217
individuals with some elementary or completed elementary education. In Mexico, the
first nine years of the educational system are considered elementary education. Sec-
ondary level includes individuals with some education after the basic level but with
no university education. The tertiary level comprises individuals with university edu-
cation, completed or incomplete, also includes individuals with one or more years of
postgraduate education.
3 We also calculated the educational distribution weighted by individuals, and it was
found that it does not differ substantially from that weighted by hours.
4 Autor et al. (1998) show that under the assumption that the elasticity of substitution
between skilled and unskilled workers equals one and the production function is Cobb-
Douglas, relative demand shifts can be represented by changes in wage bill share.
5 By conducting an international comparison Cortez (2001) shows that Mexico’s per-
formance in terms of educational expansion is poor, as the reduction in the percentage
of low-education workers is slow and the increase in the percentage of workers with
higher education is small compared to countries like South Korea and Sweden.
6 Although the average hourly income of the highly educated group rose between 1984
and 1998, the annualised change was relatively low, 0.43 per cent; however, between
1984 and 1994 the rate was higher, 1.47. This can be explained because supply of
skilled individuals fell between 1989 and 1994 and then continued to increase in the
later periods. Nevertheless, the analysis considers periods before and after 1998,
because the rest of the changes in the trend of supply and demand of skilled and
unskilled individuals, and changes in individual income Gini, appear to be stronger
around this year.
7 We also decompose income by deciles and find that the lower the income level of
individuals the fewer the years of schooling received in every period. Furthermore,
throughout the period the speed of skill upgrading is the lowest in the first quintile
whereas it is the highest in the fifth quintile. We also find that between 1998 and 2006
human capital increased faster in the last fourth quintiles compared to the previous
periods, only the first quintile showed a decrease. Information computed by deciles is
available upon request.
8 Figures computed as (exp(β) – 1)*100 where β is the coefficient on the corresponding
sector dummy variable.
9 Although Table 9.2 comprises all sources of income, in 2002 labour income accounted
for 60 per cent of total income; hence, this table can be representative of the pattern
followed by this income source over time. Moreover, we construct Table 9.2 using
labour income only and also observe that relative wages in agriculture and manufac-
turing fell more than in services.
10 The change in average wages after liberalisation is computed as (exp(b1 – b2) – 1)*100,
where b1 and b2 are the coefficients of the corresponding sector post- and pre-
liberalisation respectively.
11 In the post-liberalisation period the wage gap of agriculture and manufacturing in
relation to services is computed as (exp(b1 – b2) – 1)*100, where b1 is the coefficient of
the corresponding sector and b2 is the coefficient of the service sector post-
12 Fall post-liberalisation is computed as (exp(b1 – b2) – 1)*100, where b1 and b2 are the
coefficients of the corresponding educational level post- and pre-liberalisation
13 Marginal returns to education comparing two levels of education can be obtained as
(Exp(bupper – blower) – 1)*100, where bupper and blower are the coefficients on the education
level dummy variable for the upper and lower level respectively for a specific period.
14 Although Table 9.1 is constructed from all income sources it shows a good approxi-
mation of the trend in labour income as this income source represents 60 per cent of
total income, as noted in Note 10.
15 Note that the hypothesis applies to tertiary education in particular. This finding is
218 G. Angeles-Castro
similar to that obtained by Arbache et al. (2004) for the case of Brazil, as they con-
clude that the SETH applies to college-educated labour only.
16 By decomposing the overall income by deciles, it is found that the bottom nine deciles
lost income share and decreased average real hourly income between 1984 and 1998;
furthermore, the lower the income level, the higher is the loss. In contrast, the top
decile gained income share and increased average income in this period. Con-
sequently, the ratios of the tenth decile to the first decile, in both indicators, increased
until 1998, and they actually doubled, as they passed from 32 to 64. Information com-
puted by deciles is available upon request.
17 Arbache et al. (2004) reached similar conclusions for the case of Brazil.
18 By decomposing income by quintiles and economic sectors we observe that between
1998 and 2002 the employment share in services and manufacturing increased, but
the former had the highest increase in the first and second quintile and the latter in the
first quintile.

Acemoglu, Daron (2002) Technical Change, Inequality and the Labour Market. Journal
of Economic Literature, 40(1), pp. 7–72.
Airola, Jim and Juhn, Chinhui (2005) Wage Inequality in Post-Reform Mexico. Discus-
sion paper series, No. 1525, IZA – Institute for the Study of Labour.
Arbache, Jorge S., Dickerson, Andy, and Green, Francis (2004) Trade Liberalisation and
Wages in Developing Countries. Economic Journal, 114(493), pp. 73–96.
Autor, David H., Katz, Lawrence F., and Krueger, Alan B. (1998) Computing Inequality:
Have Computers Changed the Labour Market? Quarterly Journal of Economics,
113(4), pp. 1169–1213.
Baldwin, R.E. and Cain, G.G. (2000) Shifts in US Relative Wages: The Role of Trade,
Technology, and Factor Endowments. Review of Economics and Statistics, 82(4),
pp. 580–595.
Berman, Eli, Bound, John, and Machin, Stephen (1998) Implications of Skill Biased
Technological Change: International Evidence. Quarterly Journal of Economics,
113(4), pp. 1245–1279.
Camdessus, Michel (1998) ‘Income Distribution and Sustainable Growth: The Perspec-
tive from the IMF at Fifty’, in: Vito Tanzi and Ke-Young Chu (eds) Income Distribu-
tion and High-Quality Growth (Cambridge, MA: MIT Press), pp. xi–4.
Corden, W.M. (1993) Protection and Liberalisation: a Review of the Analytical issues.
Occasional paper, No. 54, International Monetary Fund.
Cortez, Willy W. (2001) What is Behind Increasing Wage Inequality in Mexico? World
Development, 29(11), pp. 1905–1922.
Feliciano, Zadia M. (2001) Workers and Trade Liberalisation: The Impact of Trade
Reforms in Mexico on Wages and Employment. Industrial and Labour Relations
Review, 55(1), pp. 95–115.
Feenstra, R.C. and Hanson, G.H. (1997) Foreign Direct Investment and Relative Wages:
Evidence from Mexico’s Maquiladoras. Journal of International Economics, 42,
pp. 371–394.
FitzGerald, E.V.K. (1996) ‘The New Trade Regime: Macroeconomic Behaviour and
Income Distribution in Latin America’, in: Victor Bulmer-Thomas (ed.) The New
Model in Latin America and its Impact on Income Distribution and Poverty (London:
Macmillan Press), pp. 29–52.
Flemming, J.S. and Micklewright, John (2000) ‘Income Distribution, Economic Systems
Theory and evidence in Mexico 219
and Transition’, in: Anthony B. Atkinson and Francois Bourguignon (eds) Handbook
of Income Distribution, Vol. 1 (Amsterdam: Elsevier), pp. 843–918.
Goldin, Claudia and Katz, Lawrence F. (1998) The Origins of Technology-Skills Com-
plementarity. Quarterly Journal of Economics, 113(3), pp. 693–732.
Gordon, Jim and Gupta, Poonam (2003) Understanding Indian Service Revolution, Paper
prepared for the IMF-NCAR conference, International Monetary Fund, November
2003. Online, available at:
gordon.pdf (accessed August 2008).
Haskel, J. and Slaughter, M.J. (2001) Trade, Technology and UK Wage Inequality. Eco-
nomic Journal, 111(468), pp. 163–187.
INEGI, Instituto Nacional de Estadística Geografía e Informática (1985, 1990, 1995,
1999, 2003) ENIGH, Encuesta Nacional de Ingreso y Gasto de los Hogares, 1984,
1989, 1994, 1998, 2002, CD-Rom, Mexico, DF.
Jacobsen, Peter W.F. and Giles, David E.A. (1998) Income Distribution in the United
Sates: Kuznets’ Inverted-U Hypothesis and Data non-Stationary. Journal of Interna-
tional Trade and Economic Development, 7(4), pp. 405–423.
Jones, R. and Barry, J. (1988) ‘Liberal Political Economy’, in: R. Jones (ed.) The Worlds
of Political Economy (London: Pinter Publishers), pp. 27–56.
Li, Hongyi, Squire, Lyn, and Zou, Heng-Fu (1998) Explaining International and Intertem-
poral Variations in Income Inequality. Economic Journal, 108(446), pp. 26–43.
Litwin, Carol (1998) Trade and Income Distribution in Developing Countries. Working
papers in economics, No. 9, Department of Economics, Göteborg University.
Mah, Jai S. (2002) The Impact of Globalization in Income Distribution: The Korean
Experience. Applied Economic Letters, 9(15), pp. 1007–1009.
O’Connor, D. and Lunati, M.R. (1999) Economic Opening and Demand for Skills in
Developing Countries: A Review of Theory and Evidence. Technical Papers, No. 149,
OECD Development Centre.
Pissarides, Christopher A. (1997) Learning by Trading and Returns to Human Capital in
Developing Countries. World Bank Economic Review, 11(1), pp. 17–32.
Robbins, Donald. J. (1996) HOS Hits Facts: Facts Win; Evidence on Trade and Wages in
the Developing World. Discussion paper series, No. 557, Harvard Institute for Devel-
opment, Harvard University.
Ros, Jaime and Bouillon, César (2002) ‘Mexico: Trade Liberalization, Growth, Inequality
and Poverty’, in: Rob Vos, Lance Taylor, and Ricardo Paes de Barros (eds) Economic
Liberalization, Distribution and Poverty: Latin America in the 1990s (Cheltenham:
Edward Elgar Publishing, Inc.), pp. 347–389.
Sinha, Aseema (2005) Globalization, Rising Inequality and New Insecurities in India,
Paper presented at the conference on Difference on Inequality in Developing Societies,
University of Virginia, April 2005. Online, available at:
TaskForceDiffIneqDevSinha.pdf (accessed August 2008).
Smeeding, Timothy M. (2002) Globalisation, Inequality, and the Rich Countries of the
G-20: Evidence from the Luxembourg Income Study (LIS). Working Paper, No 48,
Syracuse University, Center for Policy Research.
Tanski, Janet M. and French, Dan W. (2001) Capital Concentration and Market Power in
Mexico’s Manufacturing Industry: Has Trade Liberalization Made a Difference?
Journal of Economic Issues, 35(3), pp. 675.
Yao, Shujie (1999) On the Decomposition of Gini Coefficients by Population Class and
Income Source: a Spread Sheet Approach and Application. Applied Economics, 31(10),
pp. 1249–1264.
10 How risk factors affect growth in
A free-market liberalism approach
Francisco Venegas-Martínez

10.1 Introduction
Nothing should matter more to a country and its inhabitants than the behavior of
its rate of economic growth in the long run. In this regard, a large number of
investigations have been produced in the specialized literature for the last two
decades. Hundreds, or perhaps thousands, of theoretical and empirical studies on
many countries have highlighted the correlation between economic growth and
its determinants. However, it remains to explain the correlation between growth
and risk factors. The goal of this chapter is to stress the connection between eco-
nomic growth and the currency, market, debt, and fiscal risks.
Economic growth is quantified as the increase in the amount of the goods and
services produced by a country in a given period of time. It is, conventionally,
calculated as the percent rate of increase in real gross domestic product (GDP).
An increase in GDP of a country is generally associated with an increase in the
standard of living of (all) its inhabitants. Therefore, the issue of setting up the
determinants of growth and how risk factors impact growth are two questions of
great interest to policy makers.
A market liberal revolution is reaching most of the countries in the world, and
Mexico is not the exception. The notion of free-market liberalism emphasizes
the support of free-market economies with personal freedom and human rights.
The concept of market liberalism can be used as a synonym to economic liberal-
ism when both the economic aspects of the classical liberalism and the indi-
vidual aspects of freedom are relevant; the market-liberal order is also ethical, in
the sense that it is based on personal freedom under the principles of law.
One of the objectives of this chapter is to develop, under a free-market lib-
eralism framework, a stochastic model of endogenous growth in which the
inhabitants of a country have expectations of depreciation of the exchange rate
driven by a diffusion process combined with Poisson jumps; other investiga-
tions on stochastic models of growth can be found in Canton (2001) and Gokan
(2002). It is supposed, in our modeling, that the economy has no contingent-
claims markets to hedge against future exchange-rate depreciation. It is worth-
while to mention that a study where exchange-rate derivatives are available
can be found in Venegas-Martínez (2005a). Other studies for the Mexican case
How risk factors affect growth in Mexico 221
within a stochastic framework can be found in Venegas-Martínez (2004,
2005b, 2006, 2008).
It is supposed, in the proposed model, that an uncertain tax rate on both
wealth and taxes are governed by a geometric Brownian motion. Assuming risk
adverse agents, we examine the growth rates of consumption and output in the
presence of taxes on wealth and consumption. The production function has con-
stant return to capital (Rebelo, 1991); we combine this technology with the opti-
mizing behavior of consumers and firms to obtain the stochastic per capita
growth rates of consumption, capital, and output. This research provides a mean-
variance description of the growth rate of output, showing explicitly the under-
lying risk factors that affect growth. Moreover, the proposed model is used to
carry out a simulation experiment that reproduces the observed mean and vari-
ance of the growth rate of output in Mexico from 1930 to 2002.

10.2 The setting of the economy

In this section, we establish the main characteristics of both the economy and its
inhabitants. Let us consider a small open economy with identical (in preferences
and endowments) and infinite-lived households, with personal freedom and
human rights, in a world with a single, internationally tradable and perishable
consumption good. The individuals are both consumer and producers and share
out a technology showing a constant marginal product.
In what follows, we assume that the generic good is freely traded, and its
domestic price level, pt, is determined by the purchasing power parity condition,


where st is the foreign-currency price of the good in the rest of world, and et is
the nominal exchange rate. We will assume, for the sake of simplicity, that st = 1.
We suppose that the number of atypical movements in the exchange rate, that
is, the jumps in the exchange rate, per unit of time, follows a Poisson process Qt
with intensity h, so P(N){dQt = 1} = hdt and P(N){dQt = 0} = 1 – hdt + o(dt). In such a


Let us consider now a Brownian motion, dVt, that is, dVt ~ N(0, dt)
E[dVt] = 0 and Var[dVt] = dt. We assume that the consumer perceives that the
expected inflation rate, dpt/pt, and consequently the expected rate of deprecia-
tion, det/et, follows a geometric Brownian motion with Poisson jumps in accord-
ance with

222 F. Venegas-Martínez
where e is the mean expected rate of depreciation conditional on no jumps, sP is
the instantaneous volatility of the expected price level, and m is the mean
expected size of an exchange-rate jump. Process Vt is supposed to be independ-
ent of Qt. In what follows, e, sP, h, and m are all supposed to be positive
The agent holds real cash balances, mt = Mt/pt, where Mt is the nominal stock
of money. The stochastic rate of return of holding real cash balances, dRm, is
given by the percentage change in the price of money in terms of goods, dmt/mt.
By applying Itô’s lemma for diffusion-jump processes to the inverse of the price
level, with (10.3) as the underlying process, it can be shown that


The agent also holds capital, kt, that pays a risk-free real interest rate r, which is
constant for all terms, satisfying dkt = rktdt, where k0 is given. The representative
agent takes r as given. Let us consider now a Brownian motion dWt, that is,


We assume that the representative consumer perceives that his/her wealth is

taxed at an uncertain rate, tt, in accordance with the following stochastic


where t0 > 0 and

_ _
with κ ∈(–1, 1). Here, τ is the mean expected growth_ rate of the taxes on wealth,
sP is the volatility of the tax rate on wealth, and κ is the correlation between
changes in inflation and changes in wealth taxes. Notice that an increase in the
rate of depreciation will produce a higher depreciation in real cash balances.
This, in turn, will reduce real assets, which could lead to the fiscal authority to
modify tax rates. Processes Qt, Vt, and Wt are supposed to be pairwise
Consider a cash-in-advance constraint of the form:


where ct is consumption, and a > 0 is the average time that money must be held
to finance consumption. Condition (10.9) is critical in linking exchange-rate
dynamics with consumption. Finally, observe that
How risk factors affect growth in Mexico 223


In the sequel, we will assume that the error o(a) is negligible.

10.3 The consumer’s decision problem

In this section, we characterize the optimal decisions of a representative agent on
consumption and portfolio shares through the Hamilton–Jacobi–Bellman con-
dition (necessary condition for a maximum) of the continuous-time stochastic
dynamic programming.
The stochastic consumer’s wealth accumulation in terms of the portfolio

and consumption, ct, is given by the following stochastic system:


where dRk = dkt/kt is the return of capital, and τ̂ is a resident-based ad valorem tax
rate on consumption. By substituting (10.4), (10.5), and (10.9) into the first equa-
tion of (10.10), we find


where b = (1 + τ̂ )a–1 + r + e – s P2. The von Neumann–Morgenstern (expected)

utility at time t, Vt, of the competitive consumer is assumed to have the time-
separable form:


Notice that the agent’s subjective discount rate has been set equal to the con-
stant real international rate of interest, r, to avoid unnecessary technical
difficulties. Of course, in the real world, when the subjective discount rate is
equal to the real rate of interest is due only to coincidence, We consider the
logarithmic utility function, u(ct) = log(ct), in order to derive closed-form solu-
tions and make the analysis easy to manage. In this case, the Hamilton–
Jacobi–Bellman equation for the stochastic optimal control problem of
maximizing the agent’s lifetime expected utility subject to the intertemporal
budget constraint is:
224 F. Venegas-Martínez



is the agent’s indirect utility function (or welfare function) and Lt (xt, tt, t) is the
co-state variable. Given the exponential time discounting in (10.14), we specify
L(xt, tt, t) in a time-separable form as




Here d0, d1, and J(tt, d2, d3) are to be determined from equation (10.15). Coeffi-
cients d2 and d3 must satisfy the following conditions:


By substituting (10.14) into (10.13), we have


The first-order conditions of the intertemporal optimisation of the risk averse

representative agent lead to a time-invariant wt ≡ w, and

How risk factors affect growth in Mexico 225
Figure 10.1 shows optimal w* as a function m and h. We choose now J(tt) as a
solution of the second-order ordinary non-homogeneous differential equation


Coefficients d0 and d1 are determined from (10.15) after substituting optimal w*.
Thus, d1 = r–1, so the coefficient of log(xt) in (10.17) becomes zero, and


Logarithmic utility implies that w depends only upon the parameters deter-
mining the stochastic characteristics of the economy, and hence w is constant. In
other words, the consumer’s attitude toward currency risk is independent of his/
her wealth, that is, the resulting level of wealth at any instant has no relevance
for portfolio decisions. Moreover, due to the logarithmic utility, the correlation

Figure 10.1 Optimal w* as a function m and h (source: own estimations based on INEGI).
226 F. Venegas-Martínez
coefficient, κ, plays no role in the consumer’s optimal portfolio, it only matters
the trend and volatility components of the stochastic processes driving the
dynamics of the exchange rate and the tax policy. Finally, it is important to point
out that equation (10.18) is cubic, therefore it has at least one real root.
Notice also that from the fact that d1 = r–1, it can be shown that the solution of
(10.19) is


and a = (2τ – s2t ). Coefficients d2 and d3 are determined in such a way that J(t0) = 0
and Jʹ(t0) = 0.
Equation (10.18) is cubic with one negative and two positive roots. This can
be seen by intersecting the straight line defined by the right-hand side of (10.18)
with the graph defined by the left-hand side of (10.18). In such a case, there is
only one intersection defining a unique, perfectly viable, steady-state share of
wealth set apart for consumption such that w * ∈ (0, 1).

10.4 Wealth dynamics

Through this section, we derive the stochastic process that generates an indi-
vidual’s real wealth when the optimal rule is applied: such a stochastic process
follows a stochastic differential equation that with three components; one which
provides the physical trend, another which models small movements observed
every day (diffusion part), and another describing atypical movements (jump
part). Thus, after substituting the optimal share w * into (10.11), we get





How risk factors affect growth in Mexico 227
We also have that conditional expectation is


and the conditional variance satisfies


It can be shown that the solution of the stochastic differential equation in (10.21),
conditional on x0, is







as usual, P(×) denotes a Poisson distribution. The stationary components of the

parameters of the above distributions are:


Notice also that the conditional expectation satisfies


and the conditional variance is


Moreover, it readily follows that

228 F. Venegas-Martínez


Finally, according to (10.26), the last two equations determine the mean and var-
iance of the growth rate of real assets.

10.5 Consumption dynamics

We now study the stochastic consumption demand. In virtue of equations (10.9)
and (10.26), the stochastic process for consumption can be written as


This indicates that, in the absence of contingent-claims markets, the exchange-

rate depreciation risk has an effect on wealth via the uncertainty in xt, that is,
uncertainty changes the opportunity set faced by the consumer. On the other
hand, the depreciation risk also affects the composition of portfolio shares via its
effects on w *. Thus, a policy change will be accompanied by both wealth and
substitution effects.
On the other hand, from (10.34), we can compute the probability that, in a
given time interval, certain levels of consumption occur. It is also important to
note, regarding (10.34) and (10.12), that the assumption that the agent’s time-
preference rate is equal to the world’s interest rate does not ensure a steady-state
level of consumption. However, we do have a steady-state share of wealth set
aside for consumption.
We may conclude that uncertainty is the clue to rationalize richer consump-
tion dynamics that could not be obtained from deterministic models. Finally, in
virtue of (10.34), equations (10.32) and (10.33) determine the mean and variance
of the growth rate of consumption. Figure 10.2 shows consumption as a function
of x0 and t0.
We suppose that technology is of the form yt = Akt where A > 0. That is, the
marginal product of capital is constant and equal to A. We assume that capital
does not depreciate. The condition for profit maximisation require r = A. Since
1 – w * = kt/xt, we have yt = A(1 – w *)xt. In virtue of (10.26), we obtain
yt = A(1 – w *)x0ext. Notice first that the production–consumption ratio remains
constant according to


and due to the cash-in-advance constraint, the money–consumption ratio,

mt/ct = a, remains also constant. On the other hand, from equations (10.34) and
(10.35), we have that
How risk factors affect growth in Mexico 229

Figure 10.2 Consumption as a function of x0 and t0, (x0 in 1011 pesos of 1993) (source:
own estimations based on INEGI).

and since mt = w *xt, we obtain mt = w *x0ext. Therefore,


where, during in the instant [t, t + dt], we have qt|tt ~ N[[B[(v ]*) – tt]t, G([v w]*)
dt], dft = D(w *)dQt, and dQt ~ P(hdt). Hence, if dt = T – t, in virtue of (10.36), the
expected growth rate of output, in [t, T], satisfies


Thus, yt,T depends upon the parameters determining risk factors (uncertain fiscal
and monetary policies), which shows significant qualitative differences with
230 F. Venegas-Martínez
respect to the deterministic framework. Also, from (10.31), the variance of the
growth rate of output, in [t, T], is given by


Finally, from (10.36), the expected growth rates of consumption and real cash
balances, as well as their variances, are also determined by (10.37) and (10.38),
respectively. The above equations provide a mean-variance description of the
growth rate of output, showing explicitly the underlying risk factors that affect

10.6 Simulation exercise

The following experiment is intended to replicate, via Monte Carlo methods,
the mean and its variance of the observed annual growth rate of output,
E[dyT/yt] = E[dxT], and its variance, Var[dyT/yt] = Var[dxT], by using equations
(10.33) and (10.34), in Mexico between 1930 and 2002. In Figure 10.3, we show
the observed annual growth rate of output from 1930 to 2002.
Table 10.1 presents a vector of diffusion and jump parameter values, (e, s –1 p,
h, µ), that replicate the mean and variance of the annual growth rate. In order to
choose such a vector, we tried about 800 different feasible combinations of
parameter values. For simulation purposes, we have used a standard discrete-
time version of (10.38) with an appropriate unit of time, see, for instance, Ripley
(1987). Results are based on 10,000 iterations.

Figure 10.3 Observed growth rate from 1930 to 2002 (pesos of 1993) (source: own esti-
mations based on INEGI).
How risk factors affect growth in Mexico 231
Table 10.1 Optimal consumption shares, parameters, and estimates

ω* = 0.430004
ε = 0.300000
σ = 0.009999
h = 0.100000
μ = 0.300000
F(ω*) = 0.011026
G(ω*) 0.000019
A(ω*) = –0.004539
ω* = 0.430004
ε = 0.300000
σ = 0.009999
h = 0.100000
μ = 0.300000
F(ω*) = 0.011026
G(ω*) 0.000019
A(ω*) = –0.004539
Estimated growth rate mean = 0.0470 Estimated growth rate mean = 0.0473
Estimated growth rate variance = 0.0223 Estimated growth rate variance = 0.0222

Source: own estimations based on INEGI.

10.7 Conclusions
Most of the existing models of endogenous growth ignore uncertainty, providing
elaborate justification why uncertainty does not need to be considered. We have
shown, under a free-market liberalism framework that risk factors may lead to
significant qualitative changes in the determinants of growth in contrast with the
deterministic setting. The considerations of uncertainty in the expected dynamics
of both the exchange rate and the tax policy have led to more complex transi-
tional dynamics, but results were certainly richer.
On the other hand, it is important to mention that our investigation has pro-
vided a stochastic model of endogenous growth that explains how risk factors,
such as currency, market, debt, and fiscal risk factors, affect economic growth.
This extends the literature by including stochastic determinants of growth.
Our stochastic framework, in which a Brownian motion and a Poisson process
drive the expectations of exchange-rate jumps, and a geometric Brownian
motion guides a tax rate on wealth, has provided new elements to carry out sim-
ulation experiments and empirical research. In particular, our stochastic model
was capable of explaining the average and variance of economic growth for the
Mexican case in a given period of time.

Canton, E. (2001) ‘Fiscal Policy in a Stochastic Model of Endogenous Growth,’ Eco-
nomic Modeling, vol. 18, No. 1, pp. 19–47.
Gokan, Y. (2002) ‘Alternative Government Financing and Stochastic Endogenous
Growth,’ Journal of Economic Dynamics and Control, vol. 26, No. 3, pp. 681–706.
232 F. Venegas-Martínez
Rebelo, S. (1991) ‘Long-Run Policy Analysis and Long-Run Growth,’ Journal of Polit-
ical Economy, vol. 99, No. 3, pp. 500–21.
Ripley, B.D. (1987) Stochastic Simulation, Wiley: New York.
Venegas-Martínez, F. (2001) ‘Temporary Stabilisation: A Stochastic Analysis,’ Journal
of Economic Dynamics and Control, vol. 25, No. 9, pp. 1429–49.
Venegas-Martínez, F. (2004) ‘Reforma fiscal incierta y sus efectos en las decisiones de
consumo y portafolio: impacto en el bienestar económico,’ Problemas del Desarrollo,
Revista Latinoamericana de economía, vol. 35, No. 136, pp. 137–50.
Venegas-Martínez, F. (2005a) ‘Bayesian Inference, Prior Information on Volatility, and
Option Pricing: A Maximum Entropy Approach,’ International Journal of Theoretical
and Applied Finance, vol. 8, No. 1, pp. 1–12.
Venegas-Martínez, F. (2005b) ‘Política fiscal, estabilización de precios y mercados
incompletes,’ Estudios Económicos, vol. 20, No. 1, pp. 3–18.
Venegas-Martínez, F. (2006) ‘Stochastic Temporary Stabilisation: Undiversifiable
Devaluation and Income Risks,’ Economic Modeling, vol. 23, No. 1, pp. 157–73.
Venegas-Martínez, F. (2008) Riesgos financieros y económicos. Productos derivados y
decisiones económicas bajo incertidumbre, 2nd edn, Cengage Learning: México.
11 Anti- inflationary policy and
financial fragility
A microeconomic analysis case study
of Mexico, 1990–2004
Ignacio Perrotini-Hernández,
Blanca L. Avendaño-Vargas and
Juan Alberto Vázquez-Muñoz

11.1 Introduction
The Mexican economy has undergone a number of structural changes since the
foreign debt crisis of 1982. While the most radical reforms were introduced with
the Brady Plan (1988), trade liberalisation (TL henceforth), financial liberalisa-
tion (FL), monetary policy and exports became the driving forces of economic
The Brady Plan helped ease the debt burden and foreign saving and invest-
ment flows increased with the aid of financial deregulation and price stabilisation
policies based on a nominal exchange rate anchor through 1988–1994. Yet, the
stabilisation strategy also produced interest rate and asset price hikes, exchange
rate appreciation, deindustrialisation and slow output growth, deteriorating
saving and investment coefficients as portfolio short-term investment crowded
out productive investment and, last but not least, current account disequilibrium.
Our task in this chapter is to assess whether the monetary policy framework
of low inflation and the investment dynamics derived thereby have triggered
financial fragility in the Mexican economy.

11.2 The financial fragility hypothesis

The backbone of Keynes’s theory, according to Minsky (1975: 94), is his ‘theory
of investment and why it is so prone to fluctuate’. Minsky maintains that finan-
cial factors determine the pace of investment, which is the key to economic
Minsky’s analysis of the role of financial factors in the process of capital
investment is based on a twofold approach, namely the existence of two prices
and of a complex financial structure. The former extends Keynes’s model with a
view to construct a theory of systematic financial fragility where income flows
and the net supply of real capital assets determine the relationship between the
234 I. Perrotini-Hernández et al.
real and financial sectors of the economy. Capitalised income flows determine
investment positions and monetary policy and interest rates play a relevant part
in the process of financial instability. The supply price of investment is:


Given the term structure of interest rates, rl, . . . rn, the supply price of capital
investment will be:


Where Ki and Yi denote capital and income flows in period i respectively.

Assuming Q1 = Q2 = . . . = Qn, the capitalisation rate can be expressed as:


where expected returns (Qn) are determined by the future price of produced
goods, the volume of sales, input costs and firms’ financial costs.
Minsky uses Keynes’s (1936) concept of capitalisation of investment returns
to determine the demand price of capital assets (PKi), which depends on the
parameter of capitalisation of quasi-rents (fi)


The rate of capitalisation crucially depends on the monetary and financial con-
ditions of the economy, that is, on the capitalisation rate of the loan borrowed to
finance investment. Therefore, the price of liabilities, determined in the debt
market, bear an influence on that of capital assets. Thus, the rate of capitalisation
of Qn also depends on the level of uncertainty prevailing in financial markets.
Then, ‘the capitalisation rate of capital assets is some ratio, 0 < µ < 1, of the cap-
italisation rate of money loans’ (Minsky, 1975: 102). Let µ denote financial
uncertainty and fL the capitalisation rate of the bank loan or the financial liability
of firms, then:




The rate of capitalisation of quasi-rents depends on the performance of the finan-

cial market, whereas the capitalisation rate of the loan used to finance investment
hinges on the central bank’s monetary policy (the money supply). Since firms
Microeconomic analysis of Mexico 235
get loans in order to finance capital asset positions, the rate of interest becomes a
key variable in the process of capitalisation of Qn. In sum, there is a negative
relationship between the monetary rate of interest and f. Hence the effect of
monetary policy on investment through the effect of the interest rate on the
parameter of capitalisation:


Equation (11.5) is an unstable function – this is a ‘fundamental fact’ (Minsky,

1975) – because the monetary and financial conditions of the economy may
affect Pki through the effect of the interest rate on expected quasi-rents. The
central bank can affect the balance sheet of those firms that have been undertak-
ing debt in the credit markets with the aim of financing fixed capital asset posi-
tions. The relationship between r and fi depends on the market assessment of the
probability distribution of actually getting a certain and assured stream of
income flows, Qi, vis-à-vis the probability of getting it at a fluctuating market
rate of return, Qn, plagued by uncertainty. Investment, then, is determined by the
discrepancy between the demand price and the supply price, while the adjust-
ment process induces instability.
As for the financial structure approach, Minsky (1982) focuses on the methods
used by enterprises to finance investment. Potential sources are: (1) cash and
financial assets, (2) internal finance (after-tax and dividends profits) and (3) exter-
nal finance (loans and equity issuance). Investors consider these different methods
of finance and must include the financial cost of capital,1 apart from wage and
input costs, in the supply price of produced goods. Investors are bound to forecast
their future income streams and the specific conditions which will prevail in the
financial markets, as fixed capital investment is a long-term action, while the
means available to finance investment positions are short term in nature.
Minsky’s financial fragility hypothesis states that there exists an inherent tend-
ency for the economy to become financially fragile as a result of the negative influ-
ence of financial variables on firms’ capital structure throughout the business
cycle. He then goes on to establish a taxonomy of financial structures according to
firms’ balance sheets in terms of flows of income and liabilities and debt payments.
Thus, income sources, revenue flows from operations plus new debt (D), equal
expenditure streams – investment (I ) plus payment commitments on debt (V ):


If R < 0, the firm will face a loss; if D < 0, the firm will be repaying its debt, and
V < 0 makes the firm a net creditor. The net wealth of a firm (W) is measured as:


where A denotes total assets and B is the value of its outstanding debt. W
increases when investment is increased and/or debt gets reduced. A firm becomes
236 I. Perrotini-Hernández et al.
insolvent and goes bankrupt when W ≤ 0 and its creditors are unable to rescue the
capital involved in the unit.
A firm is said to hold hedge finance when its cash income flows from opera-
tion are expected to be larger than its payment commitments on debts. A hedge
firm can be troubled whenever R gets reduced during downswing periods and/
or V increases in a credit crisis. A firm can be characterised as speculative if
R ≥ V and R < V + I, which implies D ≥ 0, though D < I. Typically, speculative
firms run financial deficits during expansion periods when they engage in
investment opportunities that exceed their internal financial capacity. In this
case W increases, and the rate of return on investment determines whether a
speculative unit survives or goes out of business. In addition, Minsky defines
as Ponzi finance a firm that meets its cash payment commitments on debt by
augmenting the amount of debt outstanding. In this situation, R < V and D > I.
While the credit profile of a Ponzi unit depends on its ability to persuade credi-
tors that its income streams will increase in the near future, the increment in
the outstanding debt will make it harder for the former to find voluntary
lenders. By and large, business enterprises depend on expectations about future
interest rates and financial market conditions. The latter can force hedge firms
alternatively into speculative and Ponzi financing. Unless effective demand
(sales) rises and/or the interest rate falls, the probability of a firm falling into
Ponzi financing will arise.2 The financial structure of firms changes along the
business cycle, usually from hedge to Ponzi finance; hence macroeconomic
Foley (2003), following Minsky, suggests a typology in terms of the growth
rate of firms’ assets (A) and liabilities (B) and the rate of returns on assets: g = I/A
is the growth rate of assets, r* = R/A is the rate of return and i* = V/B is the inter-
est payments to debt ratio. A particular combination of i*, r* and g sets the stage
for financial fragility of firms (see Table 11.1).
The modus operandi of financial fragility can be summarised as follows (cf.
Minsky, 1982, 1986; Wolfson, 1989):

1 Expected returns (PE) depend on expected income flows of sales, which in

turn depend on effective demand. The actual flow of returns (PR) validate
the amount of outstanding debt securities. PE induces fluctuations in the rate
of capital investment.

Table 11.1 Typology of the growth rate of firms

Typology Condition

Hedge r > g > i or r > i > g

Speculative g>r>i
Ponzi i>r

Source: authors’ own elaboration.

Microeconomic analysis of Mexico 237
2 Expected returns are stable during the upswing. Since the safety margins
increase with economic expansion, refinancing of debt contracts in second-
ary markets works out properly. Therefore, business optimism produces
boom investments that tend to exceed firms’ own internal financial means;
keen investors engage into debt financing of economic activity. As the
economy gains momentum, the most aggressive firms accumulate debt;
increase their leverage rate and the potential for balance sheet destabilisa-
tion as the business cycle reaches a turning point.
3 Once the business cycle turns to a downswing PE will diminish, PR will
tend to decline with the reduction of effective demand and sales, and finan-
cial fragility sets in. Thus, financial instability results endogenously from
economic success.
4 The theory of financial instability explains how the accumulation of debt
impairs investment: while financial leverage expands aggregate demand at
the outset of the business cycle, it also alters the capital structure of those
corporations that rely heavily on loans through the effect of the interest rate
on income flows and the balance sheet of indebted firms. Finally, the influ-
ence of debt financing and higher interest payments on the level of eco-
nomic activity brings about effective demand constraints, underutilisation of
productive capacity and unemployment of the labour force.

According to Minsky, monetary policy rules can also contribute to financial

instability. Clearly, in an open economy setting where local firms borrow foreign
savings to finance economic activity, an anti-inflationary monetary policy may
add to financial fragility through interest rate and exchange rate volatility. A
fixed nominal exchange rate may anchor inflation and the net wealth of firms.
However, it also invites speculative attacks against the domestic currency and
increases the liability burden on highly indebted firms because of interest rate
hikes used to cope temporarily with speculative runs against the domestic cur-
rency. At the end of the day, as recent experience has shown, this anti-
inflationary strategy collapses and the central bank (CB), in need of an
independent monetary policy, must choose an alternative framework.
Many developing countries have adopted an inflation targeting monetary
policy framework (IT) where the interest rate has become the policy instrument
to achieve both price stability and exchange rate stability. This is the so-called
Taylor rule (Taylor, 1993).3 From a Minskyan point of view, the fact that devel-
oping economies suffer from a high pass-through effect of exchange rate fluctua-
tions on to the price level, imply that the IT model does add to financial
instability, because the CB must increase the interest rate in order to prevent
(actually, just to postpone) exchange rate depreciations and attain the expected
inflation rate. Thus, the sequence of booms and recessions along the business
cycle results from specific monetary policy strategies (Minsky, 1986). Hence
anti-inflationary monetary policies may trigger financial instability.4 The IT
model is no exception, given the inter-play between the interest rate, the
exchange rate and the rate of inflation.
238 I. Perrotini-Hernández et al.
11.3 Empirical evidence, a microeconomic analysis
Our empirical analysis is based on data for 47 non-financial firms quoted in the
Mexican stock exchange market during 1990–2004. We consider two sub-
periods, the first from 1990 to 1994 when there existed a fixed nominal exchange
rate anchor, and the second from 1995 to 2004 when an IT strategy was intro-
duced.5 Mexico shows a high pass-through effect from exchange rate (e) vari-
ations to the rate of inflation (ṗ). The IT strategy brought with it a stronger
correlation between the exchange rate and the interest rate (i) on CETES (28
The analysis of the effect of the Banco de Mexico’s anti-inflationary policy
on firms’ financial profile is focused on the real rate of interest (r) and the real
exchange rate (q), owing to the former is the relevant adjustment variable and
the latter impacts the real value of dollar-denominated liabilities.
Figure 11.1 shows the behaviour of g, r* and i* for our sample of firms as a
whole. Variables were built as follows: g = I/A, where I denotes the quarterly
increment of total assets and A is the value of total assets; r* = R/A, where R is
net income from operation plus the amount of capital invested by shareholders;
and i* = V/B, where V is the quarterly integral cost of borrowing (interest pay-
ments less interest earnings) and B is total liabilities. All variables have been
adjusted by the 2002 consumer price index.
As shown above, the firms in the sample as a whole can be characterised as
speculative during 1990–1994: g and r* performed positively, by and large
g > r*, which can be explained by the transition of the Mexican economy to a
more stable environment. The maximum value of g (17.7 per cent) was attained

Figure 11.1 Growth (g), returns (r) and interest rate (i): aggregate levels,
1990:02–2004:04 (source: authors’ own elaboration based on data from
Banco de México and Bolsa Mexicana de Valores).
Microeconomic analysis of Mexico 239
Table 11.2 Average values for the sample of firms as a whole

Variable 1990:2–2004:4 (%) 1990:2–1994:4 (%) 1995:1–2004:4 (%)

r* 0.78 2.31 0.06

r*1 0.42 1.55 –0.11
r*2 0.36 0.76 0.17
g 1.30 3.80 0.11
i* 0.54 0.79 0.42
b 1.84 4.51 0.57

Source: authors’ own calculations on the basis of data from Banco de México and the Mexican stock
exchange market.

in the last quarter of 1994. Table 11.2 shows that the returns from shareholders’
contributions to asset accumulation (r*2) is responsible for one-third of the total
rate of returns, whereas returns from operations of the firm (r*1) is responsible
for 42 per cent, signalling optimistic investors. After the financial crisis of
1994–1995, the average growth rate of firms (g) shrank from 3.80 per cent
during 1990–1994 to 0.11 per cent during 1995–2004. The rate of return, in turn,
averaged 2.31 per cent throughout the nominal exchange rate anchor period and
0.06 per cent during the IT era. The fall in r* signals the loss of momentum in
the pace of capital accumulation. On the other hand, i* remained low and stable
through 1990–1993, but increased sharply after the speculative attack against the
Mexican peso in 1994 and has declined throughout the years of the flexible
exchange rate regime (see Figure 11.1 and Table 11.2).
The rate of investment and the growth rate of private debt (b) rose during the
fixed exchange rate regime, partly because financial liberalisation enhanced the
supply of loanable funds (see Figure 11.2 and Table 11.2). As Minsky (1975,
1982) points out, an increasing rate of return spurred by financial leverage, does
generate an optimistic environment that prompts booming investment and higher
indebtedness. The financial crisis of 1994–1995 was followed by a long episode
of credit rationing which was not offset by financial sources other than the
formal banking sector: the quarterly growth rate of Mexican firms’ debt (b) aver-
aged 4.51 per cent from 1990 to 1994 and 0.57 per cent from 1995 to 2004.
The aforementioned financial crisis also produced a sharp increase in i* and a
devaluation of the exchange rate, which led to a higher b. The overall effect on
our sample of firms was a lower net wealth (W = g – b) and a faster transition
from speculative to Ponzi finance, since the higher rate of inflation that erupted
from the financial crisis eroded the real value both of total assets and total liabili-
ties (see Figure 11.3). In fact, it can be argued that attainment of price stability in
Mexico has meant a reduction of W from –0.71 per cent throughout 1990–1994
to –0.46 in 1995–2004, on the one hand, and a negative accumulation of assets
in aggregate terms, on the other.
Figures 11.4 and 11.5 show the relative composition of both firms and assets
according to a classification of hedge (h), speculative (s) and Ponzi (p) finance.
Figure 11.2 Growth (g) and firms’ indebtness (d): aggregate levels, 1990:02–2004:04
(source: authors’ own elaboration based on data from Banco de México and
Bolsa Mexicana de Valores).

Figure 11.3 Net wealth (w), growth (g) and debt (d): aggregate levels, 1990:02–2004:04
(source: authors’ own elaboration based on data from Banco de México and
Bolsa Mexicana de Valores).
Microeconomic analysis of Mexico 241
And Table 11.3 presents average values of data shown in Figures 11.4 and 11.5
for the whole period and the two sub-periods considered in the analysis. The per-
centage composition of firms and assets does not vary significantly, except for
speculative and Ponzi units during the fixed exchange rate sub-period. Interest-
ingly, the number of hedge units represented the smallest proportion from 1990
to 1994. The share of hedge units increased after the financial crisis of 1995
because several Ponzi firms were unable to recover financially and left the MSM.
This explains why the share of Ponzi units declined starting in 2003, even though
their absolute number increased during the IT sub-period.
We now turn to the analysis of the influence of BM’s anti-inflationary policy
on financial fragility. As mentioned, our analysis is focused on r and q, their
relationship being given as:


where D is a quarterly differential operator, q is the quarterly real exchange rate

depreciation, µt is white noise and bi are estimate parameters.6 A high correlation
between e and q can be confirmed throughout the IT sub-period. Thus, given this
high correlation between the nominal depreciation of e and the rate of inflation
(which implies a real depreciation), the central bank reacts, with the aim of diminish-
ing aggregate demand through higher r, via a further increase in i whenever e rises.
We estimate the effect of changes in r and q on b using the following equation:


Figure 11.4 Composition of firms according to financial structure (%), 1990:02–2004:04

(source: authors’ own elaboration based on data from Banco de México and
Bolsa Mexicana de Valores).
The size of the sample varies in each period as some firms either left the Mexican stock market or
did not provide information of their operations. h: hedge, s: speculative, p: Ponzi.
242 I. Perrotini-Hernández et al.

Figure 11.5 Composition of firms’ assets according to financial structure (%),

1990:02–2004:04 (source: authors’ own elaboration based on data from
Banco de México and Bolsa Mexicana de Valores).
The size of the sample varies in each period as some firms either left the Mexican stock market or
did not provide information of their operations. h: hedge, s: speculative, p: Ponzi.

where j refers to the j firms, Wi are the parameter estimates, vtj denote white noise
and Dr̃ represents, on the one hand, quarterly changes in r from 1990:2 to 1994:4
and, on the other, the residual from estimation of equation (11.10) from 1995:1 to
2004:4, in other words, variations in r which are unexplained by depreciations of q.
Estimation of equation (11.11) was conducted using the fixed coefficients approach
and unbalanced panel data for 47 non-financial firms. Our results are as expected:
variations in the interest rate and exchange rate depreciations increase outstanding
debt of firms (see Table 11.4).7 It is worth noting that the debt elasticity with respect
to firms’ growth (gtj) is greater than one (1.67), signalling an endogenous risk of
financial fragility. Moreover, the high and negative (–1.97) debt elasticity with
respect to firms’ rate of returns (r*tj) highlights the importance of internal finance
for capital investment. Finally, bt–1
is significant but close to zero.

Table 11.3 Average percentage composition of the companies according to their financial

Period 1990:2–2004:4 1990:2–1994:4 1995:1–2004:4

Financial structure h s P h s p h s p
By number of firms 27 30.1 42.8 23.4 39.3 37.3 28.8 25.8 45.4
By assets 26.1 33.5 40.5 22.4 45.3 32.2 27.8 27.8 44.4

Source: authors’ own calculations on the basis of data from Banco de México and the Mexican stock
exchange market.
Table 11.4 Dependent variable: b ti

Independent 1990:2–2004:4 1990:2–1994:4 1995:1–2004:4

Coefficient t-Statistics Pro.b Coefficient t-Statistics Prob. Coefficient t-Statistics Prob.

θ·t 0.25 2.34 0.02 –0.02 –0.13 0.9 0.02 0.17 0.87
r*ti 0.16 5.81 0 0.19 2.41 0.02 0.13 4.68 0
g ti –1.97 –39.73 0 –2.69 –37.65 0 –1.75 –31.17 0
bt–1 1.67 61.82 0 2.93 56.04 0 1.46 50.14 0
–0.04 –3.38 0 –0.04 –2.67 0.01 –0.04 –3.12 0
R2 adjusted 0.64 R2 adjusted 0.82 R2 adjusted 0.65
DW 2.17 DW 2.14 DW 2.16
F-Statistic 1,118.12 F-Statistic 874.6 F-Statistic 803.87

Source: authors’ own calculations on the basis of data from Banco de México and the Mexican stock exchange market.
For the sake of brevity we omit the fixed values of the coefficients.
244 I. Perrotini-Hernández et al.
The relevance of the effect of real exchange rate depreciations on firms’
indebtedness is also confirmed when we estimate equation (11.11) for sub-
periods 1990–1994 and 1995–2004, although in these cases interest rate varia-
tions became insignificant perhaps because FL and reprivatisation of the banking
sector flooded the economy with booming credit (in the first sub-period) and the
impact of interest rate variations on b was somewhat captured by depreciations
of q (during the IT sub-period). Both g tj and r*tj were higher throughout the first
sub-period than in the second one and greater than one in both cases (see Table
11.4), confirming a high propensity of Mexican non-financial firms to financial
fragility when the BM pursues anti-inflationary monetary policies, either in the
form of a nominal exchange rate anchor or an IT.
Following the same estimation method as in equation (11.11), we then estim-
ate the effect of monetary policy variables on the increment of firms’ net debt
(–w), where Yi are the parameter estimates and u tj is white noise:


Results from equation (11.12) are as expected (see Table 11.5): r and q bear a
positive correlation effect on firms’ accumulation of net debt, while r* bears a
negative relationship with net debt accumulation. However, greater effects of
variations in r and depreciations of q during 1990–1994, as opposed to those of
the IT sub-period, can be observed, possibly because of better expectations and
the acceleration of investment and debt accumulation in those years. The elastic-
ity of –w with respect to r*, both in the period as a whole and across sub-periods,
remains close to one. It is also interesting to note that, throughout the IT sub-
period, variations in the interest rate which are explained by factors other than
exchange rate depreciations are not significant.
The following ordered model with panel data estimates the probability of a
firm being hedge, speculative or Ponzi finance:


where v tj denotes the rate of debt growth of firm j which is not explained by the
independent variables in equation (11.11), and F* is a latent variable used to
order the dependent variable F as follows:

0 if

1 if

2 if

F tj represents the observed financial structure of firm j in period t: the firm is

said to be hedge if F tj = 0, speculative if F tj = 1 and Ponzi if F tj = 2. Furthermore,
we assume a ‘normal’ probability function for the determination of a specific
Table 11.5 Dependent variable: w ti

Independent 1990:2–2004:4 1990:2–1994:4 1995:1–2004:4

Coefficient t-statistics Prob. Coefficient t-statistics Prob. Coefficient t-statistics Prob.

Dr̃t 0.36 2.93 0 0.62 2.39 0.02 0.06 0.47 0.64

θ·t 0.14 4.48 0 0.5 3.73 0 0.11 3.75 0
r*ti –0.95 –30.74 0 –0.95 –10.27 0 –0.98 –31.78 0
–wt–1 –0.07 –4.24 0 –0.09 –2.45 0.01 –0.08 –3.98 0
R2 adjusted 0.29 R2 adjusted 0.13 R2 adjusted 0.38
DW 2.09 DW 2.03 DW 2.09
F-Statistic 342.72 F-Statistic 55.13 F-Statistic 358.95

Source: authors’ own calculations on the basis of data from Banco de México and the Mexican stock exchange market.
For the sake of brevity we omit the fixed values of the coefficient
Table 11.6 Dependent variable: F ti

(Unbalanced 1990:2–2004:4 1990:2–1994:4 1995:1–2004:4

No. observations: 2452 No. observations: 814 No. observations: 1708

Independent Coefficient Z-Statistic Prob. Coefficient Z-Statistic Prob. Coefficient Z-Statistic Prob.

θ·t 2.28 8.44 0.000 2.26 2.79 0.01 2.19 7.65 0.000
r*ti –14.49 –21.44 0.000 –12.1 –12.25 0.000 –14.26 –17.14 0.000
g ti 6.22 19.11 0.000 5.2 8.53 0.000 6.26 15.79 0.000
Limit values Limit values Limit values
of γi of γi of γi
Limit: γ0 –0.76 –25.09 0.000 –0.91 –16.04 0.000 –0.68 –19.32 0.000
Limit: γ1 0.18 6.52 0.000 0.32 6.43 0.000 0.11 3.45 0.000
Statistic LR 620.97 Statistic LR 186.21 Statistic LR 409.79
LR index 0.12 LR index 0.11 LR index 0.11

Source: authors’ own calculations on the basis of data from Banco de México and the Mexican stock exchange market.
Microeconomic analysis of Mexico 247
financial structure of particular firms, since variations in the firm’s debt ratio
may augment the probability of that firm falling into Ponzi finance, but the
greater the variations in the debt ratio the smaller the increment in such
Table 11.6 summarises the results from estimating equation (11.13) with the
maximum likelihood method.8 The probability of a firm falling into Ponzi
(hedge) financing increases (diminishes) with the expansion of the firm and real
exchange rate depreciations, while, conversely, such probability decreases (aug-
ments) with increments (reductions) in the firm’s rate of return. By and large,
anti-inflationary monetary policies add to Ponzi financing of firms. On the other
hand, it is difficult to determine, a priori, the influence of the relevant variables
on the probability of a firm falling into speculative finance, because, according
to empirical evidence, that effect tends to zero.
Figures 11.6 and 11.7 show the probability with which firms may be classi-
fied as hedge, speculative or Ponzi finance.
As shown in Figures 11.6 and 11.7 and Table 11.7, the average probability of
speculative finance was greatest when BM’s monetary policy focused on a fixed
nominal exchange rate anchor; the probability of Ponzi finance was second and
that of hedge third. The adoption of the IT model changed the probability distri-
bution of the various financial structures inasmuch as Ponzi finance became par-
amount among Mexican firms during 1995–2004.
All in all, it appears that according to empirical evidence for a sample set of
Mexican non-financial firms, Minsky’s financial instability hypothesis can help

Figure 11.6 Probability of hedge (h), speculative (s) and Ponzi (p) finance among firms,
1990:02–1994:04 (source: authors’ own elaboration based on data from
Banco de México and Bolsa Mexicana de Valores).
Data are ordered firm-wise.
248 I. Perrotini-Hernández et al.

Figure 11.7 Probability of hedge (h), speculative (s) and Ponzi (p) finance among firms,
1995:01–2004:04 (source: authors’ own elaboration based on data from
Banco de México and Bolsa Mexicana de Valores).
Data are ordered firm-wise.

Table 11.7 Average of the estímate probability of financial structure of firms

Period 1990:2–2004:4 1990:2–1994:4 1995:1–2004:4

Financial structure h S p h S p h s p
By number of firms 26.4 30.8 42.8 22.5 40.7 36.8 28.1 26.5 45.5

Source: authors’ own calculations on the basis of data from Banco de México and the Mexican stock
exchange market.

us understand the evolution of capital structures from hedge to speculative to

Ponzi financing when the central bank pursues an anti-inflationary monetary
policy. Therefore, inflation targeting matters for financial fragility.

11.4 Conclusion
The present chapter aimed at arguing that the structure of capital matters for eco-
nomic stability, the monetary policy framework is also relevant for financial
stability, in other words, money is not neutral and, last but not least, given a high
pass-through coefficient, the dynamics between the real exchange rate, the inter-
est rate and the rate of inflation, pursuing an inflation targeting strategy may
encourage Ponzi financing.
Using data for 47 Mexican non-financial firms quoted in the Mexican stock
exchange market from 1990 to 2004, we have assessed Minsky’s financial
Microeconomic analysis of Mexico 249
instability hypothesis. We conclude that the latter provides insights into across-
firm characteristics of financial fragility, in particular when the central bank nar-
rowly (and uniquely) pursues an inflation target with no regards whatsoever for
other economic policy targets, such as growth and employment.

1 Typically, finance for production purpose is short term and, habitually, take the form of
bank loans.
2 Fisher (1933) had described this situation, which triggers money flows due to asset
sales, in his theory of debt-deflation: firms cancel debts, thus causing inflation of the
value of money as a result of an increasing demand for money and a limited supply. As
Fisher argued, the more debt is cancelled, the more debtors owe (Minsky, 1977).
3 Ball (1998) has extended Taylor’s model for the open economy case where the CB
follows a monetary conditions index, given by a weighted average of the interest rate
and the rate of appreciation of the exchange rate, to attain the targeted rate of inflation.
4 In this scenario, both speculative and Ponzi units increase their demand for loans with
a view to refinance debt commitments, and demand is interest rate inelastic.
5 Banco de Mexico (BM) adopted an IT in 2001, though major features of such monetary
policy framework had been introduced since 1995.
6 The estimation of equation (11.10) was not significant for the fixed exchange rate sub-
period and significant for the IT sub-period. While the results from regressions are not
reported in this text, they are available from the authors upon request.
7 The estimation results from equations (11.11) and (11.12) were corroborated by the
GLS method.
8 The final estimation excludes parameters for variables r̃t and vt because our initial esti-
mation proved them statistically insignificant.

Ball, L., 1998, ‘Policy Rules for Open Economies’, National Bureau of Economic
Research, Working Paper no. w6760.
Bolsa Mexicana de Valores, Indicadores Bursátiles, various years.
Fisher, I., 1933, ‘The Debt–Deflation Theory of Great Depressions’, Econometrica, 1, pp.
Foley, D., 2003, ‘Financial Fragility in Developing Economies’, in Dutt, A. and J. Ros
(eds), Development Economies and Structuralist Macroeconomics, Essays in Honor of
Lance Taylor, Cheltenham: Edward Elgar.
Keynes, J.M., (1936), Teoría General de la Ocupación, el Interés y el Dinero, Fondo de
Cultura Económica, México.
Minsky, H.P., 1975, John Maynard Keynes, New York: Columbia University Press.
Minsky, H.P., 1977, ‘A Theory of Systematic Fragility’, in E.I. Altman and W. Sametz
(eds), Financial Crises: Institutions in a Fragile Environment, New York: Wiley.
Minsky, H.P., 1982, Can ‘It’ Happen Again? Armonk, NY: M.E. Sharpe.
Minsky, H.P., 1986, Stabilizing an Unstable Economy, New Haven and London: Yale
University Press.
Taylor, J., 1993, ‘Discretion versus Policy Rules in Practice’, Carnegie – Rochester Con-
ference Series on Public Policy, no. 39, pp. 195–214.
Wolfson, Martin, 1989, Financial Crises: Understanding the Postwar U.S. Experience,
Armonk, NY: M.E. Sharpe.
12 Technological innovation and
sectoral productivity in the
Mexican economy
Regional evidence
José Carlos Trejo-García, Humberto Ríos-Bolívar
and Ana Lilia Valderrama-Santibáñez

12.1 Introduction
There are a large number of studies in economics related to the relationship
between technological progress, innovation and economic growth. One of the pio-
neers was Robert Solow, who published his first work on this subject in 1957.
However, to date it should be recognized the difficulty of measuring the proper
role of technology and innovation in growth, especially when it comes to empiri-
cally demonstrate this fact. This has led economists to focus on the analysis of
expenditure on research and development (R&D) for innovation as a close variable
to technology and technological innovation. According to the evidence shown by
Solow, such spending contributes to technological improvements, so that invest-
ment in R&D is considered to have a significant impact on productivity growth.
Thus, the empirical analysis of the relationship between R&D for innovation
and productivity can be done through the estimate of a production function,
where technological capital and innovation are included as explicative variables.
In a production model which includes technological capital, Griliches (1979)
argues that production function includes, in addition to the usual factors of pro-
duction, another factor that may be named technological capital, technological
innovation or R&D capital.
This chapter studies the relationship between the labour productivity growth
and investment in this kind of capital for the manufacturing sector, trade and
services. For this purpose INEGI’s statistical data from the economic census of
1994 and 2004 is used. The use of cross section data for 2,438 municipalities
allows having a large number of data samples in addition to statistical informa-
tion disaggregated by economic sectors. Until now, papers in Mexico about the
relationship between productivity and R&D spending have primarily estimated
production functions. These functions determine production elasticity related to
production factors: capital and labour, leaving aside technology and innovation
factors. It should be mentioned that one of the main problems in this kind of
analysis is the need of information about the stock of research and development
capital, which is generally not available.
Innovation and productivity in Mexico 251
Therefore it was necessary to estimate the stock of capital in R&D. The per-
petual inventory method is frequently applied. That is, the capital stock for each
period is calculated from the capital stock (minus depreciation) in period t–1,
plus the capital investment in the period t. However, this difficulty can be
avoided if we estimate a transformation of the production function that requires
knowing only the R&D expenditure in each period. This is what we do in this
This requires a certain degree of novelty compared to previous research con-
ducted in Mexico, because, instead of estimating production functions where the
capital in R&D is another factor, productivity growth is directly related with the
intensity of R&D spending. To our knowledge, there is no research in this line
for Mexican economy.
The present chapter is structured as follows. The next section reviews the
literature on the subject. The third presents the theoretical model. The fourth
describes the data, variables and empirical methodology applied in the econo-
metric analysis. An econometric model is estimated in Section 5. The final
section summarizes the paper and highlights the most important conclusions.

12.2 Background
While the relationship between productivity and R&D for innovation in Mexico
has been discussed by Unger (1996) and Jasso (1998), among others, the
common feature of these papers is that they are based on the specification of a
production function as well as on the estimation of R&D–capital elasticity in the
industry. The present chapter applies a transformation of a production function
to avoid the use of technological capital stocks as an independent variable.2
This kind of analysis has been done for a wide number of countries. Some of
the most important studies are Griliches and Mairesse (1983), who analyse the
influence of R&D expenditure on productivity from individual data for the
United States (US) and France between 1973 and 1978; Clark and Griliches
(1984) study the relation between productivity growth and R&D expenditure
during the period 1970–1980, their statistical sample includes data for 924 US
manufacturing companies; Lichtenberg and Siegel (1991) use panel data to study
the relationship between R&D and productivity growth in the US industry in the
period 1972–1985. Recently, Bessen (2000) used a sample of 471 US companies
between 1983 and 1989 to get results on the relationship between productivity
and R&D expenditure.
While the primary goal of previous researches was to measure the costs for
firms for adopting the technology derived from R&D, there are other studies. For
example, Odagiri and Iwata (1986) estimated the impact of R&D expenditures
on the rate of productivity growth in Japan using data from individual companies
in two different periods: from 1966 to 1973 and from 1974 to 1982. Fecher
(1990) analyses the influence of R&D spending on productivity, from individual
data of companies from Belgium, between 1981 and 1983. Hall and Mairesse
(1995) updated their own results of previous research about the link between
252 J.C. Trejo-García et al.
productivity and R&D in the French economy. The study runs from 1980 to
1987 and involves information from 351 companies. Wakelin (2001) examines
the relationship between productivity growth and intensity of expenditure on
R&D in the UK using information provided by 170 British companies during the
years 1988–1996.
Finally, for Italy, Parisi et al. (2002) show empirical evidence of the effect of
innovation on productivity in the production process, on the one hand, and the
impact of innovations in the product, on the other. They also studied the effect of
investment in R&D on the probability of making innovations. The information
comes from 941 Italian companies and refers to 1992–1997. Estimates of the
rate of return provided for these papers are mixed. Overall, the results depend on
the way of measuring different variables included in the estimates and data
sources used in the study.

12.3 Theoretical aspects

The theoretical understanding of the relationship between productivity and R&D
expenditure for innovation is based on the model of Griliches (1979). This
model, in turn, is based on the accumulation of capital in R&D for technological
innovation as an additional factor of production, along with the usual factors of
production: physical capital and labour. In this research, the starting point for
building the model is a Cobb–Douglas production function with three production
factors: technological innovation (H ), stock of physical capital (k) and labour
(L). The aggregated production function is written as:


where subscripts i and t denote the firm and the period, respectively. Q is an
output measure (usually sales or value added of economic sectors), L represents
labour (usually the number of employees), H and K measure the stocks of
technological innovation capital and physical capital, respectively, A is a con-
stant, α, β and g are the corresponding elasticities of output related to R&D,
physical capital and labour, respectively, λ is the rate of unincorporated technical
change (exogenous changes in production technology over time that cause varia-
tions in the growth rate of productivity, common to all economic sectors), μ rep-
resents a non-observable specific effect of each economic sector, constant over
time, and ε is a random error term.
This is the Cobb–Douglas production function approach whose main charac-
teristic is factors divisibility; this allows dividing the effects of the R&D factor
and the possibility of estimating a linear model in levels, in differences or under
some kind of transformation, for instance logarithmic. The following equations,
written in logarithmic terms, result thereby:

Innovation and productivity in Mexico 253
The function in first differences:


where lower-case letters denote logarithms of each variables and Δ represents

the first difference of the specified variable.
The specific effects of the economic sector, μi, are eliminated when taking
first differences. The main disadvantage of such a specification is the need for an
appropriate measure of R&D capital stock. To avoid this problem, we can make
some changes in the Cobb–Douglas production function. Moreover, under the
assumption that the production function presents constant returns to scale related
to the standard inputs, a + b + g = 1, subtracting labour logarithm from expression
(12.2) in both sides, obtains:


where g = 1 – a – b
Substituting g, we have:


Rearranging terms:


Taking first differences in this expression obtains:


where ui = Deit and D(q – l) is the growth rate of labour productivity, D(h – l) the
growth rate of technological capital related to labour and D(k – l) is the growth
rate of capital labour. The parameter g is the elasticity of output with respect to
Moreover, the growth rate of R&D capital is calculated using the next


From elasticity g and expression (12.8) it is obtained:


where Y = (∂Q/∂H)it is the marginal productivity of capital in R&D, Rit is invest-

ment or expenditure in R&D of sector i in the period t, and R/Q is the intensity
of R&D spending or the level of technological effort.
254 J.C. Trejo-García et al.
Using equation (12.9) and a = 1 – b – g, in order to get aD(h – l)it as the growth
rate of technological capital related to labour, we rewrite it as follows:


Substituting equation (12.10) in (12.7) it is obtained:



Simplifying terms obtains:

Rearranging terms:


Having g – 1 = –a – b, expression (12.13) becomes:


Expression (12.14) can be used to estimate the R&D spending, instead of the
stock of R&D capital. Moreover, under the assumption that market operates
under competitive conditions, Y can be interpreted as the rate of return of R&D
expenditure. Equation (12.14) allows estimating the value of the rate of return of
technological capital Y. This is one of the important points of this research.
The estimate of the proposed equation is done mainly from data provided by the
Mexican economic census, elaborated by INEGI. These surveys include a set of
units listed for each municipality, for three economic sectors: manufacturing, trade
and services. Econometric estimations that use cross section data for all regions are
conducted in this chapter. Our approach is similar to that adopted by Mankiw et al.
(1992) which examines the determinants of growth in terms of R&D.
Statistical information was taken from the Municipal Information System
Database (SIMBAD), particularly:

• 1994 Economic Census,

• 2004 Economic Census.

The data fall into three sectors; manufacturing, commerce and services. Vari-
ables used are shown in Table 12.1.
Innovation and productivity in Mexico 255
Table 12.1 Variables

Name Abbreviation

Gross domestic product Q

Population N
Fixed assets K
R&D expenditure H
Technological effort R/Q
Total remunerations RT
Average employed persons L
Computers D1
Line phones D2
Internet access availability D3

Source: author’s own elaboration.

12.4 Econometric model

The chance of an economy to experience economic growth as well as improve-
ments in technology and innovation is one of the major concerns of the present
time. One premise of the theory of technological innovation is that innovation is
one of the basic elements for economic growth and for improving a nation’s
technology. In this way, it can be argued that spending on R&D could be the
indispensable factor for economies towards development.
Several empirical models have been used to demonstrate the interaction
between growth, technological change and technological innovation and more
specifically between output growth, technological change and R&D.
One of the studies that focused on such interaction was the seminal paper by
Solow (1957) on the US economy in the period 1909–1949. Solow found that
over 80 per cent of output growth per hour of labour recorded in that period was
due to technical progress. That is, of an annual average growth rate of 2.9 per
cent of real GDP, approximately 0.32 per cent of the increase was due to accu-
mulation of capital, 1.09 per cent to increases in the quantity of labour and the
remaining 1.49 per cent was attributed to technical progress.
In a subsequent study along this line of research, Denison (1974) used data
for the period 1929–1969 for the US economy and the results confirmed the find-
ings by Solow. He found that the real GDP average rate was 2.92 per cent; 0.56
per cent caused by the growth of capital, 1.34 per cent due to increases in the
quantity of labour and 1.02 per cent to technical progress.
Most of the recent empirical studies, based on neoclassical and endogenous
growth theories, emphasize the role played by R&D spending in the evolution of
technology and economic growth. The first contributions are due to Romer
(1986) and Lucas (1988), both authors agreed that spending on R&D is acquired
through formal education, informal training and labour experience; they also
argued that these factors are conducive to higher growth rates, at least during a
transition period.
256 J.C. Trejo-García et al.
Under this research framework, the purpose arose to empirically study the
role of education and human capital in the evolution of technological change and
economic growth in Mexico. In more concrete terms, the need arises to answer
questions such as: What are the components of technological progress? What is
the role of technological innovation in technological progress and output
growth? What is the role of education in technological progress? among others.
In order to respond to this set of questions, the next section presents a model
based on Solow’s model.

12.5  Econometric model specification
To provide a basis for answering the previous questions, an econometric model
represented by a production function is proposed. It establishes the relationship
between inputs and output, and indicates the maximum product that is possible
to obtain under certain combinations of inputs. As is usual in these models, we
assume that these inputs are variable, differentiable and convertible into produc-
tion. The production function is expressed as:


where Q is the product level, X is a vector of explanatory variables of the product

and Z is a vector of parameters governing the rates at which the explanatory var-
iables are transformed into product. It is assumed that A is a well-behaved, con-
tinuous and differentiable function. A functional form that meets these
characteristics is the Cobb–Douglas production function. So:


in logarithmic terms:


This expression is also known as the log-linear model. In what follows, the func-
tional form of the model, used to empirically estimate the determinants of eco-
nomic growth, is presented. A proper way to estimate the relationship between
output growth is based on the simple Solow model, where the residue accounts
for the growth that is not explained by production factors.
The function used is a production function with Hicks’ neutral technological
progress, expressed in the following equation:


Re-expressing this equation in terms of the output growth rate, the following
equation is obtained:
Innovation and productivity in Mexico 257


The model specifies that production is boosted by increase in production factors:

R&D capital, physical capital, labour and variables that can be part of the resid-
ual, such as human capital or improvements in production factors.
The general model to be considered is the determination of the output growth
rate, which is expressed in logarithmic terms and through instrumental variables
in order to approximate the variable associated with technology. The model is as


where the left term determines the logarithm rate of output growth; the first term
on the right is the intercept, the second one is the initial value of the product, its
coefficient measures the rate of convergence of the economy under study; the
following two terms specify the participation levels of capital factors, that is, of
R&D and physical capital. The remaining terms reflect the participation of not
incorporated variables into the model.
Equation (12.20) is used to determine the share of production factors and
technological change in the growth rate of output; it is also used to compare the
performance of these factors in different sectors of the economy, as discussed

12.6 Estimation and model results

To analyse the behaviour of the economy, the model (12.20) is estimated with the
weighted least squares method. Estimates are made for two general issues; on the
one hand, the rate of output growth and, on the other, the growth rate of technolo-
gical change. Because estimation with the least squares method tends to be affected
by heteroscedasticity problems, we used the weighted least squares methodology as
a way to correct this problem in the cross section estimation. The weighted term
was assigned according to the variable that could be causing problems with the var-
iance of errors. This change does not affect the calculation of the parameters.
The estimates about determinants of economic growth are classified into four
categories. For each of these four categories, there is a subdivision by economic
sectors; manufacturing, commerce and services. The first two subdivisions cor-
respond to:

1 information for municipalities with high population density (urban) and

weighted by the product;
258 J.C. Trejo-García et al.
2 information for municipalities with low population density (rural) and
weighted by the product.

We used equation (12.20) for estimating these two cases. The specific feature of
this approach lies in using the product as a weighting variable in both cases.
Both estimates were made simultaneously with the aim of carrying out a com-
parison between estimated parameters, that is, the performance of a variable
under different scenarios can be compared. In this case, the scenarios are urban
and rural. The results of these estimates are shown in Tables 12.2 and 12.3.
Table 12.2 shows the results of econometric estimates by economic sectors,
corresponding to high population density municipalities. The most relevant
descriptive statistics of the sample are shown. First, the positive growth rates (Q)
in the three sectors is indicated, with the greatest growth in the services sector.
There is also a positive role of R&D spending (H), although at very low levels.
The level of technological effort (R/Q) is positive and with high levels in all
Regarding the average labour productivity, significant differences between
sectors is also perceived. While there is some relationship between productivity
and technological effort of the sectors, that relationship is not conclusive: that is,
not all sectors with an innovative effort above average have productivity higher
than average productivity.
The descriptive analysis supports the inclusion of other determinants in the
growth of productivity, in addition to investments in R&D, which are collected
by the variable A.
Table 12.3 shows the results of econometric estimates by economic sectors,
corresponding to low population density municipalities. The most relevant

Table 12.2 Estimate for municipalities with high density of population (urban) and
weighted by the product

Descriptive statistics (1994–2004)


Probability (*)

Sector Q0 H K L R/Q A

Manufacturing –0.0021 0.0032 0.082 0.042 0.73 0.24

0.0426 0.049 0.0073 0.006 0.000 0.000
Commerce 0.027 0.0018 0.062 0.051 0.69 1.56
0.000 0.0012 0.0452 0.000 0.053 0.000
Services –0.0053 0.0729 0.141 0.097 1.73 3.036
0.000 0.000 0.000 0.000 0.000 0.000

Source: own estimations based on Economic Census.

(*) The level of probability to reject the null hypothesis is 0.05.
Innovation and productivity in Mexico 259
Table 12.3 Estimate for municipalities with low density of population (rural) and
weighted by the product

Descriptive statistics (1994–2004)


Probability (*)

Sector Q0 H K L R/Q A

Manufacturing 1.02 0.003 0.091 0.051 0.003 1.13

0.038 0.046 0.035 0.000 0.000 0.000
Commerce 1.439 0.000 0.0383 0.098 0.001 0.614
0.000 0.518 0.0652 0.000 0.064 0.000
Services 1.135 0.038 0.121 0.063 0.013 1.026
0.000 0.000 0.000 0.000 0.000 0.000

Source: own estimations based on Economic Census.

(*) The level of probability to reject the null hypothesis is 0.05.

descriptive statistics of the sample are shown. First, the positive growth rates (Q)
in the three sectors are indicated, with the greatest growth in the services sector.
There is also a positive role of R&D spending (H), although at very low levels,
even lower than in the urban sample. The level of technological effort (R/Q) is
positive and with low levels in all sectors.
Regarding the average labour productivity, significant differences between
sectors are still present and are even deeper. While there is some relationship
between productivity and technological effort of the sectors, that relationship is
not conclusive: that is, not all sectors with an innovative effort above average have
productivity higher than average productivity. As in Table 12.2, the econometric
results support the inclusion of other determinants in the growth of productivity,
in addition to investments in R&D, which are collected by the variable A.
On other hand, when statistical data is weighted by the number of registered
units, there are two possibilities for analysis:

3 information for municipalities with high population density (urban) and

weighted by the number of registered units;
4 information of the municipalities with low population density (rural) and
weighted by the number of registered units.

As in the two cases previously discussed, the possibilities (3) and (4) are esti-
mated using the model (12.20), but differing with earlier estimates since the
weighting variable is now the number of registered units. The results of these
estimates are presented in Tables 12.4 and 12.5.
Table 12.4 shows the results of econometric estimations for the three eco-
nomic sectors considered. As in the results in Table 12.2, there is also evidence
260 J.C. Trejo-García et al.
Table 12.4 Estimate for municipalities with high density of population (urban) and
weighted by the number of registered units

Descriptive statistics (1994–2004)


Probability (*)

Sector Q H K L R/Q A

Manufacturing 1.017 0.01 0.091 0.071 0.69 0.21

0.000 0.049 0.05 0.032 0.013 0.000
Commerce 1.82 0.004 0.069 0.042 0.71 0.079
0.000 0.591 0.038 0.000 0.053 0.000
Services 3.921 0.061 0.138 0.089 1.941 0.117
0.000 0.000 0.000 0.000 0.000 0.000

Source: own estimations based on Economic Census.

(*) The level of probability to reject the null hypothesis is 0.05.

Table 12.5 Estimate for municipalities with low density of population (urban) and
weighted by the number of registered units

Descriptive statistics (1994–2004)


Probability (*)

Sector Q H K L R/Q A
Manufacturing 1.021 0.051 0.0496 0.062 0.003 0.734
0.000 0.046 0.035 0.000 0.000 0.000
Commerce 1.461 0.062 0.062 0.071 0.007 0.615
0.000 0.518 0.0652 0.000 0.064 0.000
Services 1.171 0.042 0.0158 0.056 0.011 1.029
0.000 0.000 0.000 0.000 0.000 0.000

Source: own estimations based on Economic Census.

(*) The level of probability to reject the null hypothesis is 0.05.

of a positive contribution of each of the regressors; services shows the biggest

participation as long as manufacturing sector has the smallest contribution. Sig-
nificantly, there is a high share of R&D spending (H) as well as the technolo-
gical effort level (R/Q) for all the sectors.
Finally, the results of the estimates for municipalities with low population
density and weighted by the number of units surveyed are shown in Table 12.5.
This table shows the results of econometric estimations for the three economic
Innovation and productivity in Mexico 261
sectors. There is also evidence of a positive participation of each of the regres-
sors. In this case, the commerce sector has the biggest share as long as the con-
tribution of the manufacturing sector is the smallest one. Just as in the previous
cases, there is a relative high share of spending on R&D (H) as well as the
technological effort level (R/Q) for the three sectors.

12.7 Conclusions
This chapter has presented a theoretical model that relates the growth of labour
productivity with R&D spending. The model allows estimating the rate of return
of technological capital from R&D spending flows, without the need to build the
stock of research capital.
The theoretical model is specified using an econometric model of delays dis-
tributed in time. The empirical estimates are based on cross section data. Statisti-
cal information was taken from SIMBAD. In particular, economic censuses of
1994 and 2004 were used. Statistical information is for three economic sectors –
manufacturing, commerce and services – and two groups of municipalities –
urban and rural areas – according to population density.
Regarding econometric model estimation, the method of weighted least
squares was used. The empirical results achieved are consistent with theoretical
expectations, indicating that the investment in R&D by the three economic
sectors surveyed has a positive effect and in most cases statistically significant.
At first glance, it appears that the contribution of the factors over the growth
rate of output is high for urban municipalities; in consequence, it should encour-
age private and public investment in each of the production factors, especially in
R&D. Nevertheless, in practice it is found that this does not happen, since the
share of investment in this area is lowered through time.
This might be because of the high risk associated with R&D projects and the
difficulty in obtaining full benefits from innovation. This can discourage firms
about engaging in such activities, despite the high gains expected. In addition,
firms with investment intentions could find remarkable problems in financing
their investments in R&D. More so if they have already had financing problems,
especially for small and medium enterprises.
On the other hand, we have also found a significant positive relationship
between output growth and the capital–labour ratio, and between growth and
production capacity of economic sectors. This indicates that productivity
changes are factors associated with the long and short terms. Finally, the produc-
tion function for the three sectors shows diminishing returns to scale for capital
and labour, which is consistent with results from previous studies. However, the
relationship between productivity growth and technological effort does not
change very significantly when constant returns to scale condition are imposed.
The estimated model may be affected by econometric limitations. Griliches
and Mairesse (1995) suggest that an estimated production function from micro-
data presents some problems that also arise when the production function is
transformed into logarithms. Accordingly, the empirical results obtained in this
262 J.C. Trejo-García et al.
work, although they are quite reasonable, must be contemplated with caution.
Nevertheless, studies of this type are quite useful.

1 This procedure requires making assumptions about the value of the rate of depreciation
of capital and taking an initial value of this capital
2 We use investment in R&D, and the estimate will focus on determining the rate of
return of that capital, instead of its elasticity. Studies that attempt to estimate the rate of
return on R&D expenditure from individual companies are common in other countries
3 g is the production elasticity related to R&D capital. It is given by:

Bessen, J. (2000): ‘Adoption costs and the rate of return to research and development’,
Working paper 1/00, Research on Innovation, Wallingford, PA. Online, available at: (accessed 30 February 2009).
Clark, B. and Griliches, Z. (1984): ‘Productivity growth and R&D at the business level:
Results of the PIMS data base’, in Griliches, Z. (ed.): Patents and productivity, Univer-
sity of Chicago Press, Chicago, pp. 393–416.
Denison, E.F. (1974): Accounting for United States economic growth 1929–1969, Brook-
ings Institution, Washington, DC.
Fecher, F. (1990): ‘Effects directs et indirects de la R&D sur la productivité: une analyse
de l’industrie manufacturière belge’, Cahiers Économiques de Bruxelles, vol. 128, pp.
Griliches, Z. (1979): ‘Issues in assessing the contribution of research and development to
productivity growth’, Bell Journal of Economics, vol. 10, pp. 92–116.
Griliches, Z. and Mairesse, J. (1983): ‘Comparing productivity growth: An exploration of
French and U.S. industrial and firm data’, European Economic Review, vol. 21,
pp. 89–119.
Griliches, Z. and Mairesse, J. (1995): ‘Production functions: The economic search for
identification’, documento de trabajo nº 5067, NBER, Cambridge, MA.
Hall, B.H. and Mairesse, J. (1995): ‘Exploring the relationship between R&D and pro-
ductivity in French manufacturing firms’, Journal of Econometrics, vol. 65, pp.
Jasso, J. (1998): ‘Industrial organization, productivity and strategies business in Mexico’,
Working Paper 162, Division of Economics, CIDE.
Lichtenberg, F.R. and Siegel, D. (1991): ‘The impact of R&D investment on productiv-
ity: New evidence using linked R&D-LRD data’, Economic Inquiry, vol. 29,
pp. 203–228.
Lucas, R.E. Jr (1988): ‘On the mechanics of economic development’, Journal of Mone-
tary Economics, vol. 22, pp. 3–42.
Mankiw, G., Romer, D. and Weil, D. (1992): ‘A contribution to the empirics of economic
growth’, Quarterly Journal of Economics, vol. 108, pp. 407–437.
Odagiri, H. and Iwata, H. (1986): ‘The impact of R&D on productivity increase in Japa-
nese manufacturing companies’, Research Policy, vol. 15, pp. 13–19.
Parisi, M.L., Schiantarelli, F. and Sembenelli, A. (2002): ‘Productivity, innovation
Innovation and productivity in Mexico 263
creation and absoption, and R&D: Micro evidence for Italy’, Working Paper No. 526,
Department of Economics, Boston College, Chestnut Hill, MA.
Romer, Paul M. (1986): ‘Increasing returns and long run growth’, Journal of Political
Economy, vol. 94, pp. 1002–37.
Solow, R.M. (1957): ‘Technical change and the aggregate production function’, Review
of Economics and Statistics, vol. 57, pp. 312–320.
Unger, K. (1996): ‘International competitiveness and technological development: The
Mexican manufacturing industry against trade liberalization’, Mexican Economy, New
Era, Mexico, vol. V, no. 2.
Wakelin, K. (2001): ‘Productivity growth and R&D expenditure in UK manufacturing
firms’, Research Policy, vol. 30, pp. 1079–1090.
13 The robustness of Okun’s law –
evidence from Mexico
A quarterly validation, 1985.1–2006.4 1

Eduardo Loría and Leobardo de Jesús

13.1 Introduction
In 1962 Arthur Okun found a statistical regularity of great relevance for the
United States’ economy (1947.2–1960.4), which claimed that for each percent-
age point of reduction in the unemployment rate, the real GDP would grow 3.3
per cent (Okun, 1962); and inversely, for each percentage point of increase in
output, unemployment would vary –0.3 points. This regularity is commonly
known as 3:1, and also known as the Okun law, and since then has become a
concept of great importance in modern macroeconomics. It is the consequence
of relating output increase to the unemployment rate in a bi-directional way,
coming out from three specifications: first differences, output gap and fitted trend
and elasticity.
Okun’s contribution enriches the modern macroeconomic analysis because (a)
it allows to know variation of the unemployment variation in the long run, deter-
mined by structural factors such as demographics, institutions and technology, (b)
it provides a proxy of the natural rate of unemployment and (c) it identifies that
long run economic growth is the main factor that counteracts the reduction in
employment creation capacity (see Loría and Ramos, 2007). Okun’s discovery has
great importance because of its explicative capacity of economic development:

The failure to use one year’s potential fully can influence future potential
GNP: to the extent that low utilisation rates and accompanying low profits
and personal incomes hold down investment in plant, equipment, research,
housing, and education, the growth of potential GNP will be retarded.
(Okun, 1962: 2)

From the available literature, we found that since this contribution was made,
several authors have estimated different variations of this law.2 In spite of the
relevance of the subject and that in Mexico the problem of slow growth started
since the early 1980s, it is surprising that we only found three references for the
Mexican economy: Chavarín (2001), González (2002) and Loría & Ramos (2007).
Our main purpose is to estimate the three Okun models (1962) for the
Mexican economy using quarterly data (1985.1–2006.4), in order to prove that
Okun’s law and Mexico 265
unemployment constrains the long run growth, and compare our results with
those obtained with annual series (1970–2004) by Loría & Ramos (2007). There-
fore, we corroborate that in Mexico the Okun law is validated for data of differ-
ent periodicity and length. Our results indicate that there is a bi-directional
causality relationship between the unemployment rate and output growth – in its
three variants – and that Okun’s coefficient is found in the interval of 2.3–2.5,
which coincides with Loría and Ramos (2007).
In the second section we present the original Okun results. Next, we analyse
the statistical properties of quarterly data that we use and point out the differ-
ences with those used by Loría and Ramos. In the fourth section we estimate the
three Okun models with the methodology of structural time series models, using
the Kalman filter, and contrast them with the results obtained by those authors.
Additionally, we confirm the bi-directional causality in Okun’s equations and
test cointegration for model 3. Finally, we recover the main conclusion and
outline some policy recommendations.

13.2 The Okun models

Okun (1962) used three different econometric specifications to prove that there
was a robust bi-directional statistical relationship between unemployment and
economic growth for the economy of the United States (1947.2–1960.4), which
are presented in Table 13.1.
Okun’s empirical conclusion coming out from the first two estimations is that
in the long run, unemployment reduction has a more than proportional effect on
the dynamics of GDP (1/b2).3

13.3 Mexico: output and unemployment data, 1985.1–2006.4

We use quarterly GDP (Y) data (1993 prices) and the general rate of open unem-
ployment (U) reported by INEGI (2007a, b), both without seasonal adjustments.
The latter variable differs substantially from the macroeconomic unemployment

Table13.1 Okun models

Model Estimation Okun coefficient

b2 1/b2

First differences Δ Ut = 0.3 – 0.3 yt 0.3 3.3

(1) Δ Ut = β1 + β2 yt + Єt
Output gap Ut = 3.72 + 0.36 Ytb 0.36 2.8
(2) Ut = β1 + β2 Ytb + Єt
Fitted trend and elasticity In Et = 212 + 0.4 In Yt – 0.32t 0.4–0.35 2.5–2.8
(3) In Et = β1 + β2 In Yt +β3 + Єt

Source: own estimations based on INEGI.

266 E. Loría and L. de Jesús
rate (MUR)4 of Loría and Ramos (2007), estimated for annual data (1970–2004).
As a matter of example, towards 1970 the MUR was 2.03 per cent and since
1982 it began to grow until it reached 15.8 per cent in 1988, and 28.3 per cent
towards the year 2004 (see Figure 13.1); while the general unemployment rate
used here is stationary and currently has ranged from 2 to 5 per cent, although it
reached high figures in 1995 and 1996. See Figure 13.2.

13.4 Analysis and discussion

According to our main purpose, we estimated the three Okun models inversely,
thus solving a serious econometric bias problem detected by Barreto and
Howland (1993) in Okun’s seminal article. This problem consists in estimating
the current regression and afterwards solving arithmetically for the exogenous,
just by doing algebra. Therefore, it does not matter regressing U on Y or the
other way around. By doing this Okun claims that it is possible to find economic
sense in both directions. This procedure has been followed by many authors.
Accordingly, when passing directly in estimations (1) and (2) from b2 to 1/b2
Okun was able to explain – at the same statistical level – either economic growth
or unemployment.
Nevertheless, in the original Okun’s models (1 and 2) there are two variables
and the reading must be made as usual (from the right hand side to the left hand
side), and the fact of reading inversely is not only related to the causality sense
coming out from economic theory, but also – and not less important – has to do
with the properties of a joint distribution function, which refers a conditional
specification of random variables of the kind:

Figure 13.1 Mexico: macroeconomic rate of unemployment, 1970–2004 (source:

Loría and Ramos (2007)).
Okun’s law and Mexico 267

Figure 13.2 GDP, unemployment, output gap and employment rate, 1985.1–2006.4
(source: INEGI).

Barreto and Howland outline that the correct specification depends on the spe-
cific question of interest. This task determines the regression direction:

Thus Okun’s procedure [make the bi-directional reading as of b2, our aggreg-
ate] makes sense only if the underlying structure in the model is assumed to
be stable, i.e., if the parameters of the model do not change between the
sample period and the date on which the GNP gap is to be predicted. If any of
the structural parameters have changed in the intervening time, then the
sample relationship will produce biased estimates of the GNP gap.
(1993: 4)

Thus, in order to avoid the possible problem of referred bias and since our main
purpose is to prove specifically that unemployment restricts economic growth, we
choose the direct estimation for the three Okun estimations. That is, we proceeded
by the inverse specification to that of Okun’s in the following way: Y = f(U), thus
the reading is direct in terms of our hypothesis. Likewise in Loría and Ramos
(2007), we estimated the three equations through the methodology of structural
268 E. Loría and L. de Jesús
time series models, using the Kalman filter (Kalman, 1960). See results in Table
13.a.2 and the Appendix to this chapter).
One advantage of this procedure is that the estimated parameter ‘mt captures
the long run movements of the series involved as well as the effects that b2 can’t
explain’ (Loría and Ramos, 2007: 29). Empirical evidence reports that also for
this data structure and with the inverse specification of Okun, this law was
accomplished in Mexico. The value of the coefficients is congruent with the
structure of the Mexican economy: labour intensive and low productivity.
Model 3 depicts two results with high economic meaning. On the one hand,
mt indicates the actual rate of potential output: 2.6 per cent, similar to the figure
reported by Loría et al. (2008), 2.5 per cent for the span time 1980.1–2006.4.
Likewise, from the parameter of the employment rate (E) we can calculate the
output elasticity to employment (1/2.5892 = 0.386), that even with the anticipated
methodological warnings, we can take it safely since it is congruent with the
results obtained by Loría and Ramos and other applied works such as Loría
(2006) and Hernández (1998) (see Table 13.a.2).
Tests of unit roots (see Table 13.a.1 in the Appendix to this chapter) indicate
that the GPD logarithm (ln y) and that of the employment rate (ln E) are series I(1),
while the unemployment rate (U), GDP growth (y) and output gap (Yb) are I(0).5

13.5 Conclusion
We empirically corroborated that for quarterly data and with the use of structural
time series models (using the Kalman filter), Okun’s law applies in Mexico, and
the coefficient varies in the interval 2.35–2.58, which is congruent with what
Loría and Ramos (2007) estimated; and, furthermore, that these magnitudes are
adequate for an economy that suffers from high structural unemployment and
low productivity. Likewise, causality tests in the Granger sense run in a bidirec-
tional way between unemployment and output.
In order to avoid possible biases in the estimation of the slopes of the three
models, we used inverse specifications to that of Okun’s, obtaining a direct

Table 13.2 Mexico: Okun estimations

Estimation Average

Quarterly, 1985.1–2006.4 Annual, 1970–2004 (Loría and Quarterly Annual

Ramos, 2007)

(1) yt = 1.1041 µt – 2.3538 ΔUt (1) ΔUt = 2.349 µt – 0.403 yt 2.49 2.25
(2) YB = 9.5866 µt + 2.5383 Ut (2) Ut = 14.65 µt + 0.456 YtB
(3) In Yt = 2.6115 µt + 2.5892 In Et (3) In Et = 0.481 In Yt – 2.661 µt

Source: own estimations based on INEGI.

All the models were estimated with GiveWin 2.3, module STAMP 6.0 (Koopman et al. 2000).
Okun’s law and Mexico 269
reading of the Okun coefficient and can prove in a reliable manner that unem-
ployment constrains economic growth.
This evidence supports the results of other authors for different series and
periods, which allows us to use it as a good instrument of analysis and forecast-
ing of the economic cycle; it also allows us to estimate the sacrifice rate of long
run unemployment; and moreover, shows that economic policy must focus by all
means on avoiding fluctuations in growth and at the same time reducing unem-
ployment, because this way it will stimulate growth in the long run.


Table 13.a.1 Basic statistics and unit roots, 1985.1–2006.4

Mean 3.5046 14.1067 0.0067 0.7847 4.5694 0.00019 0.051

Median 3.35 14.0745 –0.0015 –0.1517 4.571 0.00055 0.0697
Std Dev. 1.0273 0.1899 0.046 4.648 0.0107 0.00532 2.8031
Skewness 1.3711 0.044 0.1151 0.1706 –1.4092 –0.3741 –0.3777
Kurtosis 5.2431 1.7397 1.8257 1.8322 5.383 3.6816 3.2045
Jarque- 46.025 5.851 5.191 5.365 49.953 3.714 2.246
0 –0.053 –0.075 –0.068 0 –0.156 –0.325
ADF –2.692 3.9131 –4.036* –3.1532 0.1003 –4.287 –7.0484
DF-GLS –1.7765 –2.618 –3.640** –3.2306 –2.513 –2.258 –6.024
PP –3.785 3.6918 –34.907*** –4.7289 3.802 –11.704 –16.3251
KPSS 1.235 202.862 0.155 0.43211 1.214 0.044 0.173

Source: own estimations based on INEGI.

Tests are non-significant at levels. ADF with four lags and intercept are valid at 90%; DF-GLS with
four lags, trend and intercept; PP with four lags and intercept, valid at 99%. KPSS with four lags and
* With intercept only; ** with 4 lags and intercept; *** with intercept; three lags

Table 13.a.2 Granger Causality Test, 1985.1–2006.4 for an unrestricted VAR(5)

VAR model Ho: not causality x2 (5) Probability

1 Δ Ut does not cause yt 21.39 0.0007

yt does not cause ΔUt 28.42 0.0000
x2 (5)
2 Ut does not cause YtB 20.92 0.0008
YtB does not cause Δ Ut 43.88 0.0000
x2 (5)
3 In Et does not cause In Yt 27.79 0.0000
In Yt does not cause In Et 17.47 0.0037

Source: own estimations based on INEGI.

Models 1 and 3 include a GDP shock dummy and an inflation dummy for 1986.3 and 1995.2.
270 E. Loría and L. de Jesús

Figure 13.a.1 Model 1 first differences.

Table 13.a.3 Mexico: Okun’s law, 1985.1–2006.4

yt = 1.1041 µt – 2.354 ΔUt

R2 0.882962
N 3.2632 (0.1956)
DW 2.3062
r –1 –0.21520 (0.8296)
–8 0.066376 (0.9471)
Q (8,6) 21.292 (0.0016)
H –34 0.74192 (0.8058)

Source: own estimations based on INEGI.

Figure 13.a.2 Model 1: first differences (source: estimates with information from INEGI).
TGDPT: GDP growth.
Okun’s law and Mexico 271

Figure 13.a.3 Diagnostic tests: correlogram, density, QQplot, cusum residual (source:
estimates with information from INEGI).

Figure 13.a.4 Model 2: output gap.

Table 13.a.4 Model 2

YB = 9.5866 µt + 2.5383 Ut
R 2
N 4.2233 (0.1210)
DW 1.749
r –1 0.086169 (0.9313)
–8 0.076475 (0.9390)
Q (8,6) 10.342 (0.1110)
H –27 1.0967 (0.4061)

Source: own estimations based on INEGI.

Figure 13.a.5 Model 2: output gap (source: estimates with information from INEGI).

Figure 13.a.6 Diagnostic tests: correlogram, density, QQplot, cusum residual (source:
estimates with information from INEGI).
Okun’s law and Mexico 273

Figure 13.a.7 Model 3: fitted trend and elasticity.

Table 13.a.5 Model 3

In Yt = 2.6115 µt + 2.5892 In Et
R2 0.993693
N 2.7109 (0.2578)
DW 1.749
r –1 0.085145 (0.9321)
–8 0.10788 (0.9141)
Q (8,6) 10.480 (0.1058)
H –34 1.1143 (0.3771)

Source: own estimations based on INEGI.

Figure 13.a.8 Model 3: fitted trend and elasticity (source: estimates with information
from INEGI).
LGDP: GDP logarithm.
274 E. Loría and L. de Jesús

Figure 13.a.9 Diagnostic tests: correlogram, density, QQplot, cusum residual (source:
estimates with information from INEGI).
LGDP: GDP logarithm.

1 We thank technical assistance from Manuel García R. and Jorge Ramírez, and com-
ments from Armando Sánchez. As usual, the responsibility is only ours.
2 Upon reviewing the literature, available works can be classified in two main categories:
(a) theoretical–empirical studies, in the sense that they review and discuss the estima-
tion methods of Okun’s law (in this respect see: Barreto and Howland, 1993; Altig et
al., 1997; Attfield and Silverstone, 1998; Sögner and Stiassny, 2000; Harris and Silver-
stone, 2001; Crespo, 2003; Friedman and Michael, 1974; Lang and de Peretti, 2002;
Prachowny, 1993; Weber, 1995; Schorderet, 2001; Knoester, 1986; Paldam, 1987), and
(b) empirical studies, whose main purpose is to estimate the Okun’s coefficients for
some countries, even at the level of states or regions, in order to know the existing
interrelations between different countries by identifying reciprocities between unem-
ployment and output (see: Abril et al., 1996; Adanu, 2002; Arias et al., 2002; Garavito,
2002; Lemois, 2003; Murillo and Usabiaga, 2002; Freeman, 2001; Lee, 2001; Moosa,
1997 and Schnabel, 2002).
3 In Loría and Ramos (op. cit.) the economic implications of the results depicted in this
table were carefully analysed. In addition, it is worth mentioning that (2) established
the natural rate of unemployment (3.72 per cent) and (3) the output elasticity to
Okun’s law and Mexico 275
where PEA = economically active population (Conapo, 2006), PO = employees (mil-
lions of persons) in the formal sector (INEGI, 2007a).
5 This way, the problem of spurious regression could only exist in model 3 and we fol-
lowed the Johansen procedure (1988) to discard it. Accordingly, with a confidence
level of 99 per cent we obtained a cointegrating vector with economic sense; statistic
trace 27.78 (24.6), adjustment coefficient –0.6275 (standard error: 0.126).

Abril, J.C., H.D. Ferullo and A. Gaínza (1996). ‘Estimación de la relación de Okun:
Argentina 1980–1996’. Facultad de Ciencias Económicas, Universidad Nacional de
Tucumán, Argentina.
Adanu, K. (2002). ‘A Cross-Province Comparison of Okun’s Coefficient for Canada’.
Department of Economics of Victoria, BC, Canada.
Altig, D., T. Fitzgerald and P. Rupert (1997). ‘Okun’s Law Revisited: Should we Worry
about Low Unemployment?’ Economic Commentary. Federal Reserve Bank of Cleve-
Arias, E., A.C. Kikut and J. Madrigal (2002). ‘Estimación de la Ley de Okun para Costa
Rica’. Departamento de Investigación Económica, Banco Central de Costa Rica.
Attfield, C. and B. Silverstone (1998). ‘Okun’s Law, Cointegration and Gap Variables’,
Journal of Macroeconomics. Vol. 20, No. 3.
Barreto, H. and F. Howland (1993). ‘There are Two Okun’s Law Relationships between
Output and Unemployment’. Wabash College, Crawfordsville, USA.
Chavarín, R. (2001). ‘El costo del desempleo medido en producto: Una revisión empírica
de la Ley de Okun para México’, El Trimestre Económico. Vol. LXVIII, No. 270,
FCE, México.
Conapo (2006). Indicadores demográficos básicos. Consejo Nacional de Población.
Online, available at: (accessed Sep-
tember 2007).
Crespo, J. (2003). ‘Okun’s Law Revisited’, Oxford Bulletin of Economics and Statistics.
Vol. 65, No. 4.
Freeman, D.G. (2001). ‘Panel Tests of Okun’s Law for Ten Industrial Countries’, Eco-
nomic Inquiry. Vol. 39, No. 4.
Friedman, B. and M.L. Wachter (1974). ‘Unemployment: Okun’s Law, Labor Force, and
Productivity’, Review of Economics and Statistics. Vol. 56, No. 2.
Garavito, C. (2002). ‘La Ley de Okun en el Perú: 1970–2000’. Pontificia Universidad
Católica de Perú.
González, J.A. (2002). ‘Labor Flexibility in Thirteen Latin American Countries and the
United States: Revisiting and Expanding Okun Coefficients’, DREDPR. Working Paper
No. 136. Stanford University.
Granger, C.W.J. (1969). ‘Investigating Causal Relations by Econometric Models and
Cross-Spectral Methods’, Econometrica. Vol. 37, No. 3.
Harris, R. and B. Silverstone (2001). ‘Testing for Asymmetry in Okun’s Law: A Cross-
Country Comparison’, Economics Bulletin. Vol. 5, No. 2.
Hernández, E. (1998). ‘Apertura comercial, productividad, empleo y contratos de trabajo
en México’, in V. Tokman and D. Martínez (ed.). Productividad y empleo en la aper-
tura económica. OIT. Lima, Perú.
INEGI (2007a). ‘Encuesta Nacional de Ocupación y Empleo’. Indicadores Económicos de
Coyuntura. Instituto Nacional de Estadística, Geografía e Informática, México. Online,
276 E. Loría and L. de Jesús
available at:
(accessed May 2007).
INEGI (2007b). ‘Sistema de Cuentas Nacionales de México’. Indicadores Económicos de
Coyuntura. Instituto Nacional de Estadística, Geografía e Informática, México. Online,
available at:
(accessed May 2007).
Johansen, S. (1988). ‘Statistical Analysis of Cointegrating Vectors’, Journal of Economic
Dynamic and Control. Vol. 12, No. 2–3.
Kalman, R. (1960). ‘A New Approach to Linear Filtering and Prediction Problems’,
Journal of Basic Engineering. Vol. 82.
Knoester, A. (1986). ‘Okun’s Law Revisited’, Weltwirtschaftliches Archiv. Vol. 122, No.
Koopman, S.J., A.C. Harvey, J.A. Doornik and N. Shephard (2000). ‘Structural Time
Series Analyser, Modeller and Predictor’. Timberlake Consultants.
Lang, D. and C. de Peretti (2002). ‘A Strong Hysteretic model for Okun’s Law: Theory
and Preliminary investigation’. Université de la Méditerranée, France.
Lee, J. (2001). ‘The Robustness of Okun’s Law: Evidence from OECD Countries’,
Journal of Macroeconomics. Vol. 22, No. 20.
Lemois, F.A. (2003). ‘Estimaciones de la Ley de Okun para Puerto Rico’. Junta de Plan-
ificación, Oficina de la Gobernadora, Estado Libre Asociado de Puerto Rico.
Loría, E. (2006). Eudoxio: modelo macroeconométrico de la economía mexicana.
UNAM. México.
Loría, E. and M. Ramos (2007). ‘La ley de Okun. Una relectura para México,
1970–2004’, Estudios Económicos. Vol. 22, No. 1. Enero-Junio. El Colegio de México,
Loría, E., M.G. Ramos and L. de Jesús (2008). ‘Producto potencial y ciclos económicos
en México, 1980.1–2006.4’, Estudios Económicos. Vol. 23, No. 1. Enero-junio. El
Colegio de México. México.
Moosa, I.A. (1997). ‘A Cross-Country Comparison of Okun’s Law Coefficient’, Journal
of Comparative Economics. Vol. 24, No. 3.
Murillo, I.P. and C. Usabiaga (2002). ‘Estimaciones de la tasa de paro de equilibrio de la
economía española a partir de la Ley de Okun’. Universidad de Extremadura y Univer-
sidad Pablo de Olavide, España.
Okun, A. (1962). ‘Potential GNP: Its Measurement and Significance’, in J. Pechman (ed.)
(1983), Economics for Policymaking. MIT Press. Cambridge, MA.
Paldam, M. (1987). ‘How Much does One Percent of Growth Change the Unemployment
Rate?’ European Economic Review. Vol. 31, No. 1–2.
Prachowny, M. (1993). ‘Okun’s Law: Theoretical Foundations and Revised Estimates’,
Review of Economics and Statistics. Vol. 75, No. 2.
Schnabel, G. (2002). ‘Output Trends and Okun’s Law’, BIS Working Papers. Monetary
and Economic Department, Bank for International Settlements, 111.
Sögner, L. and A. Stiassny (2000). ‘A Cross-Country Study on Okun’s Law’, Growth and
Employment in Europe: Sustainability and Competitiveness. University of Economics
and Business Administration, Vienna.
Schorderet, Y. (2001). ‘Revisiting Okun Law’s: An Hysteretic Perspective’, Discussion
Papers. Department of Economics, University of California, San Diego.
Weber, Christian E, (1995). ‘Cyclical Output, Cyclical Unemployment, and Okun’s Coef-
ficient: A New Approach’, Journal of Applied Econometrics. Vol. 10, No. 4.

Note: Page numbers in italics denote tables, those in bold denote figures.

adverse effects 199 ISI 136–137, 139, 142

Andean Community (CAN) 111 manufacturing exports 143
development 113, 115–116 model of post-war development 141
membership and representation 116–117 monetary policy 141
membership evolution/type of output decomposition 139
organisation 112 Peronism 136
see also gravity model; regional public capital 142
integration agreements (RIAs) real exchange rate 140
Andean Pact see Andean Community ‘Rodrigazo’ 137
(CAN) role of exports 138–139
Anderson, J.E. 119 role of state 141–144
Arbache, J.S. 198, 199, 205 stabilisation 137
Argentina summary and conclusions 145–146
business cycle 136 trade balance–GDP ratio 39
context of decline 133–134 wage share 145
debt 142–143 automatic balancing mechanism 45
debt sustainability indicators 143
decomposition of elements of demand balance of payments 17
138 balance of trade 17
development gap 41, 42, 44 Barreto, H. 267
economic performance 135–137 Barro, R. 12
exports 143 Bergstrand, J.H. 119, 155
external constraints 138–141 Bertola, G. 16
fiscal policy 140 Bessen, J. 251
foreign direct investment (FDI) 141–144 Bhagwati, J. 7, 9
foreign exchange rate 139, 140 bias, in favour of rich countries 21
GDP 32 Bidlingmaier, T. 68
Gini coefficient 144 Bolivia, exports 117
gross investment and savings 45 Bougheas , S. 127
gross investment coefficient 44–45 Bourguignon, F. 13
growth and productivity 135 Brady Plan 233
income elasticity of imports 48 Brazil
income per capita 133 devaluation 103
industrialisation 141–142 development gap 41, 42, 44
inequality 144–145 FDI inflows and exchange rate 104
inflation rate 35 GDP 32
investment composition 142 gross investment and savings 46
278 Index
Brazil continued China
gross investment coefficient 45 development gap 44, 44
income elasticity of imports 48 economic growth 31
inflation rate 35 GDP 34
inward FDI 90–103 gross investment and savings 47
inward FDI by regions 99–100 gross investment coefficient 45
inward FDI flows 97 income elasticity of imports 49
inward FDI stock 97 inflation rate 37
sectoral changes to inward FDI 96–97 poverty 11
trade balance–GDP ratio 39 price stability 37–38
see also Mercosur; regional integration trade balance–GDP ratio 41
agreements (RIAs) Clark, B. 251
Bretton Woods, effect of collapse classical economics 26–27
139–140 classical international trade theory 60
Bucheli, M. 171 Cobb–Douglas production function 252
business and finance income (B&F) Colombia, exports 117
210–212 common external tariff (CET)
Andean Community (CAN) 115–116
capital account liberalisation 48 Mercosur 112, 113
capital flows, changed structure 89 comparative advantage 52
capital goods, differentiated 61 consumer’s decision problem 223–226
capital market, liberalisation 45 consumption dynamics 228–230
‘chance’ investments 21 Convertibility 137, 141
Chen, S. 10 Convertibility Plan 139
Chile countercyclical policy 29
bilateral agreements 150, 165 crisis-proofing 29–30
data and variables 159–161 current account balance 17–18
development gap 41–42, 43 Cyrus, T.L. 121
development strategy 151
economic growth 31 Deardoff, A.V. 120
exports 150–151 debt
FDI inflows 150–155, 151 and exchange rate depreciations 244
FDI inflows by origin 154, 155 Mexico 240, 240
FDI inflows by sector 153 debt ratio 247
GDP 33 decline of labour market institutions
gross investment and savings 46 argument 215
gross investment coefficient 45 Denison, E.F. 255
income elasticity of imports 49 devaluation, Brazil 103
inflation rate 36 development gap
methodology 155–159 Argentina 41, 42, 44
national economy, sectors 150 Brazil 41, 42, 44
random effects estimation of the Chile 41–42, 43
augmented equation 164 China 44, 44
random effects estimation of the drivers 44–45
baseline equation 162 evolution of 38–50
research study 155–165 income elasticity of demand for imports
results 161–165 48–49
sources of FDI 154–155 Mexico 41, 43, 44
summary and conclusions 165 development, trade strategy for 20–23
summary statistics 161 developmental industrial policy 50–54
trade balance–GDP ratio 40 Developmentist State 52
variables 160 Direction of Trade, IMF data set 159
Chilean Foreign Investment Committee Doha trade negotiations 20–21, 168
159, 161 Dollar, D. 12, 20
Index 279
Dow Jones Index composite average and Faini, R. 16
FDI inflows 103 Fecher, F. 251
Dowrick, S. 13, 15 Ffrench-Davis, R. 44
Dunning, J.H. 83 financial fragility hypothesis 233–237
see also Mexico: financial fragility
East Asia, inward FDI 88 financial liberalisation 48
economic growth 31 firm size 142
calculation 220 firms, typology of growth rate 236
and trade liberalisation 7 flexibility, labour market 30
Economic Partnership Agreements (EPAs) Foley, D. 236
21–22 foreign debt crisis, 1982 38, 137
economy, main characteristics 221–223 as catalyst for change 28
Ecuador, exports 118 Chilean response 44
education effects in Mexico 195
changing levels with changing income foreign direct investment as percentage of
202 GDP 92
and income 201, 216 foreign direct investment (FDI) 61
and income bill share per sector 203 Argentina 141–144
and income by sector 203 determinants 84
and income distribution 200–202 growth 149
education investment, and human capital 63 inward stock as percentage of GDP 94
Edwards, S. 20 inward stocks 93
Emmert, C.F. 88 and regional integration agreements
European Union (RIAs) 82–89
Economic Partnership Agreements see also inward FDI
(EPAs) 21–22 foreign direct investment, net inflows
and Mercosur 113 Mexico, Brazil and middle income
exchange rate-based stabilisation 34 countries 91
exchange rate depreciations, and debt 244 Mexico, Brazil and rest of world 90
exchange rate fluctuations, pass-through foreign exchange rate
effect 37 Argentina 139
export performance 15 and stabilisation 139, 140, 141
exports Frankel, J. 121
Argentina 114, 138–139, 143 free-market liberalism 220
Bolivia 117 see also Mexico: free-market liberalism
Brazil 114 study
Chile 150–151 free trade, assumptions underlying 8
Colombia 117
as drivers of growth 19 GDP
Ecuador 118 Argentina 32
Paraguay 115 Brazil 32
Peru 118 Chile 33
Uruguay 115 China 34
Venezuela 119 Mexico 33
extended neoclassical growth model Gini coefficient
countries in sample 68 Argentina 144, 144
data 67–69 decomposition by income source
derivation 62–66 210–212, 211, 213
empirical application 66–69 for global inequality 13
panel data regressions of per capita international and global comparison
income 71, 76–77 14
results 69 global inequality 13
summary and conclusions 75 global price variations, simulations
extra-regional effects 85, 128 187–189
280 Index
globalisation, and income distribution growth models 9
195–196 growth performance 19–20
Goldberg, P. 11 growth rate of firms, typology 236
Goldin, C. 199, 216 Gupta, P. 198
Golley, J. 13, 15
Gordon, J. 198 Hall, B.H. 251–252
governments, role in reform 30 Hamilton, A. 10
Gradin, C. 171 Hausman Exogeneity Test 126
graphic evolution of gravity equation fixed Hausmann, R. 21
effects for Mercosur and CAN 125 Heckscher–Ohlin theorem 8, 11, 27, 121,
gravity equation for the panels of 171–172
Mercosur and CAN: fixed effects hedge financing 244, 247, 248
124 hegemonic tie 135
gravity equation for the panels of Helpman, E. 119, 121
Mercosur and CAN: random household composition 216
effects 123 Howland, F. 267
gravity model 110–112, 118–121, 155 human capital accumulation 60–61, 63–64
distance as the only regressor in the human capital, and income 76–77
gravity equation 127
exogeneity Hausman Test for distance in IADB 88
the gravity equation 126 import growth 16–17
graphic evolution of gravity equation import-substitution industrialisation (ISI)
fixed effects for Mercosur and Argentina 136–137, 139, 142
CAN 125 Mexico 195
gravity equation for the panels of Impossible Trinity proposition 34
Mercosur and CAN: fixed effects income
124 bill share by education and sector 203
gravity equation for the panels of change with changing education levels
Mercosur and CAN: random 202
effects 123 and education 201
hypotheses 119 and education by sector 203
method 122–123 income distribution 13
research study 121–127 additional forms 210–212
results 123–125 Argentina 144–145
robustness and specification 126–127 as cyclical 215–216
summary and conclusions 127–128 and education 200–202
theoretical support 119–120 and globalisation 195–196
Great Britain 10 individual 200–205
Greenaway, D. 20 and market openness 212–214, 214
Griliches, Z. 250, 251, 252, 261 and trade liberalisation 7
gross investment and savings see also inequality
Argentina 45 income distribution targets 30
Brazil 46 income elasticity of demand for imports
Chile 46 development gap 48–49
China 47 effects of trade liberalisation 16–17
Mexico 47 income elasticity of imports
gross investment coefficient 44–45 Argentina 48
Argentina 44–45 Brazil 48
Brazil 45 Chile 49
Chile 45 China 49
China 45 Mexico 49
Mexico 45 income inequality
growth and the balance of payments, trade within countries 11–12
off 18–19, 18 Uruguay 185
Index 281
industrialisation, Argentina 141–142 Kalecki, M. 135
inequality Katz, L.F. 199, 216
Argentina 144–145 Keynes, J.M. 8, 54, 233
driving factors 215 Kim, Y.-H. 87
effects of liberalisation 195–196 knowledge, diffusion 61
household vs. individual 210 knowledge spillover 63, 64
international and global 13–15 Kraay, A. 12, 20
Mexico Krugman, P. 119, 121
see Mexico: inequality study Krugman’s trade theory 9
Uruguay 171, 184, 185 Kuczynski, P.P. 29–30, 51
see also income distribution
infant country protection 22 labour costs, and inward FDI 103
inflation 31–38 labour market, flexibility 30
inflation rate law of comparative advantage 8
Argentina 35 learning-by-doing, and human capital 64,
Brazil 35 74, 78
Chile 36 Levy Yeyati, E. 88
China 37 liberalising reforms 30
Mexico 36 Lichtenberg, F.R. 251
innovation Lomé Convention 22
encouraging 21 Loría, E. 265, 267–268
Schumpeterian approach 50–51 lost decade 29, 137
institutions, building 30 Lucas, R. 60–61, 255
international capital markets 34, 37
international inequality 13 macroeconomic performance, before and
international trade, theoretical aspects after structural reform 30–38
60–62 Mah, J.S. 16
Inverse Hyperbolic Sine Function 157 Mairesse, J. 251–252, 261
inward FDI Mandelson, P. 11
Brazil, by regions 99–100, 99–100 manufacturing exports, Argentina 143
changes by home country/region 97–100 market-incentive mechanisms 30
changes by sector 95–97 market openness and income distribution
changes in inflows and FDI stock 90–95 212–214, 214
Chile 150–155, 151 market-seeking FDI 85
Chile, by origin 154, 155 Martinez-Zarzoso, I. 127
Chile, by sector 153 McKinnon’s theory 27
and Dow Jones Index composite average Medvedev, D. 83
103 Melo, O. 16
empirical evaluation 90–103 Mercosur 87–88, 91, 92, 93, 96–97, 104, 105
flows, Brazil 97 establishment 110–111
flows, Mexico 96 impact on trade flows 112–113
and labour costs 103 membership and representation 112–113
Mexico, by regions 98–99, 98 membership evolution/type of
possible explanations for increase organisation 112
88–89, 100–103 as model of integration 113
stock, Brazil 97 price changes, Uruguay 179
stock, Mexico 95 tariffs 176
see also foreign direct investment (FDI) Uruguay 170
inward potential index 101 see also gravity model; regional
Iwata, H. 251 integration agreements (RIAs)
mergers and acquisitions 90, 91, 149
Johnson, H. 8 Mexico
average hourly income per level of
Kaldoor-Verdoorn’s Law 135 education 201
282 Index
Mexico continued Mexico: free-market liberalism study
changes in income and education levels consumption dynamics 228–230
202 optimal consumption shares, parameters,
debt 240 and estimates 231
development gap 41, 43, 44 overview 220–221
effects of trade liberalisation 11 simulation exercise 230–231
FDI inflows and exchange rate 104 summary and conclusions 232
GDP 33 wealth dynamics 226–228
GDP, unemployment, output gap and Mexico: inequality study
employment rate, 1985.1–2006.4 additional forms of income distribution
267 210–212
gross investment and savings 47 data source 196
gross investment coefficient 45 decomposition of household Gini by
income elasticity of imports 49 income source 213
inflation rate 36 decomposition of income by economic
inward FDI 90–103 sectors 202–205
inward FDI by regions 98 econometric analysis 205–210
inward FDI flows 96 factors driving inequality 215
inward FDI stock 95 Gini decomposition by income source
macroeconomic rate of unemployment, 210–212, 211
1970–2004 266 hourly income and educational
observed growth rate from 1930 to 2002 attainment by sector 203
230 household vs. individual inequality 210
Okun estimations 268 income bill share by sector and
output and unemployment data 265–266 education level 203
regional integration agreements (RIAs) individual income distribution 200–205
86–87 market openness and income
sectoral changes to inward FDI 95–96 distribution 212–214, 214
trade balance–GDP ratio 40 methodology 200
see also North American Free Trade overview 195–197
Organization (NAFTA); Okun returns to education 200–202
models; regional integration returns to education: labour income 209
agreements (RIAs) returns to labour by sectors 205–210
Mexico: financial fragility sector performance: labour income 206,
average values for sample of firms 239 207
dependent variables 243, 245, 246 skill demand 204–205
empirical evidence 238–248 summary and conclusions 214–216
firms’ assets according to financial theoretical debate 197–199
structure 242 Mexico: innovation and productivity
firms’ composition according to background to study 251–252
financial structure 241, 242 data 250, 254
growth and firms’ indebtedness 240 econometric model 255–256
growth rates 239 econometric model specification 256–257
growth, returns and interest rate: estimate for rural municipalities
aggregate levels 238 weighted by product 259
investment 239 estimate for rural municipalities
methodology 238 weighted by registered units 260
net wealth, growth and debt 240 estimate for urban municipalities
overview 233 weighted by product 258
see also financial fragility hypothesis estimate for urban municipalities
private debt 239 weighted by registered units 260
probability of hedge, speculative and estimation and model results 257–261
Ponzi finance 247 methodology 251
summary and conclusions 248–249 overview 250–251
Index 283
summary and conclusions 261–262 Pacheco-López, P. 16–19, 48
theoretical aspects 252–255 Parikh, A. 17
variables 255 Parisi, M.L. 252
Milanovic, B. 12, 13 Pavcnik, N. 11
Miles, D. 171 per capita income (PCY) 13, 15
Mill, J.S. 9 Perón, Juan Domingo 135
Mincerian earning functions 205 Peru, exports 118
monetary consensus model 34–35 Pissarides, C.A. 199, 216
monetary policy, Argentina 141 Ponzi finance 236, 239, 244, 247, 248
Morrisson, C. 13 Portantiero, J.C. 135, 136
NAFTA 86–87, 91, 93, 95, 104, 105 within countries 10–11
see also regional integration agreements effects of trade liberalisation 11
(RIAs) and trade liberalisation 7
neoclassical theory of international trade Uruguay 171, 184
27, 59 Uruguay, before and after reform 184,
new growth theory 81 185
new regionalism 81, 82–83 pragmatism 48
North American Free Trade Organization price stability 31–32
(NAFTA) 86–87, 91, 93, 95, 104, China 37–38
105 Mexico 239
see also regional integration agreements primary-commodity dependence 21
(RIAs) privatisation 101
North–South integration 81–82 privatised firms, transaction values of
Nowak-Lehmann, F. 127 cross-border M&As 101
Ocampo, J.A. 28–29 arguments for 8
Odagiri, H. 251 historical context 10
Okun, A. 264, 266
Okun models 265 R&D sector, increasing human capital
analysis and discussion 266–268 productivity 61
basic statistics and unit roots, Ramirez, M.D. 88
1985.1–2006.4 269 Ramos, M. 265, 267–268
diagnostic tests 271, 272, 274 Ravallion, M. 10, 11
econometric bias 266 regional integration agreements (RIAs)
Granger Causality Test 269 Brazil 87–88
Mexico: Okun’s law, 1985.1–2006.4 270 effect on FDI inflows 93
model 1: first differences 270 evaluation of inward FDI 90–103
Model 1 first differences 270 extra-regional effects 128
model 2 271 FDI stock 93–94
model 2: output gap 271, 272 influence 89
model 3 273 inter-regional effects 85
model 3: fitted trend and elasticity 273 intra-regional effects 84–85, 128
summary and conclusions 268–269 and inward FDI 82–89
Okun’s law 264–265 Mexico 86–87
Open Regionalism 111 North–South integration 81–82
openness 20 overview 81–82
Argentina 141 possible dynamic effects 86
and income distribution 212–214 possible static effects 82–84
and inequality 196 summary and conclusions 103–106
orthodox trade theory see also Andean Community (CAN);
problems of 8–10 Mercosur; North American Free
and wage inequality 11–12 Trade Organization (NAFTA)
Oxfam 22 regional production networks 85
284 Index
regionalisation, Mercosur countries ‘tariff-jumping’ FDI 84, 85, 87
112–113 tariff reduction, poverty and inequality
rents, reduction in traded sector 199, 204, effects 189–190
214 tariff reduction, poverty and inequality
Ricardo, D. 8, 26–27, 60 effects
rise of service argument 196 see also Uruguay
rise of services 197–199, 204, 215 tariffs
Robbins, D.J. 198 impact on prices of traded goods 173
Rodriguez, F. 20 Uruguay 169
Rodrik, D. 8, 20–21, 22, 38, 51 see also Uruguay
Rojas-Suárez, L. 37 technological innovation, effects of 52–53
Romer, P. 61, 255 technological progress 26
Rossi, M. 171 technological transfer, and human capital
rules of origin (RoO) 85–86 75
technology, and income distribution 216
Sachs, J. 20 temporary adverse effects 199
Sala-i-Martin, X. 13 tequila crises 92, 93
Samuelson, P. 8 theory of technological innovation 255
Santos-Paulino, A. 15–16, 17 Thirlwall, A.P. 15–19, 48
Schiff, M. 81, 82, 87 Tinbergen, J. 155
Schumpeter’s dynamic approach 50–51, trade balance 17–18, 38
52, 54 trade balance–GDP ratio
self discovery 21 Argentina 39
services sector 89 Brazil 39
Shackle, G. 45 Chile 40
share of cross-border M&A sales in total China 41
FDI inflows in Mexico and Brazil Mexico 40
102 trade–development connection 26
Siegel, D. 251 trade flows
Sinha, A. 198–199 gravity model 110–112
skill-biased technological change (SBTC) impact of Mercosur 112–113
195–196 trade integration, effect on income 63
skill demand 204–205 trade liberalisation
skill-enhancing trade hypothesis (SETH) effects on inequality 195–196
196, 198, 199, 204, 210, 215 overview 7–8
Smith, Adam 10, 26 temporary adverse effects 199
Solow model 9 trade performance 15–19
Solow, R.M. 250, 255 trade reforms 168–169
specialisation 21, 60 see also Uruguay
speculative financing 239, 244, 247, 248 trade, static and dynamic gains 9
Spilimbergo, A. 12 trade strategy, for development 20–23
stabilisation trade-to-GDP ratio 38
Argentina 137 traded sector, rent reduction 199, 204, 214
and foreign exchange rate 139, 140, 141 traditional theory of international trade 60
stabilisation policy 29–30 transaction values of cross-border M&As
standard theory 197–198 privatised firms 101
Stiglitz, J. 7, 9, 10, 19, 22 transatlantic consensus 196
Stolper-Samuelson theorem (SST) 195, transfer income 212
197–198 trap of literacy 74, 78
structural change 38, 44–45 Treaty of Asunción 110–111
structural reform, effects on Trujillo Protocol 115–116
macroeconomic performance Tuman, J.P. 88
structuralist economic theory 38 UNCTAD 18, 83, 84, 88, 89, 91, 101
Index 285
unemployment, as grounds for protection 8 trade reform, effects of 171–177
Uruguay trade reforms 169–170
compensating variation 180, 182, 183 unit-root test: ADF 186
conclusions 189–190 unit-root test: tradable and non-tradable
consumption effect 177 prices 181
data 190 wage-price elasticities 180–183
domestic prices and labour income USA 10
Engle-Granger: cointegration test 187 Venezuela, exports 119
estimation of effects of national trade vertical FDI 85
reform 176–184 Vogt, M.G. 16
estimation of total effect 183
external trade reform, effects of Wacziard, R. 20
174–176, 184–189 wage inequality 11
income inequality 185 wage-price elasticities, Uruguay 180–183
inequality 171, 184 wage share, Argentina 145
international and domestic prices Wakelin, K. 252
174–175 Warner, A. 20
intra and extra Mercosur trade flows 170 Washington Consensus
Mercosur 170 case for industrial policy 50–53
methodology 171–176 development gap 38–50
policy implications 189–190 economic growth 31
poverty 171, 184 framework 28–30
poverty and inequality effects of inflation 31–38
liberalisation 188 macroeconomic performance 30–38
poverty, before and after reform 184, overview 27–28
185 summary and conclusions 53–54
price changes from Mercosur 179 trade liberalisation and balance 38
price transmission 184–185 wealth dynamics 226–228
prices and income 174 Wealth of Nations 10
prices co-integration 182 Welch, K. 20
prices of traded goods and price of non- Wicksell, K. 34
traded goods 173 Williamson, J. 28, 29–30, 31, 50, 51, 53,
probability of employment after Free 54
Trade Agreement with USA 188 Wincoop, E.V. 120
research study 171–189 Winters, A. 11
selection models estimation 186 Winters, L.A. 81, 82, 87
tariff structure 178 World Trade Organization (WTO) 21,
tariffs 169 22–23
tariffs and non-traded goods 177–180
tariffs and prices of traded goods 173 Yacimientos Petroliferos Fiscales (YPF)
tariffs and traded goods 176–177 141
trade openness coefficient 170